OpenCloudOS-Kernel/drivers/scsi/be2iscsi/be_main.h

1031 lines
28 KiB
C
Raw Permalink Normal View History

/* SPDX-License-Identifier: GPL-2.0-only */
/*
* Copyright 2017 Broadcom. All Rights Reserved.
* The term "Broadcom" refers to Broadcom Limited and/or its subsidiaries.
*
* Contact Information:
* linux-drivers@broadcom.com
*/
#ifndef _BEISCSI_MAIN_
#define _BEISCSI_MAIN_
#include <linux/kernel.h>
#include <linux/pci.h>
#include <linux/if_ether.h>
#include <linux/in.h>
#include <linux/ctype.h>
#include <linux/module.h>
#include <linux/aer.h>
#include <scsi/scsi.h>
#include <scsi/scsi_cmnd.h>
#include <scsi/scsi_device.h>
#include <scsi/scsi_host.h>
#include <scsi/iscsi_proto.h>
#include <scsi/libiscsi.h>
#include <scsi/scsi_transport_iscsi.h>
#define DRV_NAME "be2iscsi"
#define BUILD_STR "11.4.0.1"
#define BE_NAME "Emulex OneConnect" \
"Open-iSCSI Driver version" BUILD_STR
#define DRV_DESC BE_NAME " " "Driver"
#define BE_VENDOR_ID 0x19A2
#define ELX_VENDOR_ID 0x10DF
/* DEVICE ID's for BE2 */
#define BE_DEVICE_ID1 0x212
#define OC_DEVICE_ID1 0x702
#define OC_DEVICE_ID2 0x703
/* DEVICE ID's for BE3 */
#define BE_DEVICE_ID2 0x222
#define OC_DEVICE_ID3 0x712
/* DEVICE ID for SKH */
#define OC_SKH_ID1 0x722
#define BE2_IO_DEPTH 1024
#define BE2_MAX_SESSIONS 256
#define BE2_TMFS 16
#define BE2_NOPOUT_REQ 16
#define BE2_SGE 32
#define BE2_DEFPDU_HDR_SZ 64
#define BE2_DEFPDU_DATA_SZ 8192
#define BE2_MAX_NUM_CQ_PROC 512
#define MAX_CPUS 64U
#define BEISCSI_MAX_NUM_CPUS 7
#define BEISCSI_VER_STRLEN 32
#define BEISCSI_SGLIST_ELEMENTS 30
/**
* BE_INVLDT_CMD_TBL_SZ is 128 which is total number commands that can
* be invalidated at a time, consider it before changing the value of
* BEISCSI_CMD_PER_LUN.
*/
#define BEISCSI_CMD_PER_LUN 128 /* scsi_host->cmd_per_lun */
#define BEISCSI_MAX_SECTORS 1024 /* scsi_host->max_sectors */
#define BEISCSI_TEMPLATE_HDR_PER_CXN_SIZE 128 /* Template size per cxn */
#define BEISCSI_MAX_CMD_LEN 16 /* scsi_host->max_cmd_len */
#define BEISCSI_NUM_MAX_LUN 256 /* scsi_host->max_lun */
#define BEISCSI_MAX_FRAGS_INIT 192
#define BE_SENSE_INFO_SIZE 258
#define BE_ISCSI_PDU_HEADER_SIZE 64
#define BE_MIN_MEM_SIZE 16384
#define MAX_CMD_SZ 65536
#define IIOC_SCSI_DATA 0x05 /* Write Operation */
/**
* hardware needs the async PDU buffers to be posted in multiples of 8
* So have atleast 8 of them by default
*/
#define HWI_GET_ASYNC_PDU_CTX(phwi, ulp_num) \
(phwi->phwi_ctxt->pasync_ctx[ulp_num])
/********* Memory BAR register ************/
#define PCICFG_MEMBAR_CTRL_INT_CTRL_OFFSET 0xfc
/**
* Host Interrupt Enable, if set interrupts are enabled although "PCI Interrupt
* Disable" may still globally block interrupts in addition to individual
* interrupt masks; a mechanism for the device driver to block all interrupts
* atomically without having to arbitrate for the PCI Interrupt Disable bit
* with the OS.
*/
#define MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK (1 << 29) /* bit 29 */
/********* ISR0 Register offset **********/
#define CEV_ISR0_OFFSET 0xC18
#define CEV_ISR_SIZE 4
/**
* Macros for reading/writing a protection domain or CSR registers
* in BladeEngine.
