linux-brain/drivers/scsi/qedf/qedf_io.c

2630 lines
72 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0-only
/*
* QLogic FCoE Offload Driver
* Copyright (c) 2016-2018 Cavium Inc.
*/
#include <linux/spinlock.h>
#include <linux/vmalloc.h>
#include "qedf.h"
#include <scsi/scsi_tcq.h>
void qedf_cmd_timer_set(struct qedf_ctx *qedf, struct qedf_ioreq *io_req,
unsigned int timer_msec)
{
queue_delayed_work(qedf->timer_work_queue, &io_req->timeout_work,
msecs_to_jiffies(timer_msec));
}
static void qedf_cmd_timeout(struct work_struct *work)
{
struct qedf_ioreq *io_req =
container_of(work, struct qedf_ioreq, timeout_work.work);
struct qedf_ctx *qedf;
struct qedf_rport *fcport;
u8 op = 0;
if (io_req == NULL) {
QEDF_INFO(NULL, QEDF_LOG_IO, "io_req is NULL.\n");
return;
}
fcport = io_req->fcport;
if (io_req->fcport == NULL) {
QEDF_INFO(NULL, QEDF_LOG_IO, "fcport is NULL.\n");
return;
}
qedf = fcport->qedf;
switch (io_req->cmd_type) {
case QEDF_ABTS:
if (qedf == NULL) {
QEDF_INFO(NULL, QEDF_LOG_IO,
"qedf is NULL for ABTS xid=0x%x.\n",
io_req->xid);
return;
}
QEDF_ERR((&qedf->dbg_ctx), "ABTS timeout, xid=0x%x.\n",
io_req->xid);
/* Cleanup timed out ABTS */
qedf_initiate_cleanup(io_req, true);
complete(&io_req->abts_done);
/*
* Need to call kref_put for reference taken when initiate_abts
* was called since abts_compl won't be called now that we've
* cleaned up the task.
*/
kref_put(&io_req->refcount, qedf_release_cmd);
/* Clear in abort bit now that we're done with the command */
clear_bit(QEDF_CMD_IN_ABORT, &io_req->flags);
/*
* Now that the original I/O and the ABTS are complete see
* if we need to reconnect to the target.
*/
qedf_restart_rport(fcport);
break;
case QEDF_ELS:
if (!qedf) {
QEDF_INFO(NULL, QEDF_LOG_IO,
"qedf is NULL for ELS xid=0x%x.\n",
io_req->xid);
return;
}
/* ELS request no longer outstanding since it timed out */
clear_bit(QEDF_CMD_OUTSTANDING, &io_req->flags);
kref_get(&io_req->refcount);
/*
* Don't attempt to clean an ELS timeout as any subseqeunt
* ABTS or cleanup requests just hang. For now just free
* the resources of the original I/O and the RRQ
*/
QEDF_ERR(&(qedf->dbg_ctx), "ELS timeout, xid=0x%x.\n",
io_req->xid);
io_req->event = QEDF_IOREQ_EV_ELS_TMO;
/* Call callback function to complete command */
if (io_req->cb_func && io_req->cb_arg) {
op = io_req->cb_arg->op;
io_req->cb_func(io_req->cb_arg);
io_req->cb_arg = NULL;
}
qedf_initiate_cleanup(io_req, true);
kref_put(&io_req->refcount, qedf_release_cmd);
break;
case QEDF_SEQ_CLEANUP:
QEDF_ERR(&(qedf->dbg_ctx), "Sequence cleanup timeout, "
"xid=0x%x.\n", io_req->xid);
qedf_initiate_cleanup(io_req, true);
io_req->event = QEDF_IOREQ_EV_ELS_TMO;
qedf_process_seq_cleanup_compl(qedf, NULL, io_req);
break;
default:
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO,
"Hit default case, xid=0x%x.\n", io_req->xid);
break;
}
}
void qedf_cmd_mgr_free(struct qedf_cmd_mgr *cmgr)
{
struct io_bdt *bdt_info;
struct qedf_ctx *qedf = cmgr->qedf;
size_t bd_tbl_sz;
u16 min_xid = 0;
u16 max_xid = (FCOE_PARAMS_NUM_TASKS - 1);
int num_ios;
int i;
struct qedf_ioreq *io_req;
num_ios = max_xid - min_xid + 1;
/* Free fcoe_bdt_ctx structures */
if (!cmgr->io_bdt_pool) {
QEDF_ERR(&qedf->dbg_ctx, "io_bdt_pool is NULL.\n");
goto free_cmd_pool;
}
bd_tbl_sz = QEDF_MAX_BDS_PER_CMD * sizeof(struct scsi_sge);
for (i = 0; i < num_ios; i++) {
bdt_info = cmgr->io_bdt_pool[i];
if (bdt_info->bd_tbl) {
dma_free_coherent(&qedf->pdev->dev, bd_tbl_sz,
bdt_info->bd_tbl, bdt_info->bd_tbl_dma);
bdt_info->bd_tbl = NULL;
}
}
/* Destroy io_bdt pool */
for (i = 0; i < num_ios; i++) {
kfree(cmgr->io_bdt_pool[i]);
cmgr->io_bdt_pool[i] = NULL;
}
kfree(cmgr->io_bdt_pool);
cmgr->io_bdt_pool = NULL;
free_cmd_pool:
for (i = 0; i < num_ios; i++) {
io_req = &cmgr->cmds[i];
kfree(io_req->sgl_task_params);
kfree(io_req->task_params);
/* Make sure we free per command sense buffer */
if (io_req->sense_buffer)
dma_free_coherent(&qedf->pdev->dev,
QEDF_SCSI_SENSE_BUFFERSIZE, io_req->sense_buffer,
