OpenCloudOS-Kernel/drivers/scsi/isci/request.c

3532 lines
102 KiB
C
Raw Permalink Normal View History

/*
* This file is provided under a dual BSD/GPLv2 license. When using or
* redistributing this file, you may do so under either license.
*
* GPL LICENSE SUMMARY
*
* Copyright(c) 2008 - 2011 Intel Corporation. All rights reserved.
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of version 2 of the GNU General Public License as
* published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but
* WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
* The full GNU General Public License is included in this distribution
* in the file called LICENSE.GPL.
*
* BSD LICENSE
*
* Copyright(c) 2008 - 2011 Intel Corporation. All rights reserved.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
*
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in
* the documentation and/or other materials provided with the
* distribution.
* * Neither the name of Intel Corporation nor the names of its
* contributors may be used to endorse or promote products derived
* from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include <scsi/scsi_cmnd.h>
#include "isci.h"
#include "task.h"
#include "request.h"
#include "scu_completion_codes.h"
#include "scu_event_codes.h"
#include "sas.h"
#undef C
#define C(a) (#a)
const char *req_state_name(enum sci_base_request_states state)
{
static const char * const strings[] = REQUEST_STATES;
return strings[state];
}
#undef C
static struct scu_sgl_element_pair *to_sgl_element_pair(struct isci_request *ireq,
int idx)
{
if (idx == 0)
return &ireq->tc->sgl_pair_ab;
else if (idx == 1)
return &ireq->tc->sgl_pair_cd;
else if (idx < 0)
return NULL;
else
return &ireq->sg_table[idx - 2];
}
static dma_addr_t to_sgl_element_pair_dma(struct isci_host *ihost,
struct isci_request *ireq, u32 idx)
{
u32 offset;
if (idx == 0) {
offset = (void *) &ireq->tc->sgl_pair_ab -
(void *) &ihost->task_context_table[0];
return ihost->tc_dma + offset;
} else if (idx == 1) {
offset = (void *) &ireq->tc->sgl_pair_cd -
(void *) &ihost->task_context_table[0];
return ihost->tc_dma + offset;
}
return sci_io_request_get_dma_addr(ireq, &ireq->sg_table[idx - 2]);
}
static void init_sgl_element(struct scu_sgl_element *e, struct scatterlist *sg)
{
e->length = sg_dma_len(sg);
e->address_upper = upper_32_bits(sg_dma_address(sg));
e->address_lower = lower_32_bits(sg_dma_address(sg));
e->address_modifier = 0;
}
static void sci_request_build_sgl(struct isci_request *ireq)
{
struct isci_host *ihost = ireq->isci_host;
struct sas_task *task = isci_request_access_task(ireq);
struct scatterlist *sg = NULL;
dma_addr_t dma_addr;
u32 sg_idx = 0;
struct scu_sgl_element_pair *scu_sg = NULL;
struct scu_sgl_element_pair *prev_sg = NULL;
if (task->num_scatter > 0) {
sg = task->scatter;
while (sg) {
scu_sg = to_sgl_element_pair(ireq, sg_idx);
init_sgl_element(&scu_sg->A, sg);
sg = sg_next(sg);
if (sg) {
init_sgl_element(&scu_sg->B, sg);
sg = sg_next(sg);
} else
memset(&scu_sg->B, 0, sizeof(scu_sg->B));
if (prev_sg) {
dma_addr = to_sgl_element_pair_dma(ihost,
ireq,
sg_idx);
prev_sg->next_pair_upper =
upper_32_bits(dma_addr);
prev_sg->next_pair_lower =
lower_32_bits(dma_addr);
}
prev_sg = scu_sg;
sg_idx++;
}
} else { /* handle when no sg */
scu_sg = to_sgl_element_pair(ireq, sg_idx);
dma_addr = dma_map_single(&ihost->pdev->dev,
task->scatter,
task->total_xfer_len,
task->data_dir);
ireq->zero_scatter_daddr = dma_addr;
scu_sg->A.length = task->total_xfer_len;
scu_sg->A.address_upper = upper_32_bits(dma_addr);
scu_sg->A.address_lower = lower_32_bits(dma_addr);
}
if (scu_sg) {
scu_sg->next_pair_upper = 0;
scu_sg->next_pair_lower = 0;
}
}
static void sci_io_request_build_ssp_command_iu(struct isci_request *ireq)
{
struct ssp_cmd_iu *cmd_iu;
struct sas_task *task = isci_request_access_task(ireq);
cmd_iu = &ireq->ssp.cmd;
memcpy(cmd_iu->LUN, task->ssp_task.LUN, 8);
cmd_iu->add_cdb_len = 0;
cmd_iu->_r_a = 0;
cmd_iu->_r_b = 0;
cmd_iu->en_fburst = 0; /* unsupported */
cmd_iu->task_prio = task->ssp_task.task_prio;
cmd_iu->task_attr = task->ssp_task.task_attr;
cmd_iu->_r_c = 0;
sci_swab32_cpy(&cmd_iu->cdb, task->ssp_task.cmd->cmnd,
(task->ssp_task.cmd->cmd_len+3) / sizeof(u32));
}
static void sci_task_request_build_ssp_task_iu(struct isci_request *ireq)
{
struct ssp_task_iu *task_iu;
struct sas_task *task = isci_request_access_task(ireq);
struct isci_tmf *isci_tmf = isci_request_access_tmf(ireq);
task_iu = &ireq->ssp.tmf;
memset(task_iu, 0, sizeof(struct ssp_task_iu));
memcpy(task_iu->LUN, task->ssp_task.LUN, 8);
task_iu->task_func = isci_tmf->tmf_code;
task_iu->task_tag =
(test_bit(IREQ_TMF, &ireq->flags)) ?
isci_tmf->io_tag :
SCI_CONTROLLER_INVALID_IO_TAG;
}
/**
* This method is will fill in the SCU Task Context for any type of SSP request.
* @sci_req:
* @task_context:
*
*/
static void scu_ssp_request_construct_task_context(
struct isci_request *ireq,
struct scu_task_context *task_context)
{
dma_addr_t dma_addr;
struct isci_remote_device *idev;
struct isci_port *iport;
idev = ireq->target_device;
iport = idev->owning_port;
/* Fill in the TC with its required data */
task_context->abort = 0;
task_context->priority = 0;
task_context->initiator_request = 1;
task_context->connection_rate = idev->connection_rate;
task_context->protocol_engine_index = ISCI_PEG;
task_context->logical_port_index = iport->physical_port_index;
task_context->protocol_type = SCU_TASK_CONTEXT_PROTOCOL_SSP;
task_context->valid = SCU_TASK_CONTEXT_VALID;
task_context->context_type = SCU_TASK_CONTEXT_TYPE;
task_context->remote_node_index = idev->rnc.remote_node_index;
task_context->command_code = 0;
task_context->link_layer_control = 0;
task_context->do_not_dma_ssp_good_response = 1;
task_context->strict_ordering = 0;
task_context->control_frame = 0;
task_context->timeout_enable = 0;
task_context->block_guard_enable = 0;
task_context->address_modifier = 0;
/* task_context->type.ssp.tag = ireq->io_tag; */
task_context->task_phase = 0x01;
ireq->post_context = (SCU_CONTEXT_COMMAND_REQUEST_TYPE_POST_TC |
(ISCI_PEG << SCU_CONTEXT_COMMAND_PROTOCOL_ENGINE_GROUP_SHIFT) |
(iport->physical_port_index <<
SCU_CONTEXT_COMMAND_LOGICAL_PORT_SHIFT) |
ISCI_TAG_TCI(ireq->io_tag));
/*
* Copy the physical address for the command buffer to the
* SCU Task Context
*/
dma_addr = sci_io_request_get_dma_addr(ireq, &ireq->ssp.cmd);
task_context->command_iu_upper = upper_32_bits(dma_addr);
task_context->command_iu_lower = lower_32_bits(dma_addr);
/*
* Copy the physical address for the response buffer to the
* SCU Task Context
*/
dma_addr = sci_io_request_get_dma_addr(ireq, &ireq->ssp.rsp);
task_context->response_iu_upper = upper_32_bits(dma_addr);
task_context->response_iu_lower = lower_32_bits(dma_addr);
}
static u8 scu_bg_blk_size(struct scsi_device *sdp)
{
switch (sdp->sector_size) {
case 512:
return 0;
case 1024:
return 1;
case 4096:
return 3;
default:
return 0xff;
}
}
static u32 scu_dif_bytes(u32 len, u32 sector_size)
{
return (len >> ilog2(sector_size)) * 8;
}
static void scu_ssp_ireq_dif_insert(struct isci_request *ireq, u8 type, u8 op)
{
struct scu_task_context *tc = ireq->tc;
struct scsi_cmnd *scmd = ireq->ttype_ptr.io_task_ptr->uldd_task;
u8 blk_sz = scu_bg_blk_size(scmd->device);
tc->block_guard_enable = 1;
tc->blk_prot_en = 1;
tc->blk_sz = blk_sz;
/* DIF write insert */
tc->blk_prot_func = 0x2;
tc->transfer_length_bytes += scu_dif_bytes(tc->transfer_length_bytes,
scmd->device->sector_size);
/* always init to 0, used by hw */
tc->interm_crc_val = 0;
tc->init_crc_seed = 0;
tc->app_tag_verify = 0;
tc->app_tag_gen = 0;
tc->ref_tag_seed_verify = 0;
/* always init to same as bg_blk_sz */
tc->UD_bytes_immed_val = scmd->device->sector_size;
tc->reserved_DC_0 = 0;
/* always init to 8 */
tc->DIF_bytes_immed_val = 8;
tc->reserved_DC_1 = 0;
tc->bgc_blk_sz = scmd->device->sector_size;
tc->reserved_E0_0 = 0;
tc->app_tag_gen_mask = 0;
/** setup block guard control **/
tc->bgctl = 0;
/* DIF write insert */
tc->bgctl_f.op = 0x2;
tc->app_tag_verify_mask = 0;
/* must init to 0 for hw */
tc->blk_guard_err = 0;
tc->reserved_E8_0 = 0;
if ((type & SCSI_PROT_DIF_TYPE1) || (type & SCSI_PROT_DIF_TYPE2))
tc->ref_tag_seed_gen = scsi_get_lba(scmd) & 0xffffffff;
else if (type & SCSI_PROT_DIF_TYPE3)
tc->ref_tag_seed_gen = 0;
}
static void scu_ssp_ireq_dif_strip(struct isci_request *ireq, u8 type, u8 op)
{
struct scu_task_context *tc = ireq->tc;
struct scsi_cmnd *scmd = ireq->ttype_ptr.io_task_ptr->uldd_task;
u8 blk_sz = scu_bg_blk_size(scmd->device);
tc->block_guard_enable = 1;
tc->blk_prot_en = 1;
tc->blk_sz = blk_sz;
/* DIF read strip */
tc->blk_prot_func = 0x1;
tc->transfer_length_bytes += scu_dif_bytes(tc->transfer_length_bytes,
scmd->device->sector_size);
/* always init to 0, used by hw */
tc->interm_crc_val = 0;
tc->init_crc_seed = 0;
tc->app_tag_verify = 0;
tc->app_tag_gen = 0;
if ((type & SCSI_PROT_DIF_TYPE1) || (type & SCSI_PROT_DIF_TYPE2))
tc->ref_tag_seed_verify = scsi_get_lba(scmd) & 0xffffffff;
else if (type & SCSI_PROT_DIF_TYPE3)
tc->ref_tag_seed_verify = 0;
/* always init to same as bg_blk_sz */
tc->UD_bytes_immed_val = scmd->device->sector_size;
tc->reserved_DC_0 = 0;
/* always init to 8 */
tc->DIF_bytes_immed_val = 8;
tc->reserved_DC_1 = 0;
tc->bgc_blk_sz = scmd->device->sector_size;
tc->reserved_E0_0 = 0;
tc->app_tag_gen_mask = 0;
/** setup block guard control **/
tc->bgctl = 0;
/* DIF read strip */
tc->bgctl_f.crc_verify = 1;
tc->bgctl_f.op = 0x1;
if ((type & SCSI_PROT_DIF_TYPE1) || (type & SCSI_PROT_DIF_TYPE2)) {
tc->bgctl_f.ref_tag_chk = 1;
tc->bgctl_f.app_f_detect = 1;
} else if (type & SCSI_PROT_DIF_TYPE3)
tc->bgctl_f.app_ref_f_detect = 1;
tc->app_tag_verify_mask = 0;
/* must init to 0 for hw */
tc->blk_guard_err = 0;
tc->reserved_E8_0 = 0;
tc->ref_tag_seed_gen = 0;
}
/**
* This method is will fill in the SCU Task Context for a SSP IO request.
* @sci_req:
*
*/
static void scu_ssp_io_request_construct_task_context(struct isci_request *ireq,
enum dma_data_direction dir,
u32 len)
{
struct scu_task_context *task_context = ireq->tc;
struct sas_task *sas_task = ireq->ttype_ptr.io_task_ptr;
struct scsi_cmnd *scmd = sas_task->uldd_task;
u8 prot_type = scsi_get_prot_type(scmd);
u8 prot_op = scsi_get_prot_op(scmd);
scu_ssp_request_construct_task_context(ireq, task_context);
task_context->ssp_command_iu_length =
sizeof(struct ssp_cmd_iu) / sizeof(u32);
task_context->type.ssp.frame_type = SSP_COMMAND;
switch (dir) {
case DMA_FROM_DEVICE:
case DMA_NONE:
default:
task_context->task_type = SCU_TASK_TYPE_IOREAD;
break;
case DMA_TO_DEVICE:
task_context->task_type = SCU_TASK_TYPE_IOWRITE;
break;
}
task_context->transfer_length_bytes = len;
if (task_context->transfer_length_bytes > 0)
sci_request_build_sgl(ireq);
if (prot_type != SCSI_PROT_DIF_TYPE0) {
if (prot_op == SCSI_PROT_READ_STRIP)
scu_ssp_ireq_dif_strip(ireq, prot_type, prot_op);
else if (prot_op == SCSI_PROT_WRITE_INSERT)
scu_ssp_ireq_dif_insert(ireq, prot_type, prot_op);
}
}
/**
* This method will fill in the SCU Task Context for a SSP Task request. The
* following important settings are utilized: -# priority ==
* SCU_TASK_PRIORITY_HIGH. This ensures that the task request is issued
* ahead of other task destined for the same Remote Node. -# task_type ==
* SCU_TASK_TYPE_IOREAD. This simply indicates that a normal request type
* (i.e. non-raw frame) is being utilized to perform task management. -#
* control_frame == 1. This ensures that the proper endianess is set so
* that the bytes are transmitted in the right order for a task frame.
* @sci_req: This parameter specifies the task request object being
* constructed.
*
*/
static void scu_ssp_task_request_construct_task_context(struct isci_request *ireq)
{
struct scu_task_context *task_context = ireq->tc;
scu_ssp_request_construct_task_context(ireq, task_context);
task_context->control_frame = 1;
task_context->priority = SCU_TASK_PRIORITY_HIGH;
task_context->task_type = SCU_TASK_TYPE_RAW_FRAME;
task_context->transfer_length_bytes = 0;
task_context->type.ssp.frame_type = SSP_TASK;
task_context->ssp_command_iu_length =
sizeof(struct ssp_task_iu) / sizeof(u32);
}
/**
* This method is will fill in the SCU Task Context for any type of SATA
* request. This is called from the various SATA constructors.
* @sci_req: The general IO request object which is to be used in
* constructing the SCU task context.
* @task_context: The buffer pointer for the SCU task context which is being
* constructed.
*
* The general io request construction is complete. The buffer assignment for
* the command buffer is complete. none Revisit task context construction to
* determine what is common for SSP/SMP/STP task context structures.