*/
#define DB_TXULP0_OFFSET 0x40
#define DB_RXULP0_OFFSET 0xA0
/********* Event Q door bell *************/
#define DB_EQ_OFFSET DB_CQ_OFFSET
#define DB_EQ_RING_ID_LOW_MASK 0x1FF /* bits 0 - 8 */
/* Clear the interrupt for this eq */
#define DB_EQ_CLR_SHIFT (9) /* bit 9 */
/* Must be 1 */
#define DB_EQ_EVNT_SHIFT (10) /* bit 10 */
/* Higher Order EQ_ID bit */
#define DB_EQ_RING_ID_HIGH_MASK 0x1F /* bits 11 - 15 */
#define DB_EQ_HIGH_SET_SHIFT 11
#define DB_EQ_HIGH_FEILD_SHIFT 9
/* Number of event entries processed */
#define DB_EQ_NUM_POPPED_SHIFT (16) /* bits 16 - 28 */
/* Rearm bit */
#define DB_EQ_REARM_SHIFT (29) /* bit 29 */
/********* Compl Q door bell *************/
#define DB_CQ_OFFSET 0x120
#define DB_CQ_RING_ID_LOW_MASK 0x3FF /* bits 0 - 9 */
/* Higher Order CQ_ID bit */
#define DB_CQ_RING_ID_HIGH_MASK 0x1F /* bits 11 - 15 */
#define DB_CQ_HIGH_SET_SHIFT 11
#define DB_CQ_HIGH_FEILD_SHIFT 10
/* Number of event entries processed */
#define DB_CQ_NUM_POPPED_SHIFT (16) /* bits 16 - 28 */
/* Rearm bit */
#define DB_CQ_REARM_SHIFT (29) /* bit 29 */
#define GET_HWI_CONTROLLER_WS(pc) (pc->phwi_ctrlr)
#define HWI_GET_DEF_BUFQ_ID(pc, ulp_num) (((struct hwi_controller *)\
(GET_HWI_CONTROLLER_WS(pc)))->default_pdu_data[ulp_num].id)
#define HWI_GET_DEF_HDRQ_ID(pc, ulp_num) (((struct hwi_controller *)\
(GET_HWI_CONTROLLER_WS(pc)))->default_pdu_hdr[ulp_num].id)
#define PAGES_REQUIRED(x) \
((x < PAGE_SIZE) ? 1 : ((x + PAGE_SIZE - 1) / PAGE_SIZE))
#define MEM_DESCR_OFFSET 8
#define BEISCSI_DEFQ_HDR 1
#define BEISCSI_DEFQ_DATA 0
enum be_mem_enum {
HWI_MEM_ADDN_CONTEXT,
HWI_MEM_WRB,
HWI_MEM_WRBH,
HWI_MEM_SGLH,
HWI_MEM_SGE,
HWI_MEM_TEMPLATE_HDR_ULP0,
HWI_MEM_ASYNC_HEADER_BUF_ULP0, /* 6 */
HWI_MEM_ASYNC_DATA_BUF_ULP0,
HWI_MEM_ASYNC_HEADER_RING_ULP0,
HWI_MEM_ASYNC_DATA_RING_ULP0,
HWI_MEM_ASYNC_HEADER_HANDLE_ULP0,
HWI_MEM_ASYNC_DATA_HANDLE_ULP0, /* 11 */
HWI_MEM_ASYNC_PDU_CONTEXT_ULP0,
HWI_MEM_TEMPLATE_HDR_ULP1,
HWI_MEM_ASYNC_HEADER_BUF_ULP1, /* 14 */
HWI_MEM_ASYNC_DATA_BUF_ULP1,
HWI_MEM_ASYNC_HEADER_RING_ULP1,
HWI_MEM_ASYNC_DATA_RING_ULP1,
HWI_MEM_ASYNC_HEADER_HANDLE_ULP1,
HWI_MEM_ASYNC_DATA_HANDLE_ULP1, /* 19 */
HWI_MEM_ASYNC_PDU_CONTEXT_ULP1,
ISCSI_MEM_GLOBAL_HEADER,
SE_MEM_MAX
};
struct be_bus_address32 {
unsigned int address_lo;
unsigned int address_hi;
};
struct be_bus_address64 {
unsigned long long address;
};
struct be_bus_address {
union {
struct be_bus_address32 a32;
struct be_bus_address64 a64;
} u;
};
struct mem_array {
struct be_bus_address bus_address; /* Bus address of location */
void *virtual_address; /* virtual address to the location */
unsigned int size; /* Size required by memory block */
};
struct be_mem_descriptor {
unsigned int size_in_bytes; /* Size required by memory block */
unsigned int num_elements;
struct mem_array *mem_array;
};
struct sgl_handle {
unsigned int sgl_index;
unsigned int type;
unsigned int cid;
struct iscsi_task *task;
struct iscsi_sge *pfrag;
};
struct hba_parameters {
unsigned int ios_per_ctrl;
unsigned int cxns_per_ctrl;
unsigned int icds_per_ctrl;
unsigned int num_sge_per_io;
unsigned int defpdu_hdr_sz;
unsigned int defpdu_data_sz;
unsigned int num_cq_entries;
unsigned int num_eq_entries;
unsigned int wrbs_per_cxn;
unsigned int hwi_ws_sz;
};
#define BEISCSI_GET_ULP_FROM_CRI(phwi_ctrlr, cri) \
(phwi_ctrlr->wrb_context[cri].ulp_num)
struct hwi_wrb_context {
spinlock_t wrb_lock;
struct wrb_handle **pwrb_handle_base;
struct wrb_handle **pwrb_handle_basestd;
struct iscsi_wrb *plast_wrb;
unsigned short alloc_index;
unsigned short free_index;
unsigned short wrb_handles_available;
unsigned short cid;
uint8_t ulp_num; /* ULP to which CID binded */
uint32_t doorbell_offset;
};
struct ulp_cid_info {
unsigned short *cid_array;
unsigned short avlbl_cids;
unsigned short cid_alloc;
unsigned short cid_free;
};
#include "be.h"
#define chip_be2(phba) (phba->generation == BE_GEN2)
#define chip_be3_r(phba) (phba->generation == BE_GEN3)
#define is_chip_be2_be3r(phba) (chip_be3_r(phba) || (chip_be2(phba)))