io_req->sense_buffer_dma);
cancel_delayed_work_sync(&io_req->rrq_work);
}
/* Free command manager itself */
vfree(cmgr);
}
static void qedf_handle_rrq(struct work_struct *work)
{
struct qedf_ioreq *io_req =
container_of(work, struct qedf_ioreq, rrq_work.work);
scsi: qedf: Cleanup rrq_work after QEDF_CMD_OUTSTANDING is cleared Here is the relevant logs for the problem we are solving: qedf_flush_active_ios:1707]:3: Flush active i/o's num=0x17 fcport=0xffff948168fbcc80 port_id=0x550200 scsi_id=0. qedf_flush_active_ios:1708]:3: Locking flush mutex. qedf_flush_active_ios:1758]:3: Not outstanding, xid=0xaaf, cmd_type=3 refcount=1. qedf_flush_active_ios:1896]:3: Flushed 0x16 I/Os, active=0x1. qedf_flush_active_ios:1901]:3: Flushed 0x16 I/Os, active=0x1 cnt=60. qedf_send_rrq:295]:3: Sending RRQ orig io = ffffb48b8f7d7158, orig_xid = 0xaaf qedf_initiate_els:37]:3: Sending ELS qedf_initiate_els:68]:3: initiate_els els_req = 0xffffb48b8f6d3098 cb_arg = ffff948fd5e4de80 xid = 4c6 qedf_init_mp_req:2172]:3: Entered. qedf_init_mp_task:727]:3: Initializing MP task for cmd_type=4 qedf_initiate_els:134]:3: Ringing doorbell for ELS req qedf_flush_active_ios:1901]:3: Flushed 0x16 I/Os, active=0x2 cnt=20. qedf_cmd_timeout:96]:3: ELS timeout, xid=0x4c6. qedf_rrq_compl:186]:3: Entered. qedf_rrq_compl:204]:3: rrq_compl: orig io = ffffb48b8f7d7158, orig xid = 0xaaf, rrq_xid = 0x4c6, refcount=1 qedf_flush_active_ios:1935]:3: Unlocking flush mutex. qedf_upload_connection:1579]:3: Uploading connection port_id=550200. We found an ABTS command for which CMD_OUTSTANDING was cleared (line 3). For this command, delayed send_rrq was queued, but would take 10 secs to execute. Adding capability to detect that (based on io_req->state that is being introduced), and attempt to cancel rrq_work. If we succeed, we drop the reference and free the io_req. If we cannot, then the els will get sent out and we will wait for 10 secs for it to complete. Signed-off-by: Shyam Sundar <ssundar@marvell.com> Signed-off-by: Saurav Kashyap <skashyap@marvell.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-03-26 16:38:55 +09:00
atomic_set(&io_req->state, QEDFC_CMD_ST_RRQ_ACTIVE);
qedf_send_rrq(io_req);
}
struct qedf_cmd_mgr *qedf_cmd_mgr_alloc(struct qedf_ctx *qedf)
{
struct qedf_cmd_mgr *cmgr;
struct io_bdt *bdt_info;
struct qedf_ioreq *io_req;
u16 xid;
int i;
int num_ios;
u16 min_xid = 0;
u16 max_xid = (FCOE_PARAMS_NUM_TASKS - 1);
/* Make sure num_queues is already set before calling this function */
if (!qedf->num_queues) {
QEDF_ERR(&(qedf->dbg_ctx), "num_queues is not set.\n");
return NULL;
}
if (max_xid <= min_xid || max_xid == FC_XID_UNKNOWN) {
QEDF_WARN(&(qedf->dbg_ctx), "Invalid min_xid 0x%x and "
"max_xid 0x%x.\n", min_xid, max_xid);
return NULL;
}
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_DISC, "min xid 0x%x, max xid "
"0x%x.\n", min_xid, max_xid);
num_ios = max_xid - min_xid + 1;
cmgr = vzalloc(sizeof(struct qedf_cmd_mgr));
if (!cmgr) {
QEDF_WARN(&(qedf->dbg_ctx), "Failed to alloc cmd mgr.\n");
return NULL;
}
cmgr->qedf = qedf;
spin_lock_init(&cmgr->lock);
/*
* Initialize I/O request fields.
*/
xid = 0;
for (i = 0; i < num_ios; i++) {
io_req = &cmgr->cmds[i];
INIT_DELAYED_WORK(&io_req->timeout_work, qedf_cmd_timeout);
io_req->xid = xid++;
INIT_DELAYED_WORK(&io_req->rrq_work, qedf_handle_rrq);
/* Allocate DMA memory to hold sense buffer */
io_req->sense_buffer = dma_alloc_coherent(&qedf->pdev->dev,
QEDF_SCSI_SENSE_BUFFERSIZE, &io_req->sense_buffer_dma,
GFP_KERNEL);
if (!io_req->sense_buffer) {
QEDF_ERR(&qedf->dbg_ctx,
"Failed to alloc sense buffer.\n");
goto mem_err;
}
/* Allocate task parameters to pass to f/w init funcions */
io_req->task_params = kzalloc(sizeof(*io_req->task_params),
GFP_KERNEL);
if (!io_req->task_params) {
QEDF_ERR(&(qedf->dbg_ctx),
"Failed to allocate task_params for xid=0x%x\n",
i);
goto mem_err;
}
/*
* Allocate scatter/gather list info to pass to f/w init
* functions.