*/
static void scu_sata_request_construct_task_context(
struct isci_request *ireq,
struct scu_task_context *task_context)
{
dma_addr_t dma_addr;
struct isci_remote_device *idev;
struct isci_port *iport;
idev = ireq->target_device;
iport = idev->owning_port;
/* Fill in the TC with its required data */
task_context->abort = 0;
task_context->priority = SCU_TASK_PRIORITY_NORMAL;
task_context->initiator_request = 1;
task_context->connection_rate = idev->connection_rate;
task_context->protocol_engine_index = ISCI_PEG;
task_context->logical_port_index = iport->physical_port_index;
task_context->protocol_type = SCU_TASK_CONTEXT_PROTOCOL_STP;
task_context->valid = SCU_TASK_CONTEXT_VALID;
task_context->context_type = SCU_TASK_CONTEXT_TYPE;
task_context->remote_node_index = idev->rnc.remote_node_index;
task_context->command_code = 0;
task_context->link_layer_control = 0;
task_context->do_not_dma_ssp_good_response = 1;
task_context->strict_ordering = 0;
task_context->control_frame = 0;
task_context->timeout_enable = 0;
task_context->block_guard_enable = 0;
task_context->address_modifier = 0;
task_context->task_phase = 0x01;
task_context->ssp_command_iu_length =
(sizeof(struct host_to_dev_fis) - sizeof(u32)) / sizeof(u32);
/* Set the first word of the H2D REG FIS */
task_context->type.words[0] = *(u32 *)&ireq->stp.cmd;
ireq->post_context = (SCU_CONTEXT_COMMAND_REQUEST_TYPE_POST_TC |
(ISCI_PEG << SCU_CONTEXT_COMMAND_PROTOCOL_ENGINE_GROUP_SHIFT) |
(iport->physical_port_index <<
SCU_CONTEXT_COMMAND_LOGICAL_PORT_SHIFT) |
ISCI_TAG_TCI(ireq->io_tag));
/*
* Copy the physical address for the command buffer to the SCU Task
* Context. We must offset the command buffer by 4 bytes because the
* first 4 bytes are transfered in the body of the TC.
*/
dma_addr = sci_io_request_get_dma_addr(ireq,
((char *) &ireq->stp.cmd) +
sizeof(u32));
task_context->command_iu_upper = upper_32_bits(dma_addr);
task_context->command_iu_lower = lower_32_bits(dma_addr);
/* SATA Requests do not have a response buffer */
task_context->response_iu_upper = 0;
task_context->response_iu_lower = 0;
}
static void scu_stp_raw_request_construct_task_context(struct isci_request *ireq)
{
struct scu_task_context *task_context = ireq->tc;
scu_sata_request_construct_task_context(ireq, task_context);
task_context->control_frame = 0;
task_context->priority = SCU_TASK_PRIORITY_NORMAL;
task_context->task_type = SCU_TASK_TYPE_SATA_RAW_FRAME;
task_context->type.stp.fis_type = FIS_REGH2D;
task_context->transfer_length_bytes = sizeof(struct host_to_dev_fis) - sizeof(u32);
}
static enum sci_status sci_stp_pio_request_construct(struct isci_request *ireq,
bool copy_rx_frame)
{
struct isci_stp_request *stp_req = &ireq->stp.req;
scu_stp_raw_request_construct_task_context(ireq);
stp_req->status = 0;
stp_req->sgl.offset = 0;
stp_req->sgl.set = SCU_SGL_ELEMENT_PAIR_A;
if (copy_rx_frame) {
sci_request_build_sgl(ireq);
stp_req->sgl.index = 0;
} else {
/* The user does not want the data copied to the SGL buffer location */
stp_req->sgl.index = -1;
}
return SCI_SUCCESS;
}
/**
*
* @sci_req: This parameter specifies the request to be constructed as an
* optimized request.
* @optimized_task_type: This parameter specifies whether the request is to be
* an UDMA request or a NCQ request. - A value of 0 indicates UDMA. - A
* value of 1 indicates NCQ.
*
* This method will perform request construction common to all types of STP
* requests that are optimized by the silicon (i.e. UDMA, NCQ). This method
* returns an indication as to whether the construction was successful.
*/
static void sci_stp_optimized_request_construct(struct isci_request *ireq,
u8 optimized_task_type,
u32 len,
enum dma_data_direction dir)
{
struct scu_task_context *task_context = ireq->tc;
/* Build the STP task context structure */
scu_sata_request_construct_task_context(ireq, task_context);
/* Copy over the SGL elements */
sci_request_build_sgl(ireq);
/* Copy over the number of bytes to be transfered */
task_context->transfer_length_bytes = len;
if (dir == DMA_TO_DEVICE) {
/*
* The difference between the DMA IN and DMA OUT request task type
* values are consistent with the difference between FPDMA READ
* and FPDMA WRITE values. Add the supplied task type parameter
* to this difference to set the task type properly for this
* DATA OUT (WRITE) case. */
task_context->task_type = optimized_task_type + (SCU_TASK_TYPE_DMA_OUT
- SCU_TASK_TYPE_DMA_IN);
} else {
/*
* For the DATA IN (READ) case, simply save the supplied
* optimized task type. */
task_context->task_type = optimized_task_type;
}
}
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
static void sci_atapi_construct(struct isci_request *ireq)
{
struct host_to_dev_fis *h2d_fis = &ireq->stp.cmd;
struct sas_task *task;
/* To simplify the implementation we take advantage of the
* silicon's partial acceleration of atapi protocol (dma data
* transfers), so we promote all commands to dma protocol. This
* breaks compatibility with ATA_HORKAGE_ATAPI_MOD16_DMA drives.
*/
h2d_fis->features |= ATAPI_PKT_DMA;
scu_stp_raw_request_construct_task_context(ireq);
task = isci_request_access_task(ireq);
if (task->data_dir == DMA_NONE)
task->total_xfer_len = 0;
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
/* clear the response so we can detect arrivial of an
* unsolicited h2d fis
*/
ireq->stp.rsp.fis_type = 0;
}
static enum sci_status
sci_io_request_construct_sata(struct isci_request *ireq,
u32 len,
enum dma_data_direction dir,
bool copy)
{
enum sci_status status = SCI_SUCCESS;
struct sas_task *task = isci_request_access_task(ireq);
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
struct domain_device *dev = ireq->target_device->domain_dev;
/* check for management protocols */
if (test_bit(IREQ_TMF, &ireq->flags)) {
struct isci_tmf *tmf = isci_request_access_tmf(ireq);
dev_err(&ireq->owning_controller->pdev->dev,
"%s: Request 0x%p received un-handled SAT "
"management protocol 0x%x.\n",
__func__, ireq, tmf->tmf_code);
return SCI_FAILURE;
}
if (!sas_protocol_ata(task->task_proto)) {
dev_err(&ireq->owning_controller->pdev->dev,
"%s: Non-ATA protocol in SATA path: 0x%x\n",
__func__,
task->task_proto);
return SCI_FAILURE;
}
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
/* ATAPI */
if (dev->sata_dev.class == ATA_DEV_ATAPI &&
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
task->ata_task.fis.command == ATA_CMD_PACKET) {
sci_atapi_construct(ireq);
return SCI_SUCCESS;
}
/* non data */
if (task->data_dir == DMA_NONE) {
scu_stp_raw_request_construct_task_context(ireq);
return SCI_SUCCESS;
}
/* NCQ */
if (task->ata_task.use_ncq) {
sci_stp_optimized_request_construct(ireq,
SCU_TASK_TYPE_FPDMAQ_READ,
len, dir);
return SCI_SUCCESS;
}
/* DMA */
if (task->ata_task.dma_xfer) {
sci_stp_optimized_request_construct(ireq,
SCU_TASK_TYPE_DMA_IN,
len, dir);
return SCI_SUCCESS;
} else /* PIO */
return sci_stp_pio_request_construct(ireq, copy);
return status;
}
static enum sci_status sci_io_request_construct_basic_ssp(struct isci_request *ireq)
{
struct sas_task *task = isci_request_access_task(ireq);
ireq->protocol = SAS_PROTOCOL_SSP;
scu_ssp_io_request_construct_task_context(ireq,
task->data_dir,
task->total_xfer_len);
sci_io_request_build_ssp_command_iu(ireq);
sci_change_state(&ireq->sm, SCI_REQ_CONSTRUCTED);
return SCI_SUCCESS;
}
enum sci_status sci_task_request_construct_ssp(
struct isci_request *ireq)
{
/* Construct the SSP Task SCU Task Context */
scu_ssp_task_request_construct_task_context(ireq);
/* Fill in the SSP Task IU */
sci_task_request_build_ssp_task_iu(ireq);
sci_change_state(&ireq->sm, SCI_REQ_CONSTRUCTED);
return SCI_SUCCESS;
}
static enum sci_status sci_io_request_construct_basic_sata(struct isci_request *ireq)
{
enum sci_status status;
bool copy = false;
struct sas_task *task = isci_request_access_task(ireq);
ireq->protocol = SAS_PROTOCOL_STP;
copy = (task->data_dir == DMA_NONE) ? false : true;
status = sci_io_request_construct_sata(ireq,
task->total_xfer_len,
task->data_dir,
copy);
if (status == SCI_SUCCESS)
sci_change_state(&ireq->sm, SCI_REQ_CONSTRUCTED);
return status;
}
/**
* sci_req_tx_bytes - bytes transferred when reply underruns request
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
* @ireq: request that was terminated early
*/
#define SCU_TASK_CONTEXT_SRAM 0x200000
static u32 sci_req_tx_bytes(struct isci_request *ireq)
{
struct isci_host *ihost = ireq->owning_controller;
u32 ret_val = 0;
if (readl(&ihost->smu_registers->address_modifier) == 0) {
void __iomem *scu_reg_base = ihost->scu_registers;
/* get the bytes of data from the Address == BAR1 + 20002Ch + (256*TCi) where
* BAR1 is the scu_registers
* 0x20002C = 0x200000 + 0x2c
* = start of task context SRAM + offset of (type.ssp.data_offset)
* TCi is the io_tag of struct sci_request
*/
ret_val = readl(scu_reg_base +
(SCU_TASK_CONTEXT_SRAM + offsetof(struct scu_task_context, type.ssp.data_offset)) +
((sizeof(struct scu_task_context)) * ISCI_TAG_TCI(ireq->io_tag)));
}
return ret_val;
}
enum sci_status sci_request_start(struct isci_request *ireq)
{
enum sci_base_request_states state;
struct scu_task_context *tc = ireq->tc;
struct isci_host *ihost = ireq->owning_controller;
state = ireq->sm.current_state_id;
if (state != SCI_REQ_CONSTRUCTED) {
dev_warn(&ihost->pdev->dev,
"%s: SCIC IO Request requested to start while in wrong "
"state %d\n", __func__, state);
return SCI_FAILURE_INVALID_STATE;
}
tc->task_index = ISCI_TAG_TCI(ireq->io_tag);
switch (tc->protocol_type) {
case SCU_TASK_CONTEXT_PROTOCOL_SMP:
case SCU_TASK_CONTEXT_PROTOCOL_SSP:
/* SSP/SMP Frame */
tc->type.ssp.tag = ireq->io_tag;
tc->type.ssp.target_port_transfer_tag = 0xFFFF;
break;
case SCU_TASK_CONTEXT_PROTOCOL_STP:
/* STP/SATA Frame
* tc->type.stp.ncq_tag = ireq->ncq_tag;
*/
break;
case SCU_TASK_CONTEXT_PROTOCOL_NONE:
/* / @todo When do we set no protocol type? */
break;
default:
/* This should never happen since we build the IO
* requests */
break;
}
/* Add to the post_context the io tag value */
ireq->post_context |= ISCI_TAG_TCI(ireq->io_tag);
/* Everything is good go ahead and change state */
sci_change_state(&ireq->sm, SCI_REQ_STARTED);
return SCI_SUCCESS;
}
enum sci_status
sci_io_request_terminate(struct isci_request *ireq)
{
enum sci_base_request_states state;
state = ireq->sm.current_state_id;
switch (state) {
case SCI_REQ_CONSTRUCTED:
/* Set to make sure no HW terminate posting is done: */
set_bit(IREQ_TC_ABORT_POSTED, &ireq->flags);
ireq->scu_status = SCU_TASK_DONE_TASK_ABORT;
ireq->sci_status = SCI_FAILURE_IO_TERMINATED;
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
return SCI_SUCCESS;
case SCI_REQ_STARTED:
case SCI_REQ_TASK_WAIT_TC_COMP:
case SCI_REQ_SMP_WAIT_RESP:
case SCI_REQ_SMP_WAIT_TC_COMP:
case SCI_REQ_STP_UDMA_WAIT_TC_COMP:
case SCI_REQ_STP_UDMA_WAIT_D2H:
case SCI_REQ_STP_NON_DATA_WAIT_H2D:
case SCI_REQ_STP_NON_DATA_WAIT_D2H:
case SCI_REQ_STP_PIO_WAIT_H2D:
case SCI_REQ_STP_PIO_WAIT_FRAME:
case SCI_REQ_STP_PIO_DATA_IN:
case SCI_REQ_STP_PIO_DATA_OUT:
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
case SCI_REQ_ATAPI_WAIT_H2D:
case SCI_REQ_ATAPI_WAIT_PIO_SETUP:
case SCI_REQ_ATAPI_WAIT_D2H:
case SCI_REQ_ATAPI_WAIT_TC_COMP:
/* Fall through and change state to ABORTING... */
case SCI_REQ_TASK_WAIT_TC_RESP:
/* The task frame was already confirmed to have been
* sent by the SCU HW. Since the state machine is
* now only waiting for the task response itself,
* abort the request and complete it immediately
* and don't wait for the task response.
*/
sci_change_state(&ireq->sm, SCI_REQ_ABORTING);
/* Fall through - and handle like ABORTING... */
case SCI_REQ_ABORTING:
if (!isci_remote_device_is_safe_to_abort(ireq->target_device))
set_bit(IREQ_PENDING_ABORT, &ireq->flags);
else
clear_bit(IREQ_PENDING_ABORT, &ireq->flags);
/* If the request is only waiting on the remote device
* suspension, return SUCCESS so the caller will wait too.
*/
return SCI_SUCCESS;
case SCI_REQ_COMPLETED:
default:
dev_warn(&ireq->owning_controller->pdev->dev,
"%s: SCIC IO Request requested to abort while in wrong "
"state %d\n", __func__, ireq->sm.current_state_id);
break;
}
return SCI_FAILURE_INVALID_STATE;
}
enum sci_status sci_request_complete(struct isci_request *ireq)
{
enum sci_base_request_states state;
struct isci_host *ihost = ireq->owning_controller;
state = ireq->sm.current_state_id;
if (WARN_ONCE(state != SCI_REQ_COMPLETED,
"isci: request completion from wrong state (%s)\n",
req_state_name(state)))
return SCI_FAILURE_INVALID_STATE;
if (ireq->saved_rx_frame_index != SCU_INVALID_FRAME_INDEX)
sci_controller_release_frame(ihost,
ireq->saved_rx_frame_index);
/* XXX can we just stop the machine and remove the 'final' state? */
sci_change_state(&ireq->sm, SCI_REQ_FINAL);
return SCI_SUCCESS;
}
enum sci_status sci_io_request_event_handler(struct isci_request *ireq,
u32 event_code)
{
enum sci_base_request_states state;
struct isci_host *ihost = ireq->owning_controller;
state = ireq->sm.current_state_id;
if (state != SCI_REQ_STP_PIO_DATA_IN) {
dev_warn(&ihost->pdev->dev, "%s: (%x) in wrong state %s\n",
__func__, event_code, req_state_name(state));
return SCI_FAILURE_INVALID_STATE;
}
switch (scu_get_event_specifier(event_code)) {
case SCU_TASK_DONE_CRC_ERR << SCU_EVENT_SPECIFIC_CODE_SHIFT:
/* We are waiting for data and the SCU has R_ERR the data frame.
* Go back to waiting for the D2H Register FIS
*/
sci_change_state(&ireq->sm, SCI_REQ_STP_PIO_WAIT_FRAME);
return SCI_SUCCESS;
default:
dev_err(&ihost->pdev->dev,
"%s: pio request unexpected event %#x\n",
__func__, event_code);
/* TODO Should we fail the PIO request when we get an
* unexpected event?