#define BEISCSI_ULP0 0
#define BEISCSI_ULP1 1
#define BEISCSI_ULP_COUNT 2
#define BEISCSI_ULP0_LOADED 0x01
#define BEISCSI_ULP1_LOADED 0x02
#define BEISCSI_ULP_AVLBL_CID(phba, ulp_num) \
(((struct ulp_cid_info *)phba->cid_array_info[ulp_num])->avlbl_cids)
#define BEISCSI_ULP0_AVLBL_CID(phba) \
BEISCSI_ULP_AVLBL_CID(phba, BEISCSI_ULP0)
#define BEISCSI_ULP1_AVLBL_CID(phba) \
BEISCSI_ULP_AVLBL_CID(phba, BEISCSI_ULP1)
struct beiscsi_hba {
struct hba_parameters params;
struct hwi_controller *phwi_ctrlr;
unsigned int mem_req[SE_MEM_MAX];
/* PCI BAR mapped addresses */
u8 __iomem *csr_va; /* CSR */
u8 __iomem *db_va; /* Door Bell */
u8 __iomem *pci_va; /* PCI Config */
/* PCI representation of our HBA */
struct pci_dev *pcidev;
unsigned int num_cpus;
unsigned int nxt_cqid;
char *msi_name[MAX_CPUS];
struct be_mem_descriptor *init_mem;
unsigned short io_sgl_alloc_index;
unsigned short io_sgl_free_index;
unsigned short io_sgl_hndl_avbl;
struct sgl_handle **io_sgl_hndl_base;
unsigned short eh_sgl_alloc_index;
unsigned short eh_sgl_free_index;
unsigned short eh_sgl_hndl_avbl;
struct sgl_handle **eh_sgl_hndl_base;
spinlock_t io_sgl_lock;
spinlock_t mgmt_sgl_lock;
spinlock_t async_pdu_lock;
struct list_head hba_queue;
#define BE_MAX_SESSION 2048
#define BE_INVALID_CID 0xffff
#define BE_SET_CID_TO_CRI(cri_index, cid) \
(phba->cid_to_cri_map[cid] = cri_index)
#define BE_GET_CRI_FROM_CID(cid) (phba->cid_to_cri_map[cid])
unsigned short cid_to_cri_map[BE_MAX_SESSION];
struct ulp_cid_info *cid_array_info[BEISCSI_ULP_COUNT];
struct iscsi_endpoint **ep_array;
struct beiscsi_conn **conn_table;
struct Scsi_Host *shost;
struct iscsi_iface *ipv4_iface;
struct iscsi_iface *ipv6_iface;
struct {
/**
* group together since they are used most frequently
* for cid to cri conversion
*/
#define BEISCSI_PHYS_PORT_MAX 4
unsigned int phys_port;
/* valid values of phys_port id are 0, 1, 2, 3 */
unsigned int eqid_count;
unsigned int cqid_count;
unsigned int iscsi_cid_start[BEISCSI_ULP_COUNT];
#define BEISCSI_GET_CID_COUNT(phba, ulp_num) \
(phba->fw_config.iscsi_cid_count[ulp_num])
unsigned int iscsi_cid_count[BEISCSI_ULP_COUNT];
unsigned int iscsi_icd_count[BEISCSI_ULP_COUNT];
unsigned int iscsi_icd_start[BEISCSI_ULP_COUNT];
unsigned int iscsi_chain_start[BEISCSI_ULP_COUNT];
unsigned int iscsi_chain_count[BEISCSI_ULP_COUNT];
unsigned short iscsi_features;
uint16_t dual_ulp_aware;
unsigned long ulp_supported;
} fw_config;
unsigned long state;
#define BEISCSI_HBA_ONLINE 0
#define BEISCSI_HBA_LINK_UP 1
#define BEISCSI_HBA_BOOT_FOUND 2
#define BEISCSI_HBA_BOOT_WORK 3
#define BEISCSI_HBA_UER_SUPP 4
#define BEISCSI_HBA_PCI_ERR 5
#define BEISCSI_HBA_FW_TIMEOUT 6
#define BEISCSI_HBA_IN_UE 7
#define BEISCSI_HBA_IN_TPE 8
/* error bits */
#define BEISCSI_HBA_IN_ERR ((1 << BEISCSI_HBA_PCI_ERR) | \
(1 << BEISCSI_HBA_FW_TIMEOUT) | \
(1 << BEISCSI_HBA_IN_UE) | \
(1 << BEISCSI_HBA_IN_TPE))
u8 optic_state;
struct delayed_work eqd_update;
/* update EQ delay timer every 1000ms */
#define BEISCSI_EQD_UPDATE_INTERVAL 1000
struct timer_list hw_check;
/* check for UE every 1000ms */
#define BEISCSI_UE_DETECT_INTERVAL 1000
u32 ue2rp;
struct delayed_work recover_port;
struct work_struct sess_work;
bool mac_addr_set;
u8 mac_address[ETH_ALEN];
u8 port_name;
u8 port_speed;
char fw_ver_str[BEISCSI_VER_STRLEN];
struct workqueue_struct *wq; /* The actuak work queue */
struct be_ctrl_info ctrl;
unsigned int generation;
unsigned int interface_handle;
struct be_aic_obj aic_obj[MAX_CPUS];
unsigned int attr_log_enable;
int (*iotask_fn)(struct iscsi_task *,
struct scatterlist *sg,
uint32_t num_sg, uint32_t xferlen,
uint32_t writedir);
struct boot_struct {
int retry;
unsigned int tag;
unsigned int s_handle;
struct be_dma_mem nonemb_cmd;
enum {
BEISCSI_BOOT_REOPEN_SESS = 1,
BEISCSI_BOOT_GET_SHANDLE,
BEISCSI_BOOT_GET_SINFO,
BEISCSI_BOOT_LOGOUT_SESS,
BEISCSI_BOOT_CREATE_KSET,
} action;
struct mgmt_session_info boot_sess;
struct iscsi_boot_kset *boot_kset;