*/
io_req->sgl_task_params = kzalloc(
sizeof(struct scsi_sgl_task_params), GFP_KERNEL);
if (!io_req->sgl_task_params) {
QEDF_ERR(&(qedf->dbg_ctx),
"Failed to allocate sgl_task_params for xid=0x%x\n",
i);
goto mem_err;
}
}
/* Allocate pool of io_bdts - one for each qedf_ioreq */
cmgr->io_bdt_pool = kmalloc_array(num_ios, sizeof(struct io_bdt *),
GFP_KERNEL);
if (!cmgr->io_bdt_pool) {
QEDF_WARN(&(qedf->dbg_ctx), "Failed to alloc io_bdt_pool.\n");
goto mem_err;
}
for (i = 0; i < num_ios; i++) {
cmgr->io_bdt_pool[i] = kmalloc(sizeof(struct io_bdt),
GFP_KERNEL);
if (!cmgr->io_bdt_pool[i]) {
QEDF_WARN(&(qedf->dbg_ctx),
"Failed to alloc io_bdt_pool[%d].\n", i);
goto mem_err;
}
}
for (i = 0; i < num_ios; i++) {
bdt_info = cmgr->io_bdt_pool[i];
bdt_info->bd_tbl = dma_alloc_coherent(&qedf->pdev->dev,
QEDF_MAX_BDS_PER_CMD * sizeof(struct scsi_sge),
&bdt_info->bd_tbl_dma, GFP_KERNEL);
if (!bdt_info->bd_tbl) {
QEDF_WARN(&(qedf->dbg_ctx),
"Failed to alloc bdt_tbl[%d].\n", i);
goto mem_err;
}
}
atomic_set(&cmgr->free_list_cnt, num_ios);
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_IO,
"cmgr->free_list_cnt=%d.\n",
atomic_read(&cmgr->free_list_cnt));
return cmgr;
mem_err:
qedf_cmd_mgr_free(cmgr);
return NULL;
}
struct qedf_ioreq *qedf_alloc_cmd(struct qedf_rport *fcport, u8 cmd_type)
{
struct qedf_ctx *qedf = fcport->qedf;
struct qedf_cmd_mgr *cmd_mgr = qedf->cmd_mgr;
struct qedf_ioreq *io_req = NULL;
struct io_bdt *bd_tbl;
u16 xid;
uint32_t free_sqes;
int i;
unsigned long flags;
free_sqes = atomic_read(&fcport->free_sqes);
if (!free_sqes) {
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_IO,
"Returning NULL, free_sqes=%d.\n ",
free_sqes);
goto out_failed;
}
/* Limit the number of outstanding R/W tasks */
if ((atomic_read(&fcport->num_active_ios) >=
NUM_RW_TASKS_PER_CONNECTION)) {
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_IO,
"Returning NULL, num_active_ios=%d.\n",
atomic_read(&fcport->num_active_ios));
goto out_failed;
}
/* Limit global TIDs certain tasks */
if (atomic_read(&cmd_mgr->free_list_cnt) <= GBL_RSVD_TASKS) {
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_IO,
"Returning NULL, free_list_cnt=%d.\n",
atomic_read(&cmd_mgr->free_list_cnt));
goto out_failed;
}
spin_lock_irqsave(&cmd_mgr->lock, flags);
for (i = 0; i < FCOE_PARAMS_NUM_TASKS; i++) {
io_req = &cmd_mgr->cmds[cmd_mgr->idx];
cmd_mgr->idx++;
if (cmd_mgr->idx == FCOE_PARAMS_NUM_TASKS)
cmd_mgr->idx = 0;
/* Check to make sure command was previously freed */
if (!io_req->alloc)
break;
}
if (i == FCOE_PARAMS_NUM_TASKS) {
spin_unlock_irqrestore(&cmd_mgr->lock, flags);
goto out_failed;
}
if (test_bit(QEDF_CMD_DIRTY, &io_req->flags))
QEDF_ERR(&qedf->dbg_ctx,
"io_req found to be dirty ox_id = 0x%x.\n",
io_req->xid);
/* Clear any flags now that we've reallocated the xid */
io_req->flags = 0;
io_req->alloc = 1;
spin_unlock_irqrestore(&cmd_mgr->lock, flags);
atomic_inc(&fcport->num_active_ios);
atomic_dec(&fcport->free_sqes);
xid = io_req->xid;
atomic_dec(&cmd_mgr->free_list_cnt);
io_req->cmd_mgr = cmd_mgr;
io_req->fcport = fcport;
/* Clear any stale sc_cmd back pointer */
io_req->sc_cmd = NULL;
io_req->lun = -1;
/* Hold the io_req against deletion */
scsi: qedf: Cleanup rrq_work after QEDF_CMD_OUTSTANDING is cleared Here is the relevant logs for the problem we are solving: qedf_flush_active_ios:1707]:3: Flush active i/o's num=0x17 fcport=0xffff948168fbcc80 port_id=0x550200 scsi_id=0. qedf_flush_active_ios:1708]:3: Locking flush mutex. qedf_flush_active_ios:1758]:3: Not outstanding, xid=0xaaf, cmd_type=3 refcount=1. qedf_flush_active_ios:1896]:3: Flushed 0x16 I/Os, active=0x1. qedf_flush_active_ios:1901]:3: Flushed 0x16 I/Os, active=0x1 cnt=60. qedf_send_rrq:295]:3: Sending RRQ orig io = ffffb48b8f7d7158, orig_xid = 0xaaf qedf_initiate_els:37]:3: Sending ELS qedf_initiate_els:68]:3: initiate_els els_req = 0xffffb48b8f6d3098 cb_arg = ffff948fd5e4de80 xid = 4c6 qedf_init_mp_req:2172]:3: Entered. qedf_init_mp_task:727]:3: Initializing MP task for cmd_type=4 qedf_initiate_els:134]:3: Ringing doorbell for ELS req qedf_flush_active_ios:1901]:3: Flushed 0x16 I/Os, active=0x2 cnt=20. qedf_cmd_timeout:96]:3: ELS timeout, xid=0x4c6. qedf_rrq_compl:186]:3: Entered. qedf_rrq_compl:204]:3: rrq_compl: orig io = ffffb48b8f7d7158, orig xid = 0xaaf, rrq_xid = 0x4c6, refcount=1 qedf_flush_active_ios:1935]:3: Unlocking flush mutex. qedf_upload_connection:1579]:3: Uploading connection port_id=550200. We found an ABTS command for which CMD_OUTSTANDING was cleared (line 3). For this command, delayed send_rrq was queued, but would take 10 secs to execute. Adding capability to detect that (based on io_req->state that is being introduced), and attempt to cancel rrq_work. If we succeed, we drop the reference and free the io_req. If we cannot, then the els will get sent out and we will wait for 10 secs for it to complete. Signed-off-by: Shyam Sundar <ssundar@marvell.com> Signed-off-by: Saurav Kashyap <skashyap@marvell.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-03-26 16:38:55 +09:00
kref_init(&io_req->refcount); /* ID: 001 */
atomic_set(&io_req->state, QEDFC_CMD_ST_IO_ACTIVE);
/* Bind io_bdt for this io_req */
/* Have a static link between io_req and io_bdt_pool */