*/
return SCI_FAILURE;
}
}
/*
* This function copies response data for requests returning response data
* instead of sense data.
* @sci_req: This parameter specifies the request object for which to copy
* the response data.
*/
static void sci_io_request_copy_response(struct isci_request *ireq)
{
void *resp_buf;
u32 len;
struct ssp_response_iu *ssp_response;
struct isci_tmf *isci_tmf = isci_request_access_tmf(ireq);
ssp_response = &ireq->ssp.rsp;
resp_buf = &isci_tmf->resp.resp_iu;
len = min_t(u32,
SSP_RESP_IU_MAX_SIZE,
be32_to_cpu(ssp_response->response_data_len));
memcpy(resp_buf, ssp_response->resp_data, len);
}
static enum sci_status
request_started_state_tc_event(struct isci_request *ireq,
u32 completion_code)
{
struct ssp_response_iu *resp_iu;
u8 datapres;
/* TODO: Any SDMA return code of other than 0 is bad decode 0x003C0000
* to determine SDMA status
*/
switch (SCU_GET_COMPLETION_TL_STATUS(completion_code)) {
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_GOOD):
ireq->scu_status = SCU_TASK_DONE_GOOD;
ireq->sci_status = SCI_SUCCESS;
break;
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_EARLY_RESP): {
/* There are times when the SCU hardware will return an early
* response because the io request specified more data than is
* returned by the target device (mode pages, inquiry data,
* etc.). We must check the response stats to see if this is
* truly a failed request or a good request that just got
* completed early.
*/
struct ssp_response_iu *resp = &ireq->ssp.rsp;
ssize_t word_cnt = SSP_RESP_IU_MAX_SIZE / sizeof(u32);
sci_swab32_cpy(&ireq->ssp.rsp,
&ireq->ssp.rsp,
word_cnt);
if (resp->status == 0) {
ireq->scu_status = SCU_TASK_DONE_GOOD;
ireq->sci_status = SCI_SUCCESS_IO_DONE_EARLY;
} else {
ireq->scu_status = SCU_TASK_DONE_CHECK_RESPONSE;
ireq->sci_status = SCI_FAILURE_IO_RESPONSE_VALID;
}
break;
}
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_CHECK_RESPONSE): {
ssize_t word_cnt = SSP_RESP_IU_MAX_SIZE / sizeof(u32);
sci_swab32_cpy(&ireq->ssp.rsp,
&ireq->ssp.rsp,
word_cnt);
ireq->scu_status = SCU_TASK_DONE_CHECK_RESPONSE;
ireq->sci_status = SCI_FAILURE_IO_RESPONSE_VALID;
break;
}
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_RESP_LEN_ERR):
/* TODO With TASK_DONE_RESP_LEN_ERR is the response frame
* guaranteed to be received before this completion status is
* posted?
*/
resp_iu = &ireq->ssp.rsp;
datapres = resp_iu->datapres;
if (datapres == 1 || datapres == 2) {
ireq->scu_status = SCU_TASK_DONE_CHECK_RESPONSE;
ireq->sci_status = SCI_FAILURE_IO_RESPONSE_VALID;
} else {
ireq->scu_status = SCU_TASK_DONE_GOOD;
ireq->sci_status = SCI_SUCCESS;
}
break;
/* only stp device gets suspended. */
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_ACK_NAK_TO):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_LL_PERR):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_NAK_ERR):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_DATA_LEN_ERR):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_LL_ABORT_ERR):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_XR_WD_LEN):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_MAX_PLD_ERR):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_UNEXP_RESP):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_UNEXP_SDBFIS):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_REG_ERR):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_SDB_ERR):
if (ireq->protocol == SAS_PROTOCOL_STP) {
ireq->scu_status = SCU_GET_COMPLETION_TL_STATUS(completion_code) >>
SCU_COMPLETION_TL_STATUS_SHIFT;
ireq->sci_status = SCI_FAILURE_REMOTE_DEVICE_RESET_REQUIRED;
} else {
ireq->scu_status = SCU_GET_COMPLETION_TL_STATUS(completion_code) >>
SCU_COMPLETION_TL_STATUS_SHIFT;
ireq->sci_status = SCI_FAILURE_CONTROLLER_SPECIFIC_IO_ERR;
}
break;
/* both stp/ssp device gets suspended */
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_LF_ERR):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_OPEN_REJECT_WRONG_DESTINATION):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_OPEN_REJECT_RESERVED_ABANDON_1):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_OPEN_REJECT_RESERVED_ABANDON_2):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_OPEN_REJECT_RESERVED_ABANDON_3):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_OPEN_REJECT_BAD_DESTINATION):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_OPEN_REJECT_ZONE_VIOLATION):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_OPEN_REJECT_STP_RESOURCES_BUSY):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_OPEN_REJECT_PROTOCOL_NOT_SUPPORTED):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_OPEN_REJECT_CONNECTION_RATE_NOT_SUPPORTED):
ireq->scu_status = SCU_GET_COMPLETION_TL_STATUS(completion_code) >>
SCU_COMPLETION_TL_STATUS_SHIFT;
ireq->sci_status = SCI_FAILURE_REMOTE_DEVICE_RESET_REQUIRED;
break;
/* neither ssp nor stp gets suspended. */
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_NAK_CMD_ERR):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_UNEXP_XR):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_XR_IU_LEN_ERR):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_SDMA_ERR):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_OFFSET_ERR):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_EXCESS_DATA):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_SMP_RESP_TO_ERR):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_SMP_UFI_ERR):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_SMP_FRM_TYPE_ERR):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_SMP_LL_RX_ERR):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_UNEXP_DATA):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_OPEN_FAIL):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_VIIT_ENTRY_NV):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_IIT_ENTRY_NV):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_RNCNV_OUTBOUND):
default:
ireq->scu_status = SCU_GET_COMPLETION_TL_STATUS(completion_code) >>
SCU_COMPLETION_TL_STATUS_SHIFT;
ireq->sci_status = SCI_FAILURE_CONTROLLER_SPECIFIC_IO_ERR;
break;
}
/*
* TODO: This is probably wrong for ACK/NAK timeout conditions
*/
/* In all cases we will treat this as the completion of the IO req. */
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
return SCI_SUCCESS;
}
static enum sci_status
request_aborting_state_tc_event(struct isci_request *ireq,
u32 completion_code)
{
switch (SCU_GET_COMPLETION_TL_STATUS(completion_code)) {
case (SCU_TASK_DONE_GOOD << SCU_COMPLETION_TL_STATUS_SHIFT):
case (SCU_TASK_DONE_TASK_ABORT << SCU_COMPLETION_TL_STATUS_SHIFT):
ireq->scu_status = SCU_TASK_DONE_TASK_ABORT;
ireq->sci_status = SCI_FAILURE_IO_TERMINATED;
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
break;
default:
/* Unless we get some strange error wait for the task abort to complete
* TODO: Should there be a state change for this completion?
*/
break;
}
return SCI_SUCCESS;
}
static enum sci_status ssp_task_request_await_tc_event(struct isci_request *ireq,
u32 completion_code)
{
switch (SCU_GET_COMPLETION_TL_STATUS(completion_code)) {
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_GOOD):
ireq->scu_status = SCU_TASK_DONE_GOOD;
ireq->sci_status = SCI_SUCCESS;
sci_change_state(&ireq->sm, SCI_REQ_TASK_WAIT_TC_RESP);
break;
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_ACK_NAK_TO):
/* Currently, the decision is to simply allow the task request
* to timeout if the task IU wasn't received successfully.
* There is a potential for receiving multiple task responses if
* we decide to send the task IU again.
*/
dev_warn(&ireq->owning_controller->pdev->dev,
"%s: TaskRequest:0x%p CompletionCode:%x - "
"ACK/NAK timeout\n", __func__, ireq,
completion_code);
sci_change_state(&ireq->sm, SCI_REQ_TASK_WAIT_TC_RESP);
break;
default:
/*
* All other completion status cause the IO to be complete.
* If a NAK was received, then it is up to the user to retry
* the request.
*/
ireq->scu_status = SCU_NORMALIZE_COMPLETION_STATUS(completion_code);
ireq->sci_status = SCI_FAILURE_CONTROLLER_SPECIFIC_IO_ERR;
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
break;
}
return SCI_SUCCESS;
}
static enum sci_status
smp_request_await_response_tc_event(struct isci_request *ireq,
u32 completion_code)
{
switch (SCU_GET_COMPLETION_TL_STATUS(completion_code)) {
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_GOOD):
/* In the AWAIT RESPONSE state, any TC completion is
* unexpected. but if the TC has success status, we
* complete the IO anyway.
*/
ireq->scu_status = SCU_TASK_DONE_GOOD;
ireq->sci_status = SCI_SUCCESS;
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
break;
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_SMP_RESP_TO_ERR):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_SMP_UFI_ERR):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_SMP_FRM_TYPE_ERR):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_SMP_LL_RX_ERR):
/* These status has been seen in a specific LSI
* expander, which sometimes is not able to send smp
* response within 2 ms. This causes our hardware break
* the connection and set TC completion with one of
* these SMP_XXX_XX_ERR status. For these type of error,
* we ask ihost user to retry the request.
*/
ireq->scu_status = SCU_TASK_DONE_SMP_RESP_TO_ERR;
ireq->sci_status = SCI_FAILURE_RETRY_REQUIRED;
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
break;
default:
/* All other completion status cause the IO to be complete. If a NAK
* was received, then it is up to the user to retry the request
*/
ireq->scu_status = SCU_NORMALIZE_COMPLETION_STATUS(completion_code);
ireq->sci_status = SCI_FAILURE_CONTROLLER_SPECIFIC_IO_ERR;
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
break;
}
return SCI_SUCCESS;
}
static enum sci_status
smp_request_await_tc_event(struct isci_request *ireq,
u32 completion_code)
{
switch (SCU_GET_COMPLETION_TL_STATUS(completion_code)) {
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_GOOD):
ireq->scu_status = SCU_TASK_DONE_GOOD;
ireq->sci_status = SCI_SUCCESS;
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
break;
default:
/* All other completion status cause the IO to be
* complete. If a NAK was received, then it is up to
* the user to retry the request.
*/
ireq->scu_status = SCU_NORMALIZE_COMPLETION_STATUS(completion_code);
ireq->sci_status = SCI_FAILURE_CONTROLLER_SPECIFIC_IO_ERR;
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
break;
}
return SCI_SUCCESS;
}
static struct scu_sgl_element *pio_sgl_next(struct isci_stp_request *stp_req)
{
struct scu_sgl_element *sgl;
struct scu_sgl_element_pair *sgl_pair;
struct isci_request *ireq = to_ireq(stp_req);
struct isci_stp_pio_sgl *pio_sgl = &stp_req->sgl;
sgl_pair = to_sgl_element_pair(ireq, pio_sgl->index);
if (!sgl_pair)
sgl = NULL;
else if (pio_sgl->set == SCU_SGL_ELEMENT_PAIR_A) {
if (sgl_pair->B.address_lower == 0 &&
sgl_pair->B.address_upper == 0) {
sgl = NULL;
} else {
pio_sgl->set = SCU_SGL_ELEMENT_PAIR_B;
sgl = &sgl_pair->B;
}
} else {
if (sgl_pair->next_pair_lower == 0 &&
sgl_pair->next_pair_upper == 0) {
sgl = NULL;
} else {
pio_sgl->index++;
pio_sgl->set = SCU_SGL_ELEMENT_PAIR_A;
sgl_pair = to_sgl_element_pair(ireq, pio_sgl->index);
sgl = &sgl_pair->A;
}
}
return sgl;
}
static enum sci_status
stp_request_non_data_await_h2d_tc_event(struct isci_request *ireq,
u32 completion_code)
{
switch (SCU_GET_COMPLETION_TL_STATUS(completion_code)) {
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_GOOD):
ireq->scu_status = SCU_TASK_DONE_GOOD;
ireq->sci_status = SCI_SUCCESS;
sci_change_state(&ireq->sm, SCI_REQ_STP_NON_DATA_WAIT_D2H);
break;
default:
/* All other completion status cause the IO to be
* complete. If a NAK was received, then it is up to
* the user to retry the request.
*/
ireq->scu_status = SCU_NORMALIZE_COMPLETION_STATUS(completion_code);
ireq->sci_status = SCI_FAILURE_CONTROLLER_SPECIFIC_IO_ERR;
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
break;
}
return SCI_SUCCESS;
}
#define SCU_MAX_FRAME_BUFFER_SIZE 0x400 /* 1K is the maximum SCU frame data payload */
/* transmit DATA_FIS from (current sgl + offset) for input
* parameter length. current sgl and offset is alreay stored in the IO request
*/
static enum sci_status sci_stp_request_pio_data_out_trasmit_data_frame(
struct isci_request *ireq,
u32 length)
{
struct isci_stp_request *stp_req = &ireq->stp.req;
struct scu_task_context *task_context = ireq->tc;
struct scu_sgl_element_pair *sgl_pair;
struct scu_sgl_element *current_sgl;
/* Recycle the TC and reconstruct it for sending out DATA FIS containing
* for the data from current_sgl+offset for the input length
*/
sgl_pair = to_sgl_element_pair(ireq, stp_req->sgl.index);
if (stp_req->sgl.set == SCU_SGL_ELEMENT_PAIR_A)
current_sgl = &sgl_pair->A;
else
current_sgl = &sgl_pair->B;
/* update the TC */
task_context->command_iu_upper = current_sgl->address_upper;
task_context->command_iu_lower = current_sgl->address_lower;
task_context->transfer_length_bytes = length;
task_context->type.stp.fis_type = FIS_DATA;
/* send the new TC out. */
return sci_controller_continue_io(ireq);
}
static enum sci_status sci_stp_request_pio_data_out_transmit_data(struct isci_request *ireq)
{
struct isci_stp_request *stp_req = &ireq->stp.req;
struct scu_sgl_element_pair *sgl_pair;
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
enum sci_status status = SCI_SUCCESS;
struct scu_sgl_element *sgl;
u32 offset;
u32 len = 0;
offset = stp_req->sgl.offset;
sgl_pair = to_sgl_element_pair(ireq, stp_req->sgl.index);
if (WARN_ONCE(!sgl_pair, "%s: null sgl element", __func__))
return SCI_FAILURE;
if (stp_req->sgl.set == SCU_SGL_ELEMENT_PAIR_A) {
sgl = &sgl_pair->A;
len = sgl_pair->A.length - offset;
} else {
sgl = &sgl_pair->B;
len = sgl_pair->B.length - offset;
}
if (stp_req->pio_len == 0)
return SCI_SUCCESS;
if (stp_req->pio_len >= len) {
status = sci_stp_request_pio_data_out_trasmit_data_frame(ireq, len);
if (status != SCI_SUCCESS)
return status;
stp_req->pio_len -= len;
/* update the current sgl, offset and save for future */
sgl = pio_sgl_next(stp_req);
offset = 0;
} else if (stp_req->pio_len < len) {
sci_stp_request_pio_data_out_trasmit_data_frame(ireq, stp_req->pio_len);
/* Sgl offset will be adjusted and saved for future */
offset += stp_req->pio_len;
sgl->address_lower += stp_req->pio_len;
stp_req->pio_len = 0;
}
stp_req->sgl.offset = offset;
return status;
}
/**
*
* @stp_request: The request that is used for the SGL processing.
* @data_buffer: The buffer of data to be copied.
* @length: The length of the data transfer.