} boot_struct;
struct work_struct boot_work;
};
#define beiscsi_hba_in_error(phba) ((phba)->state & BEISCSI_HBA_IN_ERR)
#define beiscsi_hba_is_online(phba) \
(!beiscsi_hba_in_error((phba)) && \
test_bit(BEISCSI_HBA_ONLINE, &phba->state))
struct beiscsi_session {
struct dma_pool *bhs_pool;
};
/**
* struct beiscsi_conn - iscsi connection structure
*/
struct beiscsi_conn {
struct iscsi_conn *conn;
struct beiscsi_hba *phba;
u32 exp_statsn;
u32 doorbell_offset;
u32 beiscsi_conn_cid;
struct beiscsi_endpoint *ep;
unsigned short login_in_progress;
struct wrb_handle *plogin_wrb_handle;
struct sgl_handle *plogin_sgl_handle;
struct beiscsi_session *beiscsi_sess;
struct iscsi_task *task;
};
/* This structure is used by the chip */
struct pdu_data_out {
u32 dw[12];
};
/**
* Pseudo amap definition in which each bit of the actual structure is defined
* as a byte: used to calculate offset/shift/mask of each field
*/
struct amap_pdu_data_out {
u8 opcode[6]; /* opcode */
u8 rsvd0[2]; /* should be 0 */
u8 rsvd1[7];
u8 final_bit; /* F bit */
u8 rsvd2[16];
u8 ahs_length[8]; /* no AHS */
u8 data_len_hi[8];
u8 data_len_lo[16]; /* DataSegmentLength */
u8 lun[64];
u8 itt[32]; /* ITT; initiator task tag */
u8 ttt[32]; /* TTT; valid for R2T or 0xffffffff */
u8 rsvd3[32];
u8 exp_stat_sn[32];
u8 rsvd4[32];
u8 data_sn[32];
u8 buffer_offset[32];
u8 rsvd5[32];
};
struct be_cmd_bhs {
struct iscsi_scsi_req iscsi_hdr;
unsigned char pad1[16];
struct pdu_data_out iscsi_data_pdu;
unsigned char pad2[BE_SENSE_INFO_SIZE -
sizeof(struct pdu_data_out)];
};
struct beiscsi_io_task {
struct wrb_handle *pwrb_handle;
struct sgl_handle *psgl_handle;
struct beiscsi_conn *conn;
struct scsi_cmnd *scsi_cmnd;
int num_sg;
struct hwi_wrb_context *pwrb_context;
itt_t libiscsi_itt;
struct be_cmd_bhs *cmd_bhs;
struct be_bus_address bhs_pa;
unsigned short bhs_len;
dma_addr_t mtask_addr;
uint32_t mtask_data_count;
uint8_t wrb_type;
};
struct be_nonio_bhs {
struct iscsi_hdr iscsi_hdr;
unsigned char pad1[16];
struct pdu_data_out iscsi_data_pdu;
unsigned char pad2[BE_SENSE_INFO_SIZE -
sizeof(struct pdu_data_out)];
};
struct be_status_bhs {
struct iscsi_scsi_req iscsi_hdr;
unsigned char pad1[16];
/**
* The plus 2 below is to hold the sense info length that gets
* DMA'ed by RxULP
*/
unsigned char sense_info[BE_SENSE_INFO_SIZE];
};
struct iscsi_sge {
u32 dw[4];
};
/**
* Pseudo amap definition in which each bit of the actual structure is defined
* as a byte: used to calculate offset/shift/mask of each field
*/
struct amap_iscsi_sge {
u8 addr_hi[32];
u8 addr_lo[32];
u8 sge_offset[22]; /* DWORD 2 */
u8 rsvd0[9]; /* DWORD 2 */
u8 last_sge; /* DWORD 2 */
u8 len[17]; /* DWORD 3 */
u8 rsvd1[15]; /* DWORD 3 */
};
struct beiscsi_offload_params {
u32 dw[6];
};
#define OFFLD_PARAMS_ERL 0x00000003
#define OFFLD_PARAMS_DDE 0x00000004
#define OFFLD_PARAMS_HDE 0x00000008
#define OFFLD_PARAMS_IR2T 0x00000010
#define OFFLD_PARAMS_IMD 0x00000020
#define OFFLD_PARAMS_DATA_SEQ_INORDER 0x00000040
#define OFFLD_PARAMS_PDU_SEQ_INORDER 0x00000080
#define OFFLD_PARAMS_MAX_R2T 0x00FFFF00
/**
* Pseudo amap definition in which each bit of the actual structure is defined
* as a byte: used to calculate offset/shift/mask of each field
*/
struct amap_beiscsi_offload_params {
u8 max_burst_length[32];
u8 max_send_data_segment_length[32];
u8 first_burst_length[32];
u8 erl[2];
u8 dde[1];
u8 hde[1];
u8 ir2t[1];
u8 imd[1];
u8 data_seq_inorder[1];
u8 pdu_seq_inorder[1];
u8 max_r2t[16];
u8 pad[8];
u8 exp_statsn[32];
u8 max_recv_data_segment_length[32];
};
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 {
struct list_head link;
struct be_bus_address pa;
void *pbuffer;
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 buffer_len;
u16 index;
u16 cri;
u8 is_header;
u8 is_final;
u8 in_use;
};
#define BEISCSI_ASYNC_HDQ_SIZE(phba, ulp) \
(BEISCSI_GET_CID_COUNT((phba), (ulp)) * 2)
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 has list of async PDUs that are waiting to be processed.