bd_tbl = io_req->bd_tbl = cmd_mgr->io_bdt_pool[xid];
if (bd_tbl == NULL) {
QEDF_ERR(&(qedf->dbg_ctx), "bd_tbl is NULL, xid=%x.\n", xid);
kref_put(&io_req->refcount, qedf_release_cmd);
goto out_failed;
}
bd_tbl->io_req = io_req;
io_req->cmd_type = cmd_type;
io_req->tm_flags = 0;
/* Reset sequence offset data */
io_req->rx_buf_off = 0;
io_req->tx_buf_off = 0;
io_req->rx_id = 0xffff; /* No OX_ID */
return io_req;
out_failed:
/* Record failure for stats and return NULL to caller */
qedf->alloc_failures++;
return NULL;
}
static void qedf_free_mp_resc(struct qedf_ioreq *io_req)
{
struct qedf_mp_req *mp_req = &(io_req->mp_req);
struct qedf_ctx *qedf = io_req->fcport->qedf;
uint64_t sz = sizeof(struct scsi_sge);
/* clear tm flags */
if (mp_req->mp_req_bd) {
dma_free_coherent(&qedf->pdev->dev, sz,
mp_req->mp_req_bd, mp_req->mp_req_bd_dma);
mp_req->mp_req_bd = NULL;
}
if (mp_req->mp_resp_bd) {
dma_free_coherent(&qedf->pdev->dev, sz,
mp_req->mp_resp_bd, mp_req->mp_resp_bd_dma);
mp_req->mp_resp_bd = NULL;
}
if (mp_req->req_buf) {
dma_free_coherent(&qedf->pdev->dev, QEDF_PAGE_SIZE,
mp_req->req_buf, mp_req->req_buf_dma);
mp_req->req_buf = NULL;
}
if (mp_req->resp_buf) {
dma_free_coherent(&qedf->pdev->dev, QEDF_PAGE_SIZE,
mp_req->resp_buf, mp_req->resp_buf_dma);
mp_req->resp_buf = NULL;
}
}
void qedf_release_cmd(struct kref *ref)
{
struct qedf_ioreq *io_req =
container_of(ref, struct qedf_ioreq, refcount);
struct qedf_cmd_mgr *cmd_mgr = io_req->cmd_mgr;
struct qedf_rport *fcport = io_req->fcport;
unsigned long flags;
if (io_req->cmd_type == QEDF_SCSI_CMD) {
QEDF_WARN(&fcport->qedf->dbg_ctx,
"Cmd released called without scsi_done called, io_req %p xid=0x%x.\n",
io_req, io_req->xid);
WARN_ON(io_req->sc_cmd);
}
if (io_req->cmd_type == QEDF_ELS ||
io_req->cmd_type == QEDF_TASK_MGMT_CMD)
qedf_free_mp_resc(io_req);
atomic_inc(&cmd_mgr->free_list_cnt);
atomic_dec(&fcport->num_active_ios);
scsi: qedf: Cleanup rrq_work after QEDF_CMD_OUTSTANDING is cleared Here is the relevant logs for the problem we are solving: qedf_flush_active_ios:1707]:3: Flush active i/o's num=0x17 fcport=0xffff948168fbcc80 port_id=0x550200 scsi_id=0. qedf_flush_active_ios:1708]:3: Locking flush mutex. qedf_flush_active_ios:1758]:3: Not outstanding, xid=0xaaf, cmd_type=3 refcount=1. qedf_flush_active_ios:1896]:3: Flushed 0x16 I/Os, active=0x1. qedf_flush_active_ios:1901]:3: Flushed 0x16 I/Os, active=0x1 cnt=60. qedf_send_rrq:295]:3: Sending RRQ orig io = ffffb48b8f7d7158, orig_xid = 0xaaf qedf_initiate_els:37]:3: Sending ELS qedf_initiate_els:68]:3: initiate_els els_req = 0xffffb48b8f6d3098 cb_arg = ffff948fd5e4de80 xid = 4c6 qedf_init_mp_req:2172]:3: Entered. qedf_init_mp_task:727]:3: Initializing MP task for cmd_type=4 qedf_initiate_els:134]:3: Ringing doorbell for ELS req qedf_flush_active_ios:1901]:3: Flushed 0x16 I/Os, active=0x2 cnt=20. qedf_cmd_timeout:96]:3: ELS timeout, xid=0x4c6. qedf_rrq_compl:186]:3: Entered. qedf_rrq_compl:204]:3: rrq_compl: orig io = ffffb48b8f7d7158, orig xid = 0xaaf, rrq_xid = 0x4c6, refcount=1 qedf_flush_active_ios:1935]:3: Unlocking flush mutex. qedf_upload_connection:1579]:3: Uploading connection port_id=550200. We found an ABTS command for which CMD_OUTSTANDING was cleared (line 3). For this command, delayed send_rrq was queued, but would take 10 secs to execute. Adding capability to detect that (based on io_req->state that is being introduced), and attempt to cancel rrq_work. If we succeed, we drop the reference and free the io_req. If we cannot, then the els will get sent out and we will wait for 10 secs for it to complete. Signed-off-by: Shyam Sundar <ssundar@marvell.com> Signed-off-by: Saurav Kashyap <skashyap@marvell.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-03-26 16:38:55 +09:00
atomic_set(&io_req->state, QEDF_CMD_ST_INACTIVE);
if (atomic_read(&fcport->num_active_ios) < 0) {
QEDF_WARN(&(fcport->qedf->dbg_ctx), "active_ios < 0.\n");
WARN_ON(1);
}
/* Increment task retry identifier now that the request is released */
io_req->task_retry_identifier++;
io_req->fcport = NULL;
clear_bit(QEDF_CMD_DIRTY, &io_req->flags);
io_req->cpu = 0;
spin_lock_irqsave(&cmd_mgr->lock, flags);
io_req->fcport = NULL;
io_req->alloc = 0;
spin_unlock_irqrestore(&cmd_mgr->lock, flags);
}
static int qedf_map_sg(struct qedf_ioreq *io_req)
{
struct scsi_cmnd *sc = io_req->sc_cmd;
struct Scsi_Host *host = sc->device->host;
struct fc_lport *lport = shost_priv(host);
struct qedf_ctx *qedf = lport_priv(lport);
struct scsi_sge *bd = io_req->bd_tbl->bd_tbl;
struct scatterlist *sg;
int byte_count = 0;
int sg_count = 0;
int bd_count = 0;
u32 sg_len;
u64 addr, end_addr;
int i = 0;
sg_count = dma_map_sg(&qedf->pdev->dev, scsi_sglist(sc),
scsi_sg_count(sc), sc->sc_data_direction);
sg = scsi_sglist(sc);
io_req->sge_type = QEDF_IOREQ_UNKNOWN_SGE;
if (sg_count <= 8 || io_req->io_req_flags == QEDF_READ)
io_req->sge_type = QEDF_IOREQ_FAST_SGE;
scsi_for_each_sg(sc, sg, sg_count, i) {
sg_len = (u32)sg_dma_len(sg);
addr = (u64)sg_dma_address(sg);
end_addr = (u64)(addr + sg_len);
/*
* Intermediate s/g element so check if start and end address
* is page aligned. Only required for writes and only if the
* number of scatter/gather elements is 8 or more.