*
* Copy the data from the buffer for the length specified to the IO request SGL
* specified data region. enum sci_status
*/
static enum sci_status
sci_stp_request_pio_data_in_copy_data_buffer(struct isci_stp_request *stp_req,
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
u8 *data_buf, u32 len)
{
struct isci_request *ireq;
u8 *src_addr;
int copy_len;
struct sas_task *task;
struct scatterlist *sg;
void *kaddr;
int total_len = len;
ireq = to_ireq(stp_req);
task = isci_request_access_task(ireq);
src_addr = data_buf;
if (task->num_scatter > 0) {
sg = task->scatter;
while (total_len > 0) {
struct page *page = sg_page(sg);
copy_len = min_t(int, total_len, sg_dma_len(sg));
kaddr = kmap_atomic(page);
memcpy(kaddr + sg->offset, src_addr, copy_len);
kunmap_atomic(kaddr);
total_len -= copy_len;
src_addr += copy_len;
sg = sg_next(sg);
}
} else {
BUG_ON(task->total_xfer_len < total_len);
memcpy(task->scatter, src_addr, total_len);
}
return SCI_SUCCESS;
}
/**
*
* @sci_req: The PIO DATA IN request that is to receive the data.
* @data_buffer: The buffer to copy from.
*
* Copy the data buffer to the io request data region. enum sci_status
*/
static enum sci_status sci_stp_request_pio_data_in_copy_data(
struct isci_stp_request *stp_req,
u8 *data_buffer)
{
enum sci_status status;
/*
* If there is less than 1K remaining in the transfer request
* copy just the data for the transfer */
if (stp_req->pio_len < SCU_MAX_FRAME_BUFFER_SIZE) {
status = sci_stp_request_pio_data_in_copy_data_buffer(
stp_req, data_buffer, stp_req->pio_len);
if (status == SCI_SUCCESS)
stp_req->pio_len = 0;
} else {
/* We are transfering the whole frame so copy */
status = sci_stp_request_pio_data_in_copy_data_buffer(
stp_req, data_buffer, SCU_MAX_FRAME_BUFFER_SIZE);
if (status == SCI_SUCCESS)
stp_req->pio_len -= SCU_MAX_FRAME_BUFFER_SIZE;
}
return status;
}
static enum sci_status
stp_request_pio_await_h2d_completion_tc_event(struct isci_request *ireq,
u32 completion_code)
{
enum sci_status status = SCI_SUCCESS;
switch (SCU_GET_COMPLETION_TL_STATUS(completion_code)) {
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_GOOD):
ireq->scu_status = SCU_TASK_DONE_GOOD;
ireq->sci_status = SCI_SUCCESS;
sci_change_state(&ireq->sm, SCI_REQ_STP_PIO_WAIT_FRAME);
break;
default:
/* All other completion status cause the IO to be
* complete. If a NAK was received, then it is up to
* the user to retry the request.
*/
ireq->scu_status = SCU_NORMALIZE_COMPLETION_STATUS(completion_code);
ireq->sci_status = SCI_FAILURE_CONTROLLER_SPECIFIC_IO_ERR;
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
break;
}
return status;
}
static enum sci_status
pio_data_out_tx_done_tc_event(struct isci_request *ireq,
u32 completion_code)
{
enum sci_status status = SCI_SUCCESS;
bool all_frames_transferred = false;
struct isci_stp_request *stp_req = &ireq->stp.req;
switch (SCU_GET_COMPLETION_TL_STATUS(completion_code)) {
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_GOOD):
/* Transmit data */
if (stp_req->pio_len != 0) {
status = sci_stp_request_pio_data_out_transmit_data(ireq);
if (status == SCI_SUCCESS) {
if (stp_req->pio_len == 0)
all_frames_transferred = true;
}
} else if (stp_req->pio_len == 0) {
/*
* this will happen if the all data is written at the
* first time after the pio setup fis is received
*/
all_frames_transferred = true;
}
/* all data transferred. */
if (all_frames_transferred) {
/*
* Change the state to SCI_REQ_STP_PIO_DATA_IN
* and wait for PIO_SETUP fis / or D2H REg fis. */
sci_change_state(&ireq->sm, SCI_REQ_STP_PIO_WAIT_FRAME);
}
break;
default:
/*
* All other completion status cause the IO to be complete.
* If a NAK was received, then it is up to the user to retry
* the request.
*/
ireq->scu_status = SCU_NORMALIZE_COMPLETION_STATUS(completion_code);
ireq->sci_status = SCI_FAILURE_CONTROLLER_SPECIFIC_IO_ERR;
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
break;
}
return status;
}
static enum sci_status sci_stp_request_udma_general_frame_handler(struct isci_request *ireq,
u32 frame_index)
{
struct isci_host *ihost = ireq->owning_controller;
struct dev_to_host_fis *frame_header;
enum sci_status status;
u32 *frame_buffer;
status = sci_unsolicited_frame_control_get_header(&ihost->uf_control,
frame_index,
(void **)&frame_header);
if ((status == SCI_SUCCESS) &&
(frame_header->fis_type == FIS_REGD2H)) {
sci_unsolicited_frame_control_get_buffer(&ihost->uf_control,
frame_index,
(void **)&frame_buffer);
sci_controller_copy_sata_response(&ireq->stp.rsp,
frame_header,
frame_buffer);
}
sci_controller_release_frame(ihost, frame_index);
return status;
}
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
static enum sci_status process_unsolicited_fis(struct isci_request *ireq,
u32 frame_index)
{
struct isci_host *ihost = ireq->owning_controller;
enum sci_status status;
struct dev_to_host_fis *frame_header;
u32 *frame_buffer;
status = sci_unsolicited_frame_control_get_header(&ihost->uf_control,
frame_index,
(void **)&frame_header);
if (status != SCI_SUCCESS)
return status;
if (frame_header->fis_type != FIS_REGD2H) {
dev_err(&ireq->isci_host->pdev->dev,
"%s ERROR: invalid fis type 0x%X\n",
__func__, frame_header->fis_type);
return SCI_FAILURE;
}
sci_unsolicited_frame_control_get_buffer(&ihost->uf_control,
frame_index,
(void **)&frame_buffer);
sci_controller_copy_sata_response(&ireq->stp.rsp,
(u32 *)frame_header,
frame_buffer);
/* Frame has been decoded return it to the controller */
sci_controller_release_frame(ihost, frame_index);
return status;
}
static enum sci_status atapi_d2h_reg_frame_handler(struct isci_request *ireq,
u32 frame_index)
{
struct sas_task *task = isci_request_access_task(ireq);
enum sci_status status;
status = process_unsolicited_fis(ireq, frame_index);
if (status == SCI_SUCCESS) {
if (ireq->stp.rsp.status & ATA_ERR)
status = SCI_FAILURE_IO_RESPONSE_VALID;
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
} else {
status = SCI_FAILURE_IO_RESPONSE_VALID;
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
}
if (status != SCI_SUCCESS) {
ireq->scu_status = SCU_TASK_DONE_CHECK_RESPONSE;
ireq->sci_status = status;
} else {
ireq->scu_status = SCU_TASK_DONE_GOOD;
ireq->sci_status = SCI_SUCCESS;
}
/* the d2h ufi is the end of non-data commands */
if (task->data_dir == DMA_NONE)
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
return status;
}
static void scu_atapi_reconstruct_raw_frame_task_context(struct isci_request *ireq)
{
struct ata_device *dev = sas_to_ata_dev(ireq->target_device->domain_dev);
void *atapi_cdb = ireq->ttype_ptr.io_task_ptr->ata_task.atapi_packet;
struct scu_task_context *task_context = ireq->tc;
/* fill in the SCU Task Context for a DATA fis containing CDB in Raw Frame
* type. The TC for previous Packet fis was already there, we only need to
* change the H2D fis content.
*/
memset(&ireq->stp.cmd, 0, sizeof(struct host_to_dev_fis));
memcpy(((u8 *)&ireq->stp.cmd + sizeof(u32)), atapi_cdb, ATAPI_CDB_LEN);
memset(&(task_context->type.stp), 0, sizeof(struct stp_task_context));
task_context->type.stp.fis_type = FIS_DATA;
task_context->transfer_length_bytes = dev->cdb_len;
}
static void scu_atapi_construct_task_context(struct isci_request *ireq)
{
struct ata_device *dev = sas_to_ata_dev(ireq->target_device->domain_dev);
struct sas_task *task = isci_request_access_task(ireq);
struct scu_task_context *task_context = ireq->tc;
int cdb_len = dev->cdb_len;
/* reference: SSTL 1.13.4.2
* task_type, sata_direction
*/
if (task->data_dir == DMA_TO_DEVICE) {
task_context->task_type = SCU_TASK_TYPE_PACKET_DMA_OUT;
task_context->sata_direction = 0;
} else {
/* todo: for NO_DATA command, we need to send out raw frame. */
task_context->task_type = SCU_TASK_TYPE_PACKET_DMA_IN;
task_context->sata_direction = 1;
}
memset(&task_context->type.stp, 0, sizeof(task_context->type.stp));
task_context->type.stp.fis_type = FIS_DATA;
memset(&ireq->stp.cmd, 0, sizeof(ireq->stp.cmd));
memcpy(&ireq->stp.cmd.lbal, task->ata_task.atapi_packet, cdb_len);
task_context->ssp_command_iu_length = cdb_len / sizeof(u32);
/* task phase is set to TX_CMD */
task_context->task_phase = 0x1;
/* retry counter */
task_context->stp_retry_count = 0;
/* data transfer size. */
task_context->transfer_length_bytes = task->total_xfer_len;
/* setup sgl */
sci_request_build_sgl(ireq);
}
enum sci_status
sci_io_request_frame_handler(struct isci_request *ireq,
u32 frame_index)
{
struct isci_host *ihost = ireq->owning_controller;
struct isci_stp_request *stp_req = &ireq->stp.req;
enum sci_base_request_states state;
enum sci_status status;
ssize_t word_cnt;
state = ireq->sm.current_state_id;
switch (state) {
case SCI_REQ_STARTED: {
struct ssp_frame_hdr ssp_hdr;
void *frame_header;
sci_unsolicited_frame_control_get_header(&ihost->uf_control,
frame_index,
&frame_header);
word_cnt = sizeof(struct ssp_frame_hdr) / sizeof(u32);
sci_swab32_cpy(&ssp_hdr, frame_header, word_cnt);
if (ssp_hdr.frame_type == SSP_RESPONSE) {
struct ssp_response_iu *resp_iu;
ssize_t word_cnt = SSP_RESP_IU_MAX_SIZE / sizeof(u32);
sci_unsolicited_frame_control_get_buffer(&ihost->uf_control,
frame_index,
(void **)&resp_iu);
sci_swab32_cpy(&ireq->ssp.rsp, resp_iu, word_cnt);
resp_iu = &ireq->ssp.rsp;
if (resp_iu->datapres == 0x01 ||
resp_iu->datapres == 0x02) {
ireq->scu_status = SCU_TASK_DONE_CHECK_RESPONSE;
ireq->sci_status = SCI_FAILURE_CONTROLLER_SPECIFIC_IO_ERR;
} else {
ireq->scu_status = SCU_TASK_DONE_GOOD;
ireq->sci_status = SCI_SUCCESS;
}
} else {
/* not a response frame, why did it get forwarded? */
dev_err(&ihost->pdev->dev,
"%s: SCIC IO Request 0x%p received unexpected "
"frame %d type 0x%02x\n", __func__, ireq,
frame_index, ssp_hdr.frame_type);
}
/*
* In any case we are done with this frame buffer return it to
* the controller
*/
sci_controller_release_frame(ihost, frame_index);
return SCI_SUCCESS;
}
case SCI_REQ_TASK_WAIT_TC_RESP:
sci_io_request_copy_response(ireq);
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
sci_controller_release_frame(ihost, frame_index);
return SCI_SUCCESS;
case SCI_REQ_SMP_WAIT_RESP: {
struct sas_task *task = isci_request_access_task(ireq);
struct scatterlist *sg = &task->smp_task.smp_resp;
void *frame_header, *kaddr;
u8 *rsp;
sci_unsolicited_frame_control_get_header(&ihost->uf_control,
frame_index,
&frame_header);
kaddr = kmap_atomic(sg_page(sg));
rsp = kaddr + sg->offset;
sci_swab32_cpy(rsp, frame_header, 1);
if (rsp[0] == SMP_RESPONSE) {
void *smp_resp;
sci_unsolicited_frame_control_get_buffer(&ihost->uf_control,
frame_index,
&smp_resp);
word_cnt = (sg->length/4)-1;
if (word_cnt > 0)
word_cnt = min_t(unsigned int, word_cnt,
SCU_UNSOLICITED_FRAME_BUFFER_SIZE/4);
sci_swab32_cpy(rsp + 4, smp_resp, word_cnt);
ireq->scu_status = SCU_TASK_DONE_GOOD;
ireq->sci_status = SCI_SUCCESS;
sci_change_state(&ireq->sm, SCI_REQ_SMP_WAIT_TC_COMP);
} else {
/*
* This was not a response frame why did it get
* forwarded?
*/
dev_err(&ihost->pdev->dev,
"%s: SCIC SMP Request 0x%p received unexpected "
"frame %d type 0x%02x\n",
__func__,
ireq,
frame_index,
rsp[0]);
ireq->scu_status = SCU_TASK_DONE_SMP_FRM_TYPE_ERR;
ireq->sci_status = SCI_FAILURE_CONTROLLER_SPECIFIC_IO_ERR;
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
}
kunmap_atomic(kaddr);
sci_controller_release_frame(ihost, frame_index);
return SCI_SUCCESS;
}
case SCI_REQ_STP_UDMA_WAIT_TC_COMP:
return sci_stp_request_udma_general_frame_handler(ireq,
frame_index);
case SCI_REQ_STP_UDMA_WAIT_D2H:
/* Use the general frame handler to copy the resposne data */
status = sci_stp_request_udma_general_frame_handler(ireq, frame_index);
if (status != SCI_SUCCESS)
return status;
ireq->scu_status = SCU_TASK_DONE_CHECK_RESPONSE;
ireq->sci_status = SCI_FAILURE_IO_RESPONSE_VALID;
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
return SCI_SUCCESS;
case SCI_REQ_STP_NON_DATA_WAIT_D2H: {
struct dev_to_host_fis *frame_header;
u32 *frame_buffer;
status = sci_unsolicited_frame_control_get_header(&ihost->uf_control,
frame_index,
(void **)&frame_header);
if (status != SCI_SUCCESS) {
dev_err(&ihost->pdev->dev,
"%s: SCIC IO Request 0x%p could not get frame "
"header for frame index %d, status %x\n",
__func__,
stp_req,
frame_index,
status);
return status;
}
switch (frame_header->fis_type) {
case FIS_REGD2H:
sci_unsolicited_frame_control_get_buffer(&ihost->uf_control,
frame_index,
(void **)&frame_buffer);
sci_controller_copy_sata_response(&ireq->stp.rsp,
frame_header,
frame_buffer);
/* The command has completed with error */
ireq->scu_status = SCU_TASK_DONE_CHECK_RESPONSE;
ireq->sci_status = SCI_FAILURE_IO_RESPONSE_VALID;
break;
default:
dev_warn(&ihost->pdev->dev,
"%s: IO Request:0x%p Frame Id:%d protocol "
"violation occurred\n", __func__, stp_req,
frame_index);
ireq->scu_status = SCU_TASK_DONE_UNEXP_FIS;
ireq->sci_status = SCI_FAILURE_PROTOCOL_VIOLATION;
break;
}
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
/* Frame has been decoded return it to the controller */
sci_controller_release_frame(ihost, frame_index);
return status;
}
case SCI_REQ_STP_PIO_WAIT_FRAME: {
struct sas_task *task = isci_request_access_task(ireq);
struct dev_to_host_fis *frame_header;
u32 *frame_buffer;
status = sci_unsolicited_frame_control_get_header(&ihost->uf_control,
frame_index,
(void **)&frame_header);
if (status != SCI_SUCCESS) {
dev_err(&ihost->pdev->dev,
"%s: SCIC IO Request 0x%p could not get frame "
"header for frame index %d, status %x\n",
__func__, stp_req, frame_index, status);
return status;
}
switch (frame_header->fis_type) {
case FIS_PIO_SETUP:
/* Get from the frame buffer the PIO Setup Data */
sci_unsolicited_frame_control_get_buffer(&ihost->uf_control,
frame_index,
(void **)&frame_buffer);
/* Get the data from the PIO Setup The SCU Hardware
* returns first word in the frame_header and the rest
* of the data is in the frame buffer so we need to
* back up one dword
*/
/* transfer_count: first 16bits in the 4th dword */
stp_req->pio_len = frame_buffer[3] & 0xffff;
/* status: 4th byte in the 3rd dword */
stp_req->status = (frame_buffer[2] >> 24) & 0xff;
sci_controller_copy_sata_response(&ireq->stp.rsp,
frame_header,
frame_buffer);
ireq->stp.rsp.status = stp_req->status;
/* The next state is dependent on whether the
* request was PIO Data-in or Data out
*/
if (task->data_dir == DMA_FROM_DEVICE) {
sci_change_state(&ireq->sm, SCI_REQ_STP_PIO_DATA_IN);
} else if (task->data_dir == DMA_TO_DEVICE) {
/* Transmit data */
status = sci_stp_request_pio_data_out_transmit_data(ireq);
if (status != SCI_SUCCESS)
break;
sci_change_state(&ireq->sm, SCI_REQ_STP_PIO_DATA_OUT);
}
break;
case FIS_SETDEVBITS:
sci_change_state(&ireq->sm, SCI_REQ_STP_PIO_WAIT_FRAME);
break;
case FIS_REGD2H:
if (frame_header->status & ATA_BUSY) {
/*
* Now why is the drive sending a D2H Register
* FIS when it is still busy? Do nothing since
* we are still in the right state.
