linux-sg2042/drivers/scsi/be2iscsi/be_main.c

5871 lines
164 KiB
C
Raw Normal View History

/*
* Copyright 2017 Broadcom. All Rights Reserved.
* The term "Broadcom" refers to Broadcom Limited and/or its subsidiaries.
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License version 2
* as published by the Free Software Foundation. The full GNU General
* Public License is included in this distribution in the file called COPYING.
*
* Contact Information:
* linux-drivers@broadcom.com
*
*/
#include <linux/reboot.h>
#include <linux/delay.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 16:04:11 +08:00
#include <linux/slab.h>
#include <linux/interrupt.h>
#include <linux/blkdev.h>
#include <linux/pci.h>
#include <linux/string.h>
#include <linux/kernel.h>
#include <linux/semaphore.h>
#include <linux/iscsi_boot_sysfs.h>
#include <linux/module.h>
#include <linux/bsg-lib.h>
#include <linux/irq_poll.h>
#include <scsi/libiscsi.h>
#include <scsi/scsi_bsg_iscsi.h>
#include <scsi/scsi_netlink.h>
#include <scsi/scsi_transport_iscsi.h>
#include <scsi/scsi_transport.h>
#include <scsi/scsi_cmnd.h>
#include <scsi/scsi_device.h>
#include <scsi/scsi_host.h>
#include <scsi/scsi.h>
#include "be_main.h"
#include "be_iscsi.h"
#include "be_mgmt.h"
#include "be_cmds.h"
static unsigned int be_iopoll_budget = 10;
static unsigned int be_max_phys_size = 64;
static unsigned int enable_msix = 1;
MODULE_DESCRIPTION(DRV_DESC " " BUILD_STR);
MODULE_VERSION(BUILD_STR);
MODULE_AUTHOR("Emulex Corporation");
MODULE_LICENSE("GPL");
module_param(be_iopoll_budget, int, 0);
module_param(enable_msix, int, 0);
module_param(be_max_phys_size, uint, S_IRUGO);
MODULE_PARM_DESC(be_max_phys_size,
"Maximum Size (In Kilobytes) of physically contiguous "
"memory that can be allocated. Range is 16 - 128");
#define beiscsi_disp_param(_name)\
static ssize_t \
beiscsi_##_name##_disp(struct device *dev,\
struct device_attribute *attrib, char *buf) \
{ \
struct Scsi_Host *shost = class_to_shost(dev);\
struct beiscsi_hba *phba = iscsi_host_priv(shost); \
return snprintf(buf, PAGE_SIZE, "%d\n",\
phba->attr_##_name);\
}
#define beiscsi_change_param(_name, _minval, _maxval, _defaval)\
static int \
beiscsi_##_name##_change(struct beiscsi_hba *phba, uint32_t val)\
{\
if (val >= _minval && val <= _maxval) {\
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,\
"BA_%d : beiscsi_"#_name" updated "\
"from 0x%x ==> 0x%x\n",\
phba->attr_##_name, val); \
phba->attr_##_name = val;\
return 0;\
} \
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT, \
"BA_%d beiscsi_"#_name" attribute "\
"cannot be updated to 0x%x, "\
"range allowed is ["#_minval" - "#_maxval"]\n", val);\
return -EINVAL;\
}
#define beiscsi_store_param(_name) \
static ssize_t \
beiscsi_##_name##_store(struct device *dev,\
struct device_attribute *attr, const char *buf,\
size_t count) \
{ \
struct Scsi_Host *shost = class_to_shost(dev);\
struct beiscsi_hba *phba = iscsi_host_priv(shost);\
uint32_t param_val = 0;\
if (!isdigit(buf[0]))\
return -EINVAL;\
if (sscanf(buf, "%i", &param_val) != 1)\
return -EINVAL;\
if (beiscsi_##_name##_change(phba, param_val) == 0) \
return strlen(buf);\
else \
return -EINVAL;\
}
#define beiscsi_init_param(_name, _minval, _maxval, _defval) \
static int \
beiscsi_##_name##_init(struct beiscsi_hba *phba, uint32_t val) \
{ \
if (val >= _minval && val <= _maxval) {\
phba->attr_##_name = val;\
return 0;\
} \
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,\
"BA_%d beiscsi_"#_name" attribute " \
"cannot be updated to 0x%x, "\
"range allowed is ["#_minval" - "#_maxval"]\n", val);\
phba->attr_##_name = _defval;\
return -EINVAL;\
}
#define BEISCSI_RW_ATTR(_name, _minval, _maxval, _defval, _descp) \
static uint beiscsi_##_name = _defval;\
module_param(beiscsi_##_name, uint, S_IRUGO);\
MODULE_PARM_DESC(beiscsi_##_name, _descp);\
beiscsi_disp_param(_name)\
beiscsi_change_param(_name, _minval, _maxval, _defval)\
beiscsi_store_param(_name)\
beiscsi_init_param(_name, _minval, _maxval, _defval)\
DEVICE_ATTR(beiscsi_##_name, S_IRUGO | S_IWUSR,\
beiscsi_##_name##_disp, beiscsi_##_name##_store)
/*
* When new log level added update the
* the MAX allowed value for log_enable
*/
BEISCSI_RW_ATTR(log_enable, 0x00,
0xFF, 0x00, "Enable logging Bit Mask\n"
"\t\t\t\tInitialization Events : 0x01\n"
"\t\t\t\tMailbox Events : 0x02\n"
"\t\t\t\tMiscellaneous Events : 0x04\n"
"\t\t\t\tError Handling : 0x08\n"
"\t\t\t\tIO Path Events : 0x10\n"
"\t\t\t\tConfiguration Path : 0x20\n"
"\t\t\t\tiSCSI Protocol : 0x40\n");
DEVICE_ATTR(beiscsi_drvr_ver, S_IRUGO, beiscsi_drvr_ver_disp, NULL);
DEVICE_ATTR(beiscsi_adapter_family, S_IRUGO, beiscsi_adap_family_disp, NULL);
DEVICE_ATTR(beiscsi_fw_ver, S_IRUGO, beiscsi_fw_ver_disp, NULL);
DEVICE_ATTR(beiscsi_phys_port, S_IRUGO, beiscsi_phys_port_disp, NULL);
DEVICE_ATTR(beiscsi_active_session_count, S_IRUGO,
beiscsi_active_session_disp, NULL);
DEVICE_ATTR(beiscsi_free_session_count, S_IRUGO,
beiscsi_free_session_disp, NULL);
struct device_attribute *beiscsi_attrs[] = {
&dev_attr_beiscsi_log_enable,
&dev_attr_beiscsi_drvr_ver,
&dev_attr_beiscsi_adapter_family,
&dev_attr_beiscsi_fw_ver,
&dev_attr_beiscsi_active_session_count,
&dev_attr_beiscsi_free_session_count,
&dev_attr_beiscsi_phys_port,
NULL,
};
static char const *cqe_desc[] = {
"RESERVED_DESC",
"SOL_CMD_COMPLETE",
"SOL_CMD_KILLED_DATA_DIGEST_ERR",
"CXN_KILLED_PDU_SIZE_EXCEEDS_DSL",
"CXN_KILLED_BURST_LEN_MISMATCH",
"CXN_KILLED_AHS_RCVD",
"CXN_KILLED_HDR_DIGEST_ERR",
"CXN_KILLED_UNKNOWN_HDR",
"CXN_KILLED_STALE_ITT_TTT_RCVD",
"CXN_KILLED_INVALID_ITT_TTT_RCVD",
"CXN_KILLED_RST_RCVD",
"CXN_KILLED_TIMED_OUT",
"CXN_KILLED_RST_SENT",
"CXN_KILLED_FIN_RCVD",
"CXN_KILLED_BAD_UNSOL_PDU_RCVD",
"CXN_KILLED_BAD_WRB_INDEX_ERROR",
"CXN_KILLED_OVER_RUN_RESIDUAL",
"CXN_KILLED_UNDER_RUN_RESIDUAL",
"CMD_KILLED_INVALID_STATSN_RCVD",
"CMD_KILLED_INVALID_R2T_RCVD",
"CMD_CXN_KILLED_LUN_INVALID",
"CMD_CXN_KILLED_ICD_INVALID",
"CMD_CXN_KILLED_ITT_INVALID",
"CMD_CXN_KILLED_SEQ_OUTOFORDER",
"CMD_CXN_KILLED_INVALID_DATASN_RCVD",
"CXN_INVALIDATE_NOTIFY",
"CXN_INVALIDATE_INDEX_NOTIFY",
"CMD_INVALIDATED_NOTIFY",
"UNSOL_HDR_NOTIFY",
"UNSOL_DATA_NOTIFY",
"UNSOL_DATA_DIGEST_ERROR_NOTIFY",
"DRIVERMSG_NOTIFY",
"CXN_KILLED_CMND_DATA_NOT_ON_SAME_CONN",
"SOL_CMD_KILLED_DIF_ERR",
"CXN_KILLED_SYN_RCVD",
"CXN_KILLED_IMM_DATA_RCVD"
};
static int beiscsi_slave_configure(struct scsi_device *sdev)
{
blk_queue_max_segment_size(sdev->request_queue, 65536);
return 0;
}
static int beiscsi_eh_abort(struct scsi_cmnd *sc)
{
struct iscsi_task *abrt_task = (struct iscsi_task *)sc->SCp.ptr;
struct iscsi_cls_session *cls_session;
struct beiscsi_io_task *abrt_io_task;
struct beiscsi_conn *beiscsi_conn;
struct iscsi_session *session;
struct invldt_cmd_tbl inv_tbl;
struct beiscsi_hba *phba;
struct iscsi_conn *conn;
int rc;
cls_session = starget_to_session(scsi_target(sc->device));
session = cls_session->dd_data;
/* check if we raced, task just got cleaned up under us */
spin_lock_bh(&session->back_lock);
if (!abrt_task || !abrt_task->sc) {
spin_unlock_bh(&session->back_lock);
return SUCCESS;
}
/* get a task ref till FW processes the req for the ICD used */
__iscsi_get_task(abrt_task);
abrt_io_task = abrt_task->dd_data;
conn = abrt_task->conn;
beiscsi_conn = conn->dd_data;
phba = beiscsi_conn->phba;
/* mark WRB invalid which have been not processed by FW yet */
if (is_chip_be2_be3r(phba)) {
AMAP_SET_BITS(struct amap_iscsi_wrb, invld,
abrt_io_task->pwrb_handle->pwrb, 1);
} else {
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, invld,
abrt_io_task->pwrb_handle->pwrb, 1);
}
inv_tbl.cid = beiscsi_conn->beiscsi_conn_cid;
inv_tbl.icd = abrt_io_task->psgl_handle->sgl_index;
spin_unlock_bh(&session->back_lock);
rc = beiscsi_mgmt_invalidate_icds(phba, &inv_tbl, 1);
iscsi_put_task(abrt_task);
if (rc) {
beiscsi_log(phba, KERN_WARNING, BEISCSI_LOG_EH,
"BM_%d : sc %p invalidation failed %d\n",
sc, rc);
return FAILED;
}
return iscsi_eh_abort(sc);
}
static int beiscsi_eh_device_reset(struct scsi_cmnd *sc)
{
struct beiscsi_invldt_cmd_tbl {
struct invldt_cmd_tbl tbl[BE_INVLDT_CMD_TBL_SZ];
struct iscsi_task *task[BE_INVLDT_CMD_TBL_SZ];
} *inv_tbl;
struct iscsi_cls_session *cls_session;
struct beiscsi_conn *beiscsi_conn;
struct beiscsi_io_task *io_task;
struct iscsi_session *session;
struct beiscsi_hba *phba;
struct iscsi_conn *conn;
struct iscsi_task *task;
unsigned int i, nents;
int rc, more = 0;
cls_session = starget_to_session(scsi_target(sc->device));
session = cls_session->dd_data;
[SCSI] libiscsi: Reduce locking contention in fast path Replace the session lock with two locks, a forward lock and a backwards lock named frwd_lock and back_lock respectively. The forward lock protects resources that change while sending a request to the target, such as cmdsn, queued_cmdsn, and allocating task from the commands' pool with kfifo_out. The backward lock protects resources that change while processing a response or in error path, such as cmdsn_exp, cmdsn_max, and returning tasks to the commands' pool with kfifo_in. Under a steady state fast-path situation, that is when one or more processes/threads submit IO to an iscsi device and a single kernel upcall (e.g softirq) is dealing with processing of responses without errors, this patch eliminates the contention between the queuecommand()/request response/scsi_done() flows associated with iscsi sessions. Between the forward and the backward locks exists a strict locking hierarchy. The mutual exclusion zone protected by the forward lock can enclose the mutual exclusion zone protected by the backward lock but not vice versa. For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is a failure and __iscsi_put_task is called, the backward lock is taken while the forward lock is still taken. On the other hand, if in the RX path a nop is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu than the forward lock is released and the backward lock is taken for the duration of iscsi_send_nopout, later the backward lock is released and the forward lock is retaken. libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue. The insertion and deletion from these queues didn't corespond to the assumption taken by the new forward/backwards session locking paradigm. That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards) path, r2t is taken out from r2t queue and inserted to the r2t pool. In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t is also inserted to the r2t pool and another r2t is pulled from r2t queue. Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue to the TX path, r2t is taken from the r2t pool and inserted to the r2t queue. In order to cope with this situation, two spin locks were added, pool2queue and queue2pool. The former protects extracting from the r2t pool and inserting to the r2t queue, and the later protects the extracing from the r2t queue and inserting to the r2t pool. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> [minor fix up to apply cleanly and compile fix] Signed-off-by: Mike Christie <michaelc@cs.wisc.edu> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 14:41:38 +08:00
spin_lock_bh(&session->frwd_lock);
if (!session->leadconn || session->state != ISCSI_STATE_LOGGED_IN) {
[SCSI] libiscsi: Reduce locking contention in fast path Replace the session lock with two locks, a forward lock and a backwards lock named frwd_lock and back_lock respectively. The forward lock protects resources that change while sending a request to the target, such as cmdsn, queued_cmdsn, and allocating task from the commands' pool with kfifo_out. The backward lock protects resources that change while processing a response or in error path, such as cmdsn_exp, cmdsn_max, and returning tasks to the commands' pool with kfifo_in. Under a steady state fast-path situation, that is when one or more processes/threads submit IO to an iscsi device and a single kernel upcall (e.g softirq) is dealing with processing of responses without errors, this patch eliminates the contention between the queuecommand()/request response/scsi_done() flows associated with iscsi sessions. Between the forward and the backward locks exists a strict locking hierarchy. The mutual exclusion zone protected by the forward lock can enclose the mutual exclusion zone protected by the backward lock but not vice versa. For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is a failure and __iscsi_put_task is called, the backward lock is taken while the forward lock is still taken. On the other hand, if in the RX path a nop is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu than the forward lock is released and the backward lock is taken for the duration of iscsi_send_nopout, later the backward lock is released and the forward lock is retaken. libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue. The insertion and deletion from these queues didn't corespond to the assumption taken by the new forward/backwards session locking paradigm. That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards) path, r2t is taken out from r2t queue and inserted to the r2t pool. In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t is also inserted to the r2t pool and another r2t is pulled from r2t queue. Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue to the TX path, r2t is taken from the r2t pool and inserted to the r2t queue. In order to cope with this situation, two spin locks were added, pool2queue and queue2pool. The former protects extracting from the r2t pool and inserting to the r2t queue, and the later protects the extracing from the r2t queue and inserting to the r2t pool. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> [minor fix up to apply cleanly and compile fix] Signed-off-by: Mike Christie <michaelc@cs.wisc.edu> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 14:41:38 +08:00
spin_unlock_bh(&session->frwd_lock);
return FAILED;
}
conn = session->leadconn;
beiscsi_conn = conn->dd_data;
phba = beiscsi_conn->phba;
inv_tbl = kzalloc(sizeof(*inv_tbl), GFP_ATOMIC);
if (!inv_tbl) {
spin_unlock_bh(&session->frwd_lock);
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_EH,
"BM_%d : invldt_cmd_tbl alloc failed\n");
return FAILED;
}
nents = 0;
/* take back_lock to prevent task from getting cleaned up under us */
spin_lock(&session->back_lock);
for (i = 0; i < conn->session->cmds_max; i++) {
task = conn->session->cmds[i];
if (!task->sc)
continue;
if (sc->device->lun != task->sc->device->lun)
continue;
/**
* Can't fit in more cmds? Normally this won't happen b'coz
* BEISCSI_CMD_PER_LUN is same as BE_INVLDT_CMD_TBL_SZ.
*/
if (nents == BE_INVLDT_CMD_TBL_SZ) {
more = 1;
break;
}
/* get a task ref till FW processes the req for the ICD used */
__iscsi_get_task(task);
io_task = task->dd_data;
/* mark WRB invalid which have been not processed by FW yet */
if (is_chip_be2_be3r(phba)) {
AMAP_SET_BITS(struct amap_iscsi_wrb, invld,
io_task->pwrb_handle->pwrb, 1);
} else {
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, invld,
io_task->pwrb_handle->pwrb, 1);
}
inv_tbl->tbl[nents].cid = beiscsi_conn->beiscsi_conn_cid;
inv_tbl->tbl[nents].icd = io_task->psgl_handle->sgl_index;
inv_tbl->task[nents] = task;
nents++;
}
spin_unlock(&session->back_lock);
[SCSI] libiscsi: Reduce locking contention in fast path Replace the session lock with two locks, a forward lock and a backwards lock named frwd_lock and back_lock respectively. The forward lock protects resources that change while sending a request to the target, such as cmdsn, queued_cmdsn, and allocating task from the commands' pool with kfifo_out. The backward lock protects resources that change while processing a response or in error path, such as cmdsn_exp, cmdsn_max, and returning tasks to the commands' pool with kfifo_in. Under a steady state fast-path situation, that is when one or more processes/threads submit IO to an iscsi device and a single kernel upcall (e.g softirq) is dealing with processing of responses without errors, this patch eliminates the contention between the queuecommand()/request response/scsi_done() flows associated with iscsi sessions. Between the forward and the backward locks exists a strict locking hierarchy. The mutual exclusion zone protected by the forward lock can enclose the mutual exclusion zone protected by the backward lock but not vice versa. For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is a failure and __iscsi_put_task is called, the backward lock is taken while the forward lock is still taken. On the other hand, if in the RX path a nop is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu than the forward lock is released and the backward lock is taken for the duration of iscsi_send_nopout, later the backward lock is released and the forward lock is retaken. libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue. The insertion and deletion from these queues didn't corespond to the assumption taken by the new forward/backwards session locking paradigm. That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards) path, r2t is taken out from r2t queue and inserted to the r2t pool. In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t is also inserted to the r2t pool and another r2t is pulled from r2t queue. Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue to the TX path, r2t is taken from the r2t pool and inserted to the r2t queue. In order to cope with this situation, two spin locks were added, pool2queue and queue2pool. The former protects extracting from the r2t pool and inserting to the r2t queue, and the later protects the extracing from the r2t queue and inserting to the r2t pool. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> [minor fix up to apply cleanly and compile fix] Signed-off-by: Mike Christie <michaelc@cs.wisc.edu> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 14:41:38 +08:00
spin_unlock_bh(&session->frwd_lock);
rc = SUCCESS;
if (!nents)
goto end_reset;
if (more) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_EH,
"BM_%d : number of cmds exceeds size of invalidation table\n");
rc = FAILED;
goto end_reset;
}
if (beiscsi_mgmt_invalidate_icds(phba, &inv_tbl->tbl[0], nents)) {
beiscsi_log(phba, KERN_WARNING, BEISCSI_LOG_EH,
"BM_%d : cid %u scmds invalidation failed\n",
beiscsi_conn->beiscsi_conn_cid);
rc = FAILED;
}
end_reset:
for (i = 0; i < nents; i++)
iscsi_put_task(inv_tbl->task[i]);
kfree(inv_tbl);
if (rc == SUCCESS)
rc = iscsi_eh_device_reset(sc);
return rc;
}
/*------------------- PCI Driver operations and data ----------------- */
static const struct pci_device_id beiscsi_pci_id_table[] = {
{ PCI_DEVICE(BE_VENDOR_ID, BE_DEVICE_ID1) },
{ PCI_DEVICE(BE_VENDOR_ID, BE_DEVICE_ID2) },
{ PCI_DEVICE(BE_VENDOR_ID, OC_DEVICE_ID1) },
{ PCI_DEVICE(BE_VENDOR_ID, OC_DEVICE_ID2) },
{ PCI_DEVICE(BE_VENDOR_ID, OC_DEVICE_ID3) },
{ PCI_DEVICE(ELX_VENDOR_ID, OC_SKH_ID1) },
{ 0 }
};
MODULE_DEVICE_TABLE(pci, beiscsi_pci_id_table);
static struct scsi_host_template beiscsi_sht = {
.module = THIS_MODULE,
.name = "Emulex 10Gbe open-iscsi Initiator Driver",
.proc_name = DRV_NAME,
.queuecommand = iscsi_queuecommand,
.change_queue_depth = scsi_change_queue_depth,
.slave_configure = beiscsi_slave_configure,
.target_alloc = iscsi_target_alloc,
.eh_timed_out = iscsi_eh_cmd_timed_out,
.eh_abort_handler = beiscsi_eh_abort,
.eh_device_reset_handler = beiscsi_eh_device_reset,
.eh_target_reset_handler = iscsi_eh_session_reset,
.shost_attrs = beiscsi_attrs,
.sg_tablesize = BEISCSI_SGLIST_ELEMENTS,
.can_queue = BE2_IO_DEPTH,
.this_id = -1,
.max_sectors = BEISCSI_MAX_SECTORS,
.cmd_per_lun = BEISCSI_CMD_PER_LUN,
.use_clustering = ENABLE_CLUSTERING,
.vendor_id = SCSI_NL_VID_TYPE_PCI | BE_VENDOR_ID,
.track_queue_depth = 1,
};
static struct scsi_transport_template *beiscsi_scsi_transport;
static struct beiscsi_hba *beiscsi_hba_alloc(struct pci_dev *pcidev)
{
struct beiscsi_hba *phba;
struct Scsi_Host *shost;
shost = iscsi_host_alloc(&beiscsi_sht, sizeof(*phba), 0);
if (!shost) {
dev_err(&pcidev->dev,
"beiscsi_hba_alloc - iscsi_host_alloc failed\n");
return NULL;
}
shost->max_id = BE2_MAX_SESSIONS;
shost->max_channel = 0;
shost->max_cmd_len = BEISCSI_MAX_CMD_LEN;
shost->max_lun = BEISCSI_NUM_MAX_LUN;
shost->transportt = beiscsi_scsi_transport;
phba = iscsi_host_priv(shost);
memset(phba, 0, sizeof(*phba));
phba->shost = shost;
phba->pcidev = pci_dev_get(pcidev);
pci_set_drvdata(pcidev, phba);
phba->interface_handle = 0xFFFFFFFF;
return phba;
}
static void beiscsi_unmap_pci_function(struct beiscsi_hba *phba)
{
if (phba->csr_va) {
iounmap(phba->csr_va);
phba->csr_va = NULL;
}
if (phba->db_va) {
iounmap(phba->db_va);
phba->db_va = NULL;
}
if (phba->pci_va) {
iounmap(phba->pci_va);
phba->pci_va = NULL;
}
}
static int beiscsi_map_pci_bars(struct beiscsi_hba *phba,
struct pci_dev *pcidev)
{
u8 __iomem *addr;
int pcicfg_reg;
addr = ioremap_nocache(pci_resource_start(pcidev, 2),
pci_resource_len(pcidev, 2));
if (addr == NULL)
return -ENOMEM;
phba->ctrl.csr = addr;
phba->csr_va = addr;
phba->csr_pa.u.a64.address = pci_resource_start(pcidev, 2);
addr = ioremap_nocache(pci_resource_start(pcidev, 4), 128 * 1024);
if (addr == NULL)
goto pci_map_err;
phba->ctrl.db = addr;
phba->db_va = addr;
phba->db_pa.u.a64.address = pci_resource_start(pcidev, 4);
if (phba->generation == BE_GEN2)
pcicfg_reg = 1;
else
pcicfg_reg = 0;
addr = ioremap_nocache(pci_resource_start(pcidev, pcicfg_reg),
pci_resource_len(pcidev, pcicfg_reg));
if (addr == NULL)
goto pci_map_err;
phba->ctrl.pcicfg = addr;
phba->pci_va = addr;
phba->pci_pa.u.a64.address = pci_resource_start(pcidev, pcicfg_reg);
return 0;
pci_map_err:
beiscsi_unmap_pci_function(phba);
return -ENOMEM;
}
static int beiscsi_enable_pci(struct pci_dev *pcidev)
{
int ret;
ret = pci_enable_device(pcidev);
if (ret) {
dev_err(&pcidev->dev,
"beiscsi_enable_pci - enable device failed\n");
return ret;
}
ret = pci_request_regions(pcidev, DRV_NAME);
if (ret) {
dev_err(&pcidev->dev,
"beiscsi_enable_pci - request region failed\n");
goto pci_dev_disable;
}
pci_set_master(pcidev);
ret = pci_set_dma_mask(pcidev, DMA_BIT_MASK(64));
if (ret) {
ret = pci_set_dma_mask(pcidev, DMA_BIT_MASK(32));
if (ret) {
dev_err(&pcidev->dev, "Could not set PCI DMA Mask\n");
goto pci_region_release;
} else {
ret = pci_set_consistent_dma_mask(pcidev,
DMA_BIT_MASK(32));
}
} else {
ret = pci_set_consistent_dma_mask(pcidev, DMA_BIT_MASK(64));
if (ret) {
dev_err(&pcidev->dev, "Could not set PCI DMA Mask\n");
goto pci_region_release;
}
}
return 0;
pci_region_release:
pci_release_regions(pcidev);
pci_dev_disable:
pci_disable_device(pcidev);
return ret;
}
static int be_ctrl_init(struct beiscsi_hba *phba, struct pci_dev *pdev)
{
struct be_ctrl_info *ctrl = &phba->ctrl;
struct be_dma_mem *mbox_mem_alloc = &ctrl->mbox_mem_alloced;
struct be_dma_mem *mbox_mem_align = &ctrl->mbox_mem;
int status = 0;
ctrl->pdev = pdev;
status = beiscsi_map_pci_bars(phba, pdev);
if (status)
return status;
mbox_mem_alloc->size = sizeof(struct be_mcc_mailbox) + 16;
mbox_mem_alloc->va = pci_alloc_consistent(pdev,
mbox_mem_alloc->size,
&mbox_mem_alloc->dma);
if (!mbox_mem_alloc->va) {
beiscsi_unmap_pci_function(phba);
return -ENOMEM;
}
mbox_mem_align->size = sizeof(struct be_mcc_mailbox);
mbox_mem_align->va = PTR_ALIGN(mbox_mem_alloc->va, 16);
mbox_mem_align->dma = PTR_ALIGN(mbox_mem_alloc->dma, 16);
memset(mbox_mem_align->va, 0, sizeof(struct be_mcc_mailbox));
mutex_init(&ctrl->mbox_lock);
spin_lock_init(&phba->ctrl.mcc_lock);
return status;
}
/**
* beiscsi_get_params()- Set the config paramters
* @phba: ptr device priv structure
**/
static void beiscsi_get_params(struct beiscsi_hba *phba)
{
uint32_t total_cid_count = 0;
uint32_t total_icd_count = 0;
uint8_t ulp_num = 0;
total_cid_count = BEISCSI_GET_CID_COUNT(phba, BEISCSI_ULP0) +
BEISCSI_GET_CID_COUNT(phba, BEISCSI_ULP1);
for (ulp_num = 0; ulp_num < BEISCSI_ULP_COUNT; ulp_num++) {
uint32_t align_mask = 0;
uint32_t icd_post_per_page = 0;
uint32_t icd_count_unavailable = 0;
uint32_t icd_start = 0, icd_count = 0;
uint32_t icd_start_align = 0, icd_count_align = 0;
if (test_bit(ulp_num, &phba->fw_config.ulp_supported)) {
icd_start = phba->fw_config.iscsi_icd_start[ulp_num];
icd_count = phba->fw_config.iscsi_icd_count[ulp_num];
/* Get ICD count that can be posted on each page */
icd_post_per_page = (PAGE_SIZE / (BE2_SGE *
sizeof(struct iscsi_sge)));
align_mask = (icd_post_per_page - 1);
/* Check if icd_start is aligned ICD per page posting */
if (icd_start % icd_post_per_page) {
icd_start_align = ((icd_start +
icd_post_per_page) &
~(align_mask));
phba->fw_config.
iscsi_icd_start[ulp_num] =
icd_start_align;
}
icd_count_align = (icd_count & ~align_mask);
/* ICD discarded in the process of alignment */
if (icd_start_align)
icd_count_unavailable = ((icd_start_align -
icd_start) +
(icd_count -
icd_count_align));
/* Updated ICD count available */
phba->fw_config.iscsi_icd_count[ulp_num] = (icd_count -
icd_count_unavailable);
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_INIT,
"BM_%d : Aligned ICD values\n"
"\t ICD Start : %d\n"
"\t ICD Count : %d\n"
"\t ICD Discarded : %d\n",
phba->fw_config.
iscsi_icd_start[ulp_num],
phba->fw_config.
iscsi_icd_count[ulp_num],
icd_count_unavailable);
break;
}
}
total_icd_count = phba->fw_config.iscsi_icd_count[ulp_num];
phba->params.ios_per_ctrl = (total_icd_count -
(total_cid_count +
BE2_TMFS + BE2_NOPOUT_REQ));
phba->params.cxns_per_ctrl = total_cid_count;
phba->params.icds_per_ctrl = total_icd_count;
phba->params.num_sge_per_io = BE2_SGE;
phba->params.defpdu_hdr_sz = BE2_DEFPDU_HDR_SZ;
phba->params.defpdu_data_sz = BE2_DEFPDU_DATA_SZ;
phba->params.num_eq_entries = 1024;
phba->params.num_cq_entries = 1024;
phba->params.wrbs_per_cxn = 256;
}
static void hwi_ring_eq_db(struct beiscsi_hba *phba,
unsigned int id, unsigned int clr_interrupt,
unsigned int num_processed,
unsigned char rearm, unsigned char event)
{
u32 val = 0;
if (rearm)
val |= 1 << DB_EQ_REARM_SHIFT;
if (clr_interrupt)
val |= 1 << DB_EQ_CLR_SHIFT;
if (event)
val |= 1 << DB_EQ_EVNT_SHIFT;
val |= num_processed << DB_EQ_NUM_POPPED_SHIFT;
/* Setting lower order EQ_ID Bits */
val |= (id & DB_EQ_RING_ID_LOW_MASK);
/* Setting Higher order EQ_ID Bits */
val |= (((id >> DB_EQ_HIGH_FEILD_SHIFT) &
DB_EQ_RING_ID_HIGH_MASK)
<< DB_EQ_HIGH_SET_SHIFT);
iowrite32(val, phba->db_va + DB_EQ_OFFSET);
}
/**
* be_isr_mcc - The isr routine of the driver.
* @irq: Not used
* @dev_id: Pointer to host adapter structure
*/
static irqreturn_t be_isr_mcc(int irq, void *dev_id)
{
struct beiscsi_hba *phba;
struct be_eq_entry *eqe;
struct be_queue_info *eq;
struct be_queue_info *mcc;
unsigned int mcc_events;
struct be_eq_obj *pbe_eq;
pbe_eq = dev_id;
eq = &pbe_eq->q;
phba = pbe_eq->phba;
mcc = &phba->ctrl.mcc_obj.cq;
eqe = queue_tail_node(eq);
mcc_events = 0;
while (eqe->dw[offsetof(struct amap_eq_entry, valid) / 32]
& EQE_VALID_MASK) {
if (((eqe->dw[offsetof(struct amap_eq_entry,
resource_id) / 32] &
EQE_RESID_MASK) >> 16) == mcc->id) {
mcc_events++;
}
AMAP_SET_BITS(struct amap_eq_entry, valid, eqe, 0);
queue_tail_inc(eq);
eqe = queue_tail_node(eq);
}
if (mcc_events) {
queue_work(phba->wq, &pbe_eq->mcc_work);
hwi_ring_eq_db(phba, eq->id, 1, mcc_events, 1, 1);
}
return IRQ_HANDLED;
}
/**
* be_isr_msix - The isr routine of the driver.
* @irq: Not used
* @dev_id: Pointer to host adapter structure
*/
static irqreturn_t be_isr_msix(int irq, void *dev_id)
{
struct beiscsi_hba *phba;
struct be_queue_info *eq;
struct be_eq_obj *pbe_eq;
pbe_eq = dev_id;
eq = &pbe_eq->q;
phba = pbe_eq->phba;
/* disable interrupt till iopoll completes */
hwi_ring_eq_db(phba, eq->id, 1, 0, 0, 1);
irq_poll_sched(&pbe_eq->iopoll);
return IRQ_HANDLED;
}
/**
* be_isr - The isr routine of the driver.