* Buffers live in this list for a brief duration before they get
* processed and posted back to hardware.
* Note that we don't really need one cri_wait_queue per async_entry.
* We need one cri_wait_queue per CRI. Its easier to manage if this
* is tagged along with the async_entry.
*/
struct hd_async_entry {
struct cri_wait_queue {
unsigned short hdr_len;
unsigned int bytes_received;
unsigned int bytes_needed;
struct list_head list;
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;
/* handles posted to FW resides here */
struct hd_async_handle *header;
struct hd_async_handle *data;
};
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_buf_context {
struct be_bus_address pa_base;
void *va_base;
void *ring_base;
struct hd_async_handle *handle_base;
u32 buffer_size;
u16 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
};
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
/**
* hd_async_context is declared for each ULP supporting iSCSI function.
*/
struct hd_async_context {
struct hd_async_buf_context async_header;
struct hd_async_buf_context async_data;
u16 num_entries;
/**
* When unsol PDU is in, it needs to be chained till all the bytes are
* received and then processing is done. hd_async_entry is created
* based on the cid_count for each ULP. When unsol PDU comes in based
* on the conn_id it needs to be added to the correct async_entry wq.
* Below defined cid_to_async_cri_map is used to reterive the
* async_cri_map for a particular connection.
*
* This array is initialized after beiscsi_create_wrb_rings returns.
*
* - this method takes more memory space, fixed to 2K
* - any support for connections greater than this the array size needs
* to be incremented
*/
#define BE_GET_ASYNC_CRI_FROM_CID(cid) (pasync_ctx->cid_to_async_cri_map[cid])
unsigned short cid_to_async_cri_map[BE_MAX_SESSION];
/**
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 is a variable size array. Don`t add anything after this field!!
*/
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 *async_entry;
};
struct i_t_dpdu_cqe {
u32 dw[4];
} __packed;
/**
* Pseudo amap definition in which each bit of the actual structure is defined
* as a byte: used to calculate offset/shift/mask of each field
*/
struct amap_i_t_dpdu_cqe {
u8 db_addr_hi[32];
u8 db_addr_lo[32];
u8 code[6];
u8 cid[10];
u8 dpl[16];
u8 index[16];
u8 num_cons[10];
u8 rsvd0[4];
u8 final;
u8 valid;
} __packed;
struct amap_i_t_dpdu_cqe_v2 {
u8 db_addr_hi[32]; /* DWORD 0 */
u8 db_addr_lo[32]; /* DWORD 1 */
u8 code[6]; /* DWORD 2 */
u8 num_cons; /* DWORD 2*/
u8 rsvd0[8]; /* DWORD 2 */
u8 dpl[17]; /* DWORD 2 */
u8 index[16]; /* DWORD 3 */
u8 cid[13]; /* DWORD 3 */
u8 rsvd1; /* DWORD 3 */
u8 final; /* DWORD 3 */
u8 valid; /* DWORD 3 */
} __packed;
#define CQE_VALID_MASK 0x80000000
#define CQE_CODE_MASK 0x0000003F
#define CQE_CID_MASK 0x0000FFC0
#define EQE_VALID_MASK 0x00000001
#define EQE_MAJORCODE_MASK 0x0000000E
#define EQE_RESID_MASK 0xFFFF0000
struct be_eq_entry {
u32 dw[1];
} __packed;
/**
* Pseudo amap definition in which each bit of the actual structure is defined
* as a byte: used to calculate offset/shift/mask of each field
*/
struct amap_eq_entry {
u8 valid; /* DWORD 0 */
u8 major_code[3]; /* DWORD 0 */
u8 minor_code[12]; /* DWORD 0 */
u8 resource_id[16]; /* DWORD 0 */
} __packed;
struct cq_db {
u32 dw[1];