*/
dev_dbg(&ihost->pdev->dev,
"%s: SCIC PIO Request 0x%p received "
"D2H Register FIS with BSY status "
"0x%x\n",
__func__,
stp_req,
frame_header->status);
break;
}
sci_unsolicited_frame_control_get_buffer(&ihost->uf_control,
frame_index,
(void **)&frame_buffer);
sci_controller_copy_sata_response(&ireq->stp.rsp,
frame_header,
frame_buffer);
ireq->scu_status = SCU_TASK_DONE_CHECK_RESPONSE;
ireq->sci_status = SCI_FAILURE_IO_RESPONSE_VALID;
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
break;
default:
/* FIXME: what do we do here? */
break;
}
/* Frame is decoded return it to the controller */
sci_controller_release_frame(ihost, frame_index);
return status;
}
case SCI_REQ_STP_PIO_DATA_IN: {
struct dev_to_host_fis *frame_header;
struct sata_fis_data *frame_buffer;
status = sci_unsolicited_frame_control_get_header(&ihost->uf_control,
frame_index,
(void **)&frame_header);
if (status != SCI_SUCCESS) {
dev_err(&ihost->pdev->dev,
"%s: SCIC IO Request 0x%p could not get frame "
"header for frame index %d, status %x\n",
__func__,
stp_req,
frame_index,
status);
return status;
}
if (frame_header->fis_type != FIS_DATA) {
dev_err(&ihost->pdev->dev,
"%s: SCIC PIO Request 0x%p received frame %d "
"with fis type 0x%02x when expecting a data "
"fis.\n",
__func__,
stp_req,
frame_index,
frame_header->fis_type);
ireq->scu_status = SCU_TASK_DONE_GOOD;
ireq->sci_status = SCI_FAILURE_IO_REQUIRES_SCSI_ABORT;
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
/* Frame is decoded return it to the controller */
sci_controller_release_frame(ihost, frame_index);
return status;
}
if (stp_req->sgl.index < 0) {
ireq->saved_rx_frame_index = frame_index;
stp_req->pio_len = 0;
} else {
sci_unsolicited_frame_control_get_buffer(&ihost->uf_control,
frame_index,
(void **)&frame_buffer);
status = sci_stp_request_pio_data_in_copy_data(stp_req,
(u8 *)frame_buffer);
/* Frame is decoded return it to the controller */
sci_controller_release_frame(ihost, frame_index);
}
/* Check for the end of the transfer, are there more
* bytes remaining for this data transfer
*/
if (status != SCI_SUCCESS || stp_req->pio_len != 0)
return status;
if ((stp_req->status & ATA_BUSY) == 0) {
ireq->scu_status = SCU_TASK_DONE_CHECK_RESPONSE;
ireq->sci_status = SCI_FAILURE_IO_RESPONSE_VALID;
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
} else {
sci_change_state(&ireq->sm, SCI_REQ_STP_PIO_WAIT_FRAME);
}
return status;
}
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
case SCI_REQ_ATAPI_WAIT_PIO_SETUP: {
struct sas_task *task = isci_request_access_task(ireq);
sci_controller_release_frame(ihost, frame_index);
ireq->target_device->working_request = ireq;
if (task->data_dir == DMA_NONE) {
sci_change_state(&ireq->sm, SCI_REQ_ATAPI_WAIT_TC_COMP);
scu_atapi_reconstruct_raw_frame_task_context(ireq);
} else {
sci_change_state(&ireq->sm, SCI_REQ_ATAPI_WAIT_D2H);
scu_atapi_construct_task_context(ireq);
}
sci_controller_continue_io(ireq);
return SCI_SUCCESS;
}
case SCI_REQ_ATAPI_WAIT_D2H:
return atapi_d2h_reg_frame_handler(ireq, frame_index);
case SCI_REQ_ABORTING:
/*
* TODO: Is it even possible to get an unsolicited frame in the
* aborting state?
*/
sci_controller_release_frame(ihost, frame_index);
return SCI_SUCCESS;
default:
dev_warn(&ihost->pdev->dev,
"%s: SCIC IO Request given unexpected frame %x while "
"in state %d\n",
__func__,
frame_index,
state);
sci_controller_release_frame(ihost, frame_index);
return SCI_FAILURE_INVALID_STATE;
}
}
static enum sci_status stp_request_udma_await_tc_event(struct isci_request *ireq,
u32 completion_code)
{
enum sci_status status = SCI_SUCCESS;
switch (SCU_GET_COMPLETION_TL_STATUS(completion_code)) {
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_GOOD):
ireq->scu_status = SCU_TASK_DONE_GOOD;
ireq->sci_status = SCI_SUCCESS;
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
break;
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_UNEXP_FIS):
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_REG_ERR):
/* We must check ther response buffer to see if the D2H
* Register FIS was received before we got the TC
* completion.
*/
if (ireq->stp.rsp.fis_type == FIS_REGD2H) {
sci_remote_device_suspend(ireq->target_device,
SCI_SW_SUSPEND_NORMAL);
ireq->scu_status = SCU_TASK_DONE_CHECK_RESPONSE;
ireq->sci_status = SCI_FAILURE_IO_RESPONSE_VALID;
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
} else {
/* If we have an error completion status for the
* TC then we can expect a D2H register FIS from
* the device so we must change state to wait
* for it
*/
sci_change_state(&ireq->sm, SCI_REQ_STP_UDMA_WAIT_D2H);
}
break;
/* TODO Check to see if any of these completion status need to
* wait for the device to host register fis.
*/
/* TODO We can retry the command for SCU_TASK_DONE_CMD_LL_R_ERR
* - this comes only for B0
*/
default:
/* All other completion status cause the IO to be complete. */
ireq->scu_status = SCU_NORMALIZE_COMPLETION_STATUS(completion_code);
ireq->sci_status = SCI_FAILURE_CONTROLLER_SPECIFIC_IO_ERR;
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
break;
}
return status;
}
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
static enum sci_status atapi_raw_completion(struct isci_request *ireq, u32 completion_code,
enum sci_base_request_states next)
{
enum sci_status status = SCI_SUCCESS;
switch (SCU_GET_COMPLETION_TL_STATUS(completion_code)) {
case SCU_MAKE_COMPLETION_STATUS(SCU_TASK_DONE_GOOD):
ireq->scu_status = SCU_TASK_DONE_GOOD;
ireq->sci_status = SCI_SUCCESS;
sci_change_state(&ireq->sm, next);
break;
default:
/* All other completion status cause the IO to be complete.
* If a NAK was received, then it is up to the user to retry
* the request.
*/
ireq->scu_status = SCU_NORMALIZE_COMPLETION_STATUS(completion_code);
ireq->sci_status = SCI_FAILURE_CONTROLLER_SPECIFIC_IO_ERR;
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
break;
}
return status;
}
static enum sci_status atapi_data_tc_completion_handler(struct isci_request *ireq,
u32 completion_code)
{
struct isci_remote_device *idev = ireq->target_device;
struct dev_to_host_fis *d2h = &ireq->stp.rsp;
enum sci_status status = SCI_SUCCESS;
switch (SCU_GET_COMPLETION_TL_STATUS(completion_code)) {
case (SCU_TASK_DONE_GOOD << SCU_COMPLETION_TL_STATUS_SHIFT):
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
break;
case (SCU_TASK_DONE_UNEXP_FIS << SCU_COMPLETION_TL_STATUS_SHIFT): {
u16 len = sci_req_tx_bytes(ireq);
/* likely non-error data underrrun, workaround missing
* d2h frame from the controller
*/
if (d2h->fis_type != FIS_REGD2H) {
d2h->fis_type = FIS_REGD2H;
d2h->flags = (1 << 6);
d2h->status = 0x50;
d2h->error = 0;
d2h->lbal = 0;
d2h->byte_count_low = len & 0xff;
d2h->byte_count_high = len >> 8;
d2h->device = 0xa0;
d2h->lbal_exp = 0;
d2h->lbam_exp = 0;
d2h->lbah_exp = 0;
d2h->_r_a = 0;
d2h->sector_count = 0x3;
d2h->sector_count_exp = 0;
d2h->_r_b = 0;
d2h->_r_c = 0;
d2h->_r_d = 0;
}
ireq->scu_status = SCU_TASK_DONE_GOOD;
ireq->sci_status = SCI_SUCCESS_IO_DONE_EARLY;
status = ireq->sci_status;
/* the hw will have suspended the rnc, so complete the
* request upon pending resume
*/
sci_change_state(&idev->sm, SCI_STP_DEV_ATAPI_ERROR);
break;
}
case (SCU_TASK_DONE_EXCESS_DATA << SCU_COMPLETION_TL_STATUS_SHIFT):
/* In this case, there is no UF coming after.
* compelte the IO now.
*/
ireq->scu_status = SCU_TASK_DONE_GOOD;
ireq->sci_status = SCI_SUCCESS;
sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
break;
default:
if (d2h->fis_type == FIS_REGD2H) {
/* UF received change the device state to ATAPI_ERROR */
status = ireq->sci_status;
sci_change_state(&idev->sm, SCI_STP_DEV_ATAPI_ERROR);
} else {
/* If receiving any non-success TC status, no UF
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
* received yet, then an UF for the status fis
* is coming after (XXX: suspect this is
* actually a protocol error or a bug like the
* DONE_UNEXP_FIS case)
*/
ireq->scu_status = SCU_TASK_DONE_CHECK_RESPONSE;
ireq->sci_status = SCI_FAILURE_IO_RESPONSE_VALID;
sci_change_state(&ireq->sm, SCI_REQ_ATAPI_WAIT_D2H);
}
break;
}
return status;
}
static int sci_request_smp_completion_status_is_tx_suspend(
unsigned int completion_status)
{
switch (completion_status) {
case SCU_TASK_OPEN_REJECT_WRONG_DESTINATION:
case SCU_TASK_OPEN_REJECT_RESERVED_ABANDON_1:
case SCU_TASK_OPEN_REJECT_RESERVED_ABANDON_2:
case SCU_TASK_OPEN_REJECT_RESERVED_ABANDON_3:
case SCU_TASK_OPEN_REJECT_BAD_DESTINATION:
case SCU_TASK_OPEN_REJECT_ZONE_VIOLATION:
return 1;
}
return 0;
}
static int sci_request_smp_completion_status_is_tx_rx_suspend(
unsigned int completion_status)
{
return 0; /* There are no Tx/Rx SMP suspend conditions. */
}
static int sci_request_ssp_completion_status_is_tx_suspend(
unsigned int completion_status)
{
switch (completion_status) {
case SCU_TASK_DONE_TX_RAW_CMD_ERR:
case SCU_TASK_DONE_LF_ERR:
case SCU_TASK_OPEN_REJECT_WRONG_DESTINATION:
case SCU_TASK_OPEN_REJECT_RESERVED_ABANDON_1:
case SCU_TASK_OPEN_REJECT_RESERVED_ABANDON_2:
case SCU_TASK_OPEN_REJECT_RESERVED_ABANDON_3:
case SCU_TASK_OPEN_REJECT_BAD_DESTINATION:
case SCU_TASK_OPEN_REJECT_ZONE_VIOLATION:
case SCU_TASK_OPEN_REJECT_STP_RESOURCES_BUSY:
case SCU_TASK_OPEN_REJECT_PROTOCOL_NOT_SUPPORTED:
case SCU_TASK_OPEN_REJECT_CONNECTION_RATE_NOT_SUPPORTED:
return 1;
}
return 0;
}
static int sci_request_ssp_completion_status_is_tx_rx_suspend(
unsigned int completion_status)
{
return 0; /* There are no Tx/Rx SSP suspend conditions. */
}
static int sci_request_stpsata_completion_status_is_tx_suspend(
unsigned int completion_status)
{
switch (completion_status) {
case SCU_TASK_DONE_TX_RAW_CMD_ERR:
case SCU_TASK_DONE_LL_R_ERR:
case SCU_TASK_DONE_LL_PERR:
case SCU_TASK_DONE_REG_ERR:
case SCU_TASK_DONE_SDB_ERR:
case SCU_TASK_OPEN_REJECT_WRONG_DESTINATION:
case SCU_TASK_OPEN_REJECT_RESERVED_ABANDON_1:
case SCU_TASK_OPEN_REJECT_RESERVED_ABANDON_2:
case SCU_TASK_OPEN_REJECT_RESERVED_ABANDON_3:
case SCU_TASK_OPEN_REJECT_BAD_DESTINATION:
case SCU_TASK_OPEN_REJECT_ZONE_VIOLATION:
case SCU_TASK_OPEN_REJECT_STP_RESOURCES_BUSY:
case SCU_TASK_OPEN_REJECT_PROTOCOL_NOT_SUPPORTED:
case SCU_TASK_OPEN_REJECT_CONNECTION_RATE_NOT_SUPPORTED:
return 1;
}
return 0;
}
static int sci_request_stpsata_completion_status_is_tx_rx_suspend(
unsigned int completion_status)
{
switch (completion_status) {
case SCU_TASK_DONE_LF_ERR:
case SCU_TASK_DONE_LL_SY_TERM:
case SCU_TASK_DONE_LL_LF_TERM:
case SCU_TASK_DONE_BREAK_RCVD:
case SCU_TASK_DONE_INV_FIS_LEN:
case SCU_TASK_DONE_UNEXP_FIS:
case SCU_TASK_DONE_UNEXP_SDBFIS:
case SCU_TASK_DONE_MAX_PLD_ERR:
return 1;
}
return 0;
}
static void sci_request_handle_suspending_completions(
struct isci_request *ireq,
u32 completion_code)
{
int is_tx = 0;
int is_tx_rx = 0;
switch (ireq->protocol) {
case SAS_PROTOCOL_SMP:
is_tx = sci_request_smp_completion_status_is_tx_suspend(
completion_code);
is_tx_rx = sci_request_smp_completion_status_is_tx_rx_suspend(
completion_code);
break;
case SAS_PROTOCOL_SSP:
is_tx = sci_request_ssp_completion_status_is_tx_suspend(
completion_code);
is_tx_rx = sci_request_ssp_completion_status_is_tx_rx_suspend(
completion_code);
break;
case SAS_PROTOCOL_STP:
is_tx = sci_request_stpsata_completion_status_is_tx_suspend(
completion_code);
is_tx_rx =
sci_request_stpsata_completion_status_is_tx_rx_suspend(
completion_code);
break;
default:
dev_warn(&ireq->isci_host->pdev->dev,
"%s: request %p has no valid protocol\n",
__func__, ireq);
break;
}
if (is_tx || is_tx_rx) {
BUG_ON(is_tx && is_tx_rx);
sci_remote_node_context_suspend(
&ireq->target_device->rnc,
SCI_HW_SUSPEND,
(is_tx_rx) ? SCU_EVENT_TL_RNC_SUSPEND_TX_RX
: SCU_EVENT_TL_RNC_SUSPEND_TX);
}
}
enum sci_status
sci_io_request_tc_completion(struct isci_request *ireq,
u32 completion_code)
{
enum sci_base_request_states state;
struct isci_host *ihost = ireq->owning_controller;
state = ireq->sm.current_state_id;
/* Decode those completions that signal upcoming suspension events. */
sci_request_handle_suspending_completions(
ireq, SCU_GET_COMPLETION_TL_STATUS(completion_code));
switch (state) {
case SCI_REQ_STARTED:
return request_started_state_tc_event(ireq, completion_code);
case SCI_REQ_TASK_WAIT_TC_COMP:
return ssp_task_request_await_tc_event(ireq,
completion_code);
case SCI_REQ_SMP_WAIT_RESP:
return smp_request_await_response_tc_event(ireq,
completion_code);
case SCI_REQ_SMP_WAIT_TC_COMP:
return smp_request_await_tc_event(ireq, completion_code);
case SCI_REQ_STP_UDMA_WAIT_TC_COMP:
return stp_request_udma_await_tc_event(ireq,
completion_code);
case SCI_REQ_STP_NON_DATA_WAIT_H2D:
return stp_request_non_data_await_h2d_tc_event(ireq,
completion_code);
case SCI_REQ_STP_PIO_WAIT_H2D:
return stp_request_pio_await_h2d_completion_tc_event(ireq,
completion_code);
case SCI_REQ_STP_PIO_DATA_OUT:
return pio_data_out_tx_done_tc_event(ireq, completion_code);
case SCI_REQ_ABORTING:
return request_aborting_state_tc_event(ireq,
completion_code);
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
case SCI_REQ_ATAPI_WAIT_H2D:
return atapi_raw_completion(ireq, completion_code,
SCI_REQ_ATAPI_WAIT_PIO_SETUP);
case SCI_REQ_ATAPI_WAIT_TC_COMP:
return atapi_raw_completion(ireq, completion_code,
SCI_REQ_ATAPI_WAIT_D2H);
case SCI_REQ_ATAPI_WAIT_D2H:
return atapi_data_tc_completion_handler(ireq, completion_code);
default:
dev_warn(&ihost->pdev->dev, "%s: %x in wrong state %s\n",
__func__, completion_code, req_state_name(state));
return SCI_FAILURE_INVALID_STATE;
}
}
/**
* isci_request_process_response_iu() - This function sets the status and
* response iu, in the task struct, from the request object for the upper
* layer driver.