* @irq: Not used
* @dev_id: Pointer to host adapter structure
*/
static irqreturn_t be_isr(int irq, void *dev_id)
{
struct beiscsi_hba *phba;
struct hwi_controller *phwi_ctrlr;
struct hwi_context_memory *phwi_context;
struct be_eq_entry *eqe;
struct be_queue_info *eq;
struct be_queue_info *mcc;
unsigned int mcc_events, io_events;
struct be_ctrl_info *ctrl;
struct be_eq_obj *pbe_eq;
int isr, rearm;
phba = dev_id;
ctrl = &phba->ctrl;
isr = ioread32(ctrl->csr + CEV_ISR0_OFFSET +
(PCI_FUNC(ctrl->pdev->devfn) * CEV_ISR_SIZE));
if (!isr)
return IRQ_NONE;
phwi_ctrlr = phba->phwi_ctrlr;
phwi_context = phwi_ctrlr->phwi_ctxt;
pbe_eq = &phwi_context->be_eq[0];
eq = &phwi_context->be_eq[0].q;
mcc = &phba->ctrl.mcc_obj.cq;
eqe = queue_tail_node(eq);
io_events = 0;
mcc_events = 0;
while (eqe->dw[offsetof(struct amap_eq_entry, valid) / 32]
& EQE_VALID_MASK) {
if (((eqe->dw[offsetof(struct amap_eq_entry,
resource_id) / 32] & EQE_RESID_MASK) >> 16) == mcc->id)
mcc_events++;
else
io_events++;
AMAP_SET_BITS(struct amap_eq_entry, valid, eqe, 0);
queue_tail_inc(eq);
eqe = queue_tail_node(eq);
}
if (!io_events && !mcc_events)
return IRQ_NONE;
/* no need to rearm if interrupt is only for IOs */
rearm = 0;
if (mcc_events) {
queue_work(phba->wq, &pbe_eq->mcc_work);
/* rearm for MCCQ */
rearm = 1;
}
if (io_events)
irq_poll_sched(&pbe_eq->iopoll);
hwi_ring_eq_db(phba, eq->id, 0, (io_events + mcc_events), rearm, 1);
return IRQ_HANDLED;
}
static int beiscsi_init_irqs(struct beiscsi_hba *phba)
{
struct pci_dev *pcidev = phba->pcidev;
struct hwi_controller *phwi_ctrlr;
struct hwi_context_memory *phwi_context;
int ret, i, j;
phwi_ctrlr = phba->phwi_ctrlr;
phwi_context = phwi_ctrlr->phwi_ctxt;
if (pcidev->msix_enabled) {
for (i = 0; i < phba->num_cpus; i++) {
phba->msi_name[i] = kzalloc(BEISCSI_MSI_NAME,
GFP_KERNEL);
if (!phba->msi_name[i]) {
ret = -ENOMEM;
goto free_msix_irqs;
}
sprintf(phba->msi_name[i], "beiscsi_%02x_%02x",
phba->shost->host_no, i);
ret = request_irq(pci_irq_vector(pcidev, i),
be_isr_msix, 0, phba->msi_name[i],
&phwi_context->be_eq[i]);
if (ret) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : beiscsi_init_irqs-Failed to"
"register msix for i = %d\n",
i);
kfree(phba->msi_name[i]);
goto free_msix_irqs;
}
}
phba->msi_name[i] = kzalloc(BEISCSI_MSI_NAME, GFP_KERNEL);
if (!phba->msi_name[i]) {
ret = -ENOMEM;
goto free_msix_irqs;
}
sprintf(phba->msi_name[i], "beiscsi_mcc_%02x",
phba->shost->host_no);
ret = request_irq(pci_irq_vector(pcidev, i), be_isr_mcc, 0,
phba->msi_name[i], &phwi_context->be_eq[i]);
if (ret) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT ,
"BM_%d : beiscsi_init_irqs-"
"Failed to register beiscsi_msix_mcc\n");
kfree(phba->msi_name[i]);
goto free_msix_irqs;
}
} else {
ret = request_irq(pcidev->irq, be_isr, IRQF_SHARED,
"beiscsi", phba);
if (ret) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : beiscsi_init_irqs-"
"Failed to register irq\\n");
return ret;
}
}
return 0;
free_msix_irqs:
for (j = i - 1; j >= 0; j--) {
free_irq(pci_irq_vector(pcidev, i), &phwi_context->be_eq[j]);
kfree(phba->msi_name[j]);
}
return ret;
}
void hwi_ring_cq_db(struct beiscsi_hba *phba,
unsigned int id, unsigned int num_processed,
unsigned char rearm)
{
u32 val = 0;
if (rearm)
val |= 1 << DB_CQ_REARM_SHIFT;
val |= num_processed << DB_CQ_NUM_POPPED_SHIFT;
/* Setting lower order CQ_ID Bits */
val |= (id & DB_CQ_RING_ID_LOW_MASK);
/* Setting Higher order CQ_ID Bits */
val |= (((id >> DB_CQ_HIGH_FEILD_SHIFT) &
DB_CQ_RING_ID_HIGH_MASK)
<< DB_CQ_HIGH_SET_SHIFT);
iowrite32(val, phba->db_va + DB_CQ_OFFSET);
}
static struct sgl_handle *alloc_io_sgl_handle(struct beiscsi_hba *phba)
{
struct sgl_handle *psgl_handle;
scsi: be2iscsi: Replace _bh with _irqsave/irqrestore [ 3843.132217] WARNING: CPU: 20 PID: 1227 at kernel/softirq.c:150 __local_bh_enable_ip+0x6b/0x90 [ 3843.142815] Modules linked in: ... [ 3843.294328] CPU: 20 PID: 1227 Comm: kworker/20:1H Tainted: G E 4.8.0-rc1+ #3 [ 3843.304944] Hardware name: Dell Inc. PowerEdge R720/0X6H47, BIOS 1.4.8 10/25/2012 [ 3843.314798] Workqueue: kblockd blk_timeout_work [ 3843.321350] 0000000000000086 00000000a32f4533 ffff8802216d7bd8 ffffffff8135c3cf [ 3843.331146] 0000000000000000 0000000000000000 ffff8802216d7c18 ffffffff8108d661 [ 3843.340918] 00000096216d7c50 0000000000000200 ffff8802d07cc828 ffff8801b3632550 [ 3843.350687] Call Trace: [ 3843.354866] [<ffffffff8135c3cf>] dump_stack+0x63/0x84 [ 3843.362061] [<ffffffff8108d661>] __warn+0xd1/0xf0 [ 3843.368851] [<ffffffff8108d79d>] warn_slowpath_null+0x1d/0x20 [ 3843.376791] [<ffffffff810930eb>] __local_bh_enable_ip+0x6b/0x90 [ 3843.384903] [<ffffffff816fe7be>] _raw_spin_unlock_bh+0x1e/0x20 [ 3843.392940] [<ffffffffa085f710>] beiscsi_alloc_pdu+0x2f0/0x6e0 [be2iscsi] [ 3843.402076] [<ffffffffa06bc358>] __iscsi_conn_send_pdu+0xf8/0x370 [libiscsi] [ 3843.411549] [<ffffffffa06bc6fe>] iscsi_send_nopout+0xbe/0x110 [libiscsi] [ 3843.420639] [<ffffffffa06bd98b>] iscsi_eh_cmd_timed_out+0x29b/0x2b0 [libiscsi] [ 3843.430339] [<ffffffff814cd1de>] scsi_times_out+0x5e/0x250 [ 3843.438119] [<ffffffff813374af>] blk_rq_timed_out+0x1f/0x60 [ 3843.446009] [<ffffffff8133759d>] blk_timeout_work+0xad/0x150 [ 3843.454010] [<ffffffff810a6642>] process_one_work+0x152/0x400 [ 3843.462114] [<ffffffff810a6f35>] worker_thread+0x125/0x4b0 [ 3843.469961] [<ffffffff810a6e10>] ? rescuer_thread+0x380/0x380 [ 3843.478116] [<ffffffff810aca28>] kthread+0xd8/0xf0 [ 3843.485212] [<ffffffff816fedff>] ret_from_fork+0x1f/0x40 [ 3843.492908] [<ffffffff810ac950>] ? kthread_park+0x60/0x60 [ 3843.500715] ---[ end trace 57ec0a1d8f0dd3a0 ]--- [ 3852.328667] NMI watchdog: Watchdog detected hard LOCKUP on cpu 1Kernel panic - not syncing: Hard LOCKUP blk_timeout_work takes queue_lock spin_lock with interrupts disabled before invoking iscsi_eh_cmd_timed_out. This causes a WARN_ON_ONCE in spin_unlock_bh for wrb_lock/io_sgl_lock/mgmt_sgl_lock. CPU was kept busy in lot of bottom half work with interrupts disabled thus causing hard lock up. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Chris Leech <cleech@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-10-13 14:38:49 +08:00
unsigned long flags;
scsi: be2iscsi: Replace _bh with _irqsave/irqrestore [ 3843.132217] WARNING: CPU: 20 PID: 1227 at kernel/softirq.c:150 __local_bh_enable_ip+0x6b/0x90 [ 3843.142815] Modules linked in: ... [ 3843.294328] CPU: 20 PID: 1227 Comm: kworker/20:1H Tainted: G E 4.8.0-rc1+ #3 [ 3843.304944] Hardware name: Dell Inc. PowerEdge R720/0X6H47, BIOS 1.4.8 10/25/2012 [ 3843.314798] Workqueue: kblockd blk_timeout_work [ 3843.321350] 0000000000000086 00000000a32f4533 ffff8802216d7bd8 ffffffff8135c3cf [ 3843.331146] 0000000000000000 0000000000000000 ffff8802216d7c18 ffffffff8108d661 [ 3843.340918] 00000096216d7c50 0000000000000200 ffff8802d07cc828 ffff8801b3632550 [ 3843.350687] Call Trace: [ 3843.354866] [<ffffffff8135c3cf>] dump_stack+0x63/0x84 [ 3843.362061] [<ffffffff8108d661>] __warn+0xd1/0xf0 [ 3843.368851] [<ffffffff8108d79d>] warn_slowpath_null+0x1d/0x20 [ 3843.376791] [<ffffffff810930eb>] __local_bh_enable_ip+0x6b/0x90 [ 3843.384903] [<ffffffff816fe7be>] _raw_spin_unlock_bh+0x1e/0x20 [ 3843.392940] [<ffffffffa085f710>] beiscsi_alloc_pdu+0x2f0/0x6e0 [be2iscsi] [ 3843.402076] [<ffffffffa06bc358>] __iscsi_conn_send_pdu+0xf8/0x370 [libiscsi] [ 3843.411549] [<ffffffffa06bc6fe>] iscsi_send_nopout+0xbe/0x110 [libiscsi] [ 3843.420639] [<ffffffffa06bd98b>] iscsi_eh_cmd_timed_out+0x29b/0x2b0 [libiscsi] [ 3843.430339] [<ffffffff814cd1de>] scsi_times_out+0x5e/0x250 [ 3843.438119] [<ffffffff813374af>] blk_rq_timed_out+0x1f/0x60 [ 3843.446009] [<ffffffff8133759d>] blk_timeout_work+0xad/0x150 [ 3843.454010] [<ffffffff810a6642>] process_one_work+0x152/0x400 [ 3843.462114] [<ffffffff810a6f35>] worker_thread+0x125/0x4b0 [ 3843.469961] [<ffffffff810a6e10>] ? rescuer_thread+0x380/0x380 [ 3843.478116] [<ffffffff810aca28>] kthread+0xd8/0xf0 [ 3843.485212] [<ffffffff816fedff>] ret_from_fork+0x1f/0x40 [ 3843.492908] [<ffffffff810ac950>] ? kthread_park+0x60/0x60 [ 3843.500715] ---[ end trace 57ec0a1d8f0dd3a0 ]--- [ 3852.328667] NMI watchdog: Watchdog detected hard LOCKUP on cpu 1Kernel panic - not syncing: Hard LOCKUP blk_timeout_work takes queue_lock spin_lock with interrupts disabled before invoking iscsi_eh_cmd_timed_out. This causes a WARN_ON_ONCE in spin_unlock_bh for wrb_lock/io_sgl_lock/mgmt_sgl_lock. CPU was kept busy in lot of bottom half work with interrupts disabled thus causing hard lock up. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Chris Leech <cleech@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-10-13 14:38:49 +08:00
spin_lock_irqsave(&phba->io_sgl_lock, flags);
if (phba->io_sgl_hndl_avbl) {
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_IO,
"BM_%d : In alloc_io_sgl_handle,"
" io_sgl_alloc_index=%d\n",
phba->io_sgl_alloc_index);
psgl_handle = phba->io_sgl_hndl_base[phba->
io_sgl_alloc_index];
phba->io_sgl_hndl_base[phba->io_sgl_alloc_index] = NULL;
phba->io_sgl_hndl_avbl--;
if (phba->io_sgl_alloc_index == (phba->params.
ios_per_ctrl - 1))
phba->io_sgl_alloc_index = 0;
else
phba->io_sgl_alloc_index++;
} else
psgl_handle = NULL;
scsi: be2iscsi: Replace _bh with _irqsave/irqrestore [ 3843.132217] WARNING: CPU: 20 PID: 1227 at kernel/softirq.c:150 __local_bh_enable_ip+0x6b/0x90 [ 3843.142815] Modules linked in: ... [ 3843.294328] CPU: 20 PID: 1227 Comm: kworker/20:1H Tainted: G E 4.8.0-rc1+ #3 [ 3843.304944] Hardware name: Dell Inc. PowerEdge R720/0X6H47, BIOS 1.4.8 10/25/2012 [ 3843.314798] Workqueue: kblockd blk_timeout_work [ 3843.321350] 0000000000000086 00000000a32f4533 ffff8802216d7bd8 ffffffff8135c3cf [ 3843.331146] 0000000000000000 0000000000000000 ffff8802216d7c18 ffffffff8108d661 [ 3843.340918] 00000096216d7c50 0000000000000200 ffff8802d07cc828 ffff8801b3632550 [ 3843.350687] Call Trace: [ 3843.354866] [<ffffffff8135c3cf>] dump_stack+0x63/0x84 [ 3843.362061] [<ffffffff8108d661>] __warn+0xd1/0xf0 [ 3843.368851] [<ffffffff8108d79d>] warn_slowpath_null+0x1d/0x20 [ 3843.376791] [<ffffffff810930eb>] __local_bh_enable_ip+0x6b/0x90 [ 3843.384903] [<ffffffff816fe7be>] _raw_spin_unlock_bh+0x1e/0x20 [ 3843.392940] [<ffffffffa085f710>] beiscsi_alloc_pdu+0x2f0/0x6e0 [be2iscsi] [ 3843.402076] [<ffffffffa06bc358>] __iscsi_conn_send_pdu+0xf8/0x370 [libiscsi] [ 3843.411549] [<ffffffffa06bc6fe>] iscsi_send_nopout+0xbe/0x110 [libiscsi] [ 3843.420639] [<ffffffffa06bd98b>] iscsi_eh_cmd_timed_out+0x29b/0x2b0 [libiscsi] [ 3843.430339] [<ffffffff814cd1de>] scsi_times_out+0x5e/0x250 [ 3843.438119] [<ffffffff813374af>] blk_rq_timed_out+0x1f/0x60 [ 3843.446009] [<ffffffff8133759d>] blk_timeout_work+0xad/0x150 [ 3843.454010] [<ffffffff810a6642>] process_one_work+0x152/0x400 [ 3843.462114] [<ffffffff810a6f35>] worker_thread+0x125/0x4b0 [ 3843.469961] [<ffffffff810a6e10>] ? rescuer_thread+0x380/0x380 [ 3843.478116] [<ffffffff810aca28>] kthread+0xd8/0xf0 [ 3843.485212] [<ffffffff816fedff>] ret_from_fork+0x1f/0x40 [ 3843.492908] [<ffffffff810ac950>] ? kthread_park+0x60/0x60 [ 3843.500715] ---[ end trace 57ec0a1d8f0dd3a0 ]--- [ 3852.328667] NMI watchdog: Watchdog detected hard LOCKUP on cpu 1Kernel panic - not syncing: Hard LOCKUP blk_timeout_work takes queue_lock spin_lock with interrupts disabled before invoking iscsi_eh_cmd_timed_out. This causes a WARN_ON_ONCE in spin_unlock_bh for wrb_lock/io_sgl_lock/mgmt_sgl_lock. CPU was kept busy in lot of bottom half work with interrupts disabled thus causing hard lock up. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Chris Leech <cleech@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-10-13 14:38:49 +08:00
spin_unlock_irqrestore(&phba->io_sgl_lock, flags);
return psgl_handle;
}
static void
free_io_sgl_handle(struct beiscsi_hba *phba, struct sgl_handle *psgl_handle)
{
scsi: be2iscsi: Replace _bh with _irqsave/irqrestore [ 3843.132217] WARNING: CPU: 20 PID: 1227 at kernel/softirq.c:150 __local_bh_enable_ip+0x6b/0x90 [ 3843.142815] Modules linked in: ... [ 3843.294328] CPU: 20 PID: 1227 Comm: kworker/20:1H Tainted: G E 4.8.0-rc1+ #3 [ 3843.304944] Hardware name: Dell Inc. PowerEdge R720/0X6H47, BIOS 1.4.8 10/25/2012 [ 3843.314798] Workqueue: kblockd blk_timeout_work [ 3843.321350] 0000000000000086 00000000a32f4533 ffff8802216d7bd8 ffffffff8135c3cf [ 3843.331146] 0000000000000000 0000000000000000 ffff8802216d7c18 ffffffff8108d661 [ 3843.340918] 00000096216d7c50 0000000000000200 ffff8802d07cc828 ffff8801b3632550 [ 3843.350687] Call Trace: [ 3843.354866] [<ffffffff8135c3cf>] dump_stack+0x63/0x84 [ 3843.362061] [<ffffffff8108d661>] __warn+0xd1/0xf0 [ 3843.368851] [<ffffffff8108d79d>] warn_slowpath_null+0x1d/0x20 [ 3843.376791] [<ffffffff810930eb>] __local_bh_enable_ip+0x6b/0x90 [ 3843.384903] [<ffffffff816fe7be>] _raw_spin_unlock_bh+0x1e/0x20 [ 3843.392940] [<ffffffffa085f710>] beiscsi_alloc_pdu+0x2f0/0x6e0 [be2iscsi] [ 3843.402076] [<ffffffffa06bc358>] __iscsi_conn_send_pdu+0xf8/0x370 [libiscsi] [ 3843.411549] [<ffffffffa06bc6fe>] iscsi_send_nopout+0xbe/0x110 [libiscsi] [ 3843.420639] [<ffffffffa06bd98b>] iscsi_eh_cmd_timed_out+0x29b/0x2b0 [libiscsi] [ 3843.430339] [<ffffffff814cd1de>] scsi_times_out+0x5e/0x250 [ 3843.438119] [<ffffffff813374af>] blk_rq_timed_out+0x1f/0x60 [ 3843.446009] [<ffffffff8133759d>] blk_timeout_work+0xad/0x150 [ 3843.454010] [<ffffffff810a6642>] process_one_work+0x152/0x400 [ 3843.462114] [<ffffffff810a6f35>] worker_thread+0x125/0x4b0 [ 3843.469961] [<ffffffff810a6e10>] ? rescuer_thread+0x380/0x380 [ 3843.478116] [<ffffffff810aca28>] kthread+0xd8/0xf0 [ 3843.485212] [<ffffffff816fedff>] ret_from_fork+0x1f/0x40 [ 3843.492908] [<ffffffff810ac950>] ? kthread_park+0x60/0x60 [ 3843.500715] ---[ end trace 57ec0a1d8f0dd3a0 ]--- [ 3852.328667] NMI watchdog: Watchdog detected hard LOCKUP on cpu 1Kernel panic - not syncing: Hard LOCKUP blk_timeout_work takes queue_lock spin_lock with interrupts disabled before invoking iscsi_eh_cmd_timed_out. This causes a WARN_ON_ONCE in spin_unlock_bh for wrb_lock/io_sgl_lock/mgmt_sgl_lock. CPU was kept busy in lot of bottom half work with interrupts disabled thus causing hard lock up. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Chris Leech <cleech@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-10-13 14:38:49 +08:00
unsigned long flags;
spin_lock_irqsave(&phba->io_sgl_lock, flags);
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_IO,
"BM_%d : In free_,io_sgl_free_index=%d\n",
phba->io_sgl_free_index);
if (phba->io_sgl_hndl_base[phba->io_sgl_free_index]) {
/*
* this can happen if clean_task is called on a task that
* failed in xmit_task or alloc_pdu.
*/
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_IO,
"BM_%d : Double Free in IO SGL io_sgl_free_index=%d,"
"value there=%p\n", phba->io_sgl_free_index,
phba->io_sgl_hndl_base
[phba->io_sgl_free_index]);
scsi: be2iscsi: Replace _bh with _irqsave/irqrestore [ 3843.132217] WARNING: CPU: 20 PID: 1227 at kernel/softirq.c:150 __local_bh_enable_ip+0x6b/0x90 [ 3843.142815] Modules linked in: ... [ 3843.294328] CPU: 20 PID: 1227 Comm: kworker/20:1H Tainted: G E 4.8.0-rc1+ #3 [ 3843.304944] Hardware name: Dell Inc. PowerEdge R720/0X6H47, BIOS 1.4.8 10/25/2012 [ 3843.314798] Workqueue: kblockd blk_timeout_work [ 3843.321350] 0000000000000086 00000000a32f4533 ffff8802216d7bd8 ffffffff8135c3cf [ 3843.331146] 0000000000000000 0000000000000000 ffff8802216d7c18 ffffffff8108d661 [ 3843.340918] 00000096216d7c50 0000000000000200 ffff8802d07cc828 ffff8801b3632550 [ 3843.350687] Call Trace: [ 3843.354866] [<ffffffff8135c3cf>] dump_stack+0x63/0x84 [ 3843.362061] [<ffffffff8108d661>] __warn+0xd1/0xf0 [ 3843.368851] [<ffffffff8108d79d>] warn_slowpath_null+0x1d/0x20 [ 3843.376791] [<ffffffff810930eb>] __local_bh_enable_ip+0x6b/0x90 [ 3843.384903] [<ffffffff816fe7be>] _raw_spin_unlock_bh+0x1e/0x20 [ 3843.392940] [<ffffffffa085f710>] beiscsi_alloc_pdu+0x2f0/0x6e0 [be2iscsi] [ 3843.402076] [<ffffffffa06bc358>] __iscsi_conn_send_pdu+0xf8/0x370 [libiscsi] [ 3843.411549] [<ffffffffa06bc6fe>] iscsi_send_nopout+0xbe/0x110 [libiscsi] [ 3843.420639] [<ffffffffa06bd98b>] iscsi_eh_cmd_timed_out+0x29b/0x2b0 [libiscsi] [ 3843.430339] [<ffffffff814cd1de>] scsi_times_out+0x5e/0x250 [ 3843.438119] [<ffffffff813374af>] blk_rq_timed_out+0x1f/0x60 [ 3843.446009] [<ffffffff8133759d>] blk_timeout_work+0xad/0x150 [ 3843.454010] [<ffffffff810a6642>] process_one_work+0x152/0x400 [ 3843.462114] [<ffffffff810a6f35>] worker_thread+0x125/0x4b0 [ 3843.469961] [<ffffffff810a6e10>] ? rescuer_thread+0x380/0x380 [ 3843.478116] [<ffffffff810aca28>] kthread+0xd8/0xf0 [ 3843.485212] [<ffffffff816fedff>] ret_from_fork+0x1f/0x40 [ 3843.492908] [<ffffffff810ac950>] ? kthread_park+0x60/0x60 [ 3843.500715] ---[ end trace 57ec0a1d8f0dd3a0 ]--- [ 3852.328667] NMI watchdog: Watchdog detected hard LOCKUP on cpu 1Kernel panic - not syncing: Hard LOCKUP blk_timeout_work takes queue_lock spin_lock with interrupts disabled before invoking iscsi_eh_cmd_timed_out. This causes a WARN_ON_ONCE in spin_unlock_bh for wrb_lock/io_sgl_lock/mgmt_sgl_lock. CPU was kept busy in lot of bottom half work with interrupts disabled thus causing hard lock up. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Chris Leech <cleech@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-10-13 14:38:49 +08:00
spin_unlock_irqrestore(&phba->io_sgl_lock, flags);
return;
}
phba->io_sgl_hndl_base[phba->io_sgl_free_index] = psgl_handle;
phba->io_sgl_hndl_avbl++;
if (phba->io_sgl_free_index == (phba->params.ios_per_ctrl - 1))
phba->io_sgl_free_index = 0;
else
phba->io_sgl_free_index++;
scsi: be2iscsi: Replace _bh with _irqsave/irqrestore [ 3843.132217] WARNING: CPU: 20 PID: 1227 at kernel/softirq.c:150 __local_bh_enable_ip+0x6b/0x90 [ 3843.142815] Modules linked in: ... [ 3843.294328] CPU: 20 PID: 1227 Comm: kworker/20:1H Tainted: G E 4.8.0-rc1+ #3 [ 3843.304944] Hardware name: Dell Inc. PowerEdge R720/0X6H47, BIOS 1.4.8 10/25/2012 [ 3843.314798] Workqueue: kblockd blk_timeout_work [ 3843.321350] 0000000000000086 00000000a32f4533 ffff8802216d7bd8 ffffffff8135c3cf [ 3843.331146] 0000000000000000 0000000000000000 ffff8802216d7c18 ffffffff8108d661 [ 3843.340918] 00000096216d7c50 0000000000000200 ffff8802d07cc828 ffff8801b3632550 [ 3843.350687] Call Trace: [ 3843.354866] [<ffffffff8135c3cf>] dump_stack+0x63/0x84 [ 3843.362061] [<ffffffff8108d661>] __warn+0xd1/0xf0 [ 3843.368851] [<ffffffff8108d79d>] warn_slowpath_null+0x1d/0x20 [ 3843.376791] [<ffffffff810930eb>] __local_bh_enable_ip+0x6b/0x90 [ 3843.384903] [<ffffffff816fe7be>] _raw_spin_unlock_bh+0x1e/0x20 [ 3843.392940] [<ffffffffa085f710>] beiscsi_alloc_pdu+0x2f0/0x6e0 [be2iscsi] [ 3843.402076] [<ffffffffa06bc358>] __iscsi_conn_send_pdu+0xf8/0x370 [libiscsi] [ 3843.411549] [<ffffffffa06bc6fe>] iscsi_send_nopout+0xbe/0x110 [libiscsi] [ 3843.420639] [<ffffffffa06bd98b>] iscsi_eh_cmd_timed_out+0x29b/0x2b0 [libiscsi] [ 3843.430339] [<ffffffff814cd1de>] scsi_times_out+0x5e/0x250 [ 3843.438119] [<ffffffff813374af>] blk_rq_timed_out+0x1f/0x60 [ 3843.446009] [<ffffffff8133759d>] blk_timeout_work+0xad/0x150 [ 3843.454010] [<ffffffff810a6642>] process_one_work+0x152/0x400 [ 3843.462114] [<ffffffff810a6f35>] worker_thread+0x125/0x4b0 [ 3843.469961] [<ffffffff810a6e10>] ? rescuer_thread+0x380/0x380 [ 3843.478116] [<ffffffff810aca28>] kthread+0xd8/0xf0 [ 3843.485212] [<ffffffff816fedff>] ret_from_fork+0x1f/0x40 [ 3843.492908] [<ffffffff810ac950>] ? kthread_park+0x60/0x60 [ 3843.500715] ---[ end trace 57ec0a1d8f0dd3a0 ]--- [ 3852.328667] NMI watchdog: Watchdog detected hard LOCKUP on cpu 1Kernel panic - not syncing: Hard LOCKUP blk_timeout_work takes queue_lock spin_lock with interrupts disabled before invoking iscsi_eh_cmd_timed_out. This causes a WARN_ON_ONCE in spin_unlock_bh for wrb_lock/io_sgl_lock/mgmt_sgl_lock. CPU was kept busy in lot of bottom half work with interrupts disabled thus causing hard lock up. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Chris Leech <cleech@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-10-13 14:38:49 +08:00
spin_unlock_irqrestore(&phba->io_sgl_lock, flags);
}
static inline struct wrb_handle *
beiscsi_get_wrb_handle(struct hwi_wrb_context *pwrb_context,
unsigned int wrbs_per_cxn)
{
struct wrb_handle *pwrb_handle;
scsi: be2iscsi: Replace _bh with _irqsave/irqrestore [ 3843.132217] WARNING: CPU: 20 PID: 1227 at kernel/softirq.c:150 __local_bh_enable_ip+0x6b/0x90 [ 3843.142815] Modules linked in: ... [ 3843.294328] CPU: 20 PID: 1227 Comm: kworker/20:1H Tainted: G E 4.8.0-rc1+ #3 [ 3843.304944] Hardware name: Dell Inc. PowerEdge R720/0X6H47, BIOS 1.4.8 10/25/2012 [ 3843.314798] Workqueue: kblockd blk_timeout_work [ 3843.321350] 0000000000000086 00000000a32f4533 ffff8802216d7bd8 ffffffff8135c3cf [ 3843.331146] 0000000000000000 0000000000000000 ffff8802216d7c18 ffffffff8108d661 [ 3843.340918] 00000096216d7c50 0000000000000200 ffff8802d07cc828 ffff8801b3632550 [ 3843.350687] Call Trace: [ 3843.354866] [<ffffffff8135c3cf>] dump_stack+0x63/0x84 [ 3843.362061] [<ffffffff8108d661>] __warn+0xd1/0xf0 [ 3843.368851] [<ffffffff8108d79d>] warn_slowpath_null+0x1d/0x20 [ 3843.376791] [<ffffffff810930eb>] __local_bh_enable_ip+0x6b/0x90 [ 3843.384903] [<ffffffff816fe7be>] _raw_spin_unlock_bh+0x1e/0x20 [ 3843.392940] [<ffffffffa085f710>] beiscsi_alloc_pdu+0x2f0/0x6e0 [be2iscsi] [ 3843.402076] [<ffffffffa06bc358>] __iscsi_conn_send_pdu+0xf8/0x370 [libiscsi] [ 3843.411549] [<ffffffffa06bc6fe>] iscsi_send_nopout+0xbe/0x110 [libiscsi] [ 3843.420639] [<ffffffffa06bd98b>] iscsi_eh_cmd_timed_out+0x29b/0x2b0 [libiscsi] [ 3843.430339] [<ffffffff814cd1de>] scsi_times_out+0x5e/0x250 [ 3843.438119] [<ffffffff813374af>] blk_rq_timed_out+0x1f/0x60 [ 3843.446009] [<ffffffff8133759d>] blk_timeout_work+0xad/0x150 [ 3843.454010] [<ffffffff810a6642>] process_one_work+0x152/0x400 [ 3843.462114] [<ffffffff810a6f35>] worker_thread+0x125/0x4b0 [ 3843.469961] [<ffffffff810a6e10>] ? rescuer_thread+0x380/0x380 [ 3843.478116] [<ffffffff810aca28>] kthread+0xd8/0xf0 [ 3843.485212] [<ffffffff816fedff>] ret_from_fork+0x1f/0x40 [ 3843.492908] [<ffffffff810ac950>] ? kthread_park+0x60/0x60 [ 3843.500715] ---[ end trace 57ec0a1d8f0dd3a0 ]--- [ 3852.328667] NMI watchdog: Watchdog detected hard LOCKUP on cpu 1Kernel panic - not syncing: Hard LOCKUP blk_timeout_work takes queue_lock spin_lock with interrupts disabled before invoking iscsi_eh_cmd_timed_out. This causes a WARN_ON_ONCE in spin_unlock_bh for wrb_lock/io_sgl_lock/mgmt_sgl_lock. CPU was kept busy in lot of bottom half work with interrupts disabled thus causing hard lock up. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Chris Leech <cleech@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-10-13 14:38:49 +08:00
unsigned long flags;
