freebsd-dev/sys/dev/isci/isci_task_request.c
Scott Long 3da2a91a57 In rare cases, a SATA drive can stop responding to commands and trigger a
reset device task request from the driver.  If the drive fails to respond
with a signature FIS, the driver would previously get into an endless retry
loop, stalling all I/O to the drive and keeping user processes stranded.
Instead, fail the i/o and invalidate the device if the task management
command times out.  This is controllable with the sysctl and tunable
hw.isci.fail_on_task_timeout
dev.isci.0.fail_on_task_timeout

The default for these is 1.

Reviewed by:	jimharris
Obtained from:	Netflix, Inc.
MFC after:	2 days
2014-06-30 01:01:54 +00:00

277 lines
9.5 KiB
C

/*-
* 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.
*
* 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 <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <dev/isci/isci.h>
#include <dev/isci/scil/scif_controller.h>
#include <dev/isci/scil/scif_user_callback.h>
/**
* @brief This user callback will inform the user that a task management
* request completed.
*
* @param[in] controller This parameter specifies the controller on
* which the task management request is completing.
* @param[in] remote_device This parameter specifies the remote device on
* which this task management request is completing.
* @param[in] task_request This parameter specifies the task management
* request that has completed.
* @param[in] completion_status This parameter specifies the results of
* the IO request operation. SCI_TASK_SUCCESS indicates
* successful completion.
*
* @return none
*/
void
scif_cb_task_request_complete(SCI_CONTROLLER_HANDLE_T controller,
SCI_REMOTE_DEVICE_HANDLE_T remote_device,
SCI_TASK_REQUEST_HANDLE_T task_request, SCI_TASK_STATUS completion_status)
{
scif_controller_complete_task(controller, remote_device, task_request);
isci_task_request_complete(controller, remote_device, task_request,
completion_status);
}
/**
* @brief This method returns the Logical Unit to be utilized for this
* task management request.
*
* @note The contents of the value returned from this callback are defined
* by the protocol standard (e.g. T10 SAS specification). Please
* refer to the transport task information unit description
* in the associated standard.
*
* @param[in] scif_user_task_request This parameter points to the user's
* task request object. It is a cookie that allows the user to
* provide the necessary information for this callback.
*
* @return This method returns the LUN associated with this request.
* @todo This should be U64?
*/
uint32_t
scif_cb_task_request_get_lun(void * scif_user_task_request)
{
/* Currently we are only doing hard resets, not LUN resets. So
* always returning 0 is OK here, since LUN doesn't matter for
* a hard device reset.
*/
return (0);
}
/**
* @brief This method returns the task management function to be utilized
* for this task request.
*
* @note The contents of the value returned from this callback are defined
* by the protocol standard (e.g. T10 SAS specification). Please
* refer to the transport task information unit description
* in the associated standard.
*
* @param[in] scif_user_task_request This parameter points to the user's
* task request object. It is a cookie that allows the user to
* provide the necessary information for this callback.
*
* @return This method returns an unsigned byte representing the task
* management function to be performed.
*/
uint8_t scif_cb_task_request_get_function(void * scif_user_task_request)
{
/* SCIL supports many types of task management functions, but this
* driver only uses HARD_RESET.
*/
return (SCI_SAS_HARD_RESET);
}
/**
* @brief This method returns the task management IO tag to be managed.
* Depending upon the task management function the value returned
* from this method may be ignored.
*
* @param[in] scif_user_task_request This parameter points to the user's
* task request object. It is a cookie that allows the user to
* provide the necessary information for this callback.
*
* @return This method returns an unsigned 16-bit word depicting the IO
* tag to be managed.
*/
uint16_t
scif_cb_task_request_get_io_tag_to_manage(void * scif_user_task_request)
{
return (0);
}
/**