* @sas_task: This parameter is the task struct from the upper layer driver.
* @resp_iu: This parameter points to the response iu of the completed request.
* @dev: This parameter specifies the linux device struct.
*
* none.
*/
static void isci_request_process_response_iu(
struct sas_task *task,
struct ssp_response_iu *resp_iu,
struct device *dev)
{
dev_dbg(dev,
"%s: resp_iu = %p "
"resp_iu->status = 0x%x,\nresp_iu->datapres = %d "
"resp_iu->response_data_len = %x, "
"resp_iu->sense_data_len = %x\nresponse data: ",
__func__,
resp_iu,
resp_iu->status,
resp_iu->datapres,
resp_iu->response_data_len,
resp_iu->sense_data_len);
task->task_status.stat = resp_iu->status;
/* libsas updates the task status fields based on the response iu. */
sas_ssp_task_response(dev, task, resp_iu);
}
/**
* isci_request_set_open_reject_status() - This function prepares the I/O
* completion for OPEN_REJECT conditions.
* @request: This parameter is the completed isci_request object.
* @response_ptr: This parameter specifies the service response for the I/O.
* @status_ptr: This parameter specifies the exec status for the I/O.
* @open_rej_reason: This parameter specifies the encoded reason for the
* abandon-class reject.
*
* none.
*/
static void isci_request_set_open_reject_status(
struct isci_request *request,
struct sas_task *task,
enum service_response *response_ptr,
enum exec_status *status_ptr,
enum sas_open_rej_reason open_rej_reason)
{
/* Task in the target is done. */
set_bit(IREQ_COMPLETE_IN_TARGET, &request->flags);
*response_ptr = SAS_TASK_UNDELIVERED;
*status_ptr = SAS_OPEN_REJECT;
task->task_status.open_rej_reason = open_rej_reason;
}
/**
* isci_request_handle_controller_specific_errors() - This function decodes
* controller-specific I/O completion error conditions.
* @request: This parameter is the completed isci_request object.
* @response_ptr: This parameter specifies the service response for the I/O.
* @status_ptr: This parameter specifies the exec status for the I/O.
*
* none.
*/
static void isci_request_handle_controller_specific_errors(
struct isci_remote_device *idev,
struct isci_request *request,
struct sas_task *task,
enum service_response *response_ptr,
enum exec_status *status_ptr)
{
unsigned int cstatus;
cstatus = request->scu_status;
dev_dbg(&request->isci_host->pdev->dev,
"%s: %p SCI_FAILURE_CONTROLLER_SPECIFIC_IO_ERR "
"- controller status = 0x%x\n",
__func__, request, cstatus);
/* Decode the controller-specific errors; most
* important is to recognize those conditions in which
* the target may still have a task outstanding that
* must be aborted.
*
* Note that there are SCU completion codes being
* named in the decode below for which SCIC has already
* done work to handle them in a way other than as
* a controller-specific completion code; these are left
* in the decode below for completeness sake.
*/
switch (cstatus) {
case SCU_TASK_DONE_DMASETUP_DIRERR:
/* Also SCU_TASK_DONE_SMP_FRM_TYPE_ERR: */
case SCU_TASK_DONE_XFERCNT_ERR:
/* Also SCU_TASK_DONE_SMP_UFI_ERR: */
if (task->task_proto == SAS_PROTOCOL_SMP) {
/* SCU_TASK_DONE_SMP_UFI_ERR == Task Done. */
*response_ptr = SAS_TASK_COMPLETE;
/* See if the device has been/is being stopped. Note
* that we ignore the quiesce state, since we are
* concerned about the actual device state.
*/
if (!idev)
*status_ptr = SAS_DEVICE_UNKNOWN;
else
*status_ptr = SAS_ABORTED_TASK;
set_bit(IREQ_COMPLETE_IN_TARGET, &request->flags);
} else {
/* Task in the target is not done. */
*response_ptr = SAS_TASK_UNDELIVERED;
if (!idev)
*status_ptr = SAS_DEVICE_UNKNOWN;
else
*status_ptr = SAM_STAT_TASK_ABORTED;
clear_bit(IREQ_COMPLETE_IN_TARGET, &request->flags);
}
break;
case SCU_TASK_DONE_CRC_ERR:
case SCU_TASK_DONE_NAK_CMD_ERR:
case SCU_TASK_DONE_EXCESS_DATA:
case SCU_TASK_DONE_UNEXP_FIS:
/* Also SCU_TASK_DONE_UNEXP_RESP: */
case SCU_TASK_DONE_VIIT_ENTRY_NV: /* TODO - conditions? */
case SCU_TASK_DONE_IIT_ENTRY_NV: /* TODO - conditions? */
case SCU_TASK_DONE_RNCNV_OUTBOUND: /* TODO - conditions? */
/* These are conditions in which the target
* has completed the task, so that no cleanup
* is necessary.
*/
*response_ptr = SAS_TASK_COMPLETE;
/* See if the device has been/is being stopped. Note
* that we ignore the quiesce state, since we are
* concerned about the actual device state.
*/
if (!idev)
*status_ptr = SAS_DEVICE_UNKNOWN;
else
*status_ptr = SAS_ABORTED_TASK;
set_bit(IREQ_COMPLETE_IN_TARGET, &request->flags);
break;
/* Note that the only open reject completion codes seen here will be
* abandon-class codes; all others are automatically retried in the SCU.
*/
case SCU_TASK_OPEN_REJECT_WRONG_DESTINATION:
isci_request_set_open_reject_status(
request, task, response_ptr, status_ptr,
SAS_OREJ_WRONG_DEST);
break;
case SCU_TASK_OPEN_REJECT_ZONE_VIOLATION:
/* Note - the return of AB0 will change when
* libsas implements detection of zone violations.
*/
isci_request_set_open_reject_status(
request, task, response_ptr, status_ptr,
SAS_OREJ_RESV_AB0);
break;
case SCU_TASK_OPEN_REJECT_RESERVED_ABANDON_1:
isci_request_set_open_reject_status(
request, task, response_ptr, status_ptr,
SAS_OREJ_RESV_AB1);
break;
case SCU_TASK_OPEN_REJECT_RESERVED_ABANDON_2:
isci_request_set_open_reject_status(
request, task, response_ptr, status_ptr,
SAS_OREJ_RESV_AB2);
break;
case SCU_TASK_OPEN_REJECT_RESERVED_ABANDON_3:
isci_request_set_open_reject_status(
request, task, response_ptr, status_ptr,
SAS_OREJ_RESV_AB3);
break;
case SCU_TASK_OPEN_REJECT_BAD_DESTINATION:
isci_request_set_open_reject_status(
request, task, response_ptr, status_ptr,
SAS_OREJ_BAD_DEST);
break;
case SCU_TASK_OPEN_REJECT_STP_RESOURCES_BUSY:
isci_request_set_open_reject_status(
request, task, response_ptr, status_ptr,
SAS_OREJ_STP_NORES);
break;
case SCU_TASK_OPEN_REJECT_PROTOCOL_NOT_SUPPORTED:
isci_request_set_open_reject_status(
request, task, response_ptr, status_ptr,
SAS_OREJ_EPROTO);
break;
case SCU_TASK_OPEN_REJECT_CONNECTION_RATE_NOT_SUPPORTED:
isci_request_set_open_reject_status(
request, task, response_ptr, status_ptr,
SAS_OREJ_CONN_RATE);
break;
case SCU_TASK_DONE_LL_R_ERR:
/* Also SCU_TASK_DONE_ACK_NAK_TO: */
case SCU_TASK_DONE_LL_PERR:
case SCU_TASK_DONE_LL_SY_TERM:
/* Also SCU_TASK_DONE_NAK_ERR:*/
case SCU_TASK_DONE_LL_LF_TERM:
/* Also SCU_TASK_DONE_DATA_LEN_ERR: */
case SCU_TASK_DONE_LL_ABORT_ERR:
case SCU_TASK_DONE_SEQ_INV_TYPE:
/* Also SCU_TASK_DONE_UNEXP_XR: */
case SCU_TASK_DONE_XR_IU_LEN_ERR:
case SCU_TASK_DONE_INV_FIS_LEN:
/* Also SCU_TASK_DONE_XR_WD_LEN: */
case SCU_TASK_DONE_SDMA_ERR:
case SCU_TASK_DONE_OFFSET_ERR:
case SCU_TASK_DONE_MAX_PLD_ERR:
case SCU_TASK_DONE_LF_ERR:
case SCU_TASK_DONE_SMP_RESP_TO_ERR: /* Escalate to dev reset? */
case SCU_TASK_DONE_SMP_LL_RX_ERR:
case SCU_TASK_DONE_UNEXP_DATA:
case SCU_TASK_DONE_UNEXP_SDBFIS:
case SCU_TASK_DONE_REG_ERR:
case SCU_TASK_DONE_SDB_ERR:
case SCU_TASK_DONE_TASK_ABORT:
default:
/* Task in the target is not done. */
*response_ptr = SAS_TASK_UNDELIVERED;
*status_ptr = SAM_STAT_TASK_ABORTED;
if (task->task_proto == SAS_PROTOCOL_SMP)
set_bit(IREQ_COMPLETE_IN_TARGET, &request->flags);
else
clear_bit(IREQ_COMPLETE_IN_TARGET, &request->flags);
break;
}
}
static void isci_process_stp_response(struct sas_task *task, struct dev_to_host_fis *fis)
{
struct task_status_struct *ts = &task->task_status;
struct ata_task_resp *resp = (void *)&ts->buf[0];
resp->frame_len = sizeof(*fis);
memcpy(resp->ending_fis, fis, sizeof(*fis));
ts->buf_valid_size = sizeof(*resp);
/* If an error is flagged let libata decode the fis */
if (ac_err_mask(fis->status))
ts->stat = SAS_PROTO_RESPONSE;
else
ts->stat = SAM_STAT_GOOD;
ts->resp = SAS_TASK_COMPLETE;
}
static void isci_request_io_request_complete(struct isci_host *ihost,
struct isci_request *request,
enum sci_io_status completion_status)
{
struct sas_task *task = isci_request_access_task(request);
struct ssp_response_iu *resp_iu;
unsigned long task_flags;
[SCSI] isci: Lookup device references through requests in completions. The LLDD needs to obtain a reference to the device through the request itself and not through the domain_device, because the domain_device.lldd_dev is set to NULL early in the lldd_dev_gone call. This relies on the fact that the isci_remote_device object is keeping a seperate reference count of outstanding requests. TODO: unify the request count tracking with the isci_remote_device kref. The failure signature of this condition looks like the following log, where the important bits are the call to lldd_dev_gone followed by a crash in isci_terminate_request_core: [ 229.151541] isci 0000:0b:00.0: isci_remote_device_gone: domain_device = ffff8801492d4800, isci_device = ffff880143c657d0, isci_port = ffff880143c63658 [ 229.166007] isci 0000:0b:00.0: isci_remote_device_stop: isci_device = ffff880143c657d0 [ 229.175317] isci 0000:0b:00.0: isci_terminate_pending_requests: idev=ffff880143c657d0 request=ffff88014741f000; task=ffff8801470f46c0 old_state=2 [ 229.189702] isci 0000:0b:00.0: isci_terminate_request_core: device = ffff880143c657d0; request = ffff88014741f000 [ 229.201339] isci 0000:0b:00.0: isci_terminate_request_core: before completion wait (ffff88014741f000/ffff880149715ad0) [ 229.213414] isci 0000:0b:00.0: sci_controller_process_completions: completion queue entry:0x8000a0e9 [ 229.214401] BUG: unable to handle kernel NULL pointer dereference at 0000000000000228 [ 229.214401] IP:jdskirvi-testlbo [<ffffffffa00a58be>] sci_request_completed_state_enter+0x50/0xafb [isci] [ 229.214401] PGD 13d19e067 PUD 13d104067 PMD 0 [ 229.214401] Oops: 0000 [#1] SMP [ 229.214401] CPU 0 x kernel: [ 226 [ 229.214401] Modules linked in: ipv6 dm_multipath uinput nouveau snd_hda_codec_realtek snd_hda_intel