scsi: be2iscsi: Replace _bh with _irqsave/irqrestore [ 3843.132217] WARNING: CPU: 20 PID: 1227 at kernel/softirq.c:150 __local_bh_enable_ip+0x6b/0x90 [ 3843.142815] Modules linked in: ... [ 3843.294328] CPU: 20 PID: 1227 Comm: kworker/20:1H Tainted: G E 4.8.0-rc1+ #3 [ 3843.304944] Hardware name: Dell Inc. PowerEdge R720/0X6H47, BIOS 1.4.8 10/25/2012 [ 3843.314798] Workqueue: kblockd blk_timeout_work [ 3843.321350] 0000000000000086 00000000a32f4533 ffff8802216d7bd8 ffffffff8135c3cf [ 3843.331146] 0000000000000000 0000000000000000 ffff8802216d7c18 ffffffff8108d661 [ 3843.340918] 00000096216d7c50 0000000000000200 ffff8802d07cc828 ffff8801b3632550 [ 3843.350687] Call Trace: [ 3843.354866] [<ffffffff8135c3cf>] dump_stack+0x63/0x84 [ 3843.362061] [<ffffffff8108d661>] __warn+0xd1/0xf0 [ 3843.368851] [<ffffffff8108d79d>] warn_slowpath_null+0x1d/0x20 [ 3843.376791] [<ffffffff810930eb>] __local_bh_enable_ip+0x6b/0x90 [ 3843.384903] [<ffffffff816fe7be>] _raw_spin_unlock_bh+0x1e/0x20 [ 3843.392940] [<ffffffffa085f710>] beiscsi_alloc_pdu+0x2f0/0x6e0 [be2iscsi] [ 3843.402076] [<ffffffffa06bc358>] __iscsi_conn_send_pdu+0xf8/0x370 [libiscsi] [ 3843.411549] [<ffffffffa06bc6fe>] iscsi_send_nopout+0xbe/0x110 [libiscsi] [ 3843.420639] [<ffffffffa06bd98b>] iscsi_eh_cmd_timed_out+0x29b/0x2b0 [libiscsi] [ 3843.430339] [<ffffffff814cd1de>] scsi_times_out+0x5e/0x250 [ 3843.438119] [<ffffffff813374af>] blk_rq_timed_out+0x1f/0x60 [ 3843.446009] [<ffffffff8133759d>] blk_timeout_work+0xad/0x150 [ 3843.454010] [<ffffffff810a6642>] process_one_work+0x152/0x400 [ 3843.462114] [<ffffffff810a6f35>] worker_thread+0x125/0x4b0 [ 3843.469961] [<ffffffff810a6e10>] ? rescuer_thread+0x380/0x380 [ 3843.478116] [<ffffffff810aca28>] kthread+0xd8/0xf0 [ 3843.485212] [<ffffffff816fedff>] ret_from_fork+0x1f/0x40 [ 3843.492908] [<ffffffff810ac950>] ? kthread_park+0x60/0x60 [ 3843.500715] ---[ end trace 57ec0a1d8f0dd3a0 ]--- [ 3852.328667] NMI watchdog: Watchdog detected hard LOCKUP on cpu 1Kernel panic - not syncing: Hard LOCKUP blk_timeout_work takes queue_lock spin_lock with interrupts disabled before invoking iscsi_eh_cmd_timed_out. This causes a WARN_ON_ONCE in spin_unlock_bh for wrb_lock/io_sgl_lock/mgmt_sgl_lock. CPU was kept busy in lot of bottom half work with interrupts disabled thus causing hard lock up. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Chris Leech <cleech@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-10-13 14:38:49 +08:00
spin_lock_irqsave(&pwrb_context->wrb_lock, flags);
if (!pwrb_context->wrb_handles_available) {
spin_unlock_irqrestore(&pwrb_context->wrb_lock, flags);
return NULL;
}
pwrb_handle = pwrb_context->pwrb_handle_base[pwrb_context->alloc_index];
pwrb_context->wrb_handles_available--;
if (pwrb_context->alloc_index == (wrbs_per_cxn - 1))
pwrb_context->alloc_index = 0;
else
pwrb_context->alloc_index++;
scsi: be2iscsi: Replace _bh with _irqsave/irqrestore [ 3843.132217] WARNING: CPU: 20 PID: 1227 at kernel/softirq.c:150 __local_bh_enable_ip+0x6b/0x90 [ 3843.142815] Modules linked in: ... [ 3843.294328] CPU: 20 PID: 1227 Comm: kworker/20:1H Tainted: G E 4.8.0-rc1+ #3 [ 3843.304944] Hardware name: Dell Inc. PowerEdge R720/0X6H47, BIOS 1.4.8 10/25/2012 [ 3843.314798] Workqueue: kblockd blk_timeout_work [ 3843.321350] 0000000000000086 00000000a32f4533 ffff8802216d7bd8 ffffffff8135c3cf [ 3843.331146] 0000000000000000 0000000000000000 ffff8802216d7c18 ffffffff8108d661 [ 3843.340918] 00000096216d7c50 0000000000000200 ffff8802d07cc828 ffff8801b3632550 [ 3843.350687] Call Trace: [ 3843.354866] [<ffffffff8135c3cf>] dump_stack+0x63/0x84 [ 3843.362061] [<ffffffff8108d661>] __warn+0xd1/0xf0 [ 3843.368851] [<ffffffff8108d79d>] warn_slowpath_null+0x1d/0x20 [ 3843.376791] [<ffffffff810930eb>] __local_bh_enable_ip+0x6b/0x90 [ 3843.384903] [<ffffffff816fe7be>] _raw_spin_unlock_bh+0x1e/0x20 [ 3843.392940] [<ffffffffa085f710>] beiscsi_alloc_pdu+0x2f0/0x6e0 [be2iscsi] [ 3843.402076] [<ffffffffa06bc358>] __iscsi_conn_send_pdu+0xf8/0x370 [libiscsi] [ 3843.411549] [<ffffffffa06bc6fe>] iscsi_send_nopout+0xbe/0x110 [libiscsi] [ 3843.420639] [<ffffffffa06bd98b>] iscsi_eh_cmd_timed_out+0x29b/0x2b0 [libiscsi] [ 3843.430339] [<ffffffff814cd1de>] scsi_times_out+0x5e/0x250 [ 3843.438119] [<ffffffff813374af>] blk_rq_timed_out+0x1f/0x60 [ 3843.446009] [<ffffffff8133759d>] blk_timeout_work+0xad/0x150 [ 3843.454010] [<ffffffff810a6642>] process_one_work+0x152/0x400 [ 3843.462114] [<ffffffff810a6f35>] worker_thread+0x125/0x4b0 [ 3843.469961] [<ffffffff810a6e10>] ? rescuer_thread+0x380/0x380 [ 3843.478116] [<ffffffff810aca28>] kthread+0xd8/0xf0 [ 3843.485212] [<ffffffff816fedff>] ret_from_fork+0x1f/0x40 [ 3843.492908] [<ffffffff810ac950>] ? kthread_park+0x60/0x60 [ 3843.500715] ---[ end trace 57ec0a1d8f0dd3a0 ]--- [ 3852.328667] NMI watchdog: Watchdog detected hard LOCKUP on cpu 1Kernel panic - not syncing: Hard LOCKUP blk_timeout_work takes queue_lock spin_lock with interrupts disabled before invoking iscsi_eh_cmd_timed_out. This causes a WARN_ON_ONCE in spin_unlock_bh for wrb_lock/io_sgl_lock/mgmt_sgl_lock. CPU was kept busy in lot of bottom half work with interrupts disabled thus causing hard lock up. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Chris Leech <cleech@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-10-13 14:38:49 +08:00
spin_unlock_irqrestore(&pwrb_context->wrb_lock, flags);
if (pwrb_handle)
memset(pwrb_handle->pwrb, 0, sizeof(*pwrb_handle->pwrb));
return pwrb_handle;
}
/**
* alloc_wrb_handle - To allocate a wrb handle
* @phba: The hba pointer
* @cid: The cid to use for allocation
* @pwrb_context: ptr to ptr to wrb context
*
* This happens under session_lock until submission to chip
*/
struct wrb_handle *alloc_wrb_handle(struct beiscsi_hba *phba, unsigned int cid,
struct hwi_wrb_context **pcontext)
{
struct hwi_wrb_context *pwrb_context;
struct hwi_controller *phwi_ctrlr;
uint16_t cri_index = BE_GET_CRI_FROM_CID(cid);
phwi_ctrlr = phba->phwi_ctrlr;
pwrb_context = &phwi_ctrlr->wrb_context[cri_index];
/* return the context address */
*pcontext = pwrb_context;
return beiscsi_get_wrb_handle(pwrb_context, phba->params.wrbs_per_cxn);
}
static inline void
beiscsi_put_wrb_handle(struct hwi_wrb_context *pwrb_context,
struct wrb_handle *pwrb_handle,
unsigned int wrbs_per_cxn)
{
scsi: be2iscsi: Replace _bh with _irqsave/irqrestore [ 3843.132217] WARNING: CPU: 20 PID: 1227 at kernel/softirq.c:150 __local_bh_enable_ip+0x6b/0x90 [ 3843.142815] Modules linked in: ... [ 3843.294328] CPU: 20 PID: 1227 Comm: kworker/20:1H Tainted: G E 4.8.0-rc1+ #3 [ 3843.304944] Hardware name: Dell Inc. PowerEdge R720/0X6H47, BIOS 1.4.8 10/25/2012 [ 3843.314798] Workqueue: kblockd blk_timeout_work [ 3843.321350] 0000000000000086 00000000a32f4533 ffff8802216d7bd8 ffffffff8135c3cf [ 3843.331146] 0000000000000000 0000000000000000 ffff8802216d7c18 ffffffff8108d661 [ 3843.340918] 00000096216d7c50 0000000000000200 ffff8802d07cc828 ffff8801b3632550 [ 3843.350687] Call Trace: [ 3843.354866] [<ffffffff8135c3cf>] dump_stack+0x63/0x84 [ 3843.362061] [<ffffffff8108d661>] __warn+0xd1/0xf0 [ 3843.368851] [<ffffffff8108d79d>] warn_slowpath_null+0x1d/0x20 [ 3843.376791] [<ffffffff810930eb>] __local_bh_enable_ip+0x6b/0x90 [ 3843.384903] [<ffffffff816fe7be>] _raw_spin_unlock_bh+0x1e/0x20 [ 3843.392940] [<ffffffffa085f710>] beiscsi_alloc_pdu+0x2f0/0x6e0 [be2iscsi] [ 3843.402076] [<ffffffffa06bc358>] __iscsi_conn_send_pdu+0xf8/0x370 [libiscsi] [ 3843.411549] [<ffffffffa06bc6fe>] iscsi_send_nopout+0xbe/0x110 [libiscsi] [ 3843.420639] [<ffffffffa06bd98b>] iscsi_eh_cmd_timed_out+0x29b/0x2b0 [libiscsi] [ 3843.430339] [<ffffffff814cd1de>] scsi_times_out+0x5e/0x250 [ 3843.438119] [<ffffffff813374af>] blk_rq_timed_out+0x1f/0x60 [ 3843.446009] [<ffffffff8133759d>] blk_timeout_work+0xad/0x150 [ 3843.454010] [<ffffffff810a6642>] process_one_work+0x152/0x400 [ 3843.462114] [<ffffffff810a6f35>] worker_thread+0x125/0x4b0 [ 3843.469961] [<ffffffff810a6e10>] ? rescuer_thread+0x380/0x380 [ 3843.478116] [<ffffffff810aca28>] kthread+0xd8/0xf0 [ 3843.485212] [<ffffffff816fedff>] ret_from_fork+0x1f/0x40 [ 3843.492908] [<ffffffff810ac950>] ? kthread_park+0x60/0x60 [ 3843.500715] ---[ end trace 57ec0a1d8f0dd3a0 ]--- [ 3852.328667] NMI watchdog: Watchdog detected hard LOCKUP on cpu 1Kernel panic - not syncing: Hard LOCKUP blk_timeout_work takes queue_lock spin_lock with interrupts disabled before invoking iscsi_eh_cmd_timed_out. This causes a WARN_ON_ONCE in spin_unlock_bh for wrb_lock/io_sgl_lock/mgmt_sgl_lock. CPU was kept busy in lot of bottom half work with interrupts disabled thus causing hard lock up. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Chris Leech <cleech@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-10-13 14:38:49 +08:00
unsigned long flags;
spin_lock_irqsave(&pwrb_context->wrb_lock, flags);
pwrb_context->pwrb_handle_base[pwrb_context->free_index] = pwrb_handle;
pwrb_context->wrb_handles_available++;
if (pwrb_context->free_index == (wrbs_per_cxn - 1))
pwrb_context->free_index = 0;
else
pwrb_context->free_index++;
pwrb_handle->pio_handle = NULL;
scsi: be2iscsi: Replace _bh with _irqsave/irqrestore [ 3843.132217] WARNING: CPU: 20 PID: 1227 at kernel/softirq.c:150 __local_bh_enable_ip+0x6b/0x90 [ 3843.142815] Modules linked in: ... [ 3843.294328] CPU: 20 PID: 1227 Comm: kworker/20:1H Tainted: G E 4.8.0-rc1+ #3 [ 3843.304944] Hardware name: Dell Inc. PowerEdge R720/0X6H47, BIOS 1.4.8 10/25/2012 [ 3843.314798] Workqueue: kblockd blk_timeout_work [ 3843.321350] 0000000000000086 00000000a32f4533 ffff8802216d7bd8 ffffffff8135c3cf [ 3843.331146] 0000000000000000 0000000000000000 ffff8802216d7c18 ffffffff8108d661 [ 3843.340918] 00000096216d7c50 0000000000000200 ffff8802d07cc828 ffff8801b3632550 [ 3843.350687] Call Trace: [ 3843.354866] [<ffffffff8135c3cf>] dump_stack+0x63/0x84 [ 3843.362061] [<ffffffff8108d661>] __warn+0xd1/0xf0 [ 3843.368851] [<ffffffff8108d79d>] warn_slowpath_null+0x1d/0x20 [ 3843.376791] [<ffffffff810930eb>] __local_bh_enable_ip+0x6b/0x90 [ 3843.384903] [<ffffffff816fe7be>] _raw_spin_unlock_bh+0x1e/0x20 [ 3843.392940] [<ffffffffa085f710>] beiscsi_alloc_pdu+0x2f0/0x6e0 [be2iscsi] [ 3843.402076] [<ffffffffa06bc358>] __iscsi_conn_send_pdu+0xf8/0x370 [libiscsi] [ 3843.411549] [<ffffffffa06bc6fe>] iscsi_send_nopout+0xbe/0x110 [libiscsi] [ 3843.420639] [<ffffffffa06bd98b>] iscsi_eh_cmd_timed_out+0x29b/0x2b0 [libiscsi] [ 3843.430339] [<ffffffff814cd1de>] scsi_times_out+0x5e/0x250 [ 3843.438119] [<ffffffff813374af>] blk_rq_timed_out+0x1f/0x60 [ 3843.446009] [<ffffffff8133759d>] blk_timeout_work+0xad/0x150 [ 3843.454010] [<ffffffff810a6642>] process_one_work+0x152/0x400 [ 3843.462114] [<ffffffff810a6f35>] worker_thread+0x125/0x4b0 [ 3843.469961] [<ffffffff810a6e10>] ? rescuer_thread+0x380/0x380 [ 3843.478116] [<ffffffff810aca28>] kthread+0xd8/0xf0 [ 3843.485212] [<ffffffff816fedff>] ret_from_fork+0x1f/0x40 [ 3843.492908] [<ffffffff810ac950>] ? kthread_park+0x60/0x60 [ 3843.500715] ---[ end trace 57ec0a1d8f0dd3a0 ]--- [ 3852.328667] NMI watchdog: Watchdog detected hard LOCKUP on cpu 1Kernel panic - not syncing: Hard LOCKUP blk_timeout_work takes queue_lock spin_lock with interrupts disabled before invoking iscsi_eh_cmd_timed_out. This causes a WARN_ON_ONCE in spin_unlock_bh for wrb_lock/io_sgl_lock/mgmt_sgl_lock. CPU was kept busy in lot of bottom half work with interrupts disabled thus causing hard lock up. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Chris Leech <cleech@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-10-13 14:38:49 +08:00
spin_unlock_irqrestore(&pwrb_context->wrb_lock, flags);
}
/**
* free_wrb_handle - To free the wrb handle back to pool
* @phba: The hba pointer
* @pwrb_context: The context to free from
* @pwrb_handle: The wrb_handle to free
*
* This happens under session_lock until submission to chip
*/
static void
free_wrb_handle(struct beiscsi_hba *phba, struct hwi_wrb_context *pwrb_context,
struct wrb_handle *pwrb_handle)
{
beiscsi_put_wrb_handle(pwrb_context,
pwrb_handle,
phba->params.wrbs_per_cxn);
beiscsi_log(phba, KERN_INFO,
BEISCSI_LOG_IO | BEISCSI_LOG_CONFIG,
"BM_%d : FREE WRB: pwrb_handle=%p free_index=0x%x"
"wrb_handles_available=%d\n",
pwrb_handle, pwrb_context->free_index,
pwrb_context->wrb_handles_available);
}
static struct sgl_handle *alloc_mgmt_sgl_handle(struct beiscsi_hba *phba)
{
struct sgl_handle *psgl_handle;
scsi: be2iscsi: Replace _bh with _irqsave/irqrestore [ 3843.132217] WARNING: CPU: 20 PID: 1227 at kernel/softirq.c:150 __local_bh_enable_ip+0x6b/0x90 [ 3843.142815] Modules linked in: ... [ 3843.294328] CPU: 20 PID: 1227 Comm: kworker/20:1H Tainted: G E 4.8.0-rc1+ #3 [ 3843.304944] Hardware name: Dell Inc. PowerEdge R720/0X6H47, BIOS 1.4.8 10/25/2012 [ 3843.314798] Workqueue: kblockd blk_timeout_work [ 3843.321350] 0000000000000086 00000000a32f4533 ffff8802216d7bd8 ffffffff8135c3cf [ 3843.331146] 0000000000000000 0000000000000000 ffff8802216d7c18 ffffffff8108d661 [ 3843.340918] 00000096216d7c50 0000000000000200 ffff8802d07cc828 ffff8801b3632550 [ 3843.350687] Call Trace: [ 3843.354866] [<ffffffff8135c3cf>] dump_stack+0x63/0x84 [ 3843.362061] [<ffffffff8108d661>] __warn+0xd1/0xf0 [ 3843.368851] [<ffffffff8108d79d>] warn_slowpath_null+0x1d/0x20 [ 3843.376791] [<ffffffff810930eb>] __local_bh_enable_ip+0x6b/0x90 [ 3843.384903] [<ffffffff816fe7be>] _raw_spin_unlock_bh+0x1e/0x20 [ 3843.392940] [<ffffffffa085f710>] beiscsi_alloc_pdu+0x2f0/0x6e0 [be2iscsi] [ 3843.402076] [<ffffffffa06bc358>] __iscsi_conn_send_pdu+0xf8/0x370 [libiscsi] [ 3843.411549] [<ffffffffa06bc6fe>] iscsi_send_nopout+0xbe/0x110 [libiscsi] [ 3843.420639] [<ffffffffa06bd98b>] iscsi_eh_cmd_timed_out+0x29b/0x2b0 [libiscsi] [ 3843.430339] [<ffffffff814cd1de>] scsi_times_out+0x5e/0x250 [ 3843.438119] [<ffffffff813374af>] blk_rq_timed_out+0x1f/0x60 [ 3843.446009] [<ffffffff8133759d>] blk_timeout_work+0xad/0x150 [ 3843.454010] [<ffffffff810a6642>] process_one_work+0x152/0x400 [ 3843.462114] [<ffffffff810a6f35>] worker_thread+0x125/0x4b0 [ 3843.469961] [<ffffffff810a6e10>] ? rescuer_thread+0x380/0x380 [ 3843.478116] [<ffffffff810aca28>] kthread+0xd8/0xf0 [ 3843.485212] [<ffffffff816fedff>] ret_from_fork+0x1f/0x40 [ 3843.492908] [<ffffffff810ac950>] ? kthread_park+0x60/0x60 [ 3843.500715] ---[ end trace 57ec0a1d8f0dd3a0 ]--- [ 3852.328667] NMI watchdog: Watchdog detected hard LOCKUP on cpu 1Kernel panic - not syncing: Hard LOCKUP blk_timeout_work takes queue_lock spin_lock with interrupts disabled before invoking iscsi_eh_cmd_timed_out. This causes a WARN_ON_ONCE in spin_unlock_bh for wrb_lock/io_sgl_lock/mgmt_sgl_lock. CPU was kept busy in lot of bottom half work with interrupts disabled thus causing hard lock up. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Chris Leech <cleech@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-10-13 14:38:49 +08:00
unsigned long flags;
scsi: be2iscsi: Replace _bh with _irqsave/irqrestore [ 3843.132217] WARNING: CPU: 20 PID: 1227 at kernel/softirq.c:150 __local_bh_enable_ip+0x6b/0x90 [ 3843.142815] Modules linked in: ... [ 3843.294328] CPU: 20 PID: 1227 Comm: kworker/20:1H Tainted: G E 4.8.0-rc1+ #3 [ 3843.304944] Hardware name: Dell Inc. PowerEdge R720/0X6H47, BIOS 1.4.8 10/25/2012 [ 3843.314798] Workqueue: kblockd blk_timeout_work [ 3843.321350] 0000000000000086 00000000a32f4533 ffff8802216d7bd8 ffffffff8135c3cf [ 3843.331146] 0000000000000000 0000000000000000 ffff8802216d7c18 ffffffff8108d661 [ 3843.340918] 00000096216d7c50 0000000000000200 ffff8802d07cc828 ffff8801b3632550 [ 3843.350687] Call Trace: [ 3843.354866] [<ffffffff8135c3cf>] dump_stack+0x63/0x84 [ 3843.362061] [<ffffffff8108d661>] __warn+0xd1/0xf0 [ 3843.368851] [<ffffffff8108d79d>] warn_slowpath_null+0x1d/0x20 [ 3843.376791] [<ffffffff810930eb>] __local_bh_enable_ip+0x6b/0x90 [ 3843.384903] [<ffffffff816fe7be>] _raw_spin_unlock_bh+0x1e/0x20 [ 3843.392940] [<ffffffffa085f710>] beiscsi_alloc_pdu+0x2f0/0x6e0 [be2iscsi] [ 3843.402076] [<ffffffffa06bc358>] __iscsi_conn_send_pdu+0xf8/0x370 [libiscsi] [ 3843.411549] [<ffffffffa06bc6fe>] iscsi_send_nopout+0xbe/0x110 [libiscsi] [ 3843.420639] [<ffffffffa06bd98b>] iscsi_eh_cmd_timed_out+0x29b/0x2b0 [libiscsi] [ 3843.430339] [<ffffffff814cd1de>] scsi_times_out+0x5e/0x250 [ 3843.438119] [<ffffffff813374af>] blk_rq_timed_out+0x1f/0x60 [ 3843.446009] [<ffffffff8133759d>] blk_timeout_work+0xad/0x150 [ 3843.454010] [<ffffffff810a6642>] process_one_work+0x152/0x400 [ 3843.462114] [<ffffffff810a6f35>] worker_thread+0x125/0x4b0 [ 3843.469961] [<ffffffff810a6e10>] ? rescuer_thread+0x380/0x380 [ 3843.478116] [<ffffffff810aca28>] kthread+0xd8/0xf0 [ 3843.485212] [<ffffffff816fedff>] ret_from_fork+0x1f/0x40 [ 3843.492908] [<ffffffff810ac950>] ? kthread_park+0x60/0x60 [ 3843.500715] ---[ end trace 57ec0a1d8f0dd3a0 ]--- [ 3852.328667] NMI watchdog: Watchdog detected hard LOCKUP on cpu 1Kernel panic - not syncing: Hard LOCKUP blk_timeout_work takes queue_lock spin_lock with interrupts disabled before invoking iscsi_eh_cmd_timed_out. This causes a WARN_ON_ONCE in spin_unlock_bh for wrb_lock/io_sgl_lock/mgmt_sgl_lock. CPU was kept busy in lot of bottom half work with interrupts disabled thus causing hard lock up. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Chris Leech <cleech@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-10-13 14:38:49 +08:00
spin_lock_irqsave(&phba->mgmt_sgl_lock, flags);
if (phba->eh_sgl_hndl_avbl) {
psgl_handle = phba->eh_sgl_hndl_base[phba->eh_sgl_alloc_index];
phba->eh_sgl_hndl_base[phba->eh_sgl_alloc_index] = NULL;
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_CONFIG,
"BM_%d : mgmt_sgl_alloc_index=%d=0x%x\n",
phba->eh_sgl_alloc_index,
phba->eh_sgl_alloc_index);
phba->eh_sgl_hndl_avbl--;
if (phba->eh_sgl_alloc_index ==
(phba->params.icds_per_ctrl - phba->params.ios_per_ctrl -
1))
phba->eh_sgl_alloc_index = 0;
else
phba->eh_sgl_alloc_index++;
} else
psgl_handle = NULL;
scsi: be2iscsi: Replace _bh with _irqsave/irqrestore [ 3843.132217] WARNING: CPU: 20 PID: 1227 at kernel/softirq.c:150 __local_bh_enable_ip+0x6b/0x90 [ 3843.142815] Modules linked in: ... [ 3843.294328] CPU: 20 PID: 1227 Comm: kworker/20:1H Tainted: G E 4.8.0-rc1+ #3 [ 3843.304944] Hardware name: Dell Inc. PowerEdge R720/0X6H47, BIOS 1.4.8 10/25/2012 [ 3843.314798] Workqueue: kblockd blk_timeout_work [ 3843.321350] 0000000000000086 00000000a32f4533 ffff8802216d7bd8 ffffffff8135c3cf [ 3843.331146] 0000000000000000 0000000000000000 ffff8802216d7c18 ffffffff8108d661 [ 3843.340918] 00000096216d7c50 0000000000000200 ffff8802d07cc828 ffff8801b3632550 [ 3843.350687] Call Trace: [ 3843.354866] [<ffffffff8135c3cf>] dump_stack+0x63/0x84 [ 3843.362061] [<ffffffff8108d661>] __warn+0xd1/0xf0 [ 3843.368851] [<ffffffff8108d79d>] warn_slowpath_null+0x1d/0x20 [ 3843.376791] [<ffffffff810930eb>] __local_bh_enable_ip+0x6b/0x90 [ 3843.384903] [<ffffffff816fe7be>] _raw_spin_unlock_bh+0x1e/0x20 [ 3843.392940] [<ffffffffa085f710>] beiscsi_alloc_pdu+0x2f0/0x6e0 [be2iscsi] [ 3843.402076] [<ffffffffa06bc358>] __iscsi_conn_send_pdu+0xf8/0x370 [libiscsi] [ 3843.411549] [<ffffffffa06bc6fe>] iscsi_send_nopout+0xbe/0x110 [libiscsi] [ 3843.420639] [<ffffffffa06bd98b>] iscsi_eh_cmd_timed_out+0x29b/0x2b0 [libiscsi] [ 3843.430339] [<ffffffff814cd1de>] scsi_times_out+0x5e/0x250 [ 3843.438119] [<ffffffff813374af>] blk_rq_timed_out+0x1f/0x60 [ 3843.446009] [<ffffffff8133759d>] blk_timeout_work+0xad/0x150 [ 3843.454010] [<ffffffff810a6642>] process_one_work+0x152/0x400 [ 3843.462114] [<ffffffff810a6f35>] worker_thread+0x125/0x4b0 [ 3843.469961] [<ffffffff810a6e10>] ? rescuer_thread+0x380/0x380 [ 3843.478116] [<ffffffff810aca28>] kthread+0xd8/0xf0 [ 3843.485212] [<ffffffff816fedff>] ret_from_fork+0x1f/0x40 [ 3843.492908] [<ffffffff810ac950>] ? kthread_park+0x60/0x60 [ 3843.500715] ---[ end trace 57ec0a1d8f0dd3a0 ]--- [ 3852.328667] NMI watchdog: Watchdog detected hard LOCKUP on cpu 1Kernel panic - not syncing: Hard LOCKUP blk_timeout_work takes queue_lock spin_lock with interrupts disabled before invoking iscsi_eh_cmd_timed_out. This causes a WARN_ON_ONCE in spin_unlock_bh for wrb_lock/io_sgl_lock/mgmt_sgl_lock. CPU was kept busy in lot of bottom half work with interrupts disabled thus causing hard lock up. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Chris Leech <cleech@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-10-13 14:38:49 +08:00
spin_unlock_irqrestore(&phba->mgmt_sgl_lock, flags);
return psgl_handle;
}
void
free_mgmt_sgl_handle(struct beiscsi_hba *phba, struct sgl_handle *psgl_handle)
{
scsi: be2iscsi: Replace _bh with _irqsave/irqrestore [ 3843.132217] WARNING: CPU: 20 PID: 1227 at kernel/softirq.c:150 __local_bh_enable_ip+0x6b/0x90 [ 3843.142815] Modules linked in: ... [ 3843.294328] CPU: 20 PID: 1227 Comm: kworker/20:1H Tainted: G E 4.8.0-rc1+ #3 [ 3843.304944] Hardware name: Dell Inc. PowerEdge R720/0X6H47, BIOS 1.4.8 10/25/2012 [ 3843.314798] Workqueue: kblockd blk_timeout_work [ 3843.321350] 0000000000000086 00000000a32f4533 ffff8802216d7bd8 ffffffff8135c3cf [ 3843.331146] 0000000000000000 0000000000000000 ffff8802216d7c18 ffffffff8108d661 [ 3843.340918] 00000096216d7c50 0000000000000200 ffff8802d07cc828 ffff8801b3632550 [ 3843.350687] Call Trace: [ 3843.354866] [<ffffffff8135c3cf>] dump_stack+0x63/0x84 [ 3843.362061] [<ffffffff8108d661>] __warn+0xd1/0xf0 [ 3843.368851] [<ffffffff8108d79d>] warn_slowpath_null+0x1d/0x20 [ 3843.376791] [<ffffffff810930eb>] __local_bh_enable_ip+0x6b/0x90 [ 3843.384903] [<ffffffff816fe7be>] _raw_spin_unlock_bh+0x1e/0x20 [ 3843.392940] [<ffffffffa085f710>] beiscsi_alloc_pdu+0x2f0/0x6e0 [be2iscsi] [ 3843.402076] [<ffffffffa06bc358>] __iscsi_conn_send_pdu+0xf8/0x370 [libiscsi] [ 3843.411549] [<ffffffffa06bc6fe>] iscsi_send_nopout+0xbe/0x110 [libiscsi] [ 3843.420639] [<ffffffffa06bd98b>] iscsi_eh_cmd_timed_out+0x29b/0x2b0 [libiscsi] [ 3843.430339] [<ffffffff814cd1de>] scsi_times_out+0x5e/0x250 [ 3843.438119] [<ffffffff813374af>] blk_rq_timed_out+0x1f/0x60 [ 3843.446009] [<ffffffff8133759d>] blk_timeout_work+0xad/0x150 [ 3843.454010] [<ffffffff810a6642>] process_one_work+0x152/0x400 [ 3843.462114] [<ffffffff810a6f35>] worker_thread+0x125/0x4b0 [ 3843.469961] [<ffffffff810a6e10>] ? rescuer_thread+0x380/0x380 [ 3843.478116] [<ffffffff810aca28>] kthread+0xd8/0xf0 [ 3843.485212] [<ffffffff816fedff>] ret_from_fork+0x1f/0x40 [ 3843.492908] [<ffffffff810ac950>] ? kthread_park+0x60/0x60 [ 3843.500715] ---[ end trace 57ec0a1d8f0dd3a0 ]--- [ 3852.328667] NMI watchdog: Watchdog detected hard LOCKUP on cpu 1Kernel panic - not syncing: Hard LOCKUP blk_timeout_work takes queue_lock spin_lock with interrupts disabled before invoking iscsi_eh_cmd_timed_out. This causes a WARN_ON_ONCE in spin_unlock_bh for wrb_lock/io_sgl_lock/mgmt_sgl_lock. CPU was kept busy in lot of bottom half work with interrupts disabled thus causing hard lock up. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Chris Leech <cleech@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-10-13 14:38:49 +08:00
unsigned long flags;
spin_lock_irqsave(&phba->mgmt_sgl_lock, flags);
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_CONFIG,
"BM_%d : In free_mgmt_sgl_handle,"
"eh_sgl_free_index=%d\n",
phba->eh_sgl_free_index);
if (phba->eh_sgl_hndl_base[phba->eh_sgl_free_index]) {
/*
* this can happen if clean_task is called on a task that
* failed in xmit_task or alloc_pdu.