* @brief This callback method asks the user to provide the virtual
* address of the response data buffer for the supplied IO request.
*
* @param[in] scif_user_task_request This parameter points to the user's
* task request object. It is a cookie that allows the user to
* provide the necessary information for this callback.
*
* @return This method returns the virtual address for the response data buffer
* associated with this IO request.
*/
void *
scif_cb_task_request_get_response_data_address(void * scif_user_task_request)
{
struct ISCI_TASK_REQUEST *task_request =
(struct ISCI_TASK_REQUEST *)scif_user_task_request;
return (&task_request->sense_data);
}
/**
* @brief This callback method asks the user to provide the length of the
* response data buffer for the supplied IO request.
*
* @param[in] scif_user_task_request This parameter points to the user's
* task request object. It is a cookie that allows the user to
* provide the necessary information for this callback.
*
* @return This method returns the length of the response buffer data
* associated with this IO request.
*/
uint32_t
scif_cb_task_request_get_response_data_length(void * scif_user_task_request)
{
return (sizeof(struct scsi_sense_data));
}
void
isci_task_request_complete(SCI_CONTROLLER_HANDLE_T scif_controller,
SCI_REMOTE_DEVICE_HANDLE_T remote_device,
SCI_TASK_REQUEST_HANDLE_T task_request, SCI_TASK_STATUS completion_status)
{
struct ISCI_TASK_REQUEST *isci_task_request =
(struct ISCI_TASK_REQUEST *)sci_object_get_association(task_request);
struct ISCI_CONTROLLER *isci_controller =
(struct ISCI_CONTROLLER *)sci_object_get_association(scif_controller);
struct ISCI_REMOTE_DEVICE *isci_remote_device =
(struct ISCI_REMOTE_DEVICE *)sci_object_get_association(remote_device);
struct ISCI_REMOTE_DEVICE *pending_remote_device;
BOOL retry_task = FALSE;
union ccb *ccb = isci_task_request->ccb;
isci_remote_device->is_resetting = FALSE;
switch ((int)completion_status) {
case SCI_TASK_SUCCESS:
case SCI_TASK_FAILURE_RESPONSE_VALID:
break;
case SCI_TASK_FAILURE_INVALID_STATE:
retry_task = TRUE;
isci_log_message(0, "ISCI",
"task failure (invalid state) - retrying\n");
break;
case SCI_TASK_FAILURE_INSUFFICIENT_RESOURCES:
retry_task = TRUE;
isci_log_message(0, "ISCI",
"task failure (insufficient resources) - retrying\n");
break;
case SCI_FAILURE_TIMEOUT:
if (isci_controller->fail_on_task_timeout) {
retry_task = FALSE;
isci_log_message(0, "ISCI",
"task timeout - not retrying\n");
scif_cb_domain_device_removed(isci_controller,
isci_remote_device->domain, isci_remote_device);
} else {
retry_task = TRUE;
isci_log_message(0, "ISCI",
"task timeout - retrying\n");
}
break;
case SCI_TASK_FAILURE:
case SCI_TASK_FAILURE_UNSUPPORTED_PROTOCOL:
case SCI_TASK_FAILURE_INVALID_TAG:
case SCI_TASK_FAILURE_CONTROLLER_SPECIFIC_ERR:
case SCI_TASK_FAILURE_TERMINATED:
case SCI_TASK_FAILURE_INVALID_PARAMETER_VALUE:
isci_log_message(0, "ISCI",
"unhandled task completion code 0x%x\n", completion_status);
break;
default:
isci_log_message(0, "ISCI",
"unhandled task completion code 0x%x\n", completion_status);
break;
}
if (isci_controller->is_frozen == TRUE) {
isci_controller->is_frozen = FALSE;
xpt_release_simq(isci_controller->sim, TRUE);
}
sci_pool_put(isci_controller->request_pool,
(struct ISCI_REQUEST *)isci_task_request);
/* Make sure we release the device queue, since it may have been frozen
* if someone tried to start an I/O while the task was in progress.
*/
isci_remote_device_release_device_queue(isci_remote_device);
if (retry_task == TRUE)
isci_remote_device_reset(isci_remote_device, ccb);
else {
pending_remote_device = sci_fast_list_remove_head(
&isci_controller->pending_device_reset_list);
if (pending_remote_device != NULL) {
/* Any resets that were triggered from an XPT_RESET_DEV
* CCB are never put in the pending list if the request
* pool is empty - they are given back to CAM to be
* requeued. So we will alawys pass NULL here,
* denoting that there is no CCB associated with the
* device reset.
*/
isci_remote_device_reset(pending_remote_device, NULL);
} else if (ccb != NULL) {
/* There was a CCB associated with this reset, so mark
* it complete and return it to CAM.
*/
ccb->ccb_h.status &= ~CAM_STATUS_MASK;
ccb->ccb_h.status |= CAM_REQ_CMP;
xpt_done(ccb);
}
}
}