*/
if (io_req->sge_type == QEDF_IOREQ_UNKNOWN_SGE && (i) &&
(i != (sg_count - 1)) && sg_len < QEDF_PAGE_SIZE)
io_req->sge_type = QEDF_IOREQ_SLOW_SGE;
bd[bd_count].sge_addr.lo = cpu_to_le32(U64_LO(addr));
bd[bd_count].sge_addr.hi = cpu_to_le32(U64_HI(addr));
bd[bd_count].sge_len = cpu_to_le32(sg_len);
bd_count++;
byte_count += sg_len;
}
/* To catch a case where FAST and SLOW nothing is set, set FAST */
if (io_req->sge_type == QEDF_IOREQ_UNKNOWN_SGE)
io_req->sge_type = QEDF_IOREQ_FAST_SGE;
if (byte_count != scsi_bufflen(sc))
QEDF_ERR(&(qedf->dbg_ctx), "byte_count = %d != "
"scsi_bufflen = %d, task_id = 0x%x.\n", byte_count,
scsi_bufflen(sc), io_req->xid);
return bd_count;
}
static int qedf_build_bd_list_from_sg(struct qedf_ioreq *io_req)
{
struct scsi_cmnd *sc = io_req->sc_cmd;
struct scsi_sge *bd = io_req->bd_tbl->bd_tbl;
int bd_count;
if (scsi_sg_count(sc)) {
bd_count = qedf_map_sg(io_req);
if (bd_count == 0)
return -ENOMEM;
} else {
bd_count = 0;
bd[0].sge_addr.lo = bd[0].sge_addr.hi = 0;
bd[0].sge_len = 0;
}
io_req->bd_tbl->bd_valid = bd_count;
return 0;
}
static void qedf_build_fcp_cmnd(struct qedf_ioreq *io_req,
struct fcp_cmnd *fcp_cmnd)
{
struct scsi_cmnd *sc_cmd = io_req->sc_cmd;
/* fcp_cmnd is 32 bytes */
memset(fcp_cmnd, 0, FCP_CMND_LEN);
/* 8 bytes: SCSI LUN info */
int_to_scsilun(sc_cmd->device->lun,
(struct scsi_lun *)&fcp_cmnd->fc_lun);
/* 4 bytes: flag info */
fcp_cmnd->fc_pri_ta = 0;
fcp_cmnd->fc_tm_flags = io_req->tm_flags;
fcp_cmnd->fc_flags = io_req->io_req_flags;
fcp_cmnd->fc_cmdref = 0;
/* Populate data direction */
if (io_req->cmd_type == QEDF_TASK_MGMT_CMD) {
fcp_cmnd->fc_flags |= FCP_CFL_RDDATA;
} else {
if (sc_cmd->sc_data_direction == DMA_TO_DEVICE)
fcp_cmnd->fc_flags |= FCP_CFL_WRDATA;
else if (sc_cmd->sc_data_direction == DMA_FROM_DEVICE)
fcp_cmnd->fc_flags |= FCP_CFL_RDDATA;
}
fcp_cmnd->fc_pri_ta = FCP_PTA_SIMPLE;
/* 16 bytes: CDB information */
if (io_req->cmd_type != QEDF_TASK_MGMT_CMD)
memcpy(fcp_cmnd->fc_cdb, sc_cmd->cmnd, sc_cmd->cmd_len);
/* 4 bytes: FCP data length */
fcp_cmnd->fc_dl = htonl(io_req->data_xfer_len);
}
static void qedf_init_task(struct qedf_rport *fcport, struct fc_lport *lport,
struct qedf_ioreq *io_req, struct e4_fcoe_task_context *task_ctx,
struct fcoe_wqe *sqe)
{
enum fcoe_task_type task_type;
struct scsi_cmnd *sc_cmd = io_req->sc_cmd;
struct io_bdt *bd_tbl = io_req->bd_tbl;
u8 fcp_cmnd[32];
u32 tmp_fcp_cmnd[8];
int bd_count = 0;
struct qedf_ctx *qedf = fcport->qedf;
uint16_t cq_idx = smp_processor_id() % qedf->num_queues;
struct regpair sense_data_buffer_phys_addr;
u32 tx_io_size = 0;
u32 rx_io_size = 0;
int i, cnt;
/* Note init_initiator_rw_fcoe_task memsets the task context */
io_req->task = task_ctx;
memset(task_ctx, 0, sizeof(struct e4_fcoe_task_context));
memset(io_req->task_params, 0, sizeof(struct fcoe_task_params));
memset(io_req->sgl_task_params, 0, sizeof(struct scsi_sgl_task_params));
/* Set task type bassed on DMA directio of command */
if (io_req->cmd_type == QEDF_TASK_MGMT_CMD) {
task_type = FCOE_TASK_TYPE_READ_INITIATOR;
} else {
if (sc_cmd->sc_data_direction == DMA_TO_DEVICE) {
task_type = FCOE_TASK_TYPE_WRITE_INITIATOR;
tx_io_size = io_req->data_xfer_len;
} else {
task_type = FCOE_TASK_TYPE_READ_INITIATOR;
rx_io_size = io_req->data_xfer_len;
}
}
/* Setup the fields for fcoe_task_params */
io_req->task_params->context = task_ctx;
io_req->task_params->sqe = sqe;
io_req->task_params->task_type = task_type;
io_req->task_params->tx_io_size = tx_io_size;
io_req->task_params->rx_io_size = rx_io_size;
io_req->task_params->conn_cid = fcport->fw_cid;
io_req->task_params->itid = io_req->xid;
io_req->task_params->cq_rss_number = cq_idx;
io_req->task_params->is_tape_device = fcport->dev_type;
/* Fill in information for scatter/gather list */
if (io_req->cmd_type != QEDF_TASK_MGMT_CMD) {
bd_count = bd_tbl->bd_valid;
io_req->sgl_task_params->sgl = bd_tbl->bd_tbl;
io_req->sgl_task_params->sgl_phys_addr.lo =
U64_LO(bd_tbl->bd_tbl_dma);