*/
beiscsi_log(phba, KERN_WARNING, BEISCSI_LOG_CONFIG,
"BM_%d : Double Free in eh SGL ,"
"eh_sgl_free_index=%d\n",
phba->eh_sgl_free_index);
scsi: be2iscsi: Replace _bh with _irqsave/irqrestore [ 3843.132217] WARNING: CPU: 20 PID: 1227 at kernel/softirq.c:150 __local_bh_enable_ip+0x6b/0x90 [ 3843.142815] Modules linked in: ... [ 3843.294328] CPU: 20 PID: 1227 Comm: kworker/20:1H Tainted: G E 4.8.0-rc1+ #3 [ 3843.304944] Hardware name: Dell Inc. PowerEdge R720/0X6H47, BIOS 1.4.8 10/25/2012 [ 3843.314798] Workqueue: kblockd blk_timeout_work [ 3843.321350] 0000000000000086 00000000a32f4533 ffff8802216d7bd8 ffffffff8135c3cf [ 3843.331146] 0000000000000000 0000000000000000 ffff8802216d7c18 ffffffff8108d661 [ 3843.340918] 00000096216d7c50 0000000000000200 ffff8802d07cc828 ffff8801b3632550 [ 3843.350687] Call Trace: [ 3843.354866] [<ffffffff8135c3cf>] dump_stack+0x63/0x84 [ 3843.362061] [<ffffffff8108d661>] __warn+0xd1/0xf0 [ 3843.368851] [<ffffffff8108d79d>] warn_slowpath_null+0x1d/0x20 [ 3843.376791] [<ffffffff810930eb>] __local_bh_enable_ip+0x6b/0x90 [ 3843.384903] [<ffffffff816fe7be>] _raw_spin_unlock_bh+0x1e/0x20 [ 3843.392940] [<ffffffffa085f710>] beiscsi_alloc_pdu+0x2f0/0x6e0 [be2iscsi] [ 3843.402076] [<ffffffffa06bc358>] __iscsi_conn_send_pdu+0xf8/0x370 [libiscsi] [ 3843.411549] [<ffffffffa06bc6fe>] iscsi_send_nopout+0xbe/0x110 [libiscsi] [ 3843.420639] [<ffffffffa06bd98b>] iscsi_eh_cmd_timed_out+0x29b/0x2b0 [libiscsi] [ 3843.430339] [<ffffffff814cd1de>] scsi_times_out+0x5e/0x250 [ 3843.438119] [<ffffffff813374af>] blk_rq_timed_out+0x1f/0x60 [ 3843.446009] [<ffffffff8133759d>] blk_timeout_work+0xad/0x150 [ 3843.454010] [<ffffffff810a6642>] process_one_work+0x152/0x400 [ 3843.462114] [<ffffffff810a6f35>] worker_thread+0x125/0x4b0 [ 3843.469961] [<ffffffff810a6e10>] ? rescuer_thread+0x380/0x380 [ 3843.478116] [<ffffffff810aca28>] kthread+0xd8/0xf0 [ 3843.485212] [<ffffffff816fedff>] ret_from_fork+0x1f/0x40 [ 3843.492908] [<ffffffff810ac950>] ? kthread_park+0x60/0x60 [ 3843.500715] ---[ end trace 57ec0a1d8f0dd3a0 ]--- [ 3852.328667] NMI watchdog: Watchdog detected hard LOCKUP on cpu 1Kernel panic - not syncing: Hard LOCKUP blk_timeout_work takes queue_lock spin_lock with interrupts disabled before invoking iscsi_eh_cmd_timed_out. This causes a WARN_ON_ONCE in spin_unlock_bh for wrb_lock/io_sgl_lock/mgmt_sgl_lock. CPU was kept busy in lot of bottom half work with interrupts disabled thus causing hard lock up. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Chris Leech <cleech@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-10-13 14:38:49 +08:00
spin_unlock_irqrestore(&phba->mgmt_sgl_lock, flags);
return;
}
phba->eh_sgl_hndl_base[phba->eh_sgl_free_index] = psgl_handle;
phba->eh_sgl_hndl_avbl++;
if (phba->eh_sgl_free_index ==
(phba->params.icds_per_ctrl - phba->params.ios_per_ctrl - 1))
phba->eh_sgl_free_index = 0;
else
phba->eh_sgl_free_index++;
scsi: be2iscsi: Replace _bh with _irqsave/irqrestore [ 3843.132217] WARNING: CPU: 20 PID: 1227 at kernel/softirq.c:150 __local_bh_enable_ip+0x6b/0x90 [ 3843.142815] Modules linked in: ... [ 3843.294328] CPU: 20 PID: 1227 Comm: kworker/20:1H Tainted: G E 4.8.0-rc1+ #3 [ 3843.304944] Hardware name: Dell Inc. PowerEdge R720/0X6H47, BIOS 1.4.8 10/25/2012 [ 3843.314798] Workqueue: kblockd blk_timeout_work [ 3843.321350] 0000000000000086 00000000a32f4533 ffff8802216d7bd8 ffffffff8135c3cf [ 3843.331146] 0000000000000000 0000000000000000 ffff8802216d7c18 ffffffff8108d661 [ 3843.340918] 00000096216d7c50 0000000000000200 ffff8802d07cc828 ffff8801b3632550 [ 3843.350687] Call Trace: [ 3843.354866] [<ffffffff8135c3cf>] dump_stack+0x63/0x84 [ 3843.362061] [<ffffffff8108d661>] __warn+0xd1/0xf0 [ 3843.368851] [<ffffffff8108d79d>] warn_slowpath_null+0x1d/0x20 [ 3843.376791] [<ffffffff810930eb>] __local_bh_enable_ip+0x6b/0x90 [ 3843.384903] [<ffffffff816fe7be>] _raw_spin_unlock_bh+0x1e/0x20 [ 3843.392940] [<ffffffffa085f710>] beiscsi_alloc_pdu+0x2f0/0x6e0 [be2iscsi] [ 3843.402076] [<ffffffffa06bc358>] __iscsi_conn_send_pdu+0xf8/0x370 [libiscsi] [ 3843.411549] [<ffffffffa06bc6fe>] iscsi_send_nopout+0xbe/0x110 [libiscsi] [ 3843.420639] [<ffffffffa06bd98b>] iscsi_eh_cmd_timed_out+0x29b/0x2b0 [libiscsi] [ 3843.430339] [<ffffffff814cd1de>] scsi_times_out+0x5e/0x250 [ 3843.438119] [<ffffffff813374af>] blk_rq_timed_out+0x1f/0x60 [ 3843.446009] [<ffffffff8133759d>] blk_timeout_work+0xad/0x150 [ 3843.454010] [<ffffffff810a6642>] process_one_work+0x152/0x400 [ 3843.462114] [<ffffffff810a6f35>] worker_thread+0x125/0x4b0 [ 3843.469961] [<ffffffff810a6e10>] ? rescuer_thread+0x380/0x380 [ 3843.478116] [<ffffffff810aca28>] kthread+0xd8/0xf0 [ 3843.485212] [<ffffffff816fedff>] ret_from_fork+0x1f/0x40 [ 3843.492908] [<ffffffff810ac950>] ? kthread_park+0x60/0x60 [ 3843.500715] ---[ end trace 57ec0a1d8f0dd3a0 ]--- [ 3852.328667] NMI watchdog: Watchdog detected hard LOCKUP on cpu 1Kernel panic - not syncing: Hard LOCKUP blk_timeout_work takes queue_lock spin_lock with interrupts disabled before invoking iscsi_eh_cmd_timed_out. This causes a WARN_ON_ONCE in spin_unlock_bh for wrb_lock/io_sgl_lock/mgmt_sgl_lock. CPU was kept busy in lot of bottom half work with interrupts disabled thus causing hard lock up. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Chris Leech <cleech@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-10-13 14:38:49 +08:00
spin_unlock_irqrestore(&phba->mgmt_sgl_lock, flags);
}
static void
be_complete_io(struct beiscsi_conn *beiscsi_conn,
struct iscsi_task *task,
struct common_sol_cqe *csol_cqe)
{
struct beiscsi_io_task *io_task = task->dd_data;
struct be_status_bhs *sts_bhs =
(struct be_status_bhs *)io_task->cmd_bhs;
struct iscsi_conn *conn = beiscsi_conn->conn;
unsigned char *sense;
u32 resid = 0, exp_cmdsn, max_cmdsn;
u8 rsp, status, flags;
exp_cmdsn = csol_cqe->exp_cmdsn;
max_cmdsn = (csol_cqe->exp_cmdsn +
csol_cqe->cmd_wnd - 1);
rsp = csol_cqe->i_resp;
status = csol_cqe->i_sts;
flags = csol_cqe->i_flags;
resid = csol_cqe->res_cnt;
if (!task->sc) {
if (io_task->scsi_cmnd) {
scsi_dma_unmap(io_task->scsi_cmnd);
io_task->scsi_cmnd = NULL;
}
return;
}
task->sc->result = (DID_OK << 16) | status;
if (rsp != ISCSI_STATUS_CMD_COMPLETED) {
task->sc->result = DID_ERROR << 16;
goto unmap;
}
/* bidi not initially supported */
if (flags & (ISCSI_FLAG_CMD_UNDERFLOW | ISCSI_FLAG_CMD_OVERFLOW)) {
if (!status && (flags & ISCSI_FLAG_CMD_OVERFLOW))
task->sc->result = DID_ERROR << 16;
if (flags & ISCSI_FLAG_CMD_UNDERFLOW) {
scsi_set_resid(task->sc, resid);
if (!status && (scsi_bufflen(task->sc) - resid <
task->sc->underflow))
task->sc->result = DID_ERROR << 16;
}
}
if (status == SAM_STAT_CHECK_CONDITION) {
u16 sense_len;
unsigned short *slen = (unsigned short *)sts_bhs->sense_info;
sense = sts_bhs->sense_info + sizeof(unsigned short);
sense_len = be16_to_cpu(*slen);
memcpy(task->sc->sense_buffer, sense,
min_t(u16, sense_len, SCSI_SENSE_BUFFERSIZE));
}
if (io_task->cmd_bhs->iscsi_hdr.flags & ISCSI_FLAG_CMD_READ)
conn->rxdata_octets += resid;
unmap:
if (io_task->scsi_cmnd) {
scsi_dma_unmap(io_task->scsi_cmnd);
io_task->scsi_cmnd = NULL;
}
iscsi_complete_scsi_task(task, exp_cmdsn, max_cmdsn);
}
static void
be_complete_logout(struct beiscsi_conn *beiscsi_conn,
struct iscsi_task *task,
struct common_sol_cqe *csol_cqe)
{
struct iscsi_logout_rsp *hdr;
struct beiscsi_io_task *io_task = task->dd_data;
struct iscsi_conn *conn = beiscsi_conn->conn;
hdr = (struct iscsi_logout_rsp *)task->hdr;
hdr->opcode = ISCSI_OP_LOGOUT_RSP;
hdr->t2wait = 5;
hdr->t2retain = 0;
hdr->flags = csol_cqe->i_flags;
hdr->response = csol_cqe->i_resp;
hdr->exp_cmdsn = cpu_to_be32(csol_cqe->exp_cmdsn);
hdr->max_cmdsn = cpu_to_be32(csol_cqe->exp_cmdsn +
csol_cqe->cmd_wnd - 1);
hdr->dlength[0] = 0;
hdr->dlength[1] = 0;
hdr->dlength[2] = 0;
hdr->hlength = 0;
hdr->itt = io_task->libiscsi_itt;
__iscsi_complete_pdu(conn, (struct iscsi_hdr *)hdr, NULL, 0);
}
static void
be_complete_tmf(struct beiscsi_conn *beiscsi_conn,
struct iscsi_task *task,
struct common_sol_cqe *csol_cqe)
{
struct iscsi_tm_rsp *hdr;
struct iscsi_conn *conn = beiscsi_conn->conn;
struct beiscsi_io_task *io_task = task->dd_data;
hdr = (struct iscsi_tm_rsp *)task->hdr;
hdr->opcode = ISCSI_OP_SCSI_TMFUNC_RSP;
hdr->flags = csol_cqe->i_flags;
hdr->response = csol_cqe->i_resp;
hdr->exp_cmdsn = cpu_to_be32(csol_cqe->exp_cmdsn);
hdr->max_cmdsn = cpu_to_be32(csol_cqe->exp_cmdsn +
csol_cqe->cmd_wnd - 1);
hdr->itt = io_task->libiscsi_itt;
__iscsi_complete_pdu(conn, (struct iscsi_hdr *)hdr, NULL, 0);
}
static void
hwi_complete_drvr_msgs(struct beiscsi_conn *beiscsi_conn,
struct beiscsi_hba *phba, struct sol_cqe *psol)
{
struct hwi_wrb_context *pwrb_context;
uint16_t wrb_index, cid, cri_index;
struct hwi_controller *phwi_ctrlr;
struct wrb_handle *pwrb_handle;
struct iscsi_session *session;
struct iscsi_task *task;
phwi_ctrlr = phba->phwi_ctrlr;
if (is_chip_be2_be3r(phba)) {
wrb_index = AMAP_GET_BITS(struct amap_it_dmsg_cqe,
wrb_idx, psol);
cid = AMAP_GET_BITS(struct amap_it_dmsg_cqe,
cid, psol);
} else {
wrb_index = AMAP_GET_BITS(struct amap_it_dmsg_cqe_v2,
wrb_idx, psol);
cid = AMAP_GET_BITS(struct amap_it_dmsg_cqe_v2,
cid, psol);
}
cri_index = BE_GET_CRI_FROM_CID(cid);
pwrb_context = &phwi_ctrlr->wrb_context[cri_index];
pwrb_handle = pwrb_context->pwrb_handle_basestd[wrb_index];
session = beiscsi_conn->conn->session;
spin_lock_bh(&session->back_lock);
task = pwrb_handle->pio_handle;
if (task)
__iscsi_put_task(task);
spin_unlock_bh(&session->back_lock);
}
static void
be_complete_nopin_resp(struct beiscsi_conn *beiscsi_conn,
struct iscsi_task *task,
struct common_sol_cqe *csol_cqe)
{
struct iscsi_nopin *hdr;
struct iscsi_conn *conn = beiscsi_conn->conn;
struct beiscsi_io_task *io_task = task->dd_data;
hdr = (struct iscsi_nopin *)task->hdr;
hdr->flags = csol_cqe->i_flags;
hdr->exp_cmdsn = cpu_to_be32(csol_cqe->exp_cmdsn);
hdr->max_cmdsn = cpu_to_be32(csol_cqe->exp_cmdsn +
csol_cqe->cmd_wnd - 1);
hdr->opcode = ISCSI_OP_NOOP_IN;
hdr->itt = io_task->libiscsi_itt;
__iscsi_complete_pdu(conn, (struct iscsi_hdr *)hdr, NULL, 0);
}
static void adapter_get_sol_cqe(struct beiscsi_hba *phba,
struct sol_cqe *psol,
struct common_sol_cqe *csol_cqe)
{
if (is_chip_be2_be3r(phba)) {
csol_cqe->exp_cmdsn = AMAP_GET_BITS(struct amap_sol_cqe,
i_exp_cmd_sn, psol);
csol_cqe->res_cnt = AMAP_GET_BITS(struct amap_sol_cqe,
i_res_cnt, psol);
csol_cqe->cmd_wnd = AMAP_GET_BITS(struct amap_sol_cqe,
i_cmd_wnd, psol);
csol_cqe->wrb_index = AMAP_GET_BITS(struct amap_sol_cqe,
wrb_index, psol);
csol_cqe->cid = AMAP_GET_BITS(struct amap_sol_cqe,
cid, psol);
csol_cqe->hw_sts = AMAP_GET_BITS(struct amap_sol_cqe,
hw_sts, psol);
csol_cqe->i_resp = AMAP_GET_BITS(struct amap_sol_cqe,
i_resp, psol);
csol_cqe->i_sts = AMAP_GET_BITS(struct amap_sol_cqe,
i_sts, psol);
csol_cqe->i_flags = AMAP_GET_BITS(struct amap_sol_cqe,
i_flags, psol);
} else {
csol_cqe->exp_cmdsn = AMAP_GET_BITS(struct amap_sol_cqe_v2,
i_exp_cmd_sn, psol);
csol_cqe->res_cnt = AMAP_GET_BITS(struct amap_sol_cqe_v2,
i_res_cnt, psol);
csol_cqe->wrb_index = AMAP_GET_BITS(struct amap_sol_cqe_v2,
wrb_index, psol);
csol_cqe->cid = AMAP_GET_BITS(struct amap_sol_cqe_v2,
cid, psol);
csol_cqe->hw_sts = AMAP_GET_BITS(struct amap_sol_cqe_v2,
hw_sts, psol);
csol_cqe->cmd_wnd = AMAP_GET_BITS(struct amap_sol_cqe_v2,
i_cmd_wnd, psol);
if (AMAP_GET_BITS(struct amap_sol_cqe_v2,
cmd_cmpl, psol))
csol_cqe->i_sts = AMAP_GET_BITS(struct amap_sol_cqe_v2,
i_sts, psol);
else
csol_cqe->i_resp = AMAP_GET_BITS(struct amap_sol_cqe_v2,
i_sts, psol);
if (AMAP_GET_BITS(struct amap_sol_cqe_v2,
u, psol))
csol_cqe->i_flags = ISCSI_FLAG_CMD_UNDERFLOW;
if (AMAP_GET_BITS(struct amap_sol_cqe_v2,
o, psol))
csol_cqe->i_flags |= ISCSI_FLAG_CMD_OVERFLOW;
}
}
static void hwi_complete_cmd(struct beiscsi_conn *beiscsi_conn,
struct beiscsi_hba *phba, struct sol_cqe *psol)
{
struct iscsi_conn *conn = beiscsi_conn->conn;
struct iscsi_session *session = conn->session;
struct common_sol_cqe csol_cqe = {0};
struct hwi_wrb_context *pwrb_context;
struct hwi_controller *phwi_ctrlr;
struct wrb_handle *pwrb_handle;
struct iscsi_task *task;
uint16_t cri_index = 0;
uint8_t type;
phwi_ctrlr = phba->phwi_ctrlr;
/* Copy the elements to a common structure */
adapter_get_sol_cqe(phba, psol, &csol_cqe);
cri_index = BE_GET_CRI_FROM_CID(csol_cqe.cid);
pwrb_context = &phwi_ctrlr->wrb_context[cri_index];
pwrb_handle = pwrb_context->pwrb_handle_basestd[
csol_cqe.wrb_index];
spin_lock_bh(&session->back_lock);
task = pwrb_handle->pio_handle;
if (!task) {
spin_unlock_bh(&session->back_lock);
return;
}
type = ((struct beiscsi_io_task *)task->dd_data)->wrb_type;
switch (type) {
case HWH_TYPE_IO:
case HWH_TYPE_IO_RD:
if ((task->hdr->opcode & ISCSI_OPCODE_MASK) ==
ISCSI_OP_NOOP_OUT)
be_complete_nopin_resp(beiscsi_conn, task, &csol_cqe);
else
be_complete_io(beiscsi_conn, task, &csol_cqe);
break;
case HWH_TYPE_LOGOUT:
if ((task->hdr->opcode & ISCSI_OPCODE_MASK) == ISCSI_OP_LOGOUT)
be_complete_logout(beiscsi_conn, task, &csol_cqe);
else
be_complete_tmf(beiscsi_conn, task, &csol_cqe);
break;
case HWH_TYPE_LOGIN:
beiscsi_log(phba, KERN_ERR,
BEISCSI_LOG_CONFIG | BEISCSI_LOG_IO,
"BM_%d :\t\t No HWH_TYPE_LOGIN Expected in"
" hwi_complete_cmd- Solicited path\n");
break;
case HWH_TYPE_NOP:
be_complete_nopin_resp(beiscsi_conn, task, &csol_cqe);
break;
default:
beiscsi_log(phba, KERN_WARNING,
BEISCSI_LOG_CONFIG | BEISCSI_LOG_IO,
"BM_%d : In hwi_complete_cmd, unknown type = %d"
"wrb_index 0x%x CID 0x%x\n", type,
csol_cqe.wrb_index,
csol_cqe.cid);
break;
}
[SCSI] libiscsi: Reduce locking contention in fast path Replace the session lock with two locks, a forward lock and a backwards lock named frwd_lock and back_lock respectively. The forward lock protects resources that change while sending a request to the target, such as cmdsn, queued_cmdsn, and allocating task from the commands' pool with kfifo_out. The backward lock protects resources that change while processing a response or in error path, such as cmdsn_exp, cmdsn_max, and returning tasks to the commands' pool with kfifo_in. Under a steady state fast-path situation, that is when one or more processes/threads submit IO to an iscsi device and a single kernel upcall (e.g softirq) is dealing with processing of responses without errors, this patch eliminates the contention between the queuecommand()/request response/scsi_done() flows associated with iscsi sessions. Between the forward and the backward locks exists a strict locking hierarchy. The mutual exclusion zone protected by the forward lock can enclose the mutual exclusion zone protected by the backward lock but not vice versa. For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is a failure and __iscsi_put_task is called, the backward lock is taken while the forward lock is still taken. On the other hand, if in the RX path a nop is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu than the forward lock is released and the backward lock is taken for the duration of iscsi_send_nopout, later the backward lock is released and the forward lock is retaken. libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue. The insertion and deletion from these queues didn't corespond to the assumption taken by the new forward/backwards session locking paradigm. That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards) path, r2t is taken out from r2t queue and inserted to the r2t pool. In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t is also inserted to the r2t pool and another r2t is pulled from r2t queue. Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue to the TX path, r2t is taken from the r2t pool and inserted to the r2t queue. In order to cope with this situation, two spin locks were added, pool2queue and queue2pool. The former protects extracting from the r2t pool and inserting to the r2t queue, and the later protects the extracing from the r2t queue and inserting to the r2t pool. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> [minor fix up to apply cleanly and compile fix] Signed-off-by: Mike Christie <michaelc@cs.wisc.edu> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 14:41:38 +08:00
spin_unlock_bh(&session->back_lock);
}
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
/**
* ASYNC PDUs include
* a. Unsolicited NOP-In (target initiated NOP-In)
* b. ASYNC Messages
* c. Reject PDU
* d. Login response
* These headers arrive unprocessed by the EP firmware.
* iSCSI layer processes them.
*/
static unsigned int
beiscsi_complete_pdu(struct beiscsi_conn *beiscsi_conn,
struct pdu_base *phdr, void *pdata, unsigned int dlen)
{
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
struct beiscsi_hba *phba = beiscsi_conn->phba;
struct iscsi_conn *conn = beiscsi_conn->conn;
struct beiscsi_io_task *io_task;
struct iscsi_hdr *login_hdr;
struct iscsi_task *task;
u8 code;
code = AMAP_GET_BITS(struct amap_pdu_base, opcode, phdr);
switch (code) {
case ISCSI_OP_NOOP_IN:
pdata = NULL;
dlen = 0;
break;
case ISCSI_OP_ASYNC_EVENT:
break;
case ISCSI_OP_REJECT:
WARN_ON(!pdata);
WARN_ON(!(dlen == 48));
beiscsi_log(phba, KERN_ERR,
BEISCSI_LOG_CONFIG | BEISCSI_LOG_IO,
"BM_%d : In ISCSI_OP_REJECT\n");
break;
case ISCSI_OP_LOGIN_RSP:
case ISCSI_OP_TEXT_RSP:
task = conn->login_task;
io_task = task->dd_data;
login_hdr = (struct iscsi_hdr *)phdr;
login_hdr->itt = io_task->libiscsi_itt;
break;
default:
beiscsi_log(phba, KERN_WARNING,
BEISCSI_LOG_IO | BEISCSI_LOG_CONFIG,
"BM_%d : unrecognized async PDU opcode 0x%x\n",
code);
return 1;
}
__iscsi_complete_pdu(conn, (struct iscsi_hdr *)phdr, pdata, dlen);
return 0;
}
static inline void
beiscsi_hdl_put_handle(struct hd_async_context *pasync_ctx,
struct hd_async_handle *pasync_handle)
{
pasync_handle->is_final = 0;
pasync_handle->buffer_len = 0;
pasync_handle->in_use = 0;
list_del_init(&pasync_handle->link);
}
static void
beiscsi_hdl_purge_handles(struct beiscsi_hba *phba,
struct hd_async_context *pasync_ctx,
u16 cri)
{
struct hd_async_handle *pasync_handle, *tmp_handle;
struct list_head *plist;
plist = &pasync_ctx->async_entry[cri].wq.list;
list_for_each_entry_safe(pasync_handle, tmp_handle, plist, link)
beiscsi_hdl_put_handle(pasync_ctx, pasync_handle);
INIT_LIST_HEAD(&pasync_ctx->async_entry[cri].wq.list);
pasync_ctx->async_entry[cri].wq.hdr_len = 0;
pasync_ctx->async_entry[cri].wq.bytes_received = 0;
pasync_ctx->async_entry[cri].wq.bytes_needed = 0;
}
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
static struct hd_async_handle *
beiscsi_hdl_get_handle(struct beiscsi_conn *beiscsi_conn,
struct hd_async_context *pasync_ctx,
struct i_t_dpdu_cqe *pdpdu_cqe,
u8 *header)
{
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
struct beiscsi_hba *phba = beiscsi_conn->phba;
struct hd_async_handle *pasync_handle;
struct be_bus_address phys_addr;
u16 cid, code, ci, cri;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
u8 final, error = 0;
u32 dpl;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
cid = beiscsi_conn->beiscsi_conn_cid;
cri = BE_GET_ASYNC_CRI_FROM_CID(cid);
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
/**
* This function is invoked to get the right async_handle structure
* from a given DEF PDU CQ entry.
*
* - index in CQ entry gives the vertical index
* - address in CQ entry is the offset where the DMA last ended
* - final - no more notifications for this PDU
*/
if (is_chip_be2_be3r(phba)) {
dpl = AMAP_GET_BITS(struct amap_i_t_dpdu_cqe,
dpl, pdpdu_cqe);
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
ci = AMAP_GET_BITS(struct amap_i_t_dpdu_cqe,
index, pdpdu_cqe);
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
final = AMAP_GET_BITS(struct amap_i_t_dpdu_cqe,
final, pdpdu_cqe);
} else {
dpl = AMAP_GET_BITS(struct amap_i_t_dpdu_cqe_v2,
dpl, pdpdu_cqe);
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
ci = AMAP_GET_BITS(struct amap_i_t_dpdu_cqe_v2,
index, pdpdu_cqe);
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
final = AMAP_GET_BITS(struct amap_i_t_dpdu_cqe_v2,
final, pdpdu_cqe);
}
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
/**
* DB addr Hi/Lo is same for BE and SKH.
* Subtract the dataplacementlength to get to the base.
*/
phys_addr.u.a32.address_lo = AMAP_GET_BITS(struct amap_i_t_dpdu_cqe,
db_addr_lo, pdpdu_cqe);
phys_addr.u.a32.address_lo -= dpl;
phys_addr.u.a32.address_hi = AMAP_GET_BITS(struct amap_i_t_dpdu_cqe,
db_addr_hi, pdpdu_cqe);
code = AMAP_GET_BITS(struct amap_i_t_dpdu_cqe, code, pdpdu_cqe);
switch (code) {
case UNSOL_HDR_NOTIFY:
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
pasync_handle = pasync_ctx->async_entry[ci].header;
*header = 1;
break;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
case UNSOL_DATA_DIGEST_ERROR_NOTIFY:
error = 1;
case UNSOL_DATA_NOTIFY:
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
pasync_handle = pasync_ctx->async_entry[ci].data;
break;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
/* called only for above codes */
default:
return NULL;
}
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
if (pasync_handle->pa.u.a64.address != phys_addr.u.a64.address ||
pasync_handle->index != ci) {
/* driver bug - if ci does not match async handle index */
error = 1;
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_ISCSI,
"BM_%d : cid %u async PDU handle mismatch - addr in %cQE %llx at %u:addr in CQE %llx ci %u\n",
cid, pasync_handle->is_header ? 'H' : 'D',
pasync_handle->pa.u.a64.address,
pasync_handle->index,
phys_addr.u.a64.address, ci);
/* FW has stale address - attempt continuing by dropping */
}
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
/**
* DEF PDU header and data buffers with errors should be simply
* dropped as there are no consumers for it.
*/
if (error) {
beiscsi_hdl_put_handle(pasync_ctx, pasync_handle);
return NULL;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
}
if (pasync_handle->in_use || !list_empty(&pasync_handle->link)) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_ISCSI,
"BM_%d : cid %d async PDU handle in use - code %d ci %d addr %llx\n",
cid, code, ci, phys_addr.u.a64.address);
beiscsi_hdl_purge_handles(phba, pasync_ctx, cri);
}
list_del_init(&pasync_handle->link);
/**
* Each CID is associated with unique CRI.
* ASYNC_CRI_FROM_CID mapping and CRI_FROM_CID are totaly different.
**/
pasync_handle->cri = cri;
pasync_handle->is_final = final;
pasync_handle->buffer_len = dpl;
pasync_handle->in_use = 1;
return pasync_handle;
}
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
static unsigned int
beiscsi_hdl_fwd_pdu(struct beiscsi_conn *beiscsi_conn,
struct hd_async_context *pasync_ctx,
u16 cri)
{
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
struct iscsi_session *session = beiscsi_conn->conn->session;
struct hd_async_handle *pasync_handle, *plast_handle;
struct beiscsi_hba *phba = beiscsi_conn->phba;
void *phdr = NULL, *pdata = NULL;
u32 dlen = 0, status = 0;
struct list_head *plist;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
plist = &pasync_ctx->async_entry[cri].wq.list;
plast_handle = NULL;
list_for_each_entry(pasync_handle, plist, link) {
plast_handle = pasync_handle;
/* get the header, the first entry */
if (!phdr) {
phdr = pasync_handle->pbuffer;
continue;
}
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
/* use first buffer to collect all the data */
if (!pdata) {
pdata = pasync_handle->pbuffer;
dlen = pasync_handle->buffer_len;
continue;
}
if (!pasync_handle->buffer_len ||
(dlen + pasync_handle->buffer_len) >
pasync_ctx->async_data.buffer_size)
break;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
memcpy(pdata + dlen, pasync_handle->pbuffer,
pasync_handle->buffer_len);
dlen += pasync_handle->buffer_len;
}
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
if (!plast_handle->is_final) {
/* last handle should have final PDU notification from FW */
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_ISCSI,
"BM_%d : cid %u %p fwd async PDU opcode %x with last handle missing - HL%u:DN%u:DR%u\n",
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
beiscsi_conn->beiscsi_conn_cid, plast_handle,
AMAP_GET_BITS(struct amap_pdu_base, opcode, phdr),
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
pasync_ctx->async_entry[cri].wq.hdr_len,
pasync_ctx->async_entry[cri].wq.bytes_needed,
pasync_ctx->async_entry[cri].wq.bytes_received);
}
spin_lock_bh(&session->back_lock);
status = beiscsi_complete_pdu(beiscsi_conn, phdr, pdata, dlen);
spin_unlock_bh(&session->back_lock);
beiscsi_hdl_purge_handles(phba, pasync_ctx, cri);
return status;
}
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
static unsigned int
beiscsi_hdl_gather_pdu(struct beiscsi_conn *beiscsi_conn,
struct hd_async_context *pasync_ctx,
struct hd_async_handle *pasync_handle)
{
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
unsigned int bytes_needed = 0, status = 0;
u16 cri = pasync_handle->cri;
struct cri_wait_queue *wq;
struct beiscsi_hba *phba;
struct pdu_base *ppdu;
char *err = "";
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
phba = beiscsi_conn->phba;
wq = &pasync_ctx->async_entry[cri].wq;
if (pasync_handle->is_header) {
/* check if PDU hdr is rcv'd when old hdr not completed */
if (wq->hdr_len) {
err = "incomplete";
goto drop_pdu;
}
ppdu = pasync_handle->pbuffer;
bytes_needed = AMAP_GET_BITS(struct amap_pdu_base,
data_len_hi, ppdu);
bytes_needed <<= 16;
bytes_needed |= be16_to_cpu(AMAP_GET_BITS(struct amap_pdu_base,
data_len_lo, ppdu));
wq->hdr_len = pasync_handle->buffer_len;
wq->bytes_received = 0;
wq->bytes_needed = bytes_needed;
list_add_tail(&pasync_handle->link, &wq->list);
if (!bytes_needed)
status = beiscsi_hdl_fwd_pdu(beiscsi_conn,
pasync_ctx, cri);
} else {
/* check if data received has header and is needed */
if (!wq->hdr_len || !wq->bytes_needed) {
err = "header less";
goto drop_pdu;
}
wq->bytes_received += pasync_handle->buffer_len;
/* Something got overwritten? Better catch it here. */
if (wq->bytes_received > wq->bytes_needed) {
err = "overflow";
goto drop_pdu;
}
list_add_tail(&pasync_handle->link, &wq->list);
if (wq->bytes_received == wq->bytes_needed)
status = beiscsi_hdl_fwd_pdu(beiscsi_conn,
pasync_ctx, cri);
}
return status;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
drop_pdu:
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_ISCSI,
"BM_%d : cid %u async PDU %s - def-%c:HL%u:DN%u:DR%u\n",
beiscsi_conn->beiscsi_conn_cid, err,
pasync_handle->is_header ? 'H' : 'D',
wq->hdr_len, wq->bytes_needed,
pasync_handle->buffer_len);
/* discard this handle */
beiscsi_hdl_put_handle(pasync_ctx, pasync_handle);
/* free all the other handles in cri_wait_queue */
beiscsi_hdl_purge_handles(phba, pasync_ctx, cri);
/* try continuing */
return status;
}
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
static void
beiscsi_hdq_post_handles(struct beiscsi_hba *phba,
u8 header, u8 ulp_num, u16 nbuf)
{
struct hd_async_handle *pasync_handle;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
struct hd_async_context *pasync_ctx;
struct hwi_controller *phwi_ctrlr;
struct phys_addr *pasync_sge;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
u32 ring_id, doorbell = 0;
u32 doorbell_offset;
u16 prod, pi;
phwi_ctrlr = phba->phwi_ctrlr;
pasync_ctx = HWI_GET_ASYNC_PDU_CTX(phwi_ctrlr, ulp_num);
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
if (header) {
pasync_sge = pasync_ctx->async_header.ring_base;
pi = pasync_ctx->async_header.pi;
ring_id = phwi_ctrlr->default_pdu_hdr[ulp_num].id;
doorbell_offset = phwi_ctrlr->default_pdu_hdr[ulp_num].
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
doorbell_offset;
} else {
pasync_sge = pasync_ctx->async_data.ring_base;
pi = pasync_ctx->async_data.pi;
ring_id = phwi_ctrlr->default_pdu_data[ulp_num].id;
doorbell_offset = phwi_ctrlr->default_pdu_data[ulp_num].
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
doorbell_offset;
}
for (prod = 0; prod < nbuf; prod++) {
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
if (header)
pasync_handle = pasync_ctx->async_entry[pi].header;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
else
pasync_handle = pasync_ctx->async_entry[pi].data;
WARN_ON(pasync_handle->is_header != header);
WARN_ON(pasync_handle->index != pi);
/* setup the ring only once */
if (nbuf == pasync_ctx->num_entries) {
/* note hi is lo */
pasync_sge[pi].hi = pasync_handle->pa.u.a32.address_lo;
pasync_sge[pi].lo = pasync_handle->pa.u.a32.address_hi;
}
if (++pi == pasync_ctx->num_entries)
pi = 0;
}
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
if (header)
pasync_ctx->async_header.pi = pi;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
else
pasync_ctx->async_data.pi = pi;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
doorbell |= ring_id & DB_DEF_PDU_RING_ID_MASK;
doorbell |= 1 << DB_DEF_PDU_REARM_SHIFT;
doorbell |= 0 << DB_DEF_PDU_EVENT_SHIFT;
doorbell |= (prod & DB_DEF_PDU_CQPROC_MASK) << DB_DEF_PDU_CQPROC_SHIFT;
iowrite32(doorbell, phba->db_va + doorbell_offset);
}
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
static void
beiscsi_hdq_process_compl(struct beiscsi_conn *beiscsi_conn,
struct i_t_dpdu_cqe *pdpdu_cqe)
{
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
struct beiscsi_hba *phba = beiscsi_conn->phba;
struct hd_async_handle *pasync_handle = NULL;
struct hd_async_context *pasync_ctx;
struct hwi_controller *phwi_ctrlr;
u8 ulp_num, consumed, header = 0;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
u16 cid_cri;
phwi_ctrlr = phba->phwi_ctrlr;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
cid_cri = BE_GET_CRI_FROM_CID(beiscsi_conn->beiscsi_conn_cid);
ulp_num = BEISCSI_GET_ULP_FROM_CRI(phwi_ctrlr, cid_cri);
pasync_ctx = HWI_GET_ASYNC_PDU_CTX(phwi_ctrlr, ulp_num);
pasync_handle = beiscsi_hdl_get_handle(beiscsi_conn, pasync_ctx,
pdpdu_cqe, &header);
if (is_chip_be2_be3r(phba))
consumed = AMAP_GET_BITS(struct amap_i_t_dpdu_cqe,
num_cons, pdpdu_cqe);
else
consumed = AMAP_GET_BITS(struct amap_i_t_dpdu_cqe_v2,
num_cons, pdpdu_cqe);
if (pasync_handle)
beiscsi_hdl_gather_pdu(beiscsi_conn, pasync_ctx, pasync_handle);
/* num_cons indicates number of 8 RQEs consumed */
if (consumed)
beiscsi_hdq_post_handles(phba, header, ulp_num, 8 * consumed);
}
void beiscsi_process_mcc_cq(struct beiscsi_hba *phba)
{
struct be_queue_info *mcc_cq;
struct be_mcc_compl *mcc_compl;
unsigned int num_processed = 0;
mcc_cq = &phba->ctrl.mcc_obj.cq;
mcc_compl = queue_tail_node(mcc_cq);
mcc_compl->flags = le32_to_cpu(mcc_compl->flags);
while (mcc_compl->flags & CQE_FLAGS_VALID_MASK) {
if (beiscsi_hba_in_error(phba))
return;
if (num_processed >= 32) {
hwi_ring_cq_db(phba, mcc_cq->id,
num_processed, 0);
num_processed = 0;
}
if (mcc_compl->flags & CQE_FLAGS_ASYNC_MASK) {
beiscsi_process_async_event(phba, mcc_compl);
} else if (mcc_compl->flags & CQE_FLAGS_COMPLETED_MASK) {
beiscsi_process_mcc_compl(&phba->ctrl, mcc_compl);
}
mcc_compl->flags = 0;
queue_tail_inc(mcc_cq);
mcc_compl = queue_tail_node(mcc_cq);
mcc_compl->flags = le32_to_cpu(mcc_compl->flags);
num_processed++;
}
if (num_processed > 0)
hwi_ring_cq_db(phba, mcc_cq->id, num_processed, 1);
}
static void beiscsi_mcc_work(struct work_struct *work)
{
struct be_eq_obj *pbe_eq;
struct beiscsi_hba *phba;
pbe_eq = container_of(work, struct be_eq_obj, mcc_work);
phba = pbe_eq->phba;
beiscsi_process_mcc_cq(phba);
/* rearm EQ for further interrupts */
if (!beiscsi_hba_in_error(phba))
hwi_ring_eq_db(phba, pbe_eq->q.id, 0, 0, 1, 1);
}
/**
* beiscsi_process_cq()- Process the Completion Queue
* @pbe_eq: Event Q on which the Completion has come
* @budget: Max number of events to processed
*
* return
* Number of Completion Entries processed.