} __packed;
/**
* Pseudo amap definition in which each bit of the actual structure is defined
* as a byte: used to calculate offset/shift/mask of each field
*/
struct amap_cq_db {
u8 qid[10];
u8 event[1];
u8 rsvd0[5];
u8 num_popped[13];
u8 rearm[1];
u8 rsvd1[2];
} __packed;
void beiscsi_process_eq(struct beiscsi_hba *phba);
struct iscsi_wrb {
u32 dw[16];
} __packed;
#define WRB_TYPE_MASK 0xF0000000
#define SKH_WRB_TYPE_OFFSET 27
#define BE_WRB_TYPE_OFFSET 28
#define ADAPTER_SET_WRB_TYPE(pwrb, wrb_type, type_offset) \
(pwrb->dw[0] |= (wrb_type << type_offset))
/**
* Pseudo amap definition in which each bit of the actual structure is defined
* as a byte: used to calculate offset/shift/mask of each field
*/
struct amap_iscsi_wrb {
u8 lun[14]; /* DWORD 0 */
u8 lt; /* DWORD 0 */
u8 invld; /* DWORD 0 */
u8 wrb_idx[8]; /* DWORD 0 */
u8 dsp; /* DWORD 0 */
u8 dmsg; /* DWORD 0 */
u8 undr_run; /* DWORD 0 */
u8 over_run; /* DWORD 0 */
u8 type[4]; /* DWORD 0 */
u8 ptr2nextwrb[8]; /* DWORD 1 */
u8 r2t_exp_dtl[24]; /* DWORD 1 */
u8 sgl_icd_idx[12]; /* DWORD 2 */
u8 rsvd0[20]; /* DWORD 2 */
u8 exp_data_sn[32]; /* DWORD 3 */
u8 iscsi_bhs_addr_hi[32]; /* DWORD 4 */
u8 iscsi_bhs_addr_lo[32]; /* DWORD 5 */
u8 cmdsn_itt[32]; /* DWORD 6 */
u8 dif_ref_tag[32]; /* DWORD 7 */
u8 sge0_addr_hi[32]; /* DWORD 8 */
u8 sge0_addr_lo[32]; /* DWORD 9 */
u8 sge0_offset[22]; /* DWORD 10 */
u8 pbs; /* DWORD 10 */
u8 dif_mode[2]; /* DWORD 10 */
u8 rsvd1[6]; /* DWORD 10 */
u8 sge0_last; /* DWORD 10 */
u8 sge0_len[17]; /* DWORD 11 */
u8 dif_meta_tag[14]; /* DWORD 11 */
u8 sge0_in_ddr; /* DWORD 11 */
u8 sge1_addr_hi[32]; /* DWORD 12 */
u8 sge1_addr_lo[32]; /* DWORD 13 */
u8 sge1_r2t_offset[22]; /* DWORD 14 */
u8 rsvd2[9]; /* DWORD 14 */
u8 sge1_last; /* DWORD 14 */
u8 sge1_len[17]; /* DWORD 15 */
u8 ref_sgl_icd_idx[12]; /* DWORD 15 */
u8 rsvd3[2]; /* DWORD 15 */
u8 sge1_in_ddr; /* DWORD 15 */
} __packed;
struct amap_iscsi_wrb_v2 {
u8 r2t_exp_dtl[25]; /* DWORD 0 */
u8 rsvd0[2]; /* DWORD 0*/
u8 type[5]; /* DWORD 0 */
u8 ptr2nextwrb[8]; /* DWORD 1 */
u8 wrb_idx[8]; /* DWORD 1 */
u8 lun[16]; /* DWORD 1 */
u8 sgl_idx[16]; /* DWORD 2 */
u8 ref_sgl_icd_idx[16]; /* DWORD 2 */
u8 exp_data_sn[32]; /* DWORD 3 */
u8 iscsi_bhs_addr_hi[32]; /* DWORD 4 */
u8 iscsi_bhs_addr_lo[32]; /* DWORD 5 */
u8 cq_id[16]; /* DWORD 6 */
u8 rsvd1[16]; /* DWORD 6 */
u8 cmdsn_itt[32]; /* DWORD 7 */
u8 sge0_addr_hi[32]; /* DWORD 8 */
u8 sge0_addr_lo[32]; /* DWORD 9 */
u8 sge0_offset[24]; /* DWORD 10 */
u8 rsvd2[7]; /* DWORD 10 */
u8 sge0_last; /* DWORD 10 */
u8 sge0_len[17]; /* DWORD 11 */
u8 rsvd3[7]; /* DWORD 11 */
u8 diff_enbl; /* DWORD 11 */
u8 u_run; /* DWORD 11 */
u8 o_run; /* DWORD 11 */
u8 invld; /* DWORD 11 */
u8 dsp; /* DWORD 11 */
u8 dmsg; /* DWORD 11 */
u8 rsvd4; /* DWORD 11 */
u8 lt; /* DWORD 11 */
u8 sge1_addr_hi[32]; /* DWORD 12 */
u8 sge1_addr_lo[32]; /* DWORD 13 */
u8 sge1_r2t_offset[24]; /* DWORD 14 */
u8 rsvd5[7]; /* DWORD 14 */
u8 sge1_last; /* DWORD 14 */
u8 sge1_len[17]; /* DWORD 15 */
u8 rsvd6[15]; /* DWORD 15 */
} __packed;
struct wrb_handle *alloc_wrb_handle(struct beiscsi_hba *phba, unsigned int cid,
struct hwi_wrb_context **pcontext);
void
free_mgmt_sgl_handle(struct beiscsi_hba *phba, struct sgl_handle *psgl_handle);
void beiscsi_free_mgmt_task_handles(struct beiscsi_conn *beiscsi_conn,
struct iscsi_task *task);