ttm drm_kms_helper drm snd_hda_codec snd_hwdep snd_pcm snd_timer i2c_algo_bit isci snd libsas ioatdma mxm_wmi iTCO_wdt soundcore snd_page_alloc scsi_transport_sas iTCO_vendor_support wmi dca video i2c_i801 i2c_core [last unloaded: speedstep_lib] [ 229.214401] [ 229.214401] Pid: 5, comm: kworker/u:0 Not tainted 3.0.0-isci-11.7.29+ #30.353196] Buffer Intel Corporation Stoakley/Pearlcity Workstation [ 229.214401] RIP: 0010:[<ffffffffa00a58be>] I/O error on dev [<ffffffffa00a58be>] sci_request_completed_state_enter+0x50/0xafb [isci] [ 229.214401] RSP: 0018:ffff88014fc03d20 EFLAGS: 00010046 [ 229.214401] RAX: 0000000000000000 RBX: ffff88014741f000 RCX: 0000000000000000 [ 229.214401] RDX: ffffffffa00b2c90 RSI: 0000000000000017 RDI: ffff88014741f0a0 [ 229.214401] RBP: ffff88014fc03d90 R08: 0000000000000018 R09: 0000000000000000 [ 229.214401] R10: 0000000000000000 R11: ffffffff81a17d98 R12: 000000000000001d [ 229.214401] R13: ffff8801470f46c0 R14: 0000000000000000 R15: 0000000000008000 [ 229.214401] FS: 0000000000000000(0000) GS:ffff88014fc00000(0000) knlGS:0000000000000000 [ 229.214401] CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b [ 229.214401] CR2: 0000000000000228 CR3: 000000013ceaa000 CR4: 00000000000406f0 [ 229.214401] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [ 229.214401] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 [ 229.214401] Process kworker/u:0 (pid: 5, threadinfo ffff880149714000, task ffff880149718000) [ 229.214401] Call Trace: [ 229.214401] <IRQ> [ 229.214401] [<ffffffffa00aa6ce>] sci_change_state+0x4a/0x4f [isci] [ 229.214401] [<ffffffffa00a4ca6>] sci_io_request_tc_completion+0x79c/0x7a0 [isci] [ 229.214401] [<ffffffffa00acf35>] sci_controller_process_completions+0x14f/0x396 [isci] [ 229.214401] [<ffffffffa00abbda>] ? spin_lock_irq+0xe/0x10 [isci] [ 229.214401] [<ffffffffa00ad2cf>] isci_host_completion_routine+0x71/0x2be [isci] [ 229.214401] [<ffffffff8107c6b3>] ? mark_held_locks+0x52/0x70 [ 229.214401] [<ffffffff810538e8>] tasklet_action+0x90/0xf1 [ 229.214401] [<ffffffff81054050>] __do_softirq+0xe5/0x1bf [ 229.214401] [<ffffffff8106d9d1>] ? hrtimer_interrupt+0x129/0x1bb [ 229.214401] [<ffffffff814ff69c>] call_softirq+0x1c/0x30 [ 229.214401] [<ffffffff8100bb67>] do_softirq+0x4b/0xa3 [ 229.214401] [<ffffffff81053d84>] irq_exit+0x53/0xb4 [ 229.214401] [<ffffffff814fffe7>] smp_apic_timer_interrupt+0x83/0x91 [ 229.214401] [<ffffffff814fee53>] apic_timer_interrupt+0x13/0x20 [ 229.214401] <EOI> [ 229.214401] [<ffffffff814f7ad4>] ? retint_restore_args+0x13/0x13 [ 229.214401] [<ffffffff8107af29>] ? trace_hardirqs_off+0xd/0xf [ 229.214401] [<ffffffff8104ea71>] ? vprintk+0x40b/0x452 [ 229.214401] [<ffffffff814f4b5a>] printk+0x41/0x47 [ 229.214401] [<ffffffff81314484>] __dev_printk+0x78/0x7a [ 229.214401] [<ffffffff8131471e>] dev_printk+0x45/0x47 [ 229.214401] [<ffffffffa00ae2a3>] isci_terminate_request_core+0x15d/0x317 [isci] [ 229.214401] [<ffffffffa00af1ad>] isci_terminate_pending_requests+0x1a4/0x204 [isci] [ 229.214401] [<ffffffffa00229f6>] ? sas_phye_oob_error+0xc3/0xc3 [libsas] [ 229.214401] [<ffffffffa00a7d9e>] isci_remote_device_nuke_requests+0xa6/0xff [isci] [ 229.214401] [<ffffffffa00a811a>] isci_remote_device_stop+0x7c/0x166 [isci] [ 229.214401] [<ffffffffa00229f6>] ? sas_phye_oob_error+0xc3/0xc3 [libsas] [ 229.214401] [<ffffffffa00a827a>] isci_remote_device_gone+0x76/0x7e [isci] [ 229.214401] [<ffffffffa002363e>] sas_notify_lldd_dev_gone+0x34/0x36 [libsas] [ 229.214401] [<ffffffffa0023945>] sas_unregister_dev+0x57/0x9c [libsas] [ 229.214401] [<ffffffffa00239c0>] sas_unregister_domain_devices+0x36/0x65 [libsas] [ 229.214401] [<ffffffffa0022cb8>] sas_deform_port+0x72/0x1ac [libsas] [ 229.214401] [<ffffffffa00229f6>] ? sas_phye_oob_error+0xc3/0xc3 [libsas] [ 229.214401] [<ffffffffa0022a34>] sas_phye_loss_of_signal+0x3e/0x42 [libsas] Signed-off-by: Jeff Skirvin <jeffrey.d.skirvin@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-28 06:04:50 +08:00
struct isci_remote_device *idev = request->target_device;
enum service_response response = SAS_TASK_UNDELIVERED;
enum exec_status status = SAS_ABORTED_TASK;
dev_dbg(&ihost->pdev->dev,
"%s: request = %p, task = %p, "
"task->data_dir = %d completion_status = 0x%x\n",
__func__, request, task, task->data_dir, completion_status);
/* The request is done from an SCU HW perspective. */
/* This is an active request being completed from the core. */
switch (completion_status) {
case SCI_IO_FAILURE_RESPONSE_VALID:
dev_dbg(&ihost->pdev->dev,
"%s: SCI_IO_FAILURE_RESPONSE_VALID (%p/%p)\n",
__func__, request, task);
if (sas_protocol_ata(task->task_proto)) {
isci_process_stp_response(task, &request->stp.rsp);
} else if (SAS_PROTOCOL_SSP == task->task_proto) {
/* crack the iu response buffer. */
resp_iu = &request->ssp.rsp;
isci_request_process_response_iu(task, resp_iu,
&ihost->pdev->dev);
} else if (SAS_PROTOCOL_SMP == task->task_proto) {
dev_err(&ihost->pdev->dev,
"%s: SCI_IO_FAILURE_RESPONSE_VALID: "
"SAS_PROTOCOL_SMP protocol\n",
__func__);
} else
dev_err(&ihost->pdev->dev,
"%s: unknown protocol\n", __func__);
/* use the task status set in the task struct by the
* isci_request_process_response_iu call.
*/
set_bit(IREQ_COMPLETE_IN_TARGET, &request->flags);
response = task->task_status.resp;
status = task->task_status.stat;
break;
case SCI_IO_SUCCESS:
case SCI_IO_SUCCESS_IO_DONE_EARLY:
response = SAS_TASK_COMPLETE;
status = SAM_STAT_GOOD;
set_bit(IREQ_COMPLETE_IN_TARGET, &request->flags);
if (completion_status == SCI_IO_SUCCESS_IO_DONE_EARLY) {
/* This was an SSP / STP / SATA transfer.
* There is a possibility that less data than
* the maximum was transferred.
*/
u32 transferred_length = sci_req_tx_bytes(request);
task->task_status.residual
= task->total_xfer_len - transferred_length;
/* If there were residual bytes, call this an
* underrun.
*/
if (task->task_status.residual != 0)
status = SAS_DATA_UNDERRUN;
dev_dbg(&ihost->pdev->dev,
"%s: SCI_IO_SUCCESS_IO_DONE_EARLY %d\n",
__func__, status);
} else
dev_dbg(&ihost->pdev->dev, "%s: SCI_IO_SUCCESS\n",
__func__);
break;
case SCI_IO_FAILURE_TERMINATED:
dev_dbg(&ihost->pdev->dev,
"%s: SCI_IO_FAILURE_TERMINATED (%p/%p)\n",
__func__, request, task);
/* The request was terminated explicitly. */
set_bit(IREQ_COMPLETE_IN_TARGET, &request->flags);
response = SAS_TASK_UNDELIVERED;
/* See if the device has been/is being stopped. Note
* that we ignore the quiesce state, since we are
* concerned about the actual device state.
*/
if (!idev)
status = SAS_DEVICE_UNKNOWN;
else
status = SAS_ABORTED_TASK;
break;
case SCI_FAILURE_CONTROLLER_SPECIFIC_IO_ERR:
isci_request_handle_controller_specific_errors(idev, request,
task, &response,
&status);
break;
case SCI_IO_FAILURE_REMOTE_DEVICE_RESET_REQUIRED:
/* This is a special case, in that the I/O completion
* is telling us that the device needs a reset.
* In order for the device reset condition to be
* noticed, the I/O has to be handled in the error
* handler. Set the reset flag and cause the
* SCSI error thread to be scheduled.
*/
spin_lock_irqsave(&task->task_state_lock, task_flags);
task->task_state_flags |= SAS_TASK_NEED_DEV_RESET;
spin_unlock_irqrestore(&task->task_state_lock, task_flags);
/* Fail the I/O. */
response = SAS_TASK_UNDELIVERED;
status = SAM_STAT_TASK_ABORTED;
clear_bit(IREQ_COMPLETE_IN_TARGET, &request->flags);
break;
case SCI_FAILURE_RETRY_REQUIRED:
/* Fail the I/O so it can be retried. */
response = SAS_TASK_UNDELIVERED;
if (!idev)
status = SAS_DEVICE_UNKNOWN;
else
status = SAS_ABORTED_TASK;
set_bit(IREQ_COMPLETE_IN_TARGET, &request->flags);
break;
default:
/* Catch any otherwise unhandled error codes here. */
dev_dbg(&ihost->pdev->dev,
"%s: invalid completion code: 0x%x - "
"isci_request = %p\n",
__func__, completion_status, request);
response = SAS_TASK_UNDELIVERED;
/* See if the device has been/is being stopped. Note
* that we ignore the quiesce state, since we are
* concerned about the actual device state.
*/
if (!idev)
status = SAS_DEVICE_UNKNOWN;
else
status = SAS_ABORTED_TASK;
if (SAS_PROTOCOL_SMP == task->task_proto)
set_bit(IREQ_COMPLETE_IN_TARGET, &request->flags);
else
clear_bit(IREQ_COMPLETE_IN_TARGET, &request->flags);
break;
}
switch (task->task_proto) {
case SAS_PROTOCOL_SSP:
if (task->data_dir == DMA_NONE)
break;
if (task->num_scatter == 0)
/* 0 indicates a single dma address */
dma_unmap_single(&ihost->pdev->dev,
request->zero_scatter_daddr,
task->total_xfer_len, task->data_dir);
else /* unmap the sgl dma addresses */
dma_unmap_sg(&ihost->pdev->dev, task->scatter,
request->num_sg_entries, task->data_dir);
break;
case SAS_PROTOCOL_SMP: {
struct scatterlist *sg = &task->smp_task.smp_req;
struct smp_req *smp_req;
void *kaddr;
dma_unmap_sg(&ihost->pdev->dev, sg, 1, DMA_TO_DEVICE);
/* need to swab it back in case the command buffer is re-used */
kaddr = kmap_atomic(sg_page(sg));
smp_req = kaddr + sg->offset;
sci_swab32_cpy(smp_req, smp_req, sg->length / sizeof(u32));
kunmap_atomic(kaddr);
break;
}
default:
break;
}
spin_lock_irqsave(&task->task_state_lock, task_flags);
task->task_status.resp = response;
task->task_status.stat = status;
if (test_bit(IREQ_COMPLETE_IN_TARGET, &request->flags)) {
/* Normal notification (task_done) */
task->task_state_flags |= SAS_TASK_STATE_DONE;
task->task_state_flags &= ~(SAS_TASK_AT_INITIATOR |
SAS_TASK_STATE_PENDING);
}
spin_unlock_irqrestore(&task->task_state_lock, task_flags);
/* complete the io request to the core. */
sci_controller_complete_io(ihost, request->target_device, request);
/* set terminated handle so it cannot be completed or
* terminated again, and to cause any calls into abort
* task to recognize the already completed case.