**/
unsigned int beiscsi_process_cq(struct be_eq_obj *pbe_eq, int budget)
{
struct be_queue_info *cq;
struct sol_cqe *sol;
struct dmsg_cqe *dmsg;
unsigned int total = 0;
unsigned int num_processed = 0;
unsigned short code = 0, cid = 0;
uint16_t cri_index = 0;
struct beiscsi_conn *beiscsi_conn;
struct beiscsi_endpoint *beiscsi_ep;
struct iscsi_endpoint *ep;
struct beiscsi_hba *phba;
cq = pbe_eq->cq;
sol = queue_tail_node(cq);
phba = pbe_eq->phba;
while (sol->dw[offsetof(struct amap_sol_cqe, valid) / 32] &
CQE_VALID_MASK) {
if (beiscsi_hba_in_error(phba))
return 0;
be_dws_le_to_cpu(sol, sizeof(struct sol_cqe));
code = (sol->dw[offsetof(struct amap_sol_cqe, code) /
32] & CQE_CODE_MASK);
/* Get the CID */
if (is_chip_be2_be3r(phba)) {
cid = AMAP_GET_BITS(struct amap_sol_cqe, cid, sol);
} else {
if ((code == DRIVERMSG_NOTIFY) ||
(code == UNSOL_HDR_NOTIFY) ||
(code == UNSOL_DATA_NOTIFY))
cid = AMAP_GET_BITS(
struct amap_i_t_dpdu_cqe_v2,
cid, sol);
else
cid = AMAP_GET_BITS(struct amap_sol_cqe_v2,
cid, sol);
}
cri_index = BE_GET_CRI_FROM_CID(cid);
ep = phba->ep_array[cri_index];
if (ep == NULL) {
/* connection has already been freed
* just move on to next one
*/
beiscsi_log(phba, KERN_WARNING,
BEISCSI_LOG_INIT,
"BM_%d : proc cqe of disconn ep: cid %d\n",
cid);
goto proc_next_cqe;
}
beiscsi_ep = ep->dd_data;
beiscsi_conn = beiscsi_ep->conn;
/* replenish cq */
if (num_processed == 32) {
hwi_ring_cq_db(phba, cq->id, 32, 0);
num_processed = 0;
}
total++;
switch (code) {
case SOL_CMD_COMPLETE:
hwi_complete_cmd(beiscsi_conn, phba, sol);
break;
case DRIVERMSG_NOTIFY:
beiscsi_log(phba, KERN_INFO,
BEISCSI_LOG_IO | BEISCSI_LOG_CONFIG,
"BM_%d : Received %s[%d] on CID : %d\n",
cqe_desc[code], code, cid);
dmsg = (struct dmsg_cqe *)sol;
hwi_complete_drvr_msgs(beiscsi_conn, phba, sol);
break;
case UNSOL_HDR_NOTIFY:
beiscsi_log(phba, KERN_INFO,
BEISCSI_LOG_IO | BEISCSI_LOG_CONFIG,
"BM_%d : Received %s[%d] on CID : %d\n",
cqe_desc[code], code, cid);
spin_lock_bh(&phba->async_pdu_lock);
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
beiscsi_hdq_process_compl(beiscsi_conn,
(struct i_t_dpdu_cqe *)sol);
spin_unlock_bh(&phba->async_pdu_lock);
break;
case UNSOL_DATA_NOTIFY:
beiscsi_log(phba, KERN_INFO,
BEISCSI_LOG_CONFIG | BEISCSI_LOG_IO,
"BM_%d : Received %s[%d] on CID : %d\n",
cqe_desc[code], code, cid);
spin_lock_bh(&phba->async_pdu_lock);
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
beiscsi_hdq_process_compl(beiscsi_conn,
(struct i_t_dpdu_cqe *)sol);
spin_unlock_bh(&phba->async_pdu_lock);
break;
case CXN_INVALIDATE_INDEX_NOTIFY:
case CMD_INVALIDATED_NOTIFY:
case CXN_INVALIDATE_NOTIFY:
beiscsi_log(phba, KERN_ERR,
BEISCSI_LOG_IO | BEISCSI_LOG_CONFIG,
"BM_%d : Ignoring %s[%d] on CID : %d\n",
cqe_desc[code], code, cid);
break;
case CXN_KILLED_HDR_DIGEST_ERR:
case SOL_CMD_KILLED_DATA_DIGEST_ERR:
beiscsi_log(phba, KERN_ERR,
BEISCSI_LOG_CONFIG | BEISCSI_LOG_IO,
"BM_%d : Cmd Notification %s[%d] on CID : %d\n",
cqe_desc[code], code, cid);
break;
case CMD_KILLED_INVALID_STATSN_RCVD:
case CMD_KILLED_INVALID_R2T_RCVD:
case CMD_CXN_KILLED_LUN_INVALID:
case CMD_CXN_KILLED_ICD_INVALID:
case CMD_CXN_KILLED_ITT_INVALID:
case CMD_CXN_KILLED_SEQ_OUTOFORDER:
case CMD_CXN_KILLED_INVALID_DATASN_RCVD:
beiscsi_log(phba, KERN_ERR,
BEISCSI_LOG_CONFIG | BEISCSI_LOG_IO,
"BM_%d : Cmd Notification %s[%d] on CID : %d\n",
cqe_desc[code], code, cid);
break;
case UNSOL_DATA_DIGEST_ERROR_NOTIFY:
beiscsi_log(phba, KERN_ERR,
BEISCSI_LOG_IO | BEISCSI_LOG_CONFIG,
"BM_%d : Dropping %s[%d] on DPDU ring on CID : %d\n",
cqe_desc[code], code, cid);
spin_lock_bh(&phba->async_pdu_lock);
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
/* driver consumes the entry and drops the contents */
beiscsi_hdq_process_compl(beiscsi_conn,
(struct i_t_dpdu_cqe *)sol);
spin_unlock_bh(&phba->async_pdu_lock);
break;
case CXN_KILLED_PDU_SIZE_EXCEEDS_DSL:
case CXN_KILLED_BURST_LEN_MISMATCH:
case CXN_KILLED_AHS_RCVD:
case CXN_KILLED_UNKNOWN_HDR:
case CXN_KILLED_STALE_ITT_TTT_RCVD:
case CXN_KILLED_INVALID_ITT_TTT_RCVD:
case CXN_KILLED_TIMED_OUT:
case CXN_KILLED_FIN_RCVD:
case CXN_KILLED_RST_SENT:
case CXN_KILLED_RST_RCVD:
case CXN_KILLED_BAD_UNSOL_PDU_RCVD:
case CXN_KILLED_BAD_WRB_INDEX_ERROR:
case CXN_KILLED_OVER_RUN_RESIDUAL:
case CXN_KILLED_UNDER_RUN_RESIDUAL:
case CXN_KILLED_CMND_DATA_NOT_ON_SAME_CONN:
beiscsi_log(phba, KERN_ERR,
BEISCSI_LOG_IO | BEISCSI_LOG_CONFIG,
"BM_%d : Event %s[%d] received on CID : %d\n",
cqe_desc[code], code, cid);
if (beiscsi_conn)
iscsi_conn_failure(beiscsi_conn->conn,
ISCSI_ERR_CONN_FAILED);
break;
default:
beiscsi_log(phba, KERN_ERR,
BEISCSI_LOG_IO | BEISCSI_LOG_CONFIG,
"BM_%d : Invalid CQE Event Received Code : %d"
"CID 0x%x...\n",
code, cid);
break;
}
proc_next_cqe:
AMAP_SET_BITS(struct amap_sol_cqe, valid, sol, 0);
queue_tail_inc(cq);
sol = queue_tail_node(cq);
num_processed++;
if (total == budget)
break;
}
hwi_ring_cq_db(phba, cq->id, num_processed, 1);
return total;
}
static int be_iopoll(struct irq_poll *iop, int budget)
{
unsigned int ret, io_events;
struct beiscsi_hba *phba;
struct be_eq_obj *pbe_eq;
struct be_eq_entry *eqe = NULL;
struct be_queue_info *eq;
pbe_eq = container_of(iop, struct be_eq_obj, iopoll);
phba = pbe_eq->phba;
if (beiscsi_hba_in_error(phba)) {
irq_poll_complete(iop);
return 0;
}
io_events = 0;
eq = &pbe_eq->q;
eqe = queue_tail_node(eq);
while (eqe->dw[offsetof(struct amap_eq_entry, valid) / 32] &
EQE_VALID_MASK) {
AMAP_SET_BITS(struct amap_eq_entry, valid, eqe, 0);
queue_tail_inc(eq);
eqe = queue_tail_node(eq);
io_events++;
}
hwi_ring_eq_db(phba, eq->id, 1, io_events, 0, 1);
ret = beiscsi_process_cq(pbe_eq, budget);
pbe_eq->cq_count += ret;
if (ret < budget) {
irq_poll_complete(iop);
beiscsi_log(phba, KERN_INFO,
BEISCSI_LOG_CONFIG | BEISCSI_LOG_IO,
"BM_%d : rearm pbe_eq->q.id =%d ret %d\n",
pbe_eq->q.id, ret);
if (!beiscsi_hba_in_error(phba))
hwi_ring_eq_db(phba, pbe_eq->q.id, 0, 0, 1, 1);
}
return ret;
}
static void
hwi_write_sgl_v2(struct iscsi_wrb *pwrb, struct scatterlist *sg,
unsigned int num_sg, struct beiscsi_io_task *io_task)
{
struct iscsi_sge *psgl;
unsigned int sg_len, index;
unsigned int sge_len = 0;
unsigned long long addr;
struct scatterlist *l_sg;
unsigned int offset;
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, iscsi_bhs_addr_lo, pwrb,
io_task->bhs_pa.u.a32.address_lo);
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, iscsi_bhs_addr_hi, pwrb,
io_task->bhs_pa.u.a32.address_hi);
l_sg = sg;
for (index = 0; (index < num_sg) && (index < 2); index++,
sg = sg_next(sg)) {
if (index == 0) {
sg_len = sg_dma_len(sg);
addr = (u64) sg_dma_address(sg);
AMAP_SET_BITS(struct amap_iscsi_wrb_v2,
sge0_addr_lo, pwrb,
lower_32_bits(addr));
AMAP_SET_BITS(struct amap_iscsi_wrb_v2,
sge0_addr_hi, pwrb,
upper_32_bits(addr));
AMAP_SET_BITS(struct amap_iscsi_wrb_v2,
sge0_len, pwrb,
sg_len);
sge_len = sg_len;
} else {
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, sge1_r2t_offset,
pwrb, sge_len);
sg_len = sg_dma_len(sg);
addr = (u64) sg_dma_address(sg);
AMAP_SET_BITS(struct amap_iscsi_wrb_v2,
sge1_addr_lo, pwrb,
lower_32_bits(addr));
AMAP_SET_BITS(struct amap_iscsi_wrb_v2,
sge1_addr_hi, pwrb,
upper_32_bits(addr));
AMAP_SET_BITS(struct amap_iscsi_wrb_v2,
sge1_len, pwrb,
sg_len);
}
}
psgl = (struct iscsi_sge *)io_task->psgl_handle->pfrag;
memset(psgl, 0, sizeof(*psgl) * BE2_SGE);
AMAP_SET_BITS(struct amap_iscsi_sge, len, psgl, io_task->bhs_len - 2);
AMAP_SET_BITS(struct amap_iscsi_sge, addr_hi, psgl,
io_task->bhs_pa.u.a32.address_hi);
AMAP_SET_BITS(struct amap_iscsi_sge, addr_lo, psgl,
io_task->bhs_pa.u.a32.address_lo);
if (num_sg == 1) {
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, sge0_last, pwrb,
1);
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, sge1_last, pwrb,
0);
} else if (num_sg == 2) {
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, sge0_last, pwrb,
0);
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, sge1_last, pwrb,
1);
} else {
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, sge0_last, pwrb,
0);
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, sge1_last, pwrb,
0);
}
sg = l_sg;
psgl++;
psgl++;
offset = 0;
for (index = 0; index < num_sg; index++, sg = sg_next(sg), psgl++) {
sg_len = sg_dma_len(sg);
addr = (u64) sg_dma_address(sg);
AMAP_SET_BITS(struct amap_iscsi_sge, addr_lo, psgl,
lower_32_bits(addr));
AMAP_SET_BITS(struct amap_iscsi_sge, addr_hi, psgl,
upper_32_bits(addr));
AMAP_SET_BITS(struct amap_iscsi_sge, len, psgl, sg_len);
AMAP_SET_BITS(struct amap_iscsi_sge, sge_offset, psgl, offset);
AMAP_SET_BITS(struct amap_iscsi_sge, last_sge, psgl, 0);
offset += sg_len;
}
psgl--;
AMAP_SET_BITS(struct amap_iscsi_sge, last_sge, psgl, 1);
}
static void
hwi_write_sgl(struct iscsi_wrb *pwrb, struct scatterlist *sg,
unsigned int num_sg, struct beiscsi_io_task *io_task)
{
struct iscsi_sge *psgl;
unsigned int sg_len, index;
unsigned int sge_len = 0;
unsigned long long addr;
struct scatterlist *l_sg;
unsigned int offset;
AMAP_SET_BITS(struct amap_iscsi_wrb, iscsi_bhs_addr_lo, pwrb,
io_task->bhs_pa.u.a32.address_lo);
AMAP_SET_BITS(struct amap_iscsi_wrb, iscsi_bhs_addr_hi, pwrb,
io_task->bhs_pa.u.a32.address_hi);
l_sg = sg;
for (index = 0; (index < num_sg) && (index < 2); index++,
sg = sg_next(sg)) {
if (index == 0) {
sg_len = sg_dma_len(sg);
addr = (u64) sg_dma_address(sg);
AMAP_SET_BITS(struct amap_iscsi_wrb, sge0_addr_lo, pwrb,
((u32)(addr & 0xFFFFFFFF)));
AMAP_SET_BITS(struct amap_iscsi_wrb, sge0_addr_hi, pwrb,
((u32)(addr >> 32)));
AMAP_SET_BITS(struct amap_iscsi_wrb, sge0_len, pwrb,
sg_len);
sge_len = sg_len;
} else {
AMAP_SET_BITS(struct amap_iscsi_wrb, sge1_r2t_offset,
pwrb, sge_len);
sg_len = sg_dma_len(sg);
addr = (u64) sg_dma_address(sg);
AMAP_SET_BITS(struct amap_iscsi_wrb, sge1_addr_lo, pwrb,
((u32)(addr & 0xFFFFFFFF)));
AMAP_SET_BITS(struct amap_iscsi_wrb, sge1_addr_hi, pwrb,
((u32)(addr >> 32)));
AMAP_SET_BITS(struct amap_iscsi_wrb, sge1_len, pwrb,
sg_len);
}
}
psgl = (struct iscsi_sge *)io_task->psgl_handle->pfrag;
memset(psgl, 0, sizeof(*psgl) * BE2_SGE);
AMAP_SET_BITS(struct amap_iscsi_sge, len, psgl, io_task->bhs_len - 2);
AMAP_SET_BITS(struct amap_iscsi_sge, addr_hi, psgl,
io_task->bhs_pa.u.a32.address_hi);
AMAP_SET_BITS(struct amap_iscsi_sge, addr_lo, psgl,
io_task->bhs_pa.u.a32.address_lo);
if (num_sg == 1) {
AMAP_SET_BITS(struct amap_iscsi_wrb, sge0_last, pwrb,
1);
AMAP_SET_BITS(struct amap_iscsi_wrb, sge1_last, pwrb,
0);
} else if (num_sg == 2) {
AMAP_SET_BITS(struct amap_iscsi_wrb, sge0_last, pwrb,
0);
AMAP_SET_BITS(struct amap_iscsi_wrb, sge1_last, pwrb,
1);
} else {
AMAP_SET_BITS(struct amap_iscsi_wrb, sge0_last, pwrb,
0);
AMAP_SET_BITS(struct amap_iscsi_wrb, sge1_last, pwrb,
0);
}
sg = l_sg;
psgl++;
psgl++;
offset = 0;
for (index = 0; index < num_sg; index++, sg = sg_next(sg), psgl++) {
sg_len = sg_dma_len(sg);
addr = (u64) sg_dma_address(sg);
AMAP_SET_BITS(struct amap_iscsi_sge, addr_lo, psgl,
(addr & 0xFFFFFFFF));
AMAP_SET_BITS(struct amap_iscsi_sge, addr_hi, psgl,
(addr >> 32));
AMAP_SET_BITS(struct amap_iscsi_sge, len, psgl, sg_len);
AMAP_SET_BITS(struct amap_iscsi_sge, sge_offset, psgl, offset);
AMAP_SET_BITS(struct amap_iscsi_sge, last_sge, psgl, 0);
offset += sg_len;
}
psgl--;
AMAP_SET_BITS(struct amap_iscsi_sge, last_sge, psgl, 1);
}
/**
* hwi_write_buffer()- Populate the WRB with task info
* @pwrb: ptr to the WRB entry
* @task: iscsi task which is to be executed
**/
static int hwi_write_buffer(struct iscsi_wrb *pwrb, struct iscsi_task *task)
{
struct iscsi_sge *psgl;
struct beiscsi_io_task *io_task = task->dd_data;
struct beiscsi_conn *beiscsi_conn = io_task->conn;
struct beiscsi_hba *phba = beiscsi_conn->phba;
uint8_t dsp_value = 0;
io_task->bhs_len = sizeof(struct be_nonio_bhs) - 2;
AMAP_SET_BITS(struct amap_iscsi_wrb, iscsi_bhs_addr_lo, pwrb,
io_task->bhs_pa.u.a32.address_lo);
AMAP_SET_BITS(struct amap_iscsi_wrb, iscsi_bhs_addr_hi, pwrb,
io_task->bhs_pa.u.a32.address_hi);
if (task->data) {
/* Check for the data_count */
dsp_value = (task->data_count) ? 1 : 0;
if (is_chip_be2_be3r(phba))
AMAP_SET_BITS(struct amap_iscsi_wrb, dsp,
pwrb, dsp_value);
else
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, dsp,
pwrb, dsp_value);
/* Map addr only if there is data_count */
if (dsp_value) {
io_task->mtask_addr = pci_map_single(phba->pcidev,
task->data,
task->data_count,
PCI_DMA_TODEVICE);
if (pci_dma_mapping_error(phba->pcidev,
io_task->mtask_addr))
return -ENOMEM;
io_task->mtask_data_count = task->data_count;
} else
io_task->mtask_addr = 0;
AMAP_SET_BITS(struct amap_iscsi_wrb, sge0_addr_lo, pwrb,
lower_32_bits(io_task->mtask_addr));
AMAP_SET_BITS(struct amap_iscsi_wrb, sge0_addr_hi, pwrb,
upper_32_bits(io_task->mtask_addr));
AMAP_SET_BITS(struct amap_iscsi_wrb, sge0_len, pwrb,
task->data_count);
AMAP_SET_BITS(struct amap_iscsi_wrb, sge0_last, pwrb, 1);
} else {
AMAP_SET_BITS(struct amap_iscsi_wrb, dsp, pwrb, 0);
io_task->mtask_addr = 0;
}
psgl = (struct iscsi_sge *)io_task->psgl_handle->pfrag;
AMAP_SET_BITS(struct amap_iscsi_sge, len, psgl, io_task->bhs_len);
AMAP_SET_BITS(struct amap_iscsi_sge, addr_hi, psgl,
io_task->bhs_pa.u.a32.address_hi);
AMAP_SET_BITS(struct amap_iscsi_sge, addr_lo, psgl,
io_task->bhs_pa.u.a32.address_lo);
if (task->data) {
psgl++;
AMAP_SET_BITS(struct amap_iscsi_sge, addr_hi, psgl, 0);
AMAP_SET_BITS(struct amap_iscsi_sge, addr_lo, psgl, 0);
AMAP_SET_BITS(struct amap_iscsi_sge, len, psgl, 0);
AMAP_SET_BITS(struct amap_iscsi_sge, sge_offset, psgl, 0);
AMAP_SET_BITS(struct amap_iscsi_sge, rsvd0, psgl, 0);
AMAP_SET_BITS(struct amap_iscsi_sge, last_sge, psgl, 0);
psgl++;
if (task->data) {
AMAP_SET_BITS(struct amap_iscsi_sge, addr_lo, psgl,
lower_32_bits(io_task->mtask_addr));
AMAP_SET_BITS(struct amap_iscsi_sge, addr_hi, psgl,
upper_32_bits(io_task->mtask_addr));
}
AMAP_SET_BITS(struct amap_iscsi_sge, len, psgl, 0x106);
}
AMAP_SET_BITS(struct amap_iscsi_sge, last_sge, psgl, 1);
return 0;
}
/**
* beiscsi_find_mem_req()- Find mem needed
* @phba: ptr to HBA struct
**/
static void beiscsi_find_mem_req(struct beiscsi_hba *phba)
{
uint8_t mem_descr_index, ulp_num;
unsigned int num_async_pdu_buf_pages;
unsigned int num_async_pdu_data_pages, wrb_sz_per_cxn;
unsigned int num_async_pdu_buf_sgl_pages, num_async_pdu_data_sgl_pages;
phba->params.hwi_ws_sz = sizeof(struct hwi_controller);
phba->mem_req[ISCSI_MEM_GLOBAL_HEADER] = 2 *
BE_ISCSI_PDU_HEADER_SIZE;
phba->mem_req[HWI_MEM_ADDN_CONTEXT] =
sizeof(struct hwi_context_memory);
phba->mem_req[HWI_MEM_WRB] = sizeof(struct iscsi_wrb)
* (phba->params.wrbs_per_cxn)
* phba->params.cxns_per_ctrl;
wrb_sz_per_cxn = sizeof(struct wrb_handle) *
(phba->params.wrbs_per_cxn);
phba->mem_req[HWI_MEM_WRBH] = roundup_pow_of_two((wrb_sz_per_cxn) *
phba->params.cxns_per_ctrl);
phba->mem_req[HWI_MEM_SGLH] = sizeof(struct sgl_handle) *
phba->params.icds_per_ctrl;
phba->mem_req[HWI_MEM_SGE] = sizeof(struct iscsi_sge) *
phba->params.num_sge_per_io * phba->params.icds_per_ctrl;
for (ulp_num = 0; ulp_num < BEISCSI_ULP_COUNT; ulp_num++) {
if (test_bit(ulp_num, &phba->fw_config.ulp_supported)) {
num_async_pdu_buf_sgl_pages =
PAGES_REQUIRED(BEISCSI_ASYNC_HDQ_SIZE(
phba, ulp_num) *
sizeof(struct phys_addr));
num_async_pdu_buf_pages =
PAGES_REQUIRED(BEISCSI_ASYNC_HDQ_SIZE(
phba, ulp_num) *
phba->params.defpdu_hdr_sz);
num_async_pdu_data_pages =
PAGES_REQUIRED(BEISCSI_ASYNC_HDQ_SIZE(
phba, ulp_num) *
phba->params.defpdu_data_sz);
num_async_pdu_data_sgl_pages =
PAGES_REQUIRED(BEISCSI_ASYNC_HDQ_SIZE(
phba, ulp_num) *
sizeof(struct phys_addr));
mem_descr_index = (HWI_MEM_TEMPLATE_HDR_ULP0 +
(ulp_num * MEM_DESCR_OFFSET));
phba->mem_req[mem_descr_index] =
BEISCSI_GET_CID_COUNT(phba, ulp_num) *
BEISCSI_TEMPLATE_HDR_PER_CXN_SIZE;
mem_descr_index = (HWI_MEM_ASYNC_HEADER_BUF_ULP0 +
(ulp_num * MEM_DESCR_OFFSET));
phba->mem_req[mem_descr_index] =
num_async_pdu_buf_pages *
PAGE_SIZE;
mem_descr_index = (HWI_MEM_ASYNC_DATA_BUF_ULP0 +
(ulp_num * MEM_DESCR_OFFSET));
phba->mem_req[mem_descr_index] =
num_async_pdu_data_pages *
PAGE_SIZE;
mem_descr_index = (HWI_MEM_ASYNC_HEADER_RING_ULP0 +
(ulp_num * MEM_DESCR_OFFSET));
phba->mem_req[mem_descr_index] =
num_async_pdu_buf_sgl_pages *
PAGE_SIZE;
mem_descr_index = (HWI_MEM_ASYNC_DATA_RING_ULP0 +
(ulp_num * MEM_DESCR_OFFSET));
phba->mem_req[mem_descr_index] =
num_async_pdu_data_sgl_pages *
PAGE_SIZE;
mem_descr_index = (HWI_MEM_ASYNC_HEADER_HANDLE_ULP0 +
(ulp_num * MEM_DESCR_OFFSET));
phba->mem_req[mem_descr_index] =
BEISCSI_ASYNC_HDQ_SIZE(phba, ulp_num) *
sizeof(struct hd_async_handle);
mem_descr_index = (HWI_MEM_ASYNC_DATA_HANDLE_ULP0 +
(ulp_num * MEM_DESCR_OFFSET));
phba->mem_req[mem_descr_index] =
BEISCSI_ASYNC_HDQ_SIZE(phba, ulp_num) *
sizeof(struct hd_async_handle);
mem_descr_index = (HWI_MEM_ASYNC_PDU_CONTEXT_ULP0 +
(ulp_num * MEM_DESCR_OFFSET));
phba->mem_req[mem_descr_index] =
sizeof(struct hd_async_context) +
(BEISCSI_ASYNC_HDQ_SIZE(phba, ulp_num) *
sizeof(struct hd_async_entry));
}
}
}
static int beiscsi_alloc_mem(struct beiscsi_hba *phba)
{
dma_addr_t bus_add;
struct hwi_controller *phwi_ctrlr;
struct be_mem_descriptor *mem_descr;
struct mem_array *mem_arr, *mem_arr_orig;
unsigned int i, j, alloc_size, curr_alloc_size;
phba->phwi_ctrlr = kzalloc(phba->params.hwi_ws_sz, GFP_KERNEL);
if (!phba->phwi_ctrlr)
return -ENOMEM;
/* Allocate memory for wrb_context */
phwi_ctrlr = phba->phwi_ctrlr;
phwi_ctrlr->wrb_context = kzalloc(sizeof(struct hwi_wrb_context) *
phba->params.cxns_per_ctrl,
GFP_KERNEL);
if (!phwi_ctrlr->wrb_context) {
kfree(phba->phwi_ctrlr);
return -ENOMEM;
}
phba->init_mem = kcalloc(SE_MEM_MAX, sizeof(*mem_descr),
GFP_KERNEL);
if (!phba->init_mem) {
kfree(phwi_ctrlr->wrb_context);
kfree(phba->phwi_ctrlr);
return -ENOMEM;
}
mem_arr_orig = kmalloc(sizeof(*mem_arr_orig) * BEISCSI_MAX_FRAGS_INIT,
GFP_KERNEL);
if (!mem_arr_orig) {
kfree(phba->init_mem);
kfree(phwi_ctrlr->wrb_context);
kfree(phba->phwi_ctrlr);
return -ENOMEM;
}
mem_descr = phba->init_mem;
for (i = 0; i < SE_MEM_MAX; i++) {
if (!phba->mem_req[i]) {
mem_descr->mem_array = NULL;
mem_descr++;
continue;
}
j = 0;
mem_arr = mem_arr_orig;
alloc_size = phba->mem_req[i];
memset(mem_arr, 0, sizeof(struct mem_array) *
BEISCSI_MAX_FRAGS_INIT);
curr_alloc_size = min(be_max_phys_size * 1024, alloc_size);
do {
mem_arr->virtual_address = pci_alloc_consistent(
phba->pcidev,
curr_alloc_size,
&bus_add);
if (!mem_arr->virtual_address) {
if (curr_alloc_size <= BE_MIN_MEM_SIZE)
goto free_mem;
if (curr_alloc_size -
rounddown_pow_of_two(curr_alloc_size))
curr_alloc_size = rounddown_pow_of_two
(curr_alloc_size);
else
curr_alloc_size = curr_alloc_size / 2;
} else {
mem_arr->bus_address.u.
a64.address = (__u64) bus_add;
mem_arr->size = curr_alloc_size;
alloc_size -= curr_alloc_size;
curr_alloc_size = min(be_max_phys_size *
1024, alloc_size);
j++;
mem_arr++;
}
} while (alloc_size);
mem_descr->num_elements = j;
mem_descr->size_in_bytes = phba->mem_req[i];
mem_descr->mem_array = kmalloc(sizeof(*mem_arr) * j,
GFP_KERNEL);
if (!mem_descr->mem_array)
goto free_mem;
memcpy(mem_descr->mem_array, mem_arr_orig,
sizeof(struct mem_array) * j);
mem_descr++;
}
kfree(mem_arr_orig);
return 0;
free_mem:
mem_descr->num_elements = j;
while ((i) || (j)) {
for (j = mem_descr->num_elements; j > 0; j--) {
pci_free_consistent(phba->pcidev,
mem_descr->mem_array[j - 1].size,
mem_descr->mem_array[j - 1].
virtual_address,
(unsigned long)mem_descr->
mem_array[j - 1].
bus_address.u.a64.address);
}
if (i) {
i--;
kfree(mem_descr->mem_array);
mem_descr--;
}
}
kfree(mem_arr_orig);
kfree(phba->init_mem);
kfree(phba->phwi_ctrlr->wrb_context);
kfree(phba->phwi_ctrlr);
return -ENOMEM;
}
static int beiscsi_get_memory(struct beiscsi_hba *phba)
{
beiscsi_find_mem_req(phba);
return beiscsi_alloc_mem(phba);
}
static void iscsi_init_global_templates(struct beiscsi_hba *phba)
{
struct pdu_data_out *pdata_out;
struct pdu_nop_out *pnop_out;
struct be_mem_descriptor *mem_descr;
mem_descr = phba->init_mem;
mem_descr += ISCSI_MEM_GLOBAL_HEADER;
pdata_out =
(struct pdu_data_out *)mem_descr->mem_array[0].virtual_address;
memset(pdata_out, 0, BE_ISCSI_PDU_HEADER_SIZE);
AMAP_SET_BITS(struct amap_pdu_data_out, opcode, pdata_out,
IIOC_SCSI_DATA);
pnop_out =
(struct pdu_nop_out *)((unsigned char *)mem_descr->mem_array[0].
virtual_address + BE_ISCSI_PDU_HEADER_SIZE);
memset(pnop_out, 0, BE_ISCSI_PDU_HEADER_SIZE);
AMAP_SET_BITS(struct amap_pdu_nop_out, ttt, pnop_out, 0xFFFFFFFF);
AMAP_SET_BITS(struct amap_pdu_nop_out, f_bit, pnop_out, 1);
AMAP_SET_BITS(struct amap_pdu_nop_out, i_bit, pnop_out, 0);
}
static int beiscsi_init_wrb_handle(struct beiscsi_hba *phba)
{
struct be_mem_descriptor *mem_descr_wrbh, *mem_descr_wrb;
struct hwi_context_memory *phwi_ctxt;
struct wrb_handle *pwrb_handle = NULL;
struct hwi_controller *phwi_ctrlr;
struct hwi_wrb_context *pwrb_context;
struct iscsi_wrb *pwrb = NULL;
unsigned int num_cxn_wrbh = 0;
unsigned int num_cxn_wrb = 0, j, idx = 0, index;
mem_descr_wrbh = phba->init_mem;
mem_descr_wrbh += HWI_MEM_WRBH;
mem_descr_wrb = phba->init_mem;
mem_descr_wrb += HWI_MEM_WRB;
phwi_ctrlr = phba->phwi_ctrlr;
/* Allocate memory for WRBQ */
phwi_ctxt = phwi_ctrlr->phwi_ctxt;
phwi_ctxt->be_wrbq = kzalloc(sizeof(struct be_queue_info) *
phba->params.cxns_per_ctrl,
GFP_KERNEL);
if (!phwi_ctxt->be_wrbq) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : WRBQ Mem Alloc Failed\n");
return -ENOMEM;
}
for (index = 0; index < phba->params.cxns_per_ctrl; index++) {
pwrb_context = &phwi_ctrlr->wrb_context[index];
pwrb_context->pwrb_handle_base =
kzalloc(sizeof(struct wrb_handle *) *
phba->params.wrbs_per_cxn, GFP_KERNEL);
if (!pwrb_context->pwrb_handle_base) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : Mem Alloc Failed. Failing to load\n");
goto init_wrb_hndl_failed;
}
pwrb_context->pwrb_handle_basestd =
kzalloc(sizeof(struct wrb_handle *) *
phba->params.wrbs_per_cxn, GFP_KERNEL);
if (!pwrb_context->pwrb_handle_basestd) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : Mem Alloc Failed. Failing to load\n");
goto init_wrb_hndl_failed;
}
if (!num_cxn_wrbh) {
pwrb_handle =
mem_descr_wrbh->mem_array[idx].virtual_address;
num_cxn_wrbh = ((mem_descr_wrbh->mem_array[idx].size) /
((sizeof(struct wrb_handle)) *
phba->params.wrbs_per_cxn));
idx++;
}
pwrb_context->alloc_index = 0;
pwrb_context->wrb_handles_available = 0;
pwrb_context->free_index = 0;
if (num_cxn_wrbh) {
for (j = 0; j < phba->params.wrbs_per_cxn; j++) {
pwrb_context->pwrb_handle_base[j] = pwrb_handle;
pwrb_context->pwrb_handle_basestd[j] =
pwrb_handle;
pwrb_context->wrb_handles_available++;
pwrb_handle->wrb_index = j;
pwrb_handle++;
}
num_cxn_wrbh--;
}
spin_lock_init(&pwrb_context->wrb_lock);
}
idx = 0;
for (index = 0; index < phba->params.cxns_per_ctrl; index++) {
pwrb_context = &phwi_ctrlr->wrb_context[index];
if (!num_cxn_wrb) {
pwrb = mem_descr_wrb->mem_array[idx].virtual_address;
num_cxn_wrb = (mem_descr_wrb->mem_array[idx].size) /
((sizeof(struct iscsi_wrb) *
phba->params.wrbs_per_cxn));
idx++;
}
if (num_cxn_wrb) {
for (j = 0; j < phba->params.wrbs_per_cxn; j++) {
pwrb_handle = pwrb_context->pwrb_handle_base[j];
pwrb_handle->pwrb = pwrb;
pwrb++;
}
num_cxn_wrb--;
}
}
return 0;
init_wrb_hndl_failed:
for (j = index; j > 0; j--) {
pwrb_context = &phwi_ctrlr->wrb_context[j];
kfree(pwrb_context->pwrb_handle_base);
kfree(pwrb_context->pwrb_handle_basestd);
}
return -ENOMEM;
}
static int hwi_init_async_pdu_ctx(struct beiscsi_hba *phba)
{
uint8_t ulp_num;
struct hwi_controller *phwi_ctrlr;
struct hba_parameters *p = &phba->params;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
struct hd_async_context *pasync_ctx;
struct hd_async_handle *pasync_header_h, *pasync_data_h;
unsigned int index, idx, num_per_mem, num_async_data;
struct be_mem_descriptor *mem_descr;
for (ulp_num = 0; ulp_num < BEISCSI_ULP_COUNT; ulp_num++) {
if (test_bit(ulp_num, &phba->fw_config.ulp_supported)) {
/* get async_ctx for each ULP */
mem_descr = (struct be_mem_descriptor *)phba->init_mem;
mem_descr += (HWI_MEM_ASYNC_PDU_CONTEXT_ULP0 +
(ulp_num * MEM_DESCR_OFFSET));
phwi_ctrlr = phba->phwi_ctrlr;
phwi_ctrlr->phwi_ctxt->pasync_ctx[ulp_num] =
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
(struct hd_async_context *)
mem_descr->mem_array[0].virtual_address;
pasync_ctx = phwi_ctrlr->phwi_ctxt->pasync_ctx[ulp_num];
memset(pasync_ctx, 0, sizeof(*pasync_ctx));
pasync_ctx->async_entry =
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
(struct hd_async_entry *)
((long unsigned int)pasync_ctx +
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
sizeof(struct hd_async_context));
pasync_ctx->num_entries = BEISCSI_ASYNC_HDQ_SIZE(phba,
ulp_num);
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
/* setup header buffers */
mem_descr = (struct be_mem_descriptor *)phba->init_mem;
mem_descr += HWI_MEM_ASYNC_HEADER_BUF_ULP0 +
(ulp_num * MEM_DESCR_OFFSET);
if (mem_descr->mem_array[0].virtual_address) {
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_INIT,
"BM_%d : hwi_init_async_pdu_ctx"
" HWI_MEM_ASYNC_HEADER_BUF_ULP%d va=%p\n",
ulp_num,
mem_descr->mem_array[0].
virtual_address);
} else
beiscsi_log(phba, KERN_WARNING,
BEISCSI_LOG_INIT,
"BM_%d : No Virtual address for ULP : %d\n",
ulp_num);
pasync_ctx->async_header.pi = 0;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
pasync_ctx->async_header.buffer_size = p->defpdu_hdr_sz;
pasync_ctx->async_header.va_base =
mem_descr->mem_array[0].virtual_address;
pasync_ctx->async_header.pa_base.u.a64.address =
mem_descr->mem_array[0].
bus_address.u.a64.address;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
/* setup header buffer sgls */
mem_descr = (struct be_mem_descriptor *)phba->init_mem;
mem_descr += HWI_MEM_ASYNC_HEADER_RING_ULP0 +
(ulp_num * MEM_DESCR_OFFSET);
if (mem_descr->mem_array[0].virtual_address) {
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_INIT,
"BM_%d : hwi_init_async_pdu_ctx"
" HWI_MEM_ASYNC_HEADER_RING_ULP%d va=%p\n",
ulp_num,
mem_descr->mem_array[0].
virtual_address);
} else
beiscsi_log(phba, KERN_WARNING,
BEISCSI_LOG_INIT,
"BM_%d : No Virtual address for ULP : %d\n",
ulp_num);
pasync_ctx->async_header.ring_base =
mem_descr->mem_array[0].virtual_address;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
/* setup header buffer handles */
mem_descr = (struct be_mem_descriptor *)phba->init_mem;
mem_descr += HWI_MEM_ASYNC_HEADER_HANDLE_ULP0 +
(ulp_num * MEM_DESCR_OFFSET);
if (mem_descr->mem_array[0].virtual_address) {
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_INIT,
"BM_%d : hwi_init_async_pdu_ctx"
" HWI_MEM_ASYNC_HEADER_HANDLE_ULP%d va=%p\n",
ulp_num,
mem_descr->mem_array[0].
virtual_address);
} else
beiscsi_log(phba, KERN_WARNING,
BEISCSI_LOG_INIT,
"BM_%d : No Virtual address for ULP : %d\n",
ulp_num);
pasync_ctx->async_header.handle_base =
mem_descr->mem_array[0].virtual_address;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
/* setup data buffer sgls */
mem_descr = (struct be_mem_descriptor *)phba->init_mem;
mem_descr += HWI_MEM_ASYNC_DATA_RING_ULP0 +
(ulp_num * MEM_DESCR_OFFSET);
if (mem_descr->mem_array[0].virtual_address) {
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_INIT,
"BM_%d : hwi_init_async_pdu_ctx"
" HWI_MEM_ASYNC_DATA_RING_ULP%d va=%p\n",
ulp_num,
mem_descr->mem_array[0].
virtual_address);
} else
beiscsi_log(phba, KERN_WARNING,
BEISCSI_LOG_INIT,
"BM_%d : No Virtual address for ULP : %d\n",
ulp_num);
pasync_ctx->async_data.ring_base =
mem_descr->mem_array[0].virtual_address;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
/* setup data buffer handles */
mem_descr = (struct be_mem_descriptor *)phba->init_mem;
mem_descr += HWI_MEM_ASYNC_DATA_HANDLE_ULP0 +
(ulp_num * MEM_DESCR_OFFSET);
if (!mem_descr->mem_array[0].virtual_address)
beiscsi_log(phba, KERN_WARNING,
BEISCSI_LOG_INIT,
"BM_%d : No Virtual address for ULP : %d\n",
ulp_num);
pasync_ctx->async_data.handle_base =
mem_descr->mem_array[0].virtual_address;
pasync_header_h =
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
(struct hd_async_handle *)
pasync_ctx->async_header.handle_base;
pasync_data_h =
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
(struct hd_async_handle *)
pasync_ctx->async_data.handle_base;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
/* setup data buffers */
mem_descr = (struct be_mem_descriptor *)phba->init_mem;
mem_descr += HWI_MEM_ASYNC_DATA_BUF_ULP0 +
(ulp_num * MEM_DESCR_OFFSET);
if (mem_descr->mem_array[0].virtual_address) {
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_INIT,
"BM_%d : hwi_init_async_pdu_ctx"
" HWI_MEM_ASYNC_DATA_BUF_ULP%d va=%p\n",
ulp_num,
mem_descr->mem_array[0].
virtual_address);
} else
beiscsi_log(phba, KERN_WARNING,
BEISCSI_LOG_INIT,
"BM_%d : No Virtual address for ULP : %d\n",
ulp_num);
idx = 0;
pasync_ctx->async_data.pi = 0;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
pasync_ctx->async_data.buffer_size = p->defpdu_data_sz;
pasync_ctx->async_data.va_base =
mem_descr->mem_array[idx].virtual_address;
pasync_ctx->async_data.pa_base.u.a64.address =
mem_descr->mem_array[idx].
bus_address.u.a64.address;
num_async_data = ((mem_descr->mem_array[idx].size) /
phba->params.defpdu_data_sz);
num_per_mem = 0;
for (index = 0; index < BEISCSI_ASYNC_HDQ_SIZE
(phba, ulp_num); index++) {
pasync_header_h->cri = -1;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
pasync_header_h->is_header = 1;
pasync_header_h->index = index;
INIT_LIST_HEAD(&pasync_header_h->link);
pasync_header_h->pbuffer =
(void *)((unsigned long)
(pasync_ctx->
async_header.va_base) +
(p->defpdu_hdr_sz * index));
pasync_header_h->pa.u.a64.address =
pasync_ctx->async_header.pa_base.u.a64.
address + (p->defpdu_hdr_sz * index);
pasync_ctx->async_entry[index].header =
pasync_header_h;
pasync_header_h++;
INIT_LIST_HEAD(&pasync_ctx->async_entry[index].