void hwi_ring_cq_db(struct beiscsi_hba *phba,
unsigned int id, unsigned int num_processed,
unsigned char rearm);
unsigned int beiscsi_process_cq(struct be_eq_obj *pbe_eq, int budget);
void beiscsi_process_mcc_cq(struct beiscsi_hba *phba);
struct pdu_nop_out {
u32 dw[12];
};
/**
* Pseudo amap definition in which each bit of the actual structure is defined
* as a byte: used to calculate offset/shift/mask of each field
*/
struct amap_pdu_nop_out {
u8 opcode[6]; /* opcode 0x00 */
u8 i_bit; /* I Bit */
u8 x_bit; /* reserved; should be 0 */
u8 fp_bit_filler1[7];
u8 f_bit; /* always 1 */
u8 reserved1[16];
u8 ahs_length[8]; /* no AHS */
u8 data_len_hi[8];
u8 data_len_lo[16]; /* DataSegmentLength */
u8 lun[64];
u8 itt[32]; /* initiator id for ping or 0xffffffff */
u8 ttt[32]; /* target id for ping or 0xffffffff */
u8 cmd_sn[32];
u8 exp_stat_sn[32];
u8 reserved5[128];
};
#define PDUBASE_OPCODE_MASK 0x0000003F
#define PDUBASE_DATALENHI_MASK 0x0000FF00
#define PDUBASE_DATALENLO_MASK 0xFFFF0000
struct pdu_base {
u32 dw[16];
} __packed;
/**
* Pseudo amap definition in which each bit of the actual structure is defined
* as a byte: used to calculate offset/shift/mask of each field
*/
struct amap_pdu_base {
u8 opcode[6];
u8 i_bit; /* immediate bit */
u8 x_bit; /* reserved, always 0 */
u8 reserved1[24]; /* opcode-specific fields */
u8 ahs_length[8]; /* length units is 4 byte words */
u8 data_len_hi[8];
u8 data_len_lo[16]; /* DatasegmentLength */
u8 lun[64]; /* lun or opcode-specific fields */
u8 itt[32]; /* initiator task tag */
u8 reserved4[224];
};
struct iscsi_target_context_update_wrb {
u32 dw[16];
} __packed;
/**
* Pseudo amap definition in which each bit of the actual structure is defined
* as a byte: used to calculate offset/shift/mask of each field
*/
#define BE_TGT_CTX_UPDT_CMD 0x07
struct amap_iscsi_target_context_update_wrb {
u8 lun[14]; /* DWORD 0 */
u8 lt; /* DWORD 0 */
u8 invld; /* DWORD 0 */
u8 wrb_idx[8]; /* DWORD 0 */
u8 dsp; /* DWORD 0 */
u8 dmsg; /* DWORD 0 */
u8 undr_run; /* DWORD 0 */
u8 over_run; /* DWORD 0 */
u8 type[4]; /* DWORD 0 */
u8 ptr2nextwrb[8]; /* DWORD 1 */
u8 max_burst_length[19]; /* DWORD 1 */
u8 rsvd0[5]; /* DWORD 1 */
u8 rsvd1[15]; /* DWORD 2 */
u8 max_send_data_segment_length[17]; /* DWORD 2 */
u8 first_burst_length[14]; /* DWORD 3 */
u8 rsvd2[2]; /* DWORD 3 */
u8 tx_wrbindex_drv_msg[8]; /* DWORD 3 */
u8 rsvd3[5]; /* DWORD 3 */
u8 session_state[3]; /* DWORD 3 */
u8 rsvd4[16]; /* DWORD 4 */
u8 tx_jumbo; /* DWORD 4 */
u8 hde; /* DWORD 4 */
u8 dde; /* DWORD 4 */
u8 erl[2]; /* DWORD 4 */
u8 domain_id[5]; /* DWORD 4 */
u8 mode; /* DWORD 4 */
u8 imd; /* DWORD 4 */
u8 ir2t; /* DWORD 4 */
u8 notpredblq[2]; /* DWORD 4 */
u8 compltonack; /* DWORD 4 */
u8 stat_sn[32]; /* DWORD 5 */
u8 pad_buffer_addr_hi[32]; /* DWORD 6 */
u8 pad_buffer_addr_lo[32]; /* DWORD 7 */
u8 pad_addr_hi[32]; /* DWORD 8 */
u8 pad_addr_lo[32]; /* DWORD 9 */
u8 rsvd5[32]; /* DWORD 10 */
u8 rsvd6[32]; /* DWORD 11 */
u8 rsvd7[32]; /* DWORD 12 */
u8 rsvd8[32]; /* DWORD 13 */
u8 rsvd9[32]; /* DWORD 14 */
u8 rsvd10[32]; /* DWORD 15 */
} __packed;
#define BEISCSI_MAX_RECV_DATASEG_LEN (64 * 1024)
#define BEISCSI_MAX_CXNS 1
struct amap_iscsi_target_context_update_wrb_v2 {
u8 max_burst_length[24]; /* DWORD 0 */
u8 rsvd0[3]; /* DWORD 0 */
u8 type[5]; /* DWORD 0 */