*/
set_bit(IREQ_TERMINATED, &request->flags);
ireq_done(ihost, request, task);
}
static void sci_request_started_state_enter(struct sci_base_state_machine *sm)
{
struct isci_request *ireq = container_of(sm, typeof(*ireq), sm);
struct domain_device *dev = ireq->target_device->domain_dev;
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
enum sci_base_request_states state;
struct sas_task *task;
/* XXX as hch said always creating an internal sas_task for tmf
* requests would simplify the driver
*/
task = (test_bit(IREQ_TMF, &ireq->flags)) ? NULL : isci_request_access_task(ireq);
/* all unaccelerated request types (non ssp or ncq) handled with
* substates
*/
if (!task && dev->dev_type == SAS_END_DEVICE) {
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
state = SCI_REQ_TASK_WAIT_TC_COMP;
} else if (task && task->task_proto == SAS_PROTOCOL_SMP) {
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
state = SCI_REQ_SMP_WAIT_RESP;
} else if (task && sas_protocol_ata(task->task_proto) &&
!task->ata_task.use_ncq) {
if (dev->sata_dev.class == ATA_DEV_ATAPI &&
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
task->ata_task.fis.command == ATA_CMD_PACKET) {
state = SCI_REQ_ATAPI_WAIT_H2D;
} else if (task->data_dir == DMA_NONE) {
state = SCI_REQ_STP_NON_DATA_WAIT_H2D;
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
} else if (task->ata_task.dma_xfer) {
state = SCI_REQ_STP_UDMA_WAIT_TC_COMP;
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
} else /* PIO */ {
state = SCI_REQ_STP_PIO_WAIT_H2D;
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
}
} else {
/* SSP or NCQ are fully accelerated, no substates */
return;
}
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
sci_change_state(sm, state);
}
static void sci_request_completed_state_enter(struct sci_base_state_machine *sm)
{
struct isci_request *ireq = container_of(sm, typeof(*ireq), sm);
struct isci_host *ihost = ireq->owning_controller;
/* Tell the SCI_USER that the IO request is complete */
if (!test_bit(IREQ_TMF, &ireq->flags))
isci_request_io_request_complete(ihost, ireq,
ireq->sci_status);
else
isci_task_request_complete(ihost, ireq, ireq->sci_status);
}
static void sci_request_aborting_state_enter(struct sci_base_state_machine *sm)
{
struct isci_request *ireq = container_of(sm, typeof(*ireq), sm);
/* Setting the abort bit in the Task Context is required by the silicon. */
ireq->tc->abort = 1;
}
static void sci_stp_request_started_non_data_await_h2d_completion_enter(struct sci_base_state_machine *sm)
{
struct isci_request *ireq = container_of(sm, typeof(*ireq), sm);
ireq->target_device->working_request = ireq;
}
static void sci_stp_request_started_pio_await_h2d_completion_enter(struct sci_base_state_machine *sm)
{
struct isci_request *ireq = container_of(sm, typeof(*ireq), sm);
ireq->target_device->working_request = ireq;
}
static const struct sci_base_state sci_request_state_table[] = {
[SCI_REQ_INIT] = { },
[SCI_REQ_CONSTRUCTED] = { },
[SCI_REQ_STARTED] = {
.enter_state = sci_request_started_state_enter,
},
[SCI_REQ_STP_NON_DATA_WAIT_H2D] = {
.enter_state = sci_stp_request_started_non_data_await_h2d_completion_enter,
},
[SCI_REQ_STP_NON_DATA_WAIT_D2H] = { },
[SCI_REQ_STP_PIO_WAIT_H2D] = {
.enter_state = sci_stp_request_started_pio_await_h2d_completion_enter,
},
[SCI_REQ_STP_PIO_WAIT_FRAME] = { },
[SCI_REQ_STP_PIO_DATA_IN] = { },
[SCI_REQ_STP_PIO_DATA_OUT] = { },
[SCI_REQ_STP_UDMA_WAIT_TC_COMP] = { },
[SCI_REQ_STP_UDMA_WAIT_D2H] = { },
[SCI_REQ_TASK_WAIT_TC_COMP] = { },
[SCI_REQ_TASK_WAIT_TC_RESP] = { },
[SCI_REQ_SMP_WAIT_RESP] = { },
[SCI_REQ_SMP_WAIT_TC_COMP] = { },
[SCSI] isci: atapi support Based on original implementation from Jiangbi Liu and Maciej Trela. ATAPI transfers happen in two-to-three stages. The two stage atapi commands are those that include a dma data transfer. The data transfer portion of these operations is handled by the hardware packet-dma acceleration. The three-stage commands do not have a data transfer and are handled without hardware assistance in raw frame mode. stage1: transmit host-to-device fis to notify the device of an incoming atapi cdb. Upon reception of the pio-setup-fis repost the task_context to perform the dma transfer of the cdb+data (go to stage3), or repost the task_context to transmit the cdb as a raw frame (go to stage 2). stage2: wait for hardware notification of the cdb transmission and then go to stage 3. stage3: wait for the arrival of the terminating device-to-host fis and terminate the command. To keep the implementation simple we only support ATAPI packet-dma protocol (for commands with data) to avoid needing to handle the data transfer manually (like we do for SATA-PIO). This may affect compatibility for a small number of devices (see ATA_HORKAGE_ATAPI_MOD16_DMA). If the data-transfer underruns, or encounters an error the device-to-host fis is expected to arrive in the unsolicited frame queue to pass to libata for disposition. However, in the DONE_UNEXP_FIS (data underrun) case it appears we need to craft a response. In the DONE_REG_ERR case we do receive the UF and propagate it to libsas. Signed-off-by: Maciej Trela <maciej.trela@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2011-10-01 09:52:19 +08:00
[SCI_REQ_ATAPI_WAIT_H2D] = { },
[SCI_REQ_ATAPI_WAIT_PIO_SETUP] = { },
[SCI_REQ_ATAPI_WAIT_D2H] = { },
[SCI_REQ_ATAPI_WAIT_TC_COMP] = { },
[SCI_REQ_COMPLETED] = {
.enter_state = sci_request_completed_state_enter,
},
[SCI_REQ_ABORTING] = {
.enter_state = sci_request_aborting_state_enter,
},
[SCI_REQ_FINAL] = { },
};
static void
sci_general_request_construct(struct isci_host *ihost,
struct isci_remote_device *idev,
struct isci_request *ireq)
{
sci_init_sm(&ireq->sm, sci_request_state_table, SCI_REQ_INIT);
ireq->target_device = idev;
ireq->protocol = SAS_PROTOCOL_NONE;
ireq->saved_rx_frame_index = SCU_INVALID_FRAME_INDEX;
ireq->sci_status = SCI_SUCCESS;
ireq->scu_status = 0;
ireq->post_context = 0xFFFFFFFF;
}
static enum sci_status
sci_io_request_construct(struct isci_host *ihost,
struct isci_remote_device *idev,
struct isci_request *ireq)
{
struct domain_device *dev = idev->domain_dev;
enum sci_status status = SCI_SUCCESS;
/* Build the common part of the request */
sci_general_request_construct(ihost, idev, ireq);
if (idev->rnc.remote_node_index == SCIC_SDS_REMOTE_NODE_CONTEXT_INVALID_INDEX)
return SCI_FAILURE_INVALID_REMOTE_DEVICE;
if (dev->dev_type == SAS_END_DEVICE)
/* pass */;
else if (dev_is_sata(dev))
memset(&ireq->stp.cmd, 0, sizeof(ireq->stp.cmd));
else if (dev_is_expander(dev->dev_type))
/* pass */;
else
return SCI_FAILURE_UNSUPPORTED_PROTOCOL;
memset(ireq->tc, 0, offsetof(struct scu_task_context, sgl_pair_ab));
return status;
}
enum sci_status sci_task_request_construct(struct isci_host *ihost,
struct isci_remote_device *idev,
u16 io_tag, struct isci_request *ireq)
{
struct domain_device *dev = idev->domain_dev;
enum sci_status status = SCI_SUCCESS;
/* Build the common part of the request */
sci_general_request_construct(ihost, idev, ireq);
if (dev->dev_type == SAS_END_DEVICE || dev_is_sata(dev)) {
set_bit(IREQ_TMF, &ireq->flags);
memset(ireq->tc, 0, sizeof(struct scu_task_context));
/* Set the protocol indicator. */
if (dev_is_sata(dev))
ireq->protocol = SAS_PROTOCOL_STP;
else
ireq->protocol = SAS_PROTOCOL_SSP;
} else
status = SCI_FAILURE_UNSUPPORTED_PROTOCOL;
return status;
}
static enum sci_status isci_request_ssp_request_construct(
struct isci_request *request)
{
enum sci_status status;
dev_dbg(&request->isci_host->pdev->dev,
"%s: request = %p\n",
__func__,
request);
status = sci_io_request_construct_basic_ssp(request);
return status;
}
static enum sci_status isci_request_stp_request_construct(struct isci_request *ireq)
{
struct sas_task *task = isci_request_access_task(ireq);
struct host_to_dev_fis *fis = &ireq->stp.cmd;
struct ata_queued_cmd *qc = task->uldd_task;
enum sci_status status;
dev_dbg(&ireq->isci_host->pdev->dev,
"%s: ireq = %p\n",
__func__,
ireq);
memcpy(fis, &task->ata_task.fis, sizeof(struct host_to_dev_fis));
if (!task->ata_task.device_control_reg_update)
fis->flags |= 0x80;
fis->flags &= 0xF0;
status = sci_io_request_construct_basic_sata(ireq);
if (qc && (qc->tf.command == ATA_CMD_FPDMA_WRITE ||
qc->tf.command == ATA_CMD_FPDMA_READ ||
qc->tf.command == ATA_CMD_FPDMA_RECV ||
qc->tf.command == ATA_CMD_FPDMA_SEND ||
qc->tf.command == ATA_CMD_NCQ_NON_DATA)) {
fis->sector_count = qc->tag << 3;
ireq->tc->type.stp.ncq_tag = qc->tag;
}
return status;
}
static enum sci_status
sci_io_request_construct_smp(struct device *dev,
struct isci_request *ireq,
struct sas_task *task)
{
struct scatterlist *sg = &task->smp_task.smp_req;
struct isci_remote_device *idev;
struct scu_task_context *task_context;
struct isci_port *iport;
struct smp_req *smp_req;
void *kaddr;
u8 req_len;
u32 cmd;
kaddr = kmap_atomic(sg_page(sg));
smp_req = kaddr + sg->offset;
/*
* Look at the SMP requests' header fields; for certain SAS 1.x SMP
* functions under SAS 2.0, a zero request length really indicates
* a non-zero default length.
*/
if (smp_req->req_len == 0) {
switch (smp_req->func) {
case SMP_DISCOVER:
case SMP_REPORT_PHY_ERR_LOG:
case SMP_REPORT_PHY_SATA:
case SMP_REPORT_ROUTE_INFO:
smp_req->req_len = 2;
break;
case SMP_CONF_ROUTE_INFO:
case SMP_PHY_CONTROL:
case SMP_PHY_TEST_FUNCTION:
smp_req->req_len = 9;
break;
/* Default - zero is a valid default for 2.0. */
}
}
req_len = smp_req->req_len;
sci_swab32_cpy(smp_req, smp_req, sg->length / sizeof(u32));
cmd = *(u32 *) smp_req;
kunmap_atomic(kaddr);
if (!dma_map_sg(dev, sg, 1, DMA_TO_DEVICE))
return SCI_FAILURE;
ireq->protocol = SAS_PROTOCOL_SMP;
/* byte swap the smp request. */
task_context = ireq->tc;
idev = ireq->target_device;
iport = idev->owning_port;
/*
* Fill in the TC with its required data
* 00h
*/
task_context->priority = 0;
task_context->initiator_request = 1;
task_context->connection_rate = idev->connection_rate;
task_context->protocol_engine_index = ISCI_PEG;
task_context->logical_port_index = iport->physical_port_index;
task_context->protocol_type = SCU_TASK_CONTEXT_PROTOCOL_SMP;
task_context->abort = 0;
task_context->valid = SCU_TASK_CONTEXT_VALID;
task_context->context_type = SCU_TASK_CONTEXT_TYPE;
/* 04h */
task_context->remote_node_index = idev->rnc.remote_node_index;
task_context->command_code = 0;
task_context->task_type = SCU_TASK_TYPE_SMP_REQUEST;
/* 08h */
task_context->link_layer_control = 0;
task_context->do_not_dma_ssp_good_response = 1;
task_context->strict_ordering = 0;
task_context->control_frame = 1;
task_context->timeout_enable = 0;
task_context->block_guard_enable = 0;
/* 0ch */
task_context->address_modifier = 0;
/* 10h */
task_context->ssp_command_iu_length = req_len;
/* 14h */
task_context->transfer_length_bytes = 0;
/*
* 18h ~ 30h, protocol specific
* since commandIU has been build by framework at this point, we just
* copy the frist DWord from command IU to this location. */
memcpy(&task_context->type.smp, &cmd, sizeof(u32));
/*
* 40h
* "For SMP you could program it to zero. We would prefer that way
* so that done code will be consistent." - Venki
*/
task_context->task_phase = 0;
ireq->post_context = (SCU_CONTEXT_COMMAND_REQUEST_TYPE_POST_TC |
(ISCI_PEG << SCU_CONTEXT_COMMAND_PROTOCOL_ENGINE_GROUP_SHIFT) |
(iport->physical_port_index <<
SCU_CONTEXT_COMMAND_LOGICAL_PORT_SHIFT) |
ISCI_TAG_TCI(ireq->io_tag));
/*
* Copy the physical address for the command buffer to the SCU Task
* Context command buffer should not contain command header.
*/
task_context->command_iu_upper = upper_32_bits(sg_dma_address(sg));
task_context->command_iu_lower = lower_32_bits(sg_dma_address(sg) + sizeof(u32));
/* SMP response comes as UF, so no need to set response IU address. */
task_context->response_iu_upper = 0;
task_context->response_iu_lower = 0;
sci_change_state(&ireq->sm, SCI_REQ_CONSTRUCTED);
return SCI_SUCCESS;
}
/*
* isci_smp_request_build() - This function builds the smp request.
* @ireq: This parameter points to the isci_request allocated in the
* request construct function.
*
* SCI_SUCCESS on successfull completion, or specific failure code.
*/
static enum sci_status isci_smp_request_build(struct isci_request *ireq)
{
struct sas_task *task = isci_request_access_task(ireq);
struct device *dev = &ireq->isci_host->pdev->dev;
enum sci_status status = SCI_FAILURE;
status = sci_io_request_construct_smp(dev, ireq, task);
if (status != SCI_SUCCESS)
dev_dbg(&ireq->isci_host->pdev->dev,
"%s: failed with status = %d\n",
__func__,
status);
return status;
}
/**
* isci_io_request_build() - This function builds the io request object.
* @ihost: This parameter specifies the ISCI host object
* @request: This parameter points to the isci_request object allocated in the
* request construct function.
* @sci_device: This parameter is the handle for the sci core's remote device
* object that is the destination for this request.
*
* SCI_SUCCESS on successfull completion, or specific failure code.
*/
static enum sci_status isci_io_request_build(struct isci_host *ihost,
struct isci_request *request,
struct isci_remote_device *idev)
{
enum sci_status status = SCI_SUCCESS;
struct sas_task *task = isci_request_access_task(request);
dev_dbg(&ihost->pdev->dev,
"%s: idev = 0x%p; request = %p, "
"num_scatter = %d\n",
__func__,
idev,
request,
task->num_scatter);
/* map the sgl addresses, if present.
* libata does the mapping for sata devices
* before we get the request.
*/
if (task->num_scatter &&
!sas_protocol_ata(task->task_proto) &&
!(SAS_PROTOCOL_SMP & task->task_proto)) {
request->num_sg_entries = dma_map_sg(
&ihost->pdev->dev,
task->scatter,
task->num_scatter,
task->data_dir
);
if (request->num_sg_entries == 0)
return SCI_FAILURE_INSUFFICIENT_RESOURCES;
}
status = sci_io_request_construct(ihost, idev, request);
if (status != SCI_SUCCESS) {
dev_dbg(&ihost->pdev->dev,
"%s: failed request construct\n",
__func__);
return SCI_FAILURE;
}
switch (task->task_proto) {
case SAS_PROTOCOL_SMP:
status = isci_smp_request_build(request);
break;
case SAS_PROTOCOL_SSP:
status = isci_request_ssp_request_construct(request);
break;
case SAS_PROTOCOL_SATA:
case SAS_PROTOCOL_STP:
case SAS_PROTOCOL_SATA | SAS_PROTOCOL_STP:
status = isci_request_stp_request_construct(request);
break;
default:
dev_dbg(&ihost->pdev->dev,
"%s: unknown protocol\n", __func__);
return SCI_FAILURE;
}
return SCI_SUCCESS;
}
static struct isci_request *isci_request_from_tag(struct isci_host *ihost, u16 tag)
{
struct isci_request *ireq;
ireq = ihost->reqs[ISCI_TAG_TCI(tag)];
ireq->io_tag = tag;
ireq->io_request_completion = NULL;
ireq->flags = 0;
ireq->num_sg_entries = 0;
return ireq;
}
static struct isci_request *isci_io_request_from_tag(struct isci_host *ihost,
struct sas_task *task,
u16 tag)
{
struct isci_request *ireq;
ireq = isci_request_from_tag(ihost, tag);
ireq->ttype_ptr.io_task_ptr = task;
clear_bit(IREQ_TMF, &ireq->flags);
task->lldd_task = ireq;
return ireq;
}
struct isci_request *isci_tmf_request_from_tag(struct isci_host *ihost,
struct isci_tmf *isci_tmf,
u16 tag)
{
struct isci_request *ireq;
ireq = isci_request_from_tag(ihost, tag);
ireq->ttype_ptr.tmf_task_ptr = isci_tmf;
set_bit(IREQ_TMF, &ireq->flags);
return ireq;
}
int isci_request_execute(struct isci_host *ihost, struct isci_remote_device *idev,
struct sas_task *task, u16 tag)
{
enum sci_status status = SCI_FAILURE_UNSUPPORTED_PROTOCOL;
struct isci_request *ireq;
unsigned long flags;
int ret = 0;
/* do common allocation and init of request object. */
ireq = isci_io_request_from_tag(ihost, task, tag);
status = isci_io_request_build(ihost, ireq, idev);
if (status != SCI_SUCCESS) {
dev_dbg(&ihost->pdev->dev,
"%s: request_construct failed - status = 0x%x\n",
__func__,
status);
return status;
}
spin_lock_irqsave(&ihost->scic_lock, flags);
if (test_bit(IDEV_IO_NCQERROR, &idev->flags)) {
if (isci_task_is_ncq_recovery(task)) {
/* The device is in an NCQ recovery state. Issue the
* request on the task side. Note that it will
* complete on the I/O request side because the
* request was built that way (ie.
* ireq->is_task_management_request is false).
*/
status = sci_controller_start_task(ihost,
idev,
ireq);
} else {
status = SCI_FAILURE;
}
} else {
/* send the request, let the core assign the IO TAG. */
status = sci_controller_start_io(ihost, idev,
ireq);
}
if (status != SCI_SUCCESS &&
status != SCI_FAILURE_REMOTE_DEVICE_RESET_REQUIRED) {
dev_dbg(&ihost->pdev->dev,
"%s: failed request start (0x%x)\n",
__func__, status);
spin_unlock_irqrestore(&ihost->scic_lock, flags);
return status;
}
/* Either I/O started OK, or the core has signaled that
* the device needs a target reset.
*/
if (status != SCI_SUCCESS) {
/* The request did not really start in the
* hardware, so clear the request handle
* here so no terminations will be done.
*/
set_bit(IREQ_TERMINATED, &ireq->flags);
}
spin_unlock_irqrestore(&ihost->scic_lock, flags);
if (status ==
SCI_FAILURE_REMOTE_DEVICE_RESET_REQUIRED) {
/* Signal libsas that we need the SCSI error
* handler thread to work on this I/O and that
* we want a device reset.
*/
spin_lock_irqsave(&task->task_state_lock, flags);
task->task_state_flags |= SAS_TASK_NEED_DEV_RESET;
spin_unlock_irqrestore(&task->task_state_lock, flags);
/* Cause this task to be scheduled in the SCSI error
* handler thread.
*/
sas_task_abort(task);
/* Change the status, since we are holding
* the I/O until it is managed by the SCSI
* error handler.
*/
status = SCI_SUCCESS;
}
return ret;
}