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
wq.list);
pasync_data_h->cri = -1;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
pasync_data_h->is_header = 0;
pasync_data_h->index = index;
INIT_LIST_HEAD(&pasync_data_h->link);
if (!num_async_data) {
num_per_mem = 0;
idx++;
pasync_ctx->async_data.va_base =
mem_descr->mem_array[idx].
virtual_address;
pasync_ctx->async_data.pa_base.u.
a64.address =
mem_descr->mem_array[idx].
bus_address.u.a64.address;
num_async_data =
((mem_descr->mem_array[idx].
size) /
phba->params.defpdu_data_sz);
}
pasync_data_h->pbuffer =
(void *)((unsigned long)
(pasync_ctx->async_data.va_base) +
(p->defpdu_data_sz * num_per_mem));
pasync_data_h->pa.u.a64.address =
pasync_ctx->async_data.pa_base.u.a64.
address + (p->defpdu_data_sz *
num_per_mem);
num_per_mem++;
num_async_data--;
pasync_ctx->async_entry[index].data =
pasync_data_h;
pasync_data_h++;
}
}
}
return 0;
}
static int
be_sgl_create_contiguous(void *virtual_address,
u64 physical_address, u32 length,
struct be_dma_mem *sgl)
{
WARN_ON(!virtual_address);
WARN_ON(!physical_address);
WARN_ON(!length);
WARN_ON(!sgl);
sgl->va = virtual_address;
sgl->dma = (unsigned long)physical_address;
sgl->size = length;
return 0;
}
static void be_sgl_destroy_contiguous(struct be_dma_mem *sgl)
{
memset(sgl, 0, sizeof(*sgl));
}
static void
hwi_build_be_sgl_arr(struct beiscsi_hba *phba,
struct mem_array *pmem, struct be_dma_mem *sgl)
{
if (sgl->va)
be_sgl_destroy_contiguous(sgl);
be_sgl_create_contiguous(pmem->virtual_address,
pmem->bus_address.u.a64.address,
pmem->size, sgl);
}
static void
hwi_build_be_sgl_by_offset(struct beiscsi_hba *phba,
struct mem_array *pmem, struct be_dma_mem *sgl)
{
if (sgl->va)
be_sgl_destroy_contiguous(sgl);
be_sgl_create_contiguous((unsigned char *)pmem->virtual_address,
pmem->bus_address.u.a64.address,
pmem->size, sgl);
}
static int be_fill_queue(struct be_queue_info *q,
u16 len, u16 entry_size, void *vaddress)
{
struct be_dma_mem *mem = &q->dma_mem;
memset(q, 0, sizeof(*q));
q->len = len;
q->entry_size = entry_size;
mem->size = len * entry_size;
mem->va = vaddress;
if (!mem->va)
return -ENOMEM;
memset(mem->va, 0, mem->size);
return 0;
}
static int beiscsi_create_eqs(struct beiscsi_hba *phba,
struct hwi_context_memory *phwi_context)
{
int ret = -ENOMEM, eq_for_mcc;
unsigned int i, num_eq_pages;
struct be_queue_info *eq;
struct be_dma_mem *mem;
void *eq_vaddress;
dma_addr_t paddr;
num_eq_pages = PAGES_REQUIRED(phba->params.num_eq_entries * \
sizeof(struct be_eq_entry));
if (phba->pcidev->msix_enabled)
eq_for_mcc = 1;
else
eq_for_mcc = 0;
for (i = 0; i < (phba->num_cpus + eq_for_mcc); i++) {
eq = &phwi_context->be_eq[i].q;
mem = &eq->dma_mem;
phwi_context->be_eq[i].phba = phba;
eq_vaddress = pci_alloc_consistent(phba->pcidev,
num_eq_pages * PAGE_SIZE,
&paddr);
if (!eq_vaddress) {
ret = -ENOMEM;
goto create_eq_error;
}
mem->va = eq_vaddress;
ret = be_fill_queue(eq, phba->params.num_eq_entries,
sizeof(struct be_eq_entry), eq_vaddress);
if (ret) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : be_fill_queue Failed for EQ\n");
goto create_eq_error;
}
mem->dma = paddr;
ret = beiscsi_cmd_eq_create(&phba->ctrl, eq,
phwi_context->cur_eqd);
if (ret) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : beiscsi_cmd_eq_create"
"Failed for EQ\n");
goto create_eq_error;
}
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_INIT,
"BM_%d : eqid = %d\n",
phwi_context->be_eq[i].q.id);
}
return 0;
create_eq_error:
for (i = 0; i < (phba->num_cpus + eq_for_mcc); i++) {
eq = &phwi_context->be_eq[i].q;
mem = &eq->dma_mem;
if (mem->va)
pci_free_consistent(phba->pcidev, num_eq_pages
* PAGE_SIZE,
mem->va, mem->dma);
}
return ret;
}
static int beiscsi_create_cqs(struct beiscsi_hba *phba,
struct hwi_context_memory *phwi_context)
{
unsigned int i, num_cq_pages;
struct be_queue_info *cq, *eq;
struct be_dma_mem *mem;
struct be_eq_obj *pbe_eq;
void *cq_vaddress;
int ret = -ENOMEM;
dma_addr_t paddr;
num_cq_pages = PAGES_REQUIRED(phba->params.num_cq_entries * \
sizeof(struct sol_cqe));
for (i = 0; i < phba->num_cpus; i++) {
cq = &phwi_context->be_cq[i];
eq = &phwi_context->be_eq[i].q;
pbe_eq = &phwi_context->be_eq[i];
pbe_eq->cq = cq;
pbe_eq->phba = phba;
mem = &cq->dma_mem;
cq_vaddress = pci_alloc_consistent(phba->pcidev,
num_cq_pages * PAGE_SIZE,
&paddr);
if (!cq_vaddress) {
ret = -ENOMEM;
goto create_cq_error;
}
ret = be_fill_queue(cq, phba->params.num_cq_entries,
sizeof(struct sol_cqe), cq_vaddress);
if (ret) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : be_fill_queue Failed "
"for ISCSI CQ\n");
goto create_cq_error;
}
mem->dma = paddr;
ret = beiscsi_cmd_cq_create(&phba->ctrl, cq, eq, false,
false, 0);
if (ret) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : beiscsi_cmd_eq_create"
"Failed for ISCSI CQ\n");
goto create_cq_error;
}
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_INIT,
"BM_%d : iscsi cq_id is %d for eq_id %d\n"
"iSCSI CQ CREATED\n", cq->id, eq->id);
}
return 0;
create_cq_error:
for (i = 0; i < phba->num_cpus; i++) {
cq = &phwi_context->be_cq[i];
mem = &cq->dma_mem;
if (mem->va)
pci_free_consistent(phba->pcidev, num_cq_pages
* PAGE_SIZE,
mem->va, mem->dma);
}
return ret;
}
static int
beiscsi_create_def_hdr(struct beiscsi_hba *phba,
struct hwi_context_memory *phwi_context,
struct hwi_controller *phwi_ctrlr,
unsigned int def_pdu_ring_sz, uint8_t ulp_num)
{
unsigned int idx;
int ret;
struct be_queue_info *dq, *cq;
struct be_dma_mem *mem;
struct be_mem_descriptor *mem_descr;
void *dq_vaddress;
idx = 0;
dq = &phwi_context->be_def_hdrq[ulp_num];
cq = &phwi_context->be_cq[0];
mem = &dq->dma_mem;
mem_descr = phba->init_mem;
mem_descr += HWI_MEM_ASYNC_HEADER_RING_ULP0 +
(ulp_num * MEM_DESCR_OFFSET);
dq_vaddress = mem_descr->mem_array[idx].virtual_address;
ret = be_fill_queue(dq, mem_descr->mem_array[0].size /
sizeof(struct phys_addr),
sizeof(struct phys_addr), dq_vaddress);
if (ret) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : be_fill_queue Failed for DEF PDU HDR on ULP : %d\n",
ulp_num);
return ret;
}
mem->dma = (unsigned long)mem_descr->mem_array[idx].
bus_address.u.a64.address;
ret = be_cmd_create_default_pdu_queue(&phba->ctrl, cq, dq,
def_pdu_ring_sz,
phba->params.defpdu_hdr_sz,
BEISCSI_DEFQ_HDR, ulp_num);
if (ret) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : be_cmd_create_default_pdu_queue Failed DEFHDR on ULP : %d\n",
ulp_num);
return ret;
}
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_INIT,
"BM_%d : iscsi hdr def pdu id for ULP : %d is %d\n",
ulp_num,
phwi_context->be_def_hdrq[ulp_num].id);
return 0;
}
static int
beiscsi_create_def_data(struct beiscsi_hba *phba,
struct hwi_context_memory *phwi_context,
struct hwi_controller *phwi_ctrlr,
unsigned int def_pdu_ring_sz, uint8_t ulp_num)
{
unsigned int idx;
int ret;
struct be_queue_info *dataq, *cq;
struct be_dma_mem *mem;
struct be_mem_descriptor *mem_descr;
void *dq_vaddress;
idx = 0;
dataq = &phwi_context->be_def_dataq[ulp_num];
cq = &phwi_context->be_cq[0];
mem = &dataq->dma_mem;
mem_descr = phba->init_mem;
mem_descr += HWI_MEM_ASYNC_DATA_RING_ULP0 +
(ulp_num * MEM_DESCR_OFFSET);
dq_vaddress = mem_descr->mem_array[idx].virtual_address;
ret = be_fill_queue(dataq, mem_descr->mem_array[0].size /
sizeof(struct phys_addr),
sizeof(struct phys_addr), dq_vaddress);
if (ret) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : be_fill_queue Failed for DEF PDU "
"DATA on ULP : %d\n",
ulp_num);
return ret;
}
mem->dma = (unsigned long)mem_descr->mem_array[idx].
bus_address.u.a64.address;
ret = be_cmd_create_default_pdu_queue(&phba->ctrl, cq, dataq,
def_pdu_ring_sz,
phba->params.defpdu_data_sz,
BEISCSI_DEFQ_DATA, ulp_num);
if (ret) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d be_cmd_create_default_pdu_queue"
" Failed for DEF PDU DATA on ULP : %d\n",
ulp_num);
return ret;
}
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_INIT,
"BM_%d : iscsi def data id on ULP : %d is %d\n",
ulp_num,
phwi_context->be_def_dataq[ulp_num].id);
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_INIT,
"BM_%d : DEFAULT PDU DATA RING CREATED"
"on ULP : %d\n", ulp_num);
return 0;
}
static int
beiscsi_post_template_hdr(struct beiscsi_hba *phba)
{
struct be_mem_descriptor *mem_descr;
struct mem_array *pm_arr;
struct be_dma_mem sgl;
int status, ulp_num;
for (ulp_num = 0; ulp_num < BEISCSI_ULP_COUNT; ulp_num++) {
if (test_bit(ulp_num, &phba->fw_config.ulp_supported)) {
mem_descr = (struct be_mem_descriptor *)phba->init_mem;
mem_descr += HWI_MEM_TEMPLATE_HDR_ULP0 +
(ulp_num * MEM_DESCR_OFFSET);
pm_arr = mem_descr->mem_array;
hwi_build_be_sgl_arr(phba, pm_arr, &sgl);
status = be_cmd_iscsi_post_template_hdr(
&phba->ctrl, &sgl);
if (status != 0) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : Post Template HDR Failed for"
"ULP_%d\n", ulp_num);
return status;
}
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_INIT,
"BM_%d : Template HDR Pages Posted for"
"ULP_%d\n", ulp_num);
}
}
return 0;
}
static int
beiscsi_post_pages(struct beiscsi_hba *phba)
{
struct be_mem_descriptor *mem_descr;
struct mem_array *pm_arr;
unsigned int page_offset, i;
struct be_dma_mem sgl;
int status, ulp_num = 0;
mem_descr = phba->init_mem;
mem_descr += HWI_MEM_SGE;
pm_arr = mem_descr->mem_array;
for (ulp_num = 0; ulp_num < BEISCSI_ULP_COUNT; ulp_num++)
if (test_bit(ulp_num, &phba->fw_config.ulp_supported))
break;
page_offset = (sizeof(struct iscsi_sge) * phba->params.num_sge_per_io *
phba->fw_config.iscsi_icd_start[ulp_num]) / PAGE_SIZE;
for (i = 0; i < mem_descr->num_elements; i++) {
hwi_build_be_sgl_arr(phba, pm_arr, &sgl);
status = be_cmd_iscsi_post_sgl_pages(&phba->ctrl, &sgl,
page_offset,
(pm_arr->size / PAGE_SIZE));
page_offset += pm_arr->size / PAGE_SIZE;
if (status != 0) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : post sgl failed.\n");
return status;
}
pm_arr++;
}
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_INIT,
"BM_%d : POSTED PAGES\n");
return 0;
}
static void be_queue_free(struct beiscsi_hba *phba, struct be_queue_info *q)
{
struct be_dma_mem *mem = &q->dma_mem;
if (mem->va) {
pci_free_consistent(phba->pcidev, mem->size,
mem->va, mem->dma);
mem->va = NULL;
}
}
static int be_queue_alloc(struct beiscsi_hba *phba, struct be_queue_info *q,
u16 len, u16 entry_size)
{
struct be_dma_mem *mem = &q->dma_mem;
memset(q, 0, sizeof(*q));
q->len = len;
q->entry_size = entry_size;
mem->size = len * entry_size;
mem->va = pci_zalloc_consistent(phba->pcidev, mem->size, &mem->dma);
if (!mem->va)
return -ENOMEM;
return 0;
}
static int
beiscsi_create_wrb_rings(struct beiscsi_hba *phba,
struct hwi_context_memory *phwi_context,
struct hwi_controller *phwi_ctrlr)
{
unsigned int num_wrb_rings;
u64 pa_addr_lo;
unsigned int idx, num, i, ulp_num;
struct mem_array *pwrb_arr;
void *wrb_vaddr;
struct be_dma_mem sgl;
struct be_mem_descriptor *mem_descr;
struct hwi_wrb_context *pwrb_context;
int status;
uint8_t ulp_count = 0, ulp_base_num = 0;
uint16_t cid_count_ulp[BEISCSI_ULP_COUNT] = { 0 };
idx = 0;
mem_descr = phba->init_mem;
mem_descr += HWI_MEM_WRB;
pwrb_arr = kmalloc(sizeof(*pwrb_arr) * phba->params.cxns_per_ctrl,
GFP_KERNEL);
if (!pwrb_arr) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : Memory alloc failed in create wrb ring.\n");
return -ENOMEM;
}
wrb_vaddr = mem_descr->mem_array[idx].virtual_address;
pa_addr_lo = mem_descr->mem_array[idx].bus_address.u.a64.address;
num_wrb_rings = mem_descr->mem_array[idx].size /
(phba->params.wrbs_per_cxn * sizeof(struct iscsi_wrb));
for (num = 0; num < phba->params.cxns_per_ctrl; num++) {
if (num_wrb_rings) {
pwrb_arr[num].virtual_address = wrb_vaddr;
pwrb_arr[num].bus_address.u.a64.address = pa_addr_lo;
pwrb_arr[num].size = phba->params.wrbs_per_cxn *
sizeof(struct iscsi_wrb);
wrb_vaddr += pwrb_arr[num].size;
pa_addr_lo += pwrb_arr[num].size;
num_wrb_rings--;
} else {
idx++;
wrb_vaddr = mem_descr->mem_array[idx].virtual_address;
pa_addr_lo = mem_descr->mem_array[idx].\
bus_address.u.a64.address;
num_wrb_rings = mem_descr->mem_array[idx].size /
(phba->params.wrbs_per_cxn *
sizeof(struct iscsi_wrb));
pwrb_arr[num].virtual_address = wrb_vaddr;
pwrb_arr[num].bus_address.u.a64.address\
= pa_addr_lo;
pwrb_arr[num].size = phba->params.wrbs_per_cxn *
sizeof(struct iscsi_wrb);
wrb_vaddr += pwrb_arr[num].size;
pa_addr_lo += pwrb_arr[num].size;
num_wrb_rings--;
}
}
/* Get the ULP Count */
for (ulp_num = 0; ulp_num < BEISCSI_ULP_COUNT; ulp_num++)
if (test_bit(ulp_num, &phba->fw_config.ulp_supported)) {
ulp_count++;
ulp_base_num = ulp_num;
cid_count_ulp[ulp_num] =
BEISCSI_GET_CID_COUNT(phba, ulp_num);
}
for (i = 0; i < phba->params.cxns_per_ctrl; i++) {
if (ulp_count > 1) {
ulp_base_num = (ulp_base_num + 1) % BEISCSI_ULP_COUNT;
if (!cid_count_ulp[ulp_base_num])
ulp_base_num = (ulp_base_num + 1) %
BEISCSI_ULP_COUNT;
cid_count_ulp[ulp_base_num]--;
}
hwi_build_be_sgl_by_offset(phba, &pwrb_arr[i], &sgl);
status = be_cmd_wrbq_create(&phba->ctrl, &sgl,
&phwi_context->be_wrbq[i],
&phwi_ctrlr->wrb_context[i],
ulp_base_num);
if (status != 0) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : wrbq create failed.");
kfree(pwrb_arr);
return status;
}
pwrb_context = &phwi_ctrlr->wrb_context[i];
BE_SET_CID_TO_CRI(i, pwrb_context->cid);
}
kfree(pwrb_arr);
return 0;
}
static void free_wrb_handles(struct beiscsi_hba *phba)
{
unsigned int index;
struct hwi_controller *phwi_ctrlr;
struct hwi_wrb_context *pwrb_context;
phwi_ctrlr = phba->phwi_ctrlr;
for (index = 0; index < phba->params.cxns_per_ctrl; index++) {
pwrb_context = &phwi_ctrlr->wrb_context[index];
kfree(pwrb_context->pwrb_handle_base);
kfree(pwrb_context->pwrb_handle_basestd);
}
}
static void be_mcc_queues_destroy(struct beiscsi_hba *phba)
{
struct be_ctrl_info *ctrl = &phba->ctrl;
struct be_dma_mem *ptag_mem;
struct be_queue_info *q;
int i, tag;
q = &phba->ctrl.mcc_obj.q;
for (i = 0; i < MAX_MCC_CMD; i++) {
tag = i + 1;
if (!test_bit(MCC_TAG_STATE_RUNNING,
&ctrl->ptag_state[tag].tag_state))
continue;
if (test_bit(MCC_TAG_STATE_TIMEOUT,
&ctrl->ptag_state[tag].tag_state)) {
ptag_mem = &ctrl->ptag_state[tag].tag_mem_state;
if (ptag_mem->size) {
pci_free_consistent(ctrl->pdev,
ptag_mem->size,
ptag_mem->va,
ptag_mem->dma);
ptag_mem->size = 0;
}
continue;
}
/**
* If MCC is still active and waiting then wake up the process.
* We are here only because port is going offline. The process
* sees that (BEISCSI_HBA_ONLINE is cleared) and EIO error is
* returned for the operation and allocated memory cleaned up.
*/
if (waitqueue_active(&ctrl->mcc_wait[tag])) {
ctrl->mcc_tag_status[tag] = MCC_STATUS_FAILED;
ctrl->mcc_tag_status[tag] |= CQE_VALID_MASK;
wake_up_interruptible(&ctrl->mcc_wait[tag]);
/*
* Control tag info gets reinitialized in enable
* so wait for the process to clear running state.
*/
while (test_bit(MCC_TAG_STATE_RUNNING,
&ctrl->ptag_state[tag].tag_state))
schedule_timeout_uninterruptible(HZ);
}
/**
* For MCC with tag_states MCC_TAG_STATE_ASYNC and
* MCC_TAG_STATE_IGNORE nothing needs to done.
*/
}
if (q->created) {
beiscsi_cmd_q_destroy(ctrl, q, QTYPE_MCCQ);
be_queue_free(phba, q);
}
q = &phba->ctrl.mcc_obj.cq;
if (q->created) {
beiscsi_cmd_q_destroy(ctrl, q, QTYPE_CQ);
be_queue_free(phba, q);
}
}
static int be_mcc_queues_create(struct beiscsi_hba *phba,
struct hwi_context_memory *phwi_context)
{
struct be_queue_info *q, *cq;
struct be_ctrl_info *ctrl = &phba->ctrl;
/* Alloc MCC compl queue */
cq = &phba->ctrl.mcc_obj.cq;
if (be_queue_alloc(phba, cq, MCC_CQ_LEN,
sizeof(struct be_mcc_compl)))
goto err;
/* Ask BE to create MCC compl queue; */
if (phba->pcidev->msix_enabled) {
if (beiscsi_cmd_cq_create(ctrl, cq, &phwi_context->be_eq
[phba->num_cpus].q, false, true, 0))
goto mcc_cq_free;
} else {
if (beiscsi_cmd_cq_create(ctrl, cq, &phwi_context->be_eq[0].q,
false, true, 0))
goto mcc_cq_free;
}
/* Alloc MCC queue */
q = &phba->ctrl.mcc_obj.q;
if (be_queue_alloc(phba, q, MCC_Q_LEN, sizeof(struct be_mcc_wrb)))
goto mcc_cq_destroy;
/* Ask BE to create MCC queue */
if (beiscsi_cmd_mccq_create(phba, q, cq))
goto mcc_q_free;
return 0;
mcc_q_free:
be_queue_free(phba, q);
mcc_cq_destroy:
beiscsi_cmd_q_destroy(ctrl, cq, QTYPE_CQ);
mcc_cq_free:
be_queue_free(phba, cq);
err:
return -ENOMEM;
}
static void be2iscsi_enable_msix(struct beiscsi_hba *phba)
{
int nvec = 1;
switch (phba->generation) {
case BE_GEN2:
case BE_GEN3:
nvec = BEISCSI_MAX_NUM_CPUS + 1;
break;
case BE_GEN4:
nvec = phba->fw_config.eqid_count;
break;
default:
nvec = 2;
break;
}
/* if eqid_count == 1 fall back to INTX */
if (enable_msix && nvec > 1) {
const struct irq_affinity desc = { .post_vectors = 1 };
if (pci_alloc_irq_vectors_affinity(phba->pcidev, 2, nvec,
PCI_IRQ_MSIX | PCI_IRQ_AFFINITY, &desc) < 0) {
phba->num_cpus = nvec - 1;
return;
}
}
phba->num_cpus = 1;
}
static void hwi_purge_eq(struct beiscsi_hba *phba)
{
struct hwi_controller *phwi_ctrlr;
struct hwi_context_memory *phwi_context;
struct be_queue_info *eq;
struct be_eq_entry *eqe = NULL;
int i, eq_msix;
unsigned int num_processed;
if (beiscsi_hba_in_error(phba))
return;
phwi_ctrlr = phba->phwi_ctrlr;
phwi_context = phwi_ctrlr->phwi_ctxt;
if (phba->pcidev->msix_enabled)
eq_msix = 1;
else
eq_msix = 0;
for (i = 0; i < (phba->num_cpus + eq_msix); i++) {
eq = &phwi_context->be_eq[i].q;
eqe = queue_tail_node(eq);
num_processed = 0;
while (eqe->dw[offsetof(struct amap_eq_entry, valid) / 32]
& EQE_VALID_MASK) {
AMAP_SET_BITS(struct amap_eq_entry, valid, eqe, 0);
queue_tail_inc(eq);
eqe = queue_tail_node(eq);
num_processed++;
}
if (num_processed)
hwi_ring_eq_db(phba, eq->id, 1, num_processed, 1, 1);
}
}
static void hwi_cleanup_port(struct beiscsi_hba *phba)
{
struct be_queue_info *q;
struct be_ctrl_info *ctrl = &phba->ctrl;
struct hwi_controller *phwi_ctrlr;
struct hwi_context_memory *phwi_context;
int i, eq_for_mcc, ulp_num;
for (ulp_num = 0; ulp_num < BEISCSI_ULP_COUNT; ulp_num++)
if (test_bit(ulp_num, &phba->fw_config.ulp_supported))
beiscsi_cmd_iscsi_cleanup(phba, ulp_num);
/**
* Purge all EQ entries that may have been left out. This is to
* workaround a problem we've seen occasionally where driver gets an
* interrupt with EQ entry bit set after stopping the controller.
*/
hwi_purge_eq(phba);
phwi_ctrlr = phba->phwi_ctrlr;
phwi_context = phwi_ctrlr->phwi_ctxt;
be_cmd_iscsi_remove_template_hdr(ctrl);
for (i = 0; i < phba->params.cxns_per_ctrl; i++) {
q = &phwi_context->be_wrbq[i];
if (q->created)
beiscsi_cmd_q_destroy(ctrl, q, QTYPE_WRBQ);
}
kfree(phwi_context->be_wrbq);
free_wrb_handles(phba);
for (ulp_num = 0; ulp_num < BEISCSI_ULP_COUNT; ulp_num++) {
if (test_bit(ulp_num, &phba->fw_config.ulp_supported)) {
q = &phwi_context->be_def_hdrq[ulp_num];
if (q->created)
beiscsi_cmd_q_destroy(ctrl, q, QTYPE_DPDUQ);
q = &phwi_context->be_def_dataq[ulp_num];
if (q->created)
beiscsi_cmd_q_destroy(ctrl, q, QTYPE_DPDUQ);
}
}
beiscsi_cmd_q_destroy(ctrl, NULL, QTYPE_SGL);
for (i = 0; i < (phba->num_cpus); i++) {
q = &phwi_context->be_cq[i];
if (q->created) {
be_queue_free(phba, q);
beiscsi_cmd_q_destroy(ctrl, q, QTYPE_CQ);
}
}
be_mcc_queues_destroy(phba);
if (phba->pcidev->msix_enabled)
eq_for_mcc = 1;
else
eq_for_mcc = 0;
for (i = 0; i < (phba->num_cpus + eq_for_mcc); i++) {
q = &phwi_context->be_eq[i].q;
if (q->created) {
be_queue_free(phba, q);
beiscsi_cmd_q_destroy(ctrl, q, QTYPE_EQ);
}
}
/* this ensures complete FW cleanup */
beiscsi_cmd_function_reset(phba);
/* last communication, indicate driver is unloading */
beiscsi_cmd_special_wrb(&phba->ctrl, 0);
}
static int hwi_init_port(struct beiscsi_hba *phba)
{
struct hwi_controller *phwi_ctrlr;
struct hwi_context_memory *phwi_context;
unsigned int def_pdu_ring_sz;
struct be_ctrl_info *ctrl = &phba->ctrl;
int status, ulp_num;
u16 nbufs;
phwi_ctrlr = phba->phwi_ctrlr;
phwi_context = phwi_ctrlr->phwi_ctxt;
phwi_context->max_eqd = 128;
phwi_context->min_eqd = 0;
phwi_context->cur_eqd = 32;
/* set port optic state to unknown */
phba->optic_state = 0xff;
status = beiscsi_create_eqs(phba, phwi_context);
if (status != 0) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : EQ not created\n");
goto error;
}
status = be_mcc_queues_create(phba, phwi_context);
if (status != 0)
goto error;
status = beiscsi_check_supported_fw(ctrl, phba);
if (status != 0) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : Unsupported fw version\n");
goto error;
}
status = beiscsi_create_cqs(phba, phwi_context);
if (status != 0) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : CQ not created\n");
goto error;
}
for (ulp_num = 0; ulp_num < BEISCSI_ULP_COUNT; ulp_num++) {
if (test_bit(ulp_num, &phba->fw_config.ulp_supported)) {
nbufs = phwi_context->pasync_ctx[ulp_num]->num_entries;
def_pdu_ring_sz = nbufs * sizeof(struct phys_addr);
status = beiscsi_create_def_hdr(phba, phwi_context,
phwi_ctrlr,
def_pdu_ring_sz,
ulp_num);
if (status != 0) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : Default Header not created for ULP : %d\n",
ulp_num);
goto error;
}
status = beiscsi_create_def_data(phba, phwi_context,
phwi_ctrlr,
def_pdu_ring_sz,
ulp_num);
if (status != 0) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : Default Data not created for ULP : %d\n",
ulp_num);
goto error;
}
/**
* Now that the default PDU rings have been created,
* let EP know about it.
*/
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
beiscsi_hdq_post_handles(phba, BEISCSI_DEFQ_HDR,
ulp_num, nbufs);
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
beiscsi_hdq_post_handles(phba, BEISCSI_DEFQ_DATA,
ulp_num, nbufs);
}
}
status = beiscsi_post_pages(phba);
if (status != 0) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : Post SGL Pages Failed\n");
goto error;
}
status = beiscsi_post_template_hdr(phba);
if (status != 0) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : Template HDR Posting for CXN Failed\n");
}
status = beiscsi_create_wrb_rings(phba, phwi_context, phwi_ctrlr);
if (status != 0) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : WRB Rings not created\n");
goto error;
}
for (ulp_num = 0; ulp_num < BEISCSI_ULP_COUNT; ulp_num++) {
uint16_t async_arr_idx = 0;
if (test_bit(ulp_num, &phba->fw_config.ulp_supported)) {
uint16_t cri = 0;
scsi: be2iscsi: Fix async PDU handling path BUG: unable to handle kernel NULL pointer dereference at 000000000000015e IP: [<ffffffffa0081700>] hwi_get_async_handle.isra.23.constprop.39+0x90/0x1d0 [be2iscsi] PGD 0 Oops: 0000 [#1] SMP ... Call Trace: <IRQ> [<ffffffffa00818bc>] hwi_process_default_pdu_ring+0x7c/0x280 [be2iscsi] [<ffffffffa0088f51>] beiscsi_process_cq+0x321/0xb90 [be2iscsi] [<ffffffff810af028>] ? __wake_up_common+0x58/0x90 [<ffffffff810b0d84>] ? __wake_up+0x44/0x50 [<ffffffffa0089a2d>] be_iopoll+0x1d/0xb0 [be2iscsi] [<ffffffff812d1f61>] blk_iopoll_softirq+0xc1/0x100 [<ffffffff81084b0f>] __do_softirq+0xef/0x280 The symptom observed is multiple async handles get queued for same index thus causing leak in buffers posted to FW. The root cause is: - async handle is continued to be used even if it does not match the completion. - list_move operation done on already filled index. 1. Remove use of writables, host_write_ptr and ep_read_ptr. 2. Remove consumed logic to update writables. Instead, use only free_entries to do the accounting of handles to be posted back. 3. Remove busy_list, instead use simple slot to index handles. 4. Added check no data, header less and overflow to make sure all async_handles are flushed in error cases. 5. Added code to verify gathering of handles to form PDU by checking final bit before forwarding PDU. 6. Added code to catch mismatch with CQE and handle gracefully. 7. Use AMAP, traverse cri_wait_queue list to post buffers, log "async PDU" related errors. 8. Rearranged few data structures and added comments in init & processing path. 9. Added WARN_ONs to catch any HD ring corruption. Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2016-08-19 17:50:21 +08:00
struct hd_async_context *pasync_ctx;
pasync_ctx = HWI_GET_ASYNC_PDU_CTX(
phwi_ctrlr, ulp_num);
for (cri = 0; cri <
phba->params.cxns_per_ctrl; cri++) {
if (ulp_num == BEISCSI_GET_ULP_FROM_CRI
(phwi_ctrlr, cri))
pasync_ctx->cid_to_async_cri_map[
phwi_ctrlr->wrb_context[cri].cid] =
async_arr_idx++;
}
}
}
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_INIT,
"BM_%d : hwi_init_port success\n");
return 0;
error:
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : hwi_init_port failed");
hwi_cleanup_port(phba);
return status;
}
static int hwi_init_controller(struct beiscsi_hba *phba)
{
struct hwi_controller *phwi_ctrlr;
phwi_ctrlr = phba->phwi_ctrlr;
if (1 == phba->init_mem[HWI_MEM_ADDN_CONTEXT].num_elements) {
phwi_ctrlr->phwi_ctxt = (struct hwi_context_memory *)phba->
init_mem[HWI_MEM_ADDN_CONTEXT].mem_array[0].virtual_address;
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_INIT,
"BM_%d : phwi_ctrlr->phwi_ctxt=%p\n",
phwi_ctrlr->phwi_ctxt);
} else {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : HWI_MEM_ADDN_CONTEXT is more "
"than one element.Failing to load\n");
return -ENOMEM;
}
iscsi_init_global_templates(phba);
if (beiscsi_init_wrb_handle(phba))
return -ENOMEM;
if (hwi_init_async_pdu_ctx(phba)) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : hwi_init_async_pdu_ctx failed\n");
return -ENOMEM;
}
if (hwi_init_port(phba) != 0) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : hwi_init_controller failed\n");
return -ENOMEM;
}
return 0;
}
static void beiscsi_free_mem(struct beiscsi_hba *phba)
{
struct be_mem_descriptor *mem_descr;
int i, j;
mem_descr = phba->init_mem;
i = 0;
j = 0;
for (i = 0; i < SE_MEM_MAX; i++) {
for (j = mem_descr->num_elements; j > 0; j--) {
pci_free_consistent(phba->pcidev,
mem_descr->mem_array[j - 1].size,
mem_descr->mem_array[j - 1].virtual_address,
(unsigned long)mem_descr->mem_array[j - 1].
bus_address.u.a64.address);
}
kfree(mem_descr->mem_array);
mem_descr++;
}
kfree(phba->init_mem);
kfree(phba->phwi_ctrlr->wrb_context);
kfree(phba->phwi_ctrlr);
}
static int beiscsi_init_sgl_handle(struct beiscsi_hba *phba)
{
struct be_mem_descriptor *mem_descr_sglh, *mem_descr_sg;
struct sgl_handle *psgl_handle;
struct iscsi_sge *pfrag;
unsigned int arr_index, i, idx;
unsigned int ulp_icd_start, ulp_num = 0;
phba->io_sgl_hndl_avbl = 0;
phba->eh_sgl_hndl_avbl = 0;
mem_descr_sglh = phba->init_mem;
mem_descr_sglh += HWI_MEM_SGLH;
if (1 == mem_descr_sglh->num_elements) {
phba->io_sgl_hndl_base = kzalloc(sizeof(struct sgl_handle *) *
phba->params.ios_per_ctrl,
GFP_KERNEL);
if (!phba->io_sgl_hndl_base) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : Mem Alloc Failed. Failing to load\n");
return -ENOMEM;
}
phba->eh_sgl_hndl_base = kzalloc(sizeof(struct sgl_handle *) *
(phba->params.icds_per_ctrl -
phba->params.ios_per_ctrl),
GFP_KERNEL);
if (!phba->eh_sgl_hndl_base) {
kfree(phba->io_sgl_hndl_base);
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : Mem Alloc Failed. Failing to load\n");
return -ENOMEM;
}
} else {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : HWI_MEM_SGLH is more than one element."