io_req->sgl_task_params->sgl_phys_addr.hi =
U64_HI(bd_tbl->bd_tbl_dma);
io_req->sgl_task_params->num_sges = bd_count;
io_req->sgl_task_params->total_buffer_size =
scsi_bufflen(io_req->sc_cmd);
if (io_req->sge_type == QEDF_IOREQ_SLOW_SGE)
io_req->sgl_task_params->small_mid_sge = 1;
else
io_req->sgl_task_params->small_mid_sge = 0;
}
/* Fill in physical address of sense buffer */
sense_data_buffer_phys_addr.lo = U64_LO(io_req->sense_buffer_dma);
sense_data_buffer_phys_addr.hi = U64_HI(io_req->sense_buffer_dma);
/* fill FCP_CMND IU */
qedf_build_fcp_cmnd(io_req, (struct fcp_cmnd *)tmp_fcp_cmnd);
/* Swap fcp_cmnd since FC is big endian */
cnt = sizeof(struct fcp_cmnd) / sizeof(u32);
for (i = 0; i < cnt; i++) {
tmp_fcp_cmnd[i] = cpu_to_be32(tmp_fcp_cmnd[i]);
}
memcpy(fcp_cmnd, tmp_fcp_cmnd, sizeof(struct fcp_cmnd));
init_initiator_rw_fcoe_task(io_req->task_params,
io_req->sgl_task_params,
sense_data_buffer_phys_addr,
io_req->task_retry_identifier, fcp_cmnd);
/* Increment SGL type counters */
if (io_req->sge_type == QEDF_IOREQ_SLOW_SGE)
qedf->slow_sge_ios++;
else
qedf->fast_sge_ios++;
}
void qedf_init_mp_task(struct qedf_ioreq *io_req,
struct e4_fcoe_task_context *task_ctx, struct fcoe_wqe *sqe)
{
struct qedf_mp_req *mp_req = &(io_req->mp_req);
struct qedf_rport *fcport = io_req->fcport;
struct qedf_ctx *qedf = io_req->fcport->qedf;
struct fc_frame_header *fc_hdr;
struct fcoe_tx_mid_path_params task_fc_hdr;
struct scsi_sgl_task_params tx_sgl_task_params;
struct scsi_sgl_task_params rx_sgl_task_params;
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_DISC,
"Initializing MP task for cmd_type=%d\n",
io_req->cmd_type);
qedf->control_requests++;
memset(&tx_sgl_task_params, 0, sizeof(struct scsi_sgl_task_params));
memset(&rx_sgl_task_params, 0, sizeof(struct scsi_sgl_task_params));
memset(task_ctx, 0, sizeof(struct e4_fcoe_task_context));
memset(&task_fc_hdr, 0, sizeof(struct fcoe_tx_mid_path_params));
/* Setup the task from io_req for easy reference */
io_req->task = task_ctx;
/* Setup the fields for fcoe_task_params */
io_req->task_params->context = task_ctx;
io_req->task_params->sqe = sqe;
io_req->task_params->task_type = FCOE_TASK_TYPE_MIDPATH;
io_req->task_params->tx_io_size = io_req->data_xfer_len;
/* rx_io_size tells the f/w how large a response buffer we have */
io_req->task_params->rx_io_size = PAGE_SIZE;
io_req->task_params->conn_cid = fcport->fw_cid;
io_req->task_params->itid = io_req->xid;
/* Return middle path commands on CQ 0 */
io_req->task_params->cq_rss_number = 0;
io_req->task_params->is_tape_device = fcport->dev_type;
fc_hdr = &(mp_req->req_fc_hdr);
/* Set OX_ID and RX_ID based on driver task id */
fc_hdr->fh_ox_id = io_req->xid;
fc_hdr->fh_rx_id = htons(0xffff);
/* Set up FC header information */
task_fc_hdr.parameter = fc_hdr->fh_parm_offset;
task_fc_hdr.r_ctl = fc_hdr->fh_r_ctl;
task_fc_hdr.type = fc_hdr->fh_type;
task_fc_hdr.cs_ctl = fc_hdr->fh_cs_ctl;
task_fc_hdr.df_ctl = fc_hdr->fh_df_ctl;
task_fc_hdr.rx_id = fc_hdr->fh_rx_id;
task_fc_hdr.ox_id = fc_hdr->fh_ox_id;
/* Set up s/g list parameters for request buffer */
tx_sgl_task_params.sgl = mp_req->mp_req_bd;
tx_sgl_task_params.sgl_phys_addr.lo = U64_LO(mp_req->mp_req_bd_dma);
tx_sgl_task_params.sgl_phys_addr.hi = U64_HI(mp_req->mp_req_bd_dma);
tx_sgl_task_params.num_sges = 1;
/* Set PAGE_SIZE for now since sg element is that size ??? */
tx_sgl_task_params.total_buffer_size = io_req->data_xfer_len;
tx_sgl_task_params.small_mid_sge = 0;
/* Set up s/g list parameters for request buffer */
rx_sgl_task_params.sgl = mp_req->mp_resp_bd;
rx_sgl_task_params.sgl_phys_addr.lo = U64_LO(mp_req->mp_resp_bd_dma);
rx_sgl_task_params.sgl_phys_addr.hi = U64_HI(mp_req->mp_resp_bd_dma);
rx_sgl_task_params.num_sges = 1;
/* Set PAGE_SIZE for now since sg element is that size ??? */
rx_sgl_task_params.total_buffer_size = PAGE_SIZE;
rx_sgl_task_params.small_mid_sge = 0;
/*
* Last arg is 0 as previous code did not set that we wanted the
* fc header information.