u8 ptr2nextwrb[8]; /* DWORD 1 */
u8 wrb_idx[8]; /* DWORD 1 */
u8 rsvd1[16]; /* DWORD 1 */
u8 max_send_data_segment_length[24]; /* DWORD 2 */
u8 rsvd2[8]; /* DWORD 2 */
u8 first_burst_length[24]; /* DWORD 3 */
u8 rsvd3[8]; /* DOWRD 3 */
u8 max_r2t[16]; /* DWORD 4 */
u8 rsvd4; /* DWORD 4 */
u8 hde; /* DWORD 4 */
u8 dde; /* DWORD 4 */
u8 erl[2]; /* DWORD 4 */
u8 rsvd5[6]; /* DWORD 4 */
u8 imd; /* DWORD 4 */
u8 ir2t; /* DWORD 4 */
u8 rsvd6[3]; /* DWORD 4 */
u8 stat_sn[32]; /* DWORD 5 */
u8 rsvd7[32]; /* DWORD 6 */
u8 rsvd8[32]; /* DWORD 7 */
u8 max_recv_dataseg_len[24]; /* DWORD 8 */
u8 rsvd9[8]; /* DWORD 8 */
u8 rsvd10[32]; /* DWORD 9 */
u8 rsvd11[32]; /* DWORD 10 */
u8 max_cxns[16]; /* DWORD 11 */
u8 rsvd12[11]; /* DWORD 11*/
u8 invld; /* DWORD 11 */
u8 rsvd13;/* DWORD 11*/
u8 dmsg; /* DWORD 11 */
u8 data_seq_inorder; /* DWORD 11 */
u8 pdu_seq_inorder; /* DWORD 11 */
u8 rsvd14[32]; /*DWORD 12 */
u8 rsvd15[32]; /* DWORD 13 */
u8 rsvd16[32]; /* DWORD 14 */
u8 rsvd17[32]; /* DWORD 15 */
} __packed;
struct be_ring {
u32 pages; /* queue size in pages */
u32 id; /* queue id assigned by beklib */
u32 num; /* number of elements in queue */
u32 cidx; /* consumer index */
u32 pidx; /* producer index -- not used by most rings */
u32 item_size; /* size in bytes of one object */
u8 ulp_num; /* ULP to which CID binded */
u16 register_set;
u16 doorbell_format;
u32 doorbell_offset;
void *va; /* The virtual address of the ring. This
* should be last to allow 32 & 64 bit debugger
* extensions to work.
*/
};
struct hwi_controller {
struct hwi_wrb_context *wrb_context;
struct be_ring default_pdu_hdr[BEISCSI_ULP_COUNT];
struct be_ring default_pdu_data[BEISCSI_ULP_COUNT];
struct hwi_context_memory *phwi_ctxt;
};
enum hwh_type_enum {
HWH_TYPE_IO = 1,
HWH_TYPE_LOGOUT = 2,
HWH_TYPE_TMF = 3,
HWH_TYPE_NOP = 4,
HWH_TYPE_IO_RD = 5,
HWH_TYPE_LOGIN = 11,
HWH_TYPE_INVALID = 0xFFFFFFFF
};
struct wrb_handle {
unsigned short wrb_index;
struct iscsi_task *pio_handle;
struct iscsi_wrb *pwrb;
};
struct hwi_context_memory {
struct be_eq_obj be_eq[MAX_CPUS];
struct be_queue_info be_cq[MAX_CPUS - 1];
struct be_queue_info *be_wrbq;
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
/**
* Create array of ULP number for below entries as DEFQ
* will be created for both ULP if iSCSI Protocol is
* loaded on both ULP.
*/
struct be_queue_info be_def_hdrq[BEISCSI_ULP_COUNT];
struct be_queue_info be_def_dataq[BEISCSI_ULP_COUNT];
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[BEISCSI_ULP_COUNT];
};
void beiscsi_start_boot_work(struct beiscsi_hba *phba, unsigned int s_handle);
/* Logging related definitions */
#define BEISCSI_LOG_INIT 0x0001 /* Initialization events */
#define BEISCSI_LOG_MBOX 0x0002 /* Mailbox Events */
#define BEISCSI_LOG_MISC 0x0004 /* Miscllaneous Events */
#define BEISCSI_LOG_EH 0x0008 /* Error Handler */
#define BEISCSI_LOG_IO 0x0010 /* IO Code Path */
#define BEISCSI_LOG_CONFIG 0x0020 /* CONFIG Code Path */
#define BEISCSI_LOG_ISCSI 0x0040 /* SCSI/iSCSI Protocol related Logs */
#define __beiscsi_log(phba, level, fmt, arg...) \
shost_printk(level, phba->shost, fmt, __LINE__, ##arg)
#define beiscsi_log(phba, level, mask, fmt, arg...) \
do { \
uint32_t log_value = phba->attr_log_enable; \
if (((mask) & log_value) || (level[1] <= '3')) \
__beiscsi_log(phba, level, fmt, ##arg); \
} while (0);
#endif