"Failing to load\n");
return -ENOMEM;
}
arr_index = 0;
idx = 0;
while (idx < mem_descr_sglh->num_elements) {
psgl_handle = mem_descr_sglh->mem_array[idx].virtual_address;
for (i = 0; i < (mem_descr_sglh->mem_array[idx].size /
sizeof(struct sgl_handle)); i++) {
if (arr_index < phba->params.ios_per_ctrl) {
phba->io_sgl_hndl_base[arr_index] = psgl_handle;
phba->io_sgl_hndl_avbl++;
arr_index++;
} else {
phba->eh_sgl_hndl_base[arr_index -
phba->params.ios_per_ctrl] =
psgl_handle;
arr_index++;
phba->eh_sgl_hndl_avbl++;
}
psgl_handle++;
}
idx++;
}
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_INIT,
"BM_%d : phba->io_sgl_hndl_avbl=%d"
"phba->eh_sgl_hndl_avbl=%d\n",
phba->io_sgl_hndl_avbl,
phba->eh_sgl_hndl_avbl);
mem_descr_sg = phba->init_mem;
mem_descr_sg += HWI_MEM_SGE;
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_INIT,
"\n BM_%d : mem_descr_sg->num_elements=%d\n",
mem_descr_sg->num_elements);
for (ulp_num = 0; ulp_num < BEISCSI_ULP_COUNT; ulp_num++)
if (test_bit(ulp_num, &phba->fw_config.ulp_supported))
break;
ulp_icd_start = phba->fw_config.iscsi_icd_start[ulp_num];
arr_index = 0;
idx = 0;
while (idx < mem_descr_sg->num_elements) {
pfrag = mem_descr_sg->mem_array[idx].virtual_address;
for (i = 0;
i < (mem_descr_sg->mem_array[idx].size) /
(sizeof(struct iscsi_sge) * phba->params.num_sge_per_io);
i++) {
if (arr_index < phba->params.ios_per_ctrl)
psgl_handle = phba->io_sgl_hndl_base[arr_index];
else
psgl_handle = phba->eh_sgl_hndl_base[arr_index -
phba->params.ios_per_ctrl];
psgl_handle->pfrag = pfrag;
AMAP_SET_BITS(struct amap_iscsi_sge, addr_hi, pfrag, 0);
AMAP_SET_BITS(struct amap_iscsi_sge, addr_lo, pfrag, 0);
pfrag += phba->params.num_sge_per_io;
psgl_handle->sgl_index = ulp_icd_start + arr_index++;
}
idx++;
}
phba->io_sgl_free_index = 0;
phba->io_sgl_alloc_index = 0;
phba->eh_sgl_free_index = 0;
phba->eh_sgl_alloc_index = 0;
return 0;
}
static int hba_setup_cid_tbls(struct beiscsi_hba *phba)
{
int ret;
uint16_t i, ulp_num;
struct ulp_cid_info *ptr_cid_info = NULL;
for (ulp_num = 0; ulp_num < BEISCSI_ULP_COUNT; ulp_num++) {
if (test_bit(ulp_num, (void *)&phba->fw_config.ulp_supported)) {
ptr_cid_info = kzalloc(sizeof(struct ulp_cid_info),
GFP_KERNEL);
if (!ptr_cid_info) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : Failed to allocate memory"
"for ULP_CID_INFO for ULP : %d\n",
ulp_num);
ret = -ENOMEM;
goto free_memory;
}
/* Allocate memory for CID array */
ptr_cid_info->cid_array =
kcalloc(BEISCSI_GET_CID_COUNT(phba, ulp_num),
sizeof(*ptr_cid_info->cid_array),
GFP_KERNEL);
if (!ptr_cid_info->cid_array) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : Failed to allocate memory"
"for CID_ARRAY for ULP : %d\n",
ulp_num);
kfree(ptr_cid_info);
ptr_cid_info = NULL;
ret = -ENOMEM;
goto free_memory;
}
ptr_cid_info->avlbl_cids = BEISCSI_GET_CID_COUNT(
phba, ulp_num);
/* Save the cid_info_array ptr */
phba->cid_array_info[ulp_num] = ptr_cid_info;
}
}
phba->ep_array = kzalloc(sizeof(struct iscsi_endpoint *) *
phba->params.cxns_per_ctrl, GFP_KERNEL);
if (!phba->ep_array) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : Failed to allocate memory in "
"hba_setup_cid_tbls\n");
ret = -ENOMEM;
goto free_memory;
}
phba->conn_table = kzalloc(sizeof(struct beiscsi_conn *) *
phba->params.cxns_per_ctrl, GFP_KERNEL);
if (!phba->conn_table) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : Failed to allocate memory in"
"hba_setup_cid_tbls\n");
kfree(phba->ep_array);
phba->ep_array = NULL;
ret = -ENOMEM;
goto free_memory;
}
for (i = 0; i < phba->params.cxns_per_ctrl; i++) {
ulp_num = phba->phwi_ctrlr->wrb_context[i].ulp_num;
ptr_cid_info = phba->cid_array_info[ulp_num];
ptr_cid_info->cid_array[ptr_cid_info->cid_alloc++] =
phba->phwi_ctrlr->wrb_context[i].cid;
}
for (ulp_num = 0; ulp_num < BEISCSI_ULP_COUNT; ulp_num++) {
if (test_bit(ulp_num, (void *)&phba->fw_config.ulp_supported)) {
ptr_cid_info = phba->cid_array_info[ulp_num];
ptr_cid_info->cid_alloc = 0;
ptr_cid_info->cid_free = 0;
}
}
return 0;
free_memory:
for (ulp_num = 0; ulp_num < BEISCSI_ULP_COUNT; ulp_num++) {
if (test_bit(ulp_num, (void *)&phba->fw_config.ulp_supported)) {
ptr_cid_info = phba->cid_array_info[ulp_num];
if (ptr_cid_info) {
kfree(ptr_cid_info->cid_array);
kfree(ptr_cid_info);
phba->cid_array_info[ulp_num] = NULL;
}
}
}
return ret;
}
static void hwi_enable_intr(struct beiscsi_hba *phba)
{
struct be_ctrl_info *ctrl = &phba->ctrl;
struct hwi_controller *phwi_ctrlr;
struct hwi_context_memory *phwi_context;
struct be_queue_info *eq;
u8 __iomem *addr;
u32 reg, i;
u32 enabled;
phwi_ctrlr = phba->phwi_ctrlr;
phwi_context = phwi_ctrlr->phwi_ctxt;
addr = (u8 __iomem *) ((u8 __iomem *) ctrl->pcicfg +
PCICFG_MEMBAR_CTRL_INT_CTRL_OFFSET);
reg = ioread32(addr);
enabled = reg & MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
if (!enabled) {
reg |= MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_INIT,
"BM_%d : reg =x%08x addr=%p\n", reg, addr);
iowrite32(reg, addr);
}
if (!phba->pcidev->msix_enabled) {
eq = &phwi_context->be_eq[0].q;
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_INIT,
"BM_%d : eq->id=%d\n", eq->id);
hwi_ring_eq_db(phba, eq->id, 0, 0, 1, 1);
} else {
for (i = 0; i <= phba->num_cpus; i++) {
eq = &phwi_context->be_eq[i].q;
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_INIT,
"BM_%d : eq->id=%d\n", eq->id);
hwi_ring_eq_db(phba, eq->id, 0, 0, 1, 1);
}
}
}
static void hwi_disable_intr(struct beiscsi_hba *phba)
{
struct be_ctrl_info *ctrl = &phba->ctrl;
u8 __iomem *addr = ctrl->pcicfg + PCICFG_MEMBAR_CTRL_INT_CTRL_OFFSET;
u32 reg = ioread32(addr);
u32 enabled = reg & MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
if (enabled) {
reg &= ~MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
iowrite32(reg, addr);
} else
beiscsi_log(phba, KERN_WARNING, BEISCSI_LOG_INIT,
"BM_%d : In hwi_disable_intr, Already Disabled\n");
}
static int beiscsi_init_port(struct beiscsi_hba *phba)
{
int ret;
ret = hwi_init_controller(phba);
if (ret < 0) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : init controller failed\n");
return ret;
}
ret = beiscsi_init_sgl_handle(phba);
if (ret < 0) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : init sgl handles failed\n");
goto cleanup_port;
}
ret = hba_setup_cid_tbls(phba);
if (ret < 0) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : setup CID table failed\n");
kfree(phba->io_sgl_hndl_base);
kfree(phba->eh_sgl_hndl_base);
goto cleanup_port;
}
return ret;
cleanup_port:
hwi_cleanup_port(phba);
return ret;
}
static void beiscsi_cleanup_port(struct beiscsi_hba *phba)
{
struct ulp_cid_info *ptr_cid_info = NULL;
int ulp_num;
kfree(phba->io_sgl_hndl_base);
kfree(phba->eh_sgl_hndl_base);
kfree(phba->ep_array);
kfree(phba->conn_table);
for (ulp_num = 0; ulp_num < BEISCSI_ULP_COUNT; ulp_num++) {
if (test_bit(ulp_num, (void *)&phba->fw_config.ulp_supported)) {
ptr_cid_info = phba->cid_array_info[ulp_num];
if (ptr_cid_info) {
kfree(ptr_cid_info->cid_array);
kfree(ptr_cid_info);
phba->cid_array_info[ulp_num] = NULL;
}
}
}
}
/**
* beiscsi_free_mgmt_task_handles()- Free driver CXN resources
* @beiscsi_conn: ptr to the conn to be cleaned up
* @task: ptr to iscsi_task resource to be freed.
*
* Free driver mgmt resources binded to CXN.
**/
void
beiscsi_free_mgmt_task_handles(struct beiscsi_conn *beiscsi_conn,
struct iscsi_task *task)
{
struct beiscsi_io_task *io_task;
struct beiscsi_hba *phba = beiscsi_conn->phba;
struct hwi_wrb_context *pwrb_context;
struct hwi_controller *phwi_ctrlr;
uint16_t cri_index = BE_GET_CRI_FROM_CID(
beiscsi_conn->beiscsi_conn_cid);
phwi_ctrlr = phba->phwi_ctrlr;
pwrb_context = &phwi_ctrlr->wrb_context[cri_index];
io_task = task->dd_data;
if (io_task->pwrb_handle) {
free_wrb_handle(phba, pwrb_context, io_task->pwrb_handle);
io_task->pwrb_handle = NULL;
}
if (io_task->psgl_handle) {
free_mgmt_sgl_handle(phba, io_task->psgl_handle);
io_task->psgl_handle = NULL;
}
if (io_task->mtask_addr) {
pci_unmap_single(phba->pcidev,
io_task->mtask_addr,
io_task->mtask_data_count,
PCI_DMA_TODEVICE);
io_task->mtask_addr = 0;
}
}
/**
* beiscsi_cleanup_task()- Free driver resources of the task
* @task: ptr to the iscsi task
*
**/
static void beiscsi_cleanup_task(struct iscsi_task *task)
{
struct beiscsi_io_task *io_task = task->dd_data;
struct iscsi_conn *conn = task->conn;
struct beiscsi_conn *beiscsi_conn = conn->dd_data;
struct beiscsi_hba *phba = beiscsi_conn->phba;
struct beiscsi_session *beiscsi_sess = beiscsi_conn->beiscsi_sess;
struct hwi_wrb_context *pwrb_context;
struct hwi_controller *phwi_ctrlr;
uint16_t cri_index = BE_GET_CRI_FROM_CID(
beiscsi_conn->beiscsi_conn_cid);
phwi_ctrlr = phba->phwi_ctrlr;
pwrb_context = &phwi_ctrlr->wrb_context[cri_index];
if (io_task->cmd_bhs) {
dma_pool_free(beiscsi_sess->bhs_pool, io_task->cmd_bhs,
io_task->bhs_pa.u.a64.address);
io_task->cmd_bhs = NULL;
task->hdr = NULL;
}
if (task->sc) {
if (io_task->pwrb_handle) {
free_wrb_handle(phba, pwrb_context,
io_task->pwrb_handle);
io_task->pwrb_handle = NULL;
}
if (io_task->psgl_handle) {
free_io_sgl_handle(phba, io_task->psgl_handle);
io_task->psgl_handle = NULL;
}
if (io_task->scsi_cmnd) {
if (io_task->num_sg)
scsi_dma_unmap(io_task->scsi_cmnd);
io_task->scsi_cmnd = NULL;
}
} else {
if (!beiscsi_conn->login_in_progress)
beiscsi_free_mgmt_task_handles(beiscsi_conn, task);
}
}
void
beiscsi_offload_connection(struct beiscsi_conn *beiscsi_conn,
struct beiscsi_offload_params *params)
{
struct wrb_handle *pwrb_handle;
struct hwi_wrb_context *pwrb_context = NULL;
struct beiscsi_hba *phba = beiscsi_conn->phba;
struct iscsi_task *task = beiscsi_conn->task;
struct iscsi_session *session = task->conn->session;
u32 doorbell = 0;
/*
* We can always use 0 here because it is reserved by libiscsi for
* login/startup related tasks.
*/
beiscsi_conn->login_in_progress = 0;
[SCSI] libiscsi: Reduce locking contention in fast path Replace the session lock with two locks, a forward lock and a backwards lock named frwd_lock and back_lock respectively. The forward lock protects resources that change while sending a request to the target, such as cmdsn, queued_cmdsn, and allocating task from the commands' pool with kfifo_out. The backward lock protects resources that change while processing a response or in error path, such as cmdsn_exp, cmdsn_max, and returning tasks to the commands' pool with kfifo_in. Under a steady state fast-path situation, that is when one or more processes/threads submit IO to an iscsi device and a single kernel upcall (e.g softirq) is dealing with processing of responses without errors, this patch eliminates the contention between the queuecommand()/request response/scsi_done() flows associated with iscsi sessions. Between the forward and the backward locks exists a strict locking hierarchy. The mutual exclusion zone protected by the forward lock can enclose the mutual exclusion zone protected by the backward lock but not vice versa. For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is a failure and __iscsi_put_task is called, the backward lock is taken while the forward lock is still taken. On the other hand, if in the RX path a nop is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu than the forward lock is released and the backward lock is taken for the duration of iscsi_send_nopout, later the backward lock is released and the forward lock is retaken. libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue. The insertion and deletion from these queues didn't corespond to the assumption taken by the new forward/backwards session locking paradigm. That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards) path, r2t is taken out from r2t queue and inserted to the r2t pool. In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t is also inserted to the r2t pool and another r2t is pulled from r2t queue. Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue to the TX path, r2t is taken from the r2t pool and inserted to the r2t queue. In order to cope with this situation, two spin locks were added, pool2queue and queue2pool. The former protects extracting from the r2t pool and inserting to the r2t queue, and the later protects the extracing from the r2t queue and inserting to the r2t pool. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> [minor fix up to apply cleanly and compile fix] Signed-off-by: Mike Christie <michaelc@cs.wisc.edu> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 14:41:38 +08:00
spin_lock_bh(&session->back_lock);
beiscsi_cleanup_task(task);
[SCSI] libiscsi: Reduce locking contention in fast path Replace the session lock with two locks, a forward lock and a backwards lock named frwd_lock and back_lock respectively. The forward lock protects resources that change while sending a request to the target, such as cmdsn, queued_cmdsn, and allocating task from the commands' pool with kfifo_out. The backward lock protects resources that change while processing a response or in error path, such as cmdsn_exp, cmdsn_max, and returning tasks to the commands' pool with kfifo_in. Under a steady state fast-path situation, that is when one or more processes/threads submit IO to an iscsi device and a single kernel upcall (e.g softirq) is dealing with processing of responses without errors, this patch eliminates the contention between the queuecommand()/request response/scsi_done() flows associated with iscsi sessions. Between the forward and the backward locks exists a strict locking hierarchy. The mutual exclusion zone protected by the forward lock can enclose the mutual exclusion zone protected by the backward lock but not vice versa. For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is a failure and __iscsi_put_task is called, the backward lock is taken while the forward lock is still taken. On the other hand, if in the RX path a nop is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu than the forward lock is released and the backward lock is taken for the duration of iscsi_send_nopout, later the backward lock is released and the forward lock is retaken. libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue. The insertion and deletion from these queues didn't corespond to the assumption taken by the new forward/backwards session locking paradigm. That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards) path, r2t is taken out from r2t queue and inserted to the r2t pool. In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t is also inserted to the r2t pool and another r2t is pulled from r2t queue. Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue to the TX path, r2t is taken from the r2t pool and inserted to the r2t queue. In order to cope with this situation, two spin locks were added, pool2queue and queue2pool. The former protects extracting from the r2t pool and inserting to the r2t queue, and the later protects the extracing from the r2t queue and inserting to the r2t pool. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> [minor fix up to apply cleanly and compile fix] Signed-off-by: Mike Christie <michaelc@cs.wisc.edu> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 14:41:38 +08:00
spin_unlock_bh(&session->back_lock);
pwrb_handle = alloc_wrb_handle(phba, beiscsi_conn->beiscsi_conn_cid,
&pwrb_context);
/* Check for the adapter family */
if (is_chip_be2_be3r(phba))
beiscsi_offload_cxn_v0(params, pwrb_handle,
phba->init_mem,
pwrb_context);
else
beiscsi_offload_cxn_v2(params, pwrb_handle,
pwrb_context);
be_dws_le_to_cpu(pwrb_handle->pwrb,
sizeof(struct iscsi_target_context_update_wrb));
doorbell |= beiscsi_conn->beiscsi_conn_cid & DB_WRB_POST_CID_MASK;
doorbell |= (pwrb_handle->wrb_index & DB_DEF_PDU_WRB_INDEX_MASK)
<< DB_DEF_PDU_WRB_INDEX_SHIFT;
doorbell |= 1 << DB_DEF_PDU_NUM_POSTED_SHIFT;
iowrite32(doorbell, phba->db_va +
beiscsi_conn->doorbell_offset);
/*
* There is no completion for CONTEXT_UPDATE. The completion of next
* WRB posted guarantees FW's processing and DMA'ing of it.
* Use beiscsi_put_wrb_handle to put it back in the pool which makes
* sure zero'ing or reuse of the WRB only after wrbs_per_cxn.
*/
beiscsi_put_wrb_handle(pwrb_context, pwrb_handle,
phba->params.wrbs_per_cxn);
beiscsi_log(phba, KERN_INFO,
BEISCSI_LOG_IO | BEISCSI_LOG_CONFIG,
"BM_%d : put CONTEXT_UPDATE pwrb_handle=%p free_index=0x%x wrb_handles_available=%d\n",
pwrb_handle, pwrb_context->free_index,
pwrb_context->wrb_handles_available);
}
static void beiscsi_parse_pdu(struct iscsi_conn *conn, itt_t itt,
int *index, int *age)
{
*index = (int)itt;
if (age)
*age = conn->session->age;
}
/**
* beiscsi_alloc_pdu - allocates pdu and related resources
* @task: libiscsi task
* @opcode: opcode of pdu for task
*
* This is called with the session lock held. It will allocate
* the wrb and sgl if needed for the command. And it will prep
* the pdu's itt. beiscsi_parse_pdu will later translate
* the pdu itt to the libiscsi task itt.
*/
static int beiscsi_alloc_pdu(struct iscsi_task *task, uint8_t opcode)
{
struct beiscsi_io_task *io_task = task->dd_data;
struct iscsi_conn *conn = task->conn;
struct beiscsi_conn *beiscsi_conn = conn->dd_data;
struct beiscsi_hba *phba = beiscsi_conn->phba;
struct hwi_wrb_context *pwrb_context;
struct hwi_controller *phwi_ctrlr;
itt_t itt;
uint16_t cri_index = 0;
struct beiscsi_session *beiscsi_sess = beiscsi_conn->beiscsi_sess;
dma_addr_t paddr;
io_task->cmd_bhs = dma_pool_alloc(beiscsi_sess->bhs_pool,
GFP_ATOMIC, &paddr);
if (!io_task->cmd_bhs)
return -ENOMEM;
io_task->bhs_pa.u.a64.address = paddr;
io_task->libiscsi_itt = (itt_t)task->itt;
io_task->conn = beiscsi_conn;
task->hdr = (struct iscsi_hdr *)&io_task->cmd_bhs->iscsi_hdr;
task->hdr_max = sizeof(struct be_cmd_bhs);
io_task->psgl_handle = NULL;
io_task->pwrb_handle = NULL;
if (task->sc) {
io_task->psgl_handle = alloc_io_sgl_handle(phba);
if (!io_task->psgl_handle) {
beiscsi_log(phba, KERN_ERR,
BEISCSI_LOG_IO | BEISCSI_LOG_CONFIG,
"BM_%d : Alloc of IO_SGL_ICD Failed"
"for the CID : %d\n",
beiscsi_conn->beiscsi_conn_cid);
goto free_hndls;
}
io_task->pwrb_handle = alloc_wrb_handle(phba,
beiscsi_conn->beiscsi_conn_cid,
&io_task->pwrb_context);
if (!io_task->pwrb_handle) {
beiscsi_log(phba, KERN_ERR,
BEISCSI_LOG_IO | BEISCSI_LOG_CONFIG,
"BM_%d : Alloc of WRB_HANDLE Failed"
"for the CID : %d\n",
beiscsi_conn->beiscsi_conn_cid);
goto free_io_hndls;
}
} else {
io_task->scsi_cmnd = NULL;
if ((opcode & ISCSI_OPCODE_MASK) == ISCSI_OP_LOGIN) {
beiscsi_conn->task = task;
if (!beiscsi_conn->login_in_progress) {
io_task->psgl_handle = (struct sgl_handle *)
alloc_mgmt_sgl_handle(phba);
if (!io_task->psgl_handle) {
beiscsi_log(phba, KERN_ERR,
BEISCSI_LOG_IO |
BEISCSI_LOG_CONFIG,
"BM_%d : Alloc of MGMT_SGL_ICD Failed"
"for the CID : %d\n",
beiscsi_conn->
beiscsi_conn_cid);
goto free_hndls;
}
beiscsi_conn->login_in_progress = 1;
beiscsi_conn->plogin_sgl_handle =
io_task->psgl_handle;
io_task->pwrb_handle =
alloc_wrb_handle(phba,
beiscsi_conn->beiscsi_conn_cid,
&io_task->pwrb_context);
if (!io_task->pwrb_handle) {
beiscsi_log(phba, KERN_ERR,
BEISCSI_LOG_IO |
BEISCSI_LOG_CONFIG,
"BM_%d : Alloc of WRB_HANDLE Failed"
"for the CID : %d\n",
beiscsi_conn->
beiscsi_conn_cid);
goto free_mgmt_hndls;
}
beiscsi_conn->plogin_wrb_handle =
io_task->pwrb_handle;
} else {
io_task->psgl_handle =
beiscsi_conn->plogin_sgl_handle;
io_task->pwrb_handle =
beiscsi_conn->plogin_wrb_handle;
}
} else {
io_task->psgl_handle = alloc_mgmt_sgl_handle(phba);
if (!io_task->psgl_handle) {
beiscsi_log(phba, KERN_ERR,
BEISCSI_LOG_IO |
BEISCSI_LOG_CONFIG,
"BM_%d : Alloc of MGMT_SGL_ICD Failed"
"for the CID : %d\n",
beiscsi_conn->
beiscsi_conn_cid);
goto free_hndls;
}
io_task->pwrb_handle =
alloc_wrb_handle(phba,
beiscsi_conn->beiscsi_conn_cid,
&io_task->pwrb_context);
if (!io_task->pwrb_handle) {
beiscsi_log(phba, KERN_ERR,
BEISCSI_LOG_IO | BEISCSI_LOG_CONFIG,
"BM_%d : Alloc of WRB_HANDLE Failed"
"for the CID : %d\n",
beiscsi_conn->beiscsi_conn_cid);
goto free_mgmt_hndls;
}
}
}
itt = (itt_t) cpu_to_be32(((unsigned int)io_task->pwrb_handle->
wrb_index << 16) | (unsigned int)
(io_task->psgl_handle->sgl_index));
io_task->pwrb_handle->pio_handle = task;
io_task->cmd_bhs->iscsi_hdr.itt = itt;
return 0;
free_io_hndls:
free_io_sgl_handle(phba, io_task->psgl_handle);
goto free_hndls;
free_mgmt_hndls:
free_mgmt_sgl_handle(phba, io_task->psgl_handle);
io_task->psgl_handle = NULL;
free_hndls:
phwi_ctrlr = phba->phwi_ctrlr;
cri_index = BE_GET_CRI_FROM_CID(
beiscsi_conn->beiscsi_conn_cid);
pwrb_context = &phwi_ctrlr->wrb_context[cri_index];
if (io_task->pwrb_handle)
free_wrb_handle(phba, pwrb_context, io_task->pwrb_handle);
io_task->pwrb_handle = NULL;
dma_pool_free(beiscsi_sess->bhs_pool, io_task->cmd_bhs,
io_task->bhs_pa.u.a64.address);
io_task->cmd_bhs = NULL;
return -ENOMEM;
}
static int beiscsi_iotask_v2(struct iscsi_task *task, struct scatterlist *sg,
unsigned int num_sg, unsigned int xferlen,
unsigned int writedir)
{
struct beiscsi_io_task *io_task = task->dd_data;
struct iscsi_conn *conn = task->conn;
struct beiscsi_conn *beiscsi_conn = conn->dd_data;
struct beiscsi_hba *phba = beiscsi_conn->phba;
struct iscsi_wrb *pwrb = NULL;
unsigned int doorbell = 0;
pwrb = io_task->pwrb_handle->pwrb;
io_task->bhs_len = sizeof(struct be_cmd_bhs);
if (writedir) {
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, type, pwrb,
INI_WR_CMD);
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, dsp, pwrb, 1);
} else {
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, type, pwrb,
INI_RD_CMD);
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, dsp, pwrb, 0);
}
io_task->wrb_type = AMAP_GET_BITS(struct amap_iscsi_wrb_v2,
type, pwrb);
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, lun, pwrb,
cpu_to_be16(*(unsigned short *)
&io_task->cmd_bhs->iscsi_hdr.lun));
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, r2t_exp_dtl, pwrb, xferlen);
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, wrb_idx, pwrb,
io_task->pwrb_handle->wrb_index);
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, cmdsn_itt, pwrb,
be32_to_cpu(task->cmdsn));
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, sgl_idx, pwrb,
io_task->psgl_handle->sgl_index);
hwi_write_sgl_v2(pwrb, sg, num_sg, io_task);
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, ptr2nextwrb, pwrb,
io_task->pwrb_handle->wrb_index);
if (io_task->pwrb_context->plast_wrb)
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, ptr2nextwrb,
io_task->pwrb_context->plast_wrb,
io_task->pwrb_handle->wrb_index);
io_task->pwrb_context->plast_wrb = pwrb;
be_dws_le_to_cpu(pwrb, sizeof(struct iscsi_wrb));
doorbell |= beiscsi_conn->beiscsi_conn_cid & DB_WRB_POST_CID_MASK;
doorbell |= (io_task->pwrb_handle->wrb_index &
DB_DEF_PDU_WRB_INDEX_MASK) <<
DB_DEF_PDU_WRB_INDEX_SHIFT;
doorbell |= 1 << DB_DEF_PDU_NUM_POSTED_SHIFT;
iowrite32(doorbell, phba->db_va +
beiscsi_conn->doorbell_offset);
return 0;
}
static int beiscsi_iotask(struct iscsi_task *task, struct scatterlist *sg,
unsigned int num_sg, unsigned int xferlen,
unsigned int writedir)
{
struct beiscsi_io_task *io_task = task->dd_data;
struct iscsi_conn *conn = task->conn;
struct beiscsi_conn *beiscsi_conn = conn->dd_data;
struct beiscsi_hba *phba = beiscsi_conn->phba;
struct iscsi_wrb *pwrb = NULL;
unsigned int doorbell = 0;
pwrb = io_task->pwrb_handle->pwrb;
io_task->bhs_len = sizeof(struct be_cmd_bhs);
if (writedir) {
AMAP_SET_BITS(struct amap_iscsi_wrb, type, pwrb,
INI_WR_CMD);
AMAP_SET_BITS(struct amap_iscsi_wrb, dsp, pwrb, 1);
} else {
AMAP_SET_BITS(struct amap_iscsi_wrb, type, pwrb,
INI_RD_CMD);
AMAP_SET_BITS(struct amap_iscsi_wrb, dsp, pwrb, 0);
}
io_task->wrb_type = AMAP_GET_BITS(struct amap_iscsi_wrb,
type, pwrb);
AMAP_SET_BITS(struct amap_iscsi_wrb, lun, pwrb,
cpu_to_be16(*(unsigned short *)
&io_task->cmd_bhs->iscsi_hdr.lun));
AMAP_SET_BITS(struct amap_iscsi_wrb, r2t_exp_dtl, pwrb, xferlen);
AMAP_SET_BITS(struct amap_iscsi_wrb, wrb_idx, pwrb,
io_task->pwrb_handle->wrb_index);
AMAP_SET_BITS(struct amap_iscsi_wrb, cmdsn_itt, pwrb,
be32_to_cpu(task->cmdsn));
AMAP_SET_BITS(struct amap_iscsi_wrb, sgl_icd_idx, pwrb,
io_task->psgl_handle->sgl_index);
hwi_write_sgl(pwrb, sg, num_sg, io_task);
AMAP_SET_BITS(struct amap_iscsi_wrb, ptr2nextwrb, pwrb,
io_task->pwrb_handle->wrb_index);
if (io_task->pwrb_context->plast_wrb)
AMAP_SET_BITS(struct amap_iscsi_wrb, ptr2nextwrb,
io_task->pwrb_context->plast_wrb,
io_task->pwrb_handle->wrb_index);
io_task->pwrb_context->plast_wrb = pwrb;
be_dws_le_to_cpu(pwrb, sizeof(struct iscsi_wrb));
doorbell |= beiscsi_conn->beiscsi_conn_cid & DB_WRB_POST_CID_MASK;
doorbell |= (io_task->pwrb_handle->wrb_index &
DB_DEF_PDU_WRB_INDEX_MASK) << DB_DEF_PDU_WRB_INDEX_SHIFT;
doorbell |= 1 << DB_DEF_PDU_NUM_POSTED_SHIFT;
iowrite32(doorbell, phba->db_va +
beiscsi_conn->doorbell_offset);
return 0;
}
static int beiscsi_mtask(struct iscsi_task *task)
{
struct beiscsi_io_task *io_task = task->dd_data;
struct iscsi_conn *conn = task->conn;
struct beiscsi_conn *beiscsi_conn = conn->dd_data;
struct beiscsi_hba *phba = beiscsi_conn->phba;
struct iscsi_wrb *pwrb = NULL;
unsigned int doorbell = 0;
unsigned int cid;
unsigned int pwrb_typeoffset = 0;
int ret = 0;
cid = beiscsi_conn->beiscsi_conn_cid;
pwrb = io_task->pwrb_handle->pwrb;
if (is_chip_be2_be3r(phba)) {
AMAP_SET_BITS(struct amap_iscsi_wrb, cmdsn_itt, pwrb,
be32_to_cpu(task->cmdsn));
AMAP_SET_BITS(struct amap_iscsi_wrb, wrb_idx, pwrb,
io_task->pwrb_handle->wrb_index);
AMAP_SET_BITS(struct amap_iscsi_wrb, sgl_icd_idx, pwrb,
io_task->psgl_handle->sgl_index);
AMAP_SET_BITS(struct amap_iscsi_wrb, r2t_exp_dtl, pwrb,
task->data_count);
AMAP_SET_BITS(struct amap_iscsi_wrb, ptr2nextwrb, pwrb,
io_task->pwrb_handle->wrb_index);
if (io_task->pwrb_context->plast_wrb)
AMAP_SET_BITS(struct amap_iscsi_wrb, ptr2nextwrb,
io_task->pwrb_context->plast_wrb,
io_task->pwrb_handle->wrb_index);
io_task->pwrb_context->plast_wrb = pwrb;
pwrb_typeoffset = BE_WRB_TYPE_OFFSET;
} else {
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, cmdsn_itt, pwrb,
be32_to_cpu(task->cmdsn));
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, wrb_idx, pwrb,
io_task->pwrb_handle->wrb_index);
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, sgl_idx, pwrb,
io_task->psgl_handle->sgl_index);
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, r2t_exp_dtl, pwrb,
task->data_count);
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, ptr2nextwrb, pwrb,
io_task->pwrb_handle->wrb_index);
if (io_task->pwrb_context->plast_wrb)
AMAP_SET_BITS(struct amap_iscsi_wrb_v2, ptr2nextwrb,
io_task->pwrb_context->plast_wrb,
io_task->pwrb_handle->wrb_index);
io_task->pwrb_context->plast_wrb = pwrb;
pwrb_typeoffset = SKH_WRB_TYPE_OFFSET;
}
switch (task->hdr->opcode & ISCSI_OPCODE_MASK) {
case ISCSI_OP_LOGIN:
AMAP_SET_BITS(struct amap_iscsi_wrb, cmdsn_itt, pwrb, 1);
ADAPTER_SET_WRB_TYPE(pwrb, TGT_DM_CMD, pwrb_typeoffset);
ret = hwi_write_buffer(pwrb, task);
break;
case ISCSI_OP_NOOP_OUT:
if (task->hdr->ttt != ISCSI_RESERVED_TAG) {
ADAPTER_SET_WRB_TYPE(pwrb, TGT_DM_CMD, pwrb_typeoffset);
if (is_chip_be2_be3r(phba))
AMAP_SET_BITS(struct amap_iscsi_wrb,
dmsg, pwrb, 1);
else
AMAP_SET_BITS(struct amap_iscsi_wrb_v2,
dmsg, pwrb, 1);
} else {
ADAPTER_SET_WRB_TYPE(pwrb, INI_RD_CMD, pwrb_typeoffset);
if (is_chip_be2_be3r(phba))
AMAP_SET_BITS(struct amap_iscsi_wrb,
dmsg, pwrb, 0);
else
AMAP_SET_BITS(struct amap_iscsi_wrb_v2,
dmsg, pwrb, 0);
}
ret = hwi_write_buffer(pwrb, task);
break;
case ISCSI_OP_TEXT:
ADAPTER_SET_WRB_TYPE(pwrb, TGT_DM_CMD, pwrb_typeoffset);
ret = hwi_write_buffer(pwrb, task);
break;
case ISCSI_OP_SCSI_TMFUNC:
ADAPTER_SET_WRB_TYPE(pwrb, INI_TMF_CMD, pwrb_typeoffset);
ret = hwi_write_buffer(pwrb, task);
break;
case ISCSI_OP_LOGOUT:
ADAPTER_SET_WRB_TYPE(pwrb, HWH_TYPE_LOGOUT, pwrb_typeoffset);
ret = hwi_write_buffer(pwrb, task);
break;
default:
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_CONFIG,
"BM_%d : opcode =%d Not supported\n",
task->hdr->opcode & ISCSI_OPCODE_MASK);
return -EINVAL;
}
if (ret)
return ret;
/* Set the task type */
io_task->wrb_type = (is_chip_be2_be3r(phba)) ?
AMAP_GET_BITS(struct amap_iscsi_wrb, type, pwrb) :
AMAP_GET_BITS(struct amap_iscsi_wrb_v2, type, pwrb);
doorbell |= cid & DB_WRB_POST_CID_MASK;
doorbell |= (io_task->pwrb_handle->wrb_index &
DB_DEF_PDU_WRB_INDEX_MASK) << DB_DEF_PDU_WRB_INDEX_SHIFT;
doorbell |= 1 << DB_DEF_PDU_NUM_POSTED_SHIFT;
iowrite32(doorbell, phba->db_va +
beiscsi_conn->doorbell_offset);
return 0;
}
static int beiscsi_task_xmit(struct iscsi_task *task)
{
struct beiscsi_io_task *io_task = task->dd_data;
struct scsi_cmnd *sc = task->sc;
struct beiscsi_hba *phba;
struct scatterlist *sg;
int num_sg;
unsigned int writedir = 0, xferlen = 0;
phba = io_task->conn->phba;
/**
* HBA in error includes BEISCSI_HBA_FW_TIMEOUT. IO path might be
* operational if FW still gets heartbeat from EP FW. Is management
* path really needed to continue further?
*/
if (!beiscsi_hba_is_online(phba))
return -EIO;
if (!io_task->conn->login_in_progress)
task->hdr->exp_statsn = 0;
if (!sc)
return beiscsi_mtask(task);
io_task->scsi_cmnd = sc;
io_task->num_sg = 0;
num_sg = scsi_dma_map(sc);
if (num_sg < 0) {
beiscsi_log(phba, KERN_ERR,
BEISCSI_LOG_IO | BEISCSI_LOG_ISCSI,
"BM_%d : scsi_dma_map Failed "
"Driver_ITT : 0x%x ITT : 0x%x Xferlen : 0x%x\n",
be32_to_cpu(io_task->cmd_bhs->iscsi_hdr.itt),
io_task->libiscsi_itt, scsi_bufflen(sc));
return num_sg;
}
/**
* For scsi cmd task, check num_sg before unmapping in cleanup_task.
* For management task, cleanup_task checks mtask_addr before unmapping.