*/
init_initiator_midpath_unsolicited_fcoe_task(io_req->task_params,
&task_fc_hdr,
&tx_sgl_task_params,
&rx_sgl_task_params, 0);
}
/* Presumed that fcport->rport_lock is held */
u16 qedf_get_sqe_idx(struct qedf_rport *fcport)
{
uint16_t total_sqe = (fcport->sq_mem_size)/(sizeof(struct fcoe_wqe));
u16 rval;
rval = fcport->sq_prod_idx;
/* Adjust ring index */
fcport->sq_prod_idx++;
fcport->fw_sq_prod_idx++;
if (fcport->sq_prod_idx == total_sqe)
fcport->sq_prod_idx = 0;
return rval;
}
void qedf_ring_doorbell(struct qedf_rport *fcport)
{
struct fcoe_db_data dbell = { 0 };
dbell.agg_flags = 0;
dbell.params |= DB_DEST_XCM << FCOE_DB_DATA_DEST_SHIFT;
dbell.params |= DB_AGG_CMD_SET << FCOE_DB_DATA_AGG_CMD_SHIFT;
dbell.params |= DQ_XCM_FCOE_SQ_PROD_CMD <<
FCOE_DB_DATA_AGG_VAL_SEL_SHIFT;
dbell.sq_prod = fcport->fw_sq_prod_idx;
/* wmb makes sure that the BDs data is updated before updating the
* producer, otherwise FW may read old data from the BDs.
*/
wmb();
barrier();
writel(*(u32 *)&dbell, fcport->p_doorbell);
/*
* Fence required to flush the write combined buffer, since another
* CPU may write to the same doorbell address and data may be lost
* due to relaxed order nature of write combined bar.
*/
wmb();
}
static void qedf_trace_io(struct qedf_rport *fcport, struct qedf_ioreq *io_req,
int8_t direction)
{
struct qedf_ctx *qedf = fcport->qedf;
struct qedf_io_log *io_log;
struct scsi_cmnd *sc_cmd = io_req->sc_cmd;
unsigned long flags;
uint8_t op;
spin_lock_irqsave(&qedf->io_trace_lock, flags);
io_log = &qedf->io_trace_buf[qedf->io_trace_idx];
io_log->direction = direction;
io_log->task_id = io_req->xid;
io_log->port_id = fcport->rdata->ids.port_id;
io_log->lun = sc_cmd->device->lun;
io_log->op = op = sc_cmd->cmnd[0];
io_log->lba[0] = sc_cmd->cmnd[2];
io_log->lba[1] = sc_cmd->cmnd[3];
io_log->lba[2] = sc_cmd->cmnd[4];
io_log->lba[3] = sc_cmd->cmnd[5];
io_log->bufflen = scsi_bufflen(sc_cmd);
io_log->sg_count = scsi_sg_count(sc_cmd);
io_log->result = sc_cmd->result;
io_log->jiffies = jiffies;
io_log->refcount = kref_read(&io_req->refcount);
if (direction == QEDF_IO_TRACE_REQ) {
/* For requests we only care abot the submission CPU */
io_log->req_cpu = io_req->cpu;
io_log->int_cpu = 0;
io_log->rsp_cpu = 0;
} else if (direction == QEDF_IO_TRACE_RSP) {
io_log->req_cpu = io_req->cpu;
io_log->int_cpu = io_req->int_cpu;
io_log->rsp_cpu = smp_processor_id();
}
io_log->sge_type = io_req->sge_type;
qedf->io_trace_idx++;
if (qedf->io_trace_idx == QEDF_IO_TRACE_SIZE)
qedf->io_trace_idx = 0;
spin_unlock_irqrestore(&qedf->io_trace_lock, flags);
}
int qedf_post_io_req(struct qedf_rport *fcport, struct qedf_ioreq *io_req)
{
struct scsi_cmnd *sc_cmd = io_req->sc_cmd;
struct Scsi_Host *host = sc_cmd->device->host;
struct fc_lport *lport = shost_priv(host);
struct qedf_ctx *qedf = lport_priv(lport);
struct e4_fcoe_task_context *task_ctx;
u16 xid;
enum fcoe_task_type req_type = 0;
struct fcoe_wqe *sqe;
u16 sqe_idx;
/* Initialize rest of io_req fileds */
io_req->data_xfer_len = scsi_bufflen(sc_cmd);
sc_cmd->SCp.ptr = (char *)io_req;
io_req->sge_type = QEDF_IOREQ_FAST_SGE; /* Assume fast SGL by default */
/* Record which cpu this request is associated with */
io_req->cpu = smp_processor_id();
if (sc_cmd->sc_data_direction == DMA_FROM_DEVICE) {
req_type = FCOE_TASK_TYPE_READ_INITIATOR;
io_req->io_req_flags = QEDF_READ;
qedf->input_requests++;
} else if (sc_cmd->sc_data_direction == DMA_TO_DEVICE) {
req_type = FCOE_TASK_TYPE_WRITE_INITIATOR;
io_req->io_req_flags = QEDF_WRITE;
qedf->output_requests++;
} else {
io_req->io_req_flags = 0;
qedf->control_requests++;
}
xid = io_req->xid;
/* Build buffer descriptor list for firmware from sg list */
if (qedf_build_bd_list_from_sg(io_req)) {
QEDF_ERR(&(qedf->dbg_ctx), "BD list creation failed.\n");
/* Release cmd will release io_req, but sc_cmd is assigned */
io_req->sc_cmd = NULL;
kref_put(&io_req->refcount, qedf_release_cmd);