*/
io_task->num_sg = num_sg;
xferlen = scsi_bufflen(sc);
sg = scsi_sglist(sc);
if (sc->sc_data_direction == DMA_TO_DEVICE)
writedir = 1;
else
writedir = 0;
return phba->iotask_fn(task, sg, num_sg, xferlen, writedir);
}
/**
* beiscsi_bsg_request - handle bsg request from ISCSI transport
* @job: job to handle
*/
static int beiscsi_bsg_request(struct bsg_job *job)
{
struct Scsi_Host *shost;
struct beiscsi_hba *phba;
struct iscsi_bsg_request *bsg_req = job->request;
int rc = -EINVAL;
unsigned int tag;
struct be_dma_mem nonemb_cmd;
struct be_cmd_resp_hdr *resp;
struct iscsi_bsg_reply *bsg_reply = job->reply;
unsigned short status, extd_status;
shost = iscsi_job_to_shost(job);
phba = iscsi_host_priv(shost);
if (!beiscsi_hba_is_online(phba)) {
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_CONFIG,
"BM_%d : HBA in error 0x%lx\n", phba->state);
return -ENXIO;
}
switch (bsg_req->msgcode) {
case ISCSI_BSG_HST_VENDOR:
nonemb_cmd.va = pci_alloc_consistent(phba->ctrl.pdev,
job->request_payload.payload_len,
&nonemb_cmd.dma);
if (nonemb_cmd.va == NULL) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_CONFIG,
"BM_%d : Failed to allocate memory for "
"beiscsi_bsg_request\n");
return -ENOMEM;
}
tag = mgmt_vendor_specific_fw_cmd(&phba->ctrl, phba, job,
&nonemb_cmd);
if (!tag) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_CONFIG,
"BM_%d : MBX Tag Allocation Failed\n");
pci_free_consistent(phba->ctrl.pdev, nonemb_cmd.size,
nonemb_cmd.va, nonemb_cmd.dma);
return -EAGAIN;
}
rc = wait_event_interruptible_timeout(
phba->ctrl.mcc_wait[tag],
phba->ctrl.mcc_tag_status[tag],
msecs_to_jiffies(
BEISCSI_HOST_MBX_TIMEOUT));
if (!test_bit(BEISCSI_HBA_ONLINE, &phba->state)) {
clear_bit(MCC_TAG_STATE_RUNNING,
&phba->ctrl.ptag_state[tag].tag_state);
pci_free_consistent(phba->ctrl.pdev, nonemb_cmd.size,
nonemb_cmd.va, nonemb_cmd.dma);
return -EIO;
}
extd_status = (phba->ctrl.mcc_tag_status[tag] &
CQE_STATUS_ADDL_MASK) >> CQE_STATUS_ADDL_SHIFT;
status = phba->ctrl.mcc_tag_status[tag] & CQE_STATUS_MASK;
free_mcc_wrb(&phba->ctrl, tag);
resp = (struct be_cmd_resp_hdr *)nonemb_cmd.va;
sg_copy_from_buffer(job->reply_payload.sg_list,
job->reply_payload.sg_cnt,
nonemb_cmd.va, (resp->response_length
+ sizeof(*resp)));
bsg_reply->reply_payload_rcv_len = resp->response_length;
bsg_reply->result = status;
bsg_job_done(job, bsg_reply->result,
bsg_reply->reply_payload_rcv_len);
pci_free_consistent(phba->ctrl.pdev, nonemb_cmd.size,
nonemb_cmd.va, nonemb_cmd.dma);
if (status || extd_status) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_CONFIG,
"BM_%d : MBX Cmd Failed"
" status = %d extd_status = %d\n",
status, extd_status);
return -EIO;
} else {
rc = 0;
}
break;
default:
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_CONFIG,
"BM_%d : Unsupported bsg command: 0x%x\n",
bsg_req->msgcode);
break;
}
return rc;
}
static void beiscsi_hba_attrs_init(struct beiscsi_hba *phba)
{
/* Set the logging parameter */
beiscsi_log_enable_init(phba, beiscsi_log_enable);
}
void beiscsi_start_boot_work(struct beiscsi_hba *phba, unsigned int s_handle)
{
if (phba->boot_struct.boot_kset)
return;
/* skip if boot work is already in progress */
if (test_and_set_bit(BEISCSI_HBA_BOOT_WORK, &phba->state))
return;
phba->boot_struct.retry = 3;
phba->boot_struct.tag = 0;
phba->boot_struct.s_handle = s_handle;
phba->boot_struct.action = BEISCSI_BOOT_GET_SHANDLE;
schedule_work(&phba->boot_work);
}
static ssize_t beiscsi_show_boot_tgt_info(void *data, int type, char *buf)
{
struct beiscsi_hba *phba = data;
struct mgmt_session_info *boot_sess = &phba->boot_struct.boot_sess;
struct mgmt_conn_info *boot_conn = &boot_sess->conn_list[0];
char *str = buf;
int rc = -EPERM;
switch (type) {
case ISCSI_BOOT_TGT_NAME:
rc = sprintf(buf, "%.*s\n",
(int)strlen(boot_sess->target_name),
(char *)&boot_sess->target_name);
break;
case ISCSI_BOOT_TGT_IP_ADDR:
if (boot_conn->dest_ipaddr.ip_type == BEISCSI_IP_TYPE_V4)
rc = sprintf(buf, "%pI4\n",
(char *)&boot_conn->dest_ipaddr.addr);
else
rc = sprintf(str, "%pI6\n",
(char *)&boot_conn->dest_ipaddr.addr);
break;
case ISCSI_BOOT_TGT_PORT:
rc = sprintf(str, "%d\n", boot_conn->dest_port);
break;
case ISCSI_BOOT_TGT_CHAP_NAME:
rc = sprintf(str, "%.*s\n",
boot_conn->negotiated_login_options.auth_data.chap.
target_chap_name_length,
(char *)&boot_conn->negotiated_login_options.
auth_data.chap.target_chap_name);
break;
case ISCSI_BOOT_TGT_CHAP_SECRET:
rc = sprintf(str, "%.*s\n",
boot_conn->negotiated_login_options.auth_data.chap.
target_secret_length,
(char *)&boot_conn->negotiated_login_options.
auth_data.chap.target_secret);
break;
case ISCSI_BOOT_TGT_REV_CHAP_NAME:
rc = sprintf(str, "%.*s\n",
boot_conn->negotiated_login_options.auth_data.chap.
intr_chap_name_length,
(char *)&boot_conn->negotiated_login_options.
auth_data.chap.intr_chap_name);
break;
case ISCSI_BOOT_TGT_REV_CHAP_SECRET:
rc = sprintf(str, "%.*s\n",
boot_conn->negotiated_login_options.auth_data.chap.
intr_secret_length,
(char *)&boot_conn->negotiated_login_options.
auth_data.chap.intr_secret);
break;
case ISCSI_BOOT_TGT_FLAGS:
rc = sprintf(str, "2\n");
break;
case ISCSI_BOOT_TGT_NIC_ASSOC:
rc = sprintf(str, "0\n");
break;
}
return rc;
}
static ssize_t beiscsi_show_boot_ini_info(void *data, int type, char *buf)
{
struct beiscsi_hba *phba = data;
char *str = buf;
int rc = -EPERM;
switch (type) {
case ISCSI_BOOT_INI_INITIATOR_NAME:
rc = sprintf(str, "%s\n",
phba->boot_struct.boot_sess.initiator_iscsiname);
break;
}
return rc;
}
static ssize_t beiscsi_show_boot_eth_info(void *data, int type, char *buf)
{
struct beiscsi_hba *phba = data;
char *str = buf;
int rc = -EPERM;
switch (type) {
case ISCSI_BOOT_ETH_FLAGS:
rc = sprintf(str, "2\n");
break;
case ISCSI_BOOT_ETH_INDEX:
rc = sprintf(str, "0\n");
break;
case ISCSI_BOOT_ETH_MAC:
rc = beiscsi_get_macaddr(str, phba);
break;
}
return rc;
}
static umode_t beiscsi_tgt_get_attr_visibility(void *data, int type)
{
umode_t rc = 0;
switch (type) {
case ISCSI_BOOT_TGT_NAME:
case ISCSI_BOOT_TGT_IP_ADDR:
case ISCSI_BOOT_TGT_PORT:
case ISCSI_BOOT_TGT_CHAP_NAME:
case ISCSI_BOOT_TGT_CHAP_SECRET:
case ISCSI_BOOT_TGT_REV_CHAP_NAME:
case ISCSI_BOOT_TGT_REV_CHAP_SECRET:
case ISCSI_BOOT_TGT_NIC_ASSOC:
case ISCSI_BOOT_TGT_FLAGS:
rc = S_IRUGO;
break;
}
return rc;
}
static umode_t beiscsi_ini_get_attr_visibility(void *data, int type)
{
umode_t rc = 0;
switch (type) {
case ISCSI_BOOT_INI_INITIATOR_NAME:
rc = S_IRUGO;
break;
}
return rc;
}
static umode_t beiscsi_eth_get_attr_visibility(void *data, int type)
{
umode_t rc = 0;
switch (type) {
case ISCSI_BOOT_ETH_FLAGS:
case ISCSI_BOOT_ETH_MAC:
case ISCSI_BOOT_ETH_INDEX:
rc = S_IRUGO;
break;
}
return rc;
}
static void beiscsi_boot_kobj_release(void *data)
{
struct beiscsi_hba *phba = data;
scsi_host_put(phba->shost);
}
static int beiscsi_boot_create_kset(struct beiscsi_hba *phba)
{
struct boot_struct *bs = &phba->boot_struct;
struct iscsi_boot_kobj *boot_kobj;
if (bs->boot_kset) {
__beiscsi_log(phba, KERN_ERR,
"BM_%d: boot_kset already created\n");
return 0;
}
bs->boot_kset = iscsi_boot_create_host_kset(phba->shost->host_no);
if (!bs->boot_kset) {
__beiscsi_log(phba, KERN_ERR,
"BM_%d: boot_kset alloc failed\n");
return -ENOMEM;
}
/* get shost ref because the show function will refer phba */
if (!scsi_host_get(phba->shost))
goto free_kset;
boot_kobj = iscsi_boot_create_target(bs->boot_kset, 0, phba,
beiscsi_show_boot_tgt_info,
beiscsi_tgt_get_attr_visibility,
beiscsi_boot_kobj_release);
if (!boot_kobj)
goto put_shost;
if (!scsi_host_get(phba->shost))
goto free_kset;
boot_kobj = iscsi_boot_create_initiator(bs->boot_kset, 0, phba,
beiscsi_show_boot_ini_info,
beiscsi_ini_get_attr_visibility,
beiscsi_boot_kobj_release);
if (!boot_kobj)
goto put_shost;
if (!scsi_host_get(phba->shost))
goto free_kset;
boot_kobj = iscsi_boot_create_ethernet(bs->boot_kset, 0, phba,
beiscsi_show_boot_eth_info,
beiscsi_eth_get_attr_visibility,
beiscsi_boot_kobj_release);
if (!boot_kobj)
goto put_shost;
return 0;
put_shost:
scsi_host_put(phba->shost);
free_kset:
iscsi_boot_destroy_kset(bs->boot_kset);
bs->boot_kset = NULL;
return -ENOMEM;
}
static void beiscsi_boot_work(struct work_struct *work)
{
struct beiscsi_hba *phba =
container_of(work, struct beiscsi_hba, boot_work);
struct boot_struct *bs = &phba->boot_struct;
unsigned int tag = 0;
if (!beiscsi_hba_is_online(phba))
return;
beiscsi_log(phba, KERN_INFO,
BEISCSI_LOG_CONFIG | BEISCSI_LOG_MBOX,
"BM_%d : %s action %d\n",
__func__, phba->boot_struct.action);
switch (phba->boot_struct.action) {
case BEISCSI_BOOT_REOPEN_SESS:
tag = beiscsi_boot_reopen_sess(phba);
break;
case BEISCSI_BOOT_GET_SHANDLE:
tag = __beiscsi_boot_get_shandle(phba, 1);
break;
case BEISCSI_BOOT_GET_SINFO:
tag = beiscsi_boot_get_sinfo(phba);
break;
case BEISCSI_BOOT_LOGOUT_SESS:
tag = beiscsi_boot_logout_sess(phba);
break;
case BEISCSI_BOOT_CREATE_KSET:
beiscsi_boot_create_kset(phba);
/**
* updated boot_kset is made visible to all before
* ending the boot work.
*/
mb();
clear_bit(BEISCSI_HBA_BOOT_WORK, &phba->state);
return;
}
if (!tag) {
if (bs->retry--)
schedule_work(&phba->boot_work);
else
clear_bit(BEISCSI_HBA_BOOT_WORK, &phba->state);
}
}
static void beiscsi_eqd_update_work(struct work_struct *work)
{
struct hwi_context_memory *phwi_context;
struct be_set_eqd set_eqd[MAX_CPUS];
struct hwi_controller *phwi_ctrlr;
struct be_eq_obj *pbe_eq;
struct beiscsi_hba *phba;
unsigned int pps, delta;
struct be_aic_obj *aic;
int eqd, i, num = 0;
unsigned long now;
phba = container_of(work, struct beiscsi_hba, eqd_update.work);
if (!beiscsi_hba_is_online(phba))
return;
phwi_ctrlr = phba->phwi_ctrlr;
phwi_context = phwi_ctrlr->phwi_ctxt;
for (i = 0; i <= phba->num_cpus; i++) {
aic = &phba->aic_obj[i];
pbe_eq = &phwi_context->be_eq[i];
now = jiffies;
if (!aic->jiffies || time_before(now, aic->jiffies) ||
pbe_eq->cq_count < aic->eq_prev) {
aic->jiffies = now;
aic->eq_prev = pbe_eq->cq_count;
continue;
}
delta = jiffies_to_msecs(now - aic->jiffies);
pps = (((u32)(pbe_eq->cq_count - aic->eq_prev) * 1000) / delta);
eqd = (pps / 1500) << 2;
if (eqd < 8)
eqd = 0;
eqd = min_t(u32, eqd, phwi_context->max_eqd);
eqd = max_t(u32, eqd, phwi_context->min_eqd);
aic->jiffies = now;
aic->eq_prev = pbe_eq->cq_count;
if (eqd != aic->prev_eqd) {
set_eqd[num].delay_multiplier = (eqd * 65)/100;
set_eqd[num].eq_id = pbe_eq->q.id;
aic->prev_eqd = eqd;
num++;
}
}
if (num)
/* completion of this is ignored */
beiscsi_modify_eq_delay(phba, set_eqd, num);
schedule_delayed_work(&phba->eqd_update,
msecs_to_jiffies(BEISCSI_EQD_UPDATE_INTERVAL));
}
static void beiscsi_hw_tpe_check(unsigned long ptr)
{
struct beiscsi_hba *phba;
u32 wait;
phba = (struct beiscsi_hba *)ptr;
/* if not TPE, do nothing */
if (!beiscsi_detect_tpe(phba))
return;
/* wait default 4000ms before recovering */
wait = 4000;
if (phba->ue2rp > BEISCSI_UE_DETECT_INTERVAL)
wait = phba->ue2rp - BEISCSI_UE_DETECT_INTERVAL;
queue_delayed_work(phba->wq, &phba->recover_port,
msecs_to_jiffies(wait));
}
static void beiscsi_hw_health_check(unsigned long ptr)
{
struct beiscsi_hba *phba;
phba = (struct beiscsi_hba *)ptr;
beiscsi_detect_ue(phba);
if (beiscsi_detect_ue(phba)) {
__beiscsi_log(phba, KERN_ERR,
"BM_%d : port in error: %lx\n", phba->state);
/* sessions are no longer valid, so first fail the sessions */
queue_work(phba->wq, &phba->sess_work);
/* detect UER supported */
if (!test_bit(BEISCSI_HBA_UER_SUPP, &phba->state))
return;
/* modify this timer to check TPE */
phba->hw_check.function = beiscsi_hw_tpe_check;
}
mod_timer(&phba->hw_check,
jiffies + msecs_to_jiffies(BEISCSI_UE_DETECT_INTERVAL));
}
/*
* beiscsi_enable_port()- Enables the disabled port.
* Only port resources freed in disable function are reallocated.
* This is called in HBA error handling path.
*
* @phba: Instance of driver private structure
*
**/
static int beiscsi_enable_port(struct beiscsi_hba *phba)
{
struct hwi_context_memory *phwi_context;
struct hwi_controller *phwi_ctrlr;
struct be_eq_obj *pbe_eq;
int ret, i;
if (test_bit(BEISCSI_HBA_ONLINE, &phba->state)) {
__beiscsi_log(phba, KERN_ERR,
"BM_%d : %s : port is online %lx\n",
__func__, phba->state);
return 0;
}
ret = beiscsi_init_sliport(phba);
if (ret)
return ret;
be2iscsi_enable_msix(phba);
beiscsi_get_params(phba);
/* Re-enable UER. If different TPE occurs then it is recoverable. */
beiscsi_set_uer_feature(phba);
phba->shost->max_id = phba->params.cxns_per_ctrl;
phba->shost->can_queue = phba->params.ios_per_ctrl;
ret = beiscsi_init_port(phba);
if (ret < 0) {
__beiscsi_log(phba, KERN_ERR,
"BM_%d : init port failed\n");
goto disable_msix;
}
for (i = 0; i < MAX_MCC_CMD; i++) {
init_waitqueue_head(&phba->ctrl.mcc_wait[i + 1]);
phba->ctrl.mcc_tag[i] = i + 1;
phba->ctrl.mcc_tag_status[i + 1] = 0;
phba->ctrl.mcc_tag_available++;
}
phwi_ctrlr = phba->phwi_ctrlr;
phwi_context = phwi_ctrlr->phwi_ctxt;
for (i = 0; i < phba->num_cpus; i++) {
pbe_eq = &phwi_context->be_eq[i];
irq_poll_init(&pbe_eq->iopoll, be_iopoll_budget, be_iopoll);
}
i = (phba->pcidev->msix_enabled) ? i : 0;
/* Work item for MCC handling */
pbe_eq = &phwi_context->be_eq[i];
INIT_WORK(&pbe_eq->mcc_work, beiscsi_mcc_work);
ret = beiscsi_init_irqs(phba);
if (ret < 0) {
__beiscsi_log(phba, KERN_ERR,
"BM_%d : setup IRQs failed %d\n", ret);
goto cleanup_port;
}
hwi_enable_intr(phba);
/* port operational: clear all error bits */
set_bit(BEISCSI_HBA_ONLINE, &phba->state);
__beiscsi_log(phba, KERN_INFO,
"BM_%d : port online: 0x%lx\n", phba->state);
/* start hw_check timer and eqd_update work */
schedule_delayed_work(&phba->eqd_update,
msecs_to_jiffies(BEISCSI_EQD_UPDATE_INTERVAL));
/**
* Timer function gets modified for TPE detection.
* Always reinit to do health check first.
*/
phba->hw_check.function = beiscsi_hw_health_check;
mod_timer(&phba->hw_check,
jiffies + msecs_to_jiffies(BEISCSI_UE_DETECT_INTERVAL));
return 0;
cleanup_port:
for (i = 0; i < phba->num_cpus; i++) {
pbe_eq = &phwi_context->be_eq[i];
irq_poll_disable(&pbe_eq->iopoll);
}
hwi_cleanup_port(phba);
disable_msix:
pci_free_irq_vectors(phba->pcidev);
return ret;
}
/*
* beiscsi_disable_port()- Disable port and cleanup driver resources.
* This is called in HBA error handling and driver removal.
* @phba: Instance Priv structure
* @unload: indicate driver is unloading
*
* Free the OS and HW resources held by the driver
**/
static void beiscsi_disable_port(struct beiscsi_hba *phba, int unload)
{
struct hwi_context_memory *phwi_context;
struct hwi_controller *phwi_ctrlr;
struct be_eq_obj *pbe_eq;
unsigned int i;
if (!test_and_clear_bit(BEISCSI_HBA_ONLINE, &phba->state))
return;
phwi_ctrlr = phba->phwi_ctrlr;
phwi_context = phwi_ctrlr->phwi_ctxt;
hwi_disable_intr(phba);
if (phba->pcidev->msix_enabled) {
for (i = 0; i <= phba->num_cpus; i++) {
free_irq(pci_irq_vector(phba->pcidev, i),
&phwi_context->be_eq[i]);
kfree(phba->msi_name[i]);
}
} else
if (phba->pcidev->irq)
free_irq(phba->pcidev->irq, phba);
pci_free_irq_vectors(phba->pcidev);
for (i = 0; i < phba->num_cpus; i++) {
pbe_eq = &phwi_context->be_eq[i];
irq_poll_disable(&pbe_eq->iopoll);
}
cancel_delayed_work_sync(&phba->eqd_update);
cancel_work_sync(&phba->boot_work);
/* WQ might be running cancel queued mcc_work if we are not exiting */
if (!unload && beiscsi_hba_in_error(phba)) {
pbe_eq = &phwi_context->be_eq[i];
cancel_work_sync(&pbe_eq->mcc_work);
}
hwi_cleanup_port(phba);
beiscsi_cleanup_port(phba);
}
static void beiscsi_sess_work(struct work_struct *work)
{
struct beiscsi_hba *phba;
phba = container_of(work, struct beiscsi_hba, sess_work);
/*
* This work gets scheduled only in case of HBA error.
* Old sessions are gone so need to be re-established.
* iscsi_session_failure needs process context hence this work.
*/
iscsi_host_for_each_session(phba->shost, beiscsi_session_fail);
}
static void beiscsi_recover_port(struct work_struct *work)
{
struct beiscsi_hba *phba;
phba = container_of(work, struct beiscsi_hba, recover_port.work);
beiscsi_disable_port(phba, 0);
beiscsi_enable_port(phba);
}
static pci_ers_result_t beiscsi_eeh_err_detected(struct pci_dev *pdev,
pci_channel_state_t state)
{
struct beiscsi_hba *phba = NULL;
phba = (struct beiscsi_hba *)pci_get_drvdata(pdev);
set_bit(BEISCSI_HBA_PCI_ERR, &phba->state);
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : EEH error detected\n");
/* first stop UE detection when PCI error detected */
del_timer_sync(&phba->hw_check);
cancel_delayed_work_sync(&phba->recover_port);
/* sessions are no longer valid, so first fail the sessions */
iscsi_host_for_each_session(phba->shost, beiscsi_session_fail);
beiscsi_disable_port(phba, 0);
if (state == pci_channel_io_perm_failure) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : EEH : State PERM Failure");
return PCI_ERS_RESULT_DISCONNECT;
}
pci_disable_device(pdev);
/* The error could cause the FW to trigger a flash debug dump.
* Resetting the card while flash dump is in progress
* can cause it not to recover; wait for it to finish.
* Wait only for first function as it is needed only once per
* adapter.
**/
if (pdev->devfn == 0)
ssleep(30);
return PCI_ERS_RESULT_NEED_RESET;
}
static pci_ers_result_t beiscsi_eeh_reset(struct pci_dev *pdev)
{
struct beiscsi_hba *phba = NULL;
int status = 0;
phba = (struct beiscsi_hba *)pci_get_drvdata(pdev);
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : EEH Reset\n");
status = pci_enable_device(pdev);
if (status)
return PCI_ERS_RESULT_DISCONNECT;
pci_set_master(pdev);
pci_set_power_state(pdev, PCI_D0);
pci_restore_state(pdev);
status = beiscsi_check_fw_rdy(phba);
if (status) {
beiscsi_log(phba, KERN_WARNING, BEISCSI_LOG_INIT,
"BM_%d : EEH Reset Completed\n");
} else {
beiscsi_log(phba, KERN_WARNING, BEISCSI_LOG_INIT,
"BM_%d : EEH Reset Completion Failure\n");
return PCI_ERS_RESULT_DISCONNECT;
}
pci_cleanup_aer_uncorrect_error_status(pdev);
return PCI_ERS_RESULT_RECOVERED;
}
static void beiscsi_eeh_resume(struct pci_dev *pdev)
{
struct beiscsi_hba *phba;
int ret;
phba = (struct beiscsi_hba *)pci_get_drvdata(pdev);
pci_save_state(pdev);
ret = beiscsi_enable_port(phba);
if (ret)
__beiscsi_log(phba, KERN_ERR,
"BM_%d : AER EEH resume failed\n");
}
static int beiscsi_dev_probe(struct pci_dev *pcidev,
const struct pci_device_id *id)
{
struct hwi_context_memory *phwi_context;
struct hwi_controller *phwi_ctrlr;
struct beiscsi_hba *phba = NULL;
struct be_eq_obj *pbe_eq;
unsigned int s_handle;
char wq_name[20];
int ret, i;
ret = beiscsi_enable_pci(pcidev);
if (ret < 0) {
dev_err(&pcidev->dev,
"beiscsi_dev_probe - Failed to enable pci device\n");
return ret;
}
phba = beiscsi_hba_alloc(pcidev);
if (!phba) {
dev_err(&pcidev->dev,
"beiscsi_dev_probe - Failed in beiscsi_hba_alloc\n");
ret = -ENOMEM;
goto disable_pci;
}
/* Enable EEH reporting */
ret = pci_enable_pcie_error_reporting(pcidev);
if (ret)
beiscsi_log(phba, KERN_WARNING, BEISCSI_LOG_INIT,
"BM_%d : PCIe Error Reporting "
"Enabling Failed\n");
pci_save_state(pcidev);
/* Initialize Driver configuration Paramters */
beiscsi_hba_attrs_init(phba);
phba->mac_addr_set = false;
switch (pcidev->device) {
case BE_DEVICE_ID1:
case OC_DEVICE_ID1:
case OC_DEVICE_ID2:
phba->generation = BE_GEN2;
phba->iotask_fn = beiscsi_iotask;
dev_warn(&pcidev->dev,
"Obsolete/Unsupported BE2 Adapter Family\n");
break;
case BE_DEVICE_ID2:
case OC_DEVICE_ID3:
phba->generation = BE_GEN3;
phba->iotask_fn = beiscsi_iotask;
break;
case OC_SKH_ID1:
phba->generation = BE_GEN4;
phba->iotask_fn = beiscsi_iotask_v2;
break;
default:
phba->generation = 0;
}
ret = be_ctrl_init(phba, pcidev);
if (ret) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : be_ctrl_init failed\n");
goto hba_free;
}
ret = beiscsi_init_sliport(phba);
if (ret)
goto hba_free;
spin_lock_init(&phba->io_sgl_lock);
spin_lock_init(&phba->mgmt_sgl_lock);
spin_lock_init(&phba->async_pdu_lock);
ret = beiscsi_get_fw_config(&phba->ctrl, phba);
if (ret != 0) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : Error getting fw config\n");
goto free_port;
}
beiscsi_get_port_name(&phba->ctrl, phba);
beiscsi_get_params(phba);
beiscsi_set_uer_feature(phba);
be2iscsi_enable_msix(phba);
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_INIT,
"BM_%d : num_cpus = %d\n",
phba->num_cpus);
phba->shost->max_id = phba->params.cxns_per_ctrl;
phba->shost->can_queue = phba->params.ios_per_ctrl;
ret = beiscsi_get_memory(phba);
if (ret < 0) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : alloc host mem failed\n");
goto free_port;
}
ret = beiscsi_init_port(phba);
if (ret < 0) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : init port failed\n");
beiscsi_free_mem(phba);
goto free_port;
}
for (i = 0; i < MAX_MCC_CMD; i++) {
init_waitqueue_head(&phba->ctrl.mcc_wait[i + 1]);
phba->ctrl.mcc_tag[i] = i + 1;
phba->ctrl.mcc_tag_status[i + 1] = 0;
phba->ctrl.mcc_tag_available++;
memset(&phba->ctrl.ptag_state[i].tag_mem_state, 0,
sizeof(struct be_dma_mem));
}
phba->ctrl.mcc_alloc_index = phba->ctrl.mcc_free_index = 0;
snprintf(wq_name, sizeof(wq_name), "beiscsi_%02x_wq",
phba->shost->host_no);
phba->wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 1, wq_name);
if (!phba->wq) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : beiscsi_dev_probe-"
"Failed to allocate work queue\n");
ret = -ENOMEM;
goto free_twq;
}
INIT_DELAYED_WORK(&phba->eqd_update, beiscsi_eqd_update_work);
phwi_ctrlr = phba->phwi_ctrlr;
phwi_context = phwi_ctrlr->phwi_ctxt;
for (i = 0; i < phba->num_cpus; i++) {
pbe_eq = &phwi_context->be_eq[i];
irq_poll_init(&pbe_eq->iopoll, be_iopoll_budget, be_iopoll);
}
i = (phba->pcidev->msix_enabled) ? i : 0;
/* Work item for MCC handling */
pbe_eq = &phwi_context->be_eq[i];
INIT_WORK(&pbe_eq->mcc_work, beiscsi_mcc_work);
ret = beiscsi_init_irqs(phba);
if (ret < 0) {
beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
"BM_%d : beiscsi_dev_probe-"
"Failed to beiscsi_init_irqs\n");
goto free_blkenbld;
}
hwi_enable_intr(phba);
ret = iscsi_host_add(phba->shost, &phba->pcidev->dev);
if (ret)
goto free_blkenbld;
/* set online bit after port is operational */
set_bit(BEISCSI_HBA_ONLINE, &phba->state);
__beiscsi_log(phba, KERN_INFO,
"BM_%d : port online: 0x%lx\n", phba->state);
INIT_WORK(&phba->boot_work, beiscsi_boot_work);
ret = beiscsi_boot_get_shandle(phba, &s_handle);
if (ret > 0) {
beiscsi_start_boot_work(phba, s_handle);
/**
* Set this bit after starting the work to let
* probe handle it first.
* ASYNC event can too schedule this work.
*/
set_bit(BEISCSI_HBA_BOOT_FOUND, &phba->state);
}
beiscsi_iface_create_default(phba);
schedule_delayed_work(&phba->eqd_update,
msecs_to_jiffies(BEISCSI_EQD_UPDATE_INTERVAL));
INIT_WORK(&phba->sess_work, beiscsi_sess_work);
INIT_DELAYED_WORK(&phba->recover_port, beiscsi_recover_port);
/**
* Start UE detection here. UE before this will cause stall in probe
* and eventually fail the probe.
*/
init_timer(&phba->hw_check);
phba->hw_check.function = beiscsi_hw_health_check;
phba->hw_check.data = (unsigned long)phba;
mod_timer(&phba->hw_check,
jiffies + msecs_to_jiffies(BEISCSI_UE_DETECT_INTERVAL));
beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_INIT,
"\n\n\n BM_%d : SUCCESS - DRIVER LOADED\n\n\n");
return 0;
free_blkenbld:
destroy_workqueue(phba->wq);
for (i = 0; i < phba->num_cpus; i++) {
pbe_eq = &phwi_context->be_eq[i];
irq_poll_disable(&pbe_eq->iopoll);
}
free_twq:
hwi_cleanup_port(phba);
beiscsi_cleanup_port(phba);
beiscsi_free_mem(phba);
free_port:
pci_free_consistent(phba->pcidev,
phba->ctrl.mbox_mem_alloced.size,
phba->ctrl.mbox_mem_alloced.va,
phba->ctrl.mbox_mem_alloced.dma);
beiscsi_unmap_pci_function(phba);
hba_free:
pci_disable_msix(phba->pcidev);
pci_dev_put(phba->pcidev);
iscsi_host_free(phba->shost);
pci_set_drvdata(pcidev, NULL);
disable_pci:
pci_release_regions(pcidev);
pci_disable_device(pcidev);
return ret;
}
static void beiscsi_remove(struct pci_dev *pcidev)
{
struct beiscsi_hba *phba = NULL;
phba = pci_get_drvdata(pcidev);
if (!phba) {
dev_err(&pcidev->dev, "beiscsi_remove called with no phba\n");
return;
}
/* first stop UE detection before unloading */
del_timer_sync(&phba->hw_check);
cancel_delayed_work_sync(&phba->recover_port);
cancel_work_sync(&phba->sess_work);
beiscsi_iface_destroy_default(phba);
iscsi_host_remove(phba->shost);
beiscsi_disable_port(phba, 1);
/* after cancelling boot_work */
iscsi_boot_destroy_kset(phba->boot_struct.boot_kset);
/* free all resources */
destroy_workqueue(phba->wq);
beiscsi_free_mem(phba);
/* ctrl uninit */
beiscsi_unmap_pci_function(phba);
pci_free_consistent(phba->pcidev,
phba->ctrl.mbox_mem_alloced.size,
phba->ctrl.mbox_mem_alloced.va,
phba->ctrl.mbox_mem_alloced.dma);
pci_dev_put(phba->pcidev);
iscsi_host_free(phba->shost);
pci_disable_pcie_error_reporting(pcidev);
pci_set_drvdata(pcidev, NULL);
pci_release_regions(pcidev);
pci_disable_device(pcidev);
}
static struct pci_error_handlers beiscsi_eeh_handlers = {
.error_detected = beiscsi_eeh_err_detected,
.slot_reset = beiscsi_eeh_reset,
.resume = beiscsi_eeh_resume,
};
struct iscsi_transport beiscsi_iscsi_transport = {
.owner = THIS_MODULE,
.name = DRV_NAME,
.caps = CAP_RECOVERY_L0 | CAP_HDRDGST | CAP_TEXT_NEGO |
CAP_MULTI_R2T | CAP_DATADGST | CAP_DATA_PATH_OFFLOAD,
.create_session = beiscsi_session_create,
.destroy_session = beiscsi_session_destroy,
.create_conn = beiscsi_conn_create,
.bind_conn = beiscsi_conn_bind,
.destroy_conn = iscsi_conn_teardown,
.attr_is_visible = beiscsi_attr_is_visible,
.set_iface_param = beiscsi_iface_set_param,
.get_iface_param = beiscsi_iface_get_param,
.set_param = beiscsi_set_param,
.get_conn_param = iscsi_conn_get_param,
.get_session_param = iscsi_session_get_param,
.get_host_param = beiscsi_get_host_param,
.start_conn = beiscsi_conn_start,
.stop_conn = iscsi_conn_stop,
.send_pdu = iscsi_conn_send_pdu,
.xmit_task = beiscsi_task_xmit,
.cleanup_task = beiscsi_cleanup_task,
.alloc_pdu = beiscsi_alloc_pdu,
.parse_pdu_itt = beiscsi_parse_pdu,
.get_stats = beiscsi_conn_get_stats,
.get_ep_param = beiscsi_ep_get_param,
.ep_connect = beiscsi_ep_connect,
.ep_poll = beiscsi_ep_poll,
.ep_disconnect = beiscsi_ep_disconnect,
.session_recovery_timedout = iscsi_session_recovery_timedout,
.bsg_request = beiscsi_bsg_request,
};
static struct pci_driver beiscsi_pci_driver = {
.name = DRV_NAME,
.probe = beiscsi_dev_probe,
.remove = beiscsi_remove,
.id_table = beiscsi_pci_id_table,
.err_handler = &beiscsi_eeh_handlers
};
static int __init beiscsi_module_init(void)
{
int ret;
beiscsi_scsi_transport =
iscsi_register_transport(&beiscsi_iscsi_transport);
if (!beiscsi_scsi_transport) {
printk(KERN_ERR
"beiscsi_module_init - Unable to register beiscsi transport.\n");
return -ENOMEM;
}
printk(KERN_INFO "In beiscsi_module_init, tt=%p\n",
&beiscsi_iscsi_transport);
ret = pci_register_driver(&beiscsi_pci_driver);
if (ret) {
printk(KERN_ERR
"beiscsi_module_init - Unable to register beiscsi pci driver.\n");
goto unregister_iscsi_transport;
}
return 0;
unregister_iscsi_transport:
iscsi_unregister_transport(&beiscsi_iscsi_transport);
return ret;
}
static void __exit beiscsi_module_exit(void)
{
pci_unregister_driver(&beiscsi_pci_driver);
iscsi_unregister_transport(&beiscsi_iscsi_transport);
}
module_init(beiscsi_module_init);
module_exit(beiscsi_module_exit);