return -EAGAIN;
}
if (!test_bit(QEDF_RPORT_SESSION_READY, &fcport->flags) ||
test_bit(QEDF_RPORT_UPLOADING_CONNECTION, &fcport->flags)) {
QEDF_ERR(&(qedf->dbg_ctx), "Session not offloaded yet.\n");
/* Release cmd will release io_req, but sc_cmd is assigned */
io_req->sc_cmd = NULL;
kref_put(&io_req->refcount, qedf_release_cmd);
scsi: qedf: Add missing return in qedf_post_io_req() in the fcport offload check Fixes the following crash as the return was missing from the check if an fcport is offloaded. If we hit this code we continue to try to post an invalid task which can lead to the crash: [30259.616411] [0000:61:00.3]:[qedf_post_io_req:989]:3: Session not offloaded yet. [30259.616413] [0000:61:00.3]:[qedf_upload_connection:1340]:3: Uploading connection port_id=490020. [30259.623769] BUG: unable to handle kernel NULL pointer dereference at 0000000000000198 [30259.631645] IP: [<ffffffffc035b1ed>] qedf_init_task.isra.16+0x3d/0x450 [qedf] [30259.638816] PGD 0 [30259.640841] Oops: 0000 [#1] SMP [30259.644098] Modules linked in: fuse xt_CHECKSUM iptable_mangle ipt_MASQUERADE nf_nat_masquerade_ipv4 iptable_nat nf_nat_ipv4 nf_nat nf_conntrack_ipv4 nf_defrag_ipv4 xt_conntrack nf_conntrack ipt_REJECT nf_reject_ipv4 tun bridge stp llc ebtable_filter ebtables devlink ip6table_filter ip6_tables iptable_filter vfat fat ib_isert iscsi_target_mod ib_srpt target_core_mod ib_srp scsi_transport_srp ib_ipoib ib_ucm ib_umad dm_service_time skx_edac intel_powerclamp coretemp intel_rapl iosf_mbi kvm_intel kvm irqbypass crc32_pclmul ghash_clmulni_intel aesni_intel rpcrdma sunrpc rdma_ucm ib_uverbs lrw gf128mul ib_iser rdma_cm iw_cm ib_cm libiscsi scsi_transport_iscsi qedr(OE) glue_helper ablk_helper cryptd ib_core dm_round_robin joydev pcspkr ipmi_ssif ses enclosure ipmi_si ipmi_devintf ipmi_msghandler mei_me [30259.715529] mei sg hpilo hpwdt shpchp wmi lpc_ich acpi_power_meter dm_multipath ip_tables xfs libcrc32c sd_mod crc_t10dif crct10dif_generic uas usb_storage mgag200 qedf(OE) i2c_algo_bit libfcoe drm_kms_helper libfc syscopyarea sysfillrect scsi_transport_fc qede(OE) sysimgblt fb_sys_fops ptp ttm pps_core drm qed(OE) smartpqi crct10dif_pclmul crct10dif_common crc32c_intel i2c_core scsi_transport_sas scsi_tgt dm_mirror dm_region_hash dm_log dm_mod [30259.754237] CPU: 9 PID: 977 Comm: kdmwork-253:7 Kdump: loaded Tainted: G W OE ------------ 3.10.0-862.el7.x86_64 #1 [30259.765664] Hardware name: HPE Synergy 480 Gen10/Synergy 480 Gen10 Compute Module, BIOS I42 04/04/2018 [30259.775000] task: ffff8c801efd0000 ti: ffff8c801efd8000 task.ti: ffff8c801efd8000 [30259.782505] RIP: 0010:[<ffffffffc035b1ed>] [<ffffffffc035b1ed>] qedf_init_task.isra.16+0x3d/0x450 [qedf] [30259.792116] RSP: 0018:ffff8c801efdbbb0 EFLAGS: 00010046 [30259.797444] RAX: 0000000000000000 RBX: ffffa7f1450948d8 RCX: ffff8c7fe5bc40c8 [30259.804600] RDX: ffff8c800715b300 RSI: ffffa7f1450948d8 RDI: ffff8c80169c2480 [30259.811755] RBP: ffff8c801efdbc30 R08: 00000000000000ae R09: ffff8c800a314540 [30259.818911] R10: ffff8c7fe5bc40c8 R11: ffff8c801efdb8ae R12: 0000000000000000 [30259.826068] R13: ffff8c800715b300 R14: ffff8c80169c2480 R15: ffff8c8005da28e0 [30259.833223] FS: 0000000000000000(0000) GS:ffff8c803f840000(0000) knlGS:0000000000000000 [30259.841338] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [30259.847100] CR2: 0000000000000198 CR3: 000000081242e000 CR4: 00000000007607e0 [30259.854256] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [30259.861412] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 [30259.868568] PKRU: 00000000 [30259.871278] Call Trace: [30259.873737] [<ffffffffc035c948>] qedf_post_io_req+0x148/0x680 [qedf] [30259.880201] [<ffffffffc035d070>] qedf_queuecommand+0x1f0/0x240 [qedf] [30259.886749] [<ffffffffa329b050>] scsi_dispatch_cmd+0xb0/0x240 [30259.892600] [<ffffffffa32a45bc>] scsi_request_fn+0x4cc/0x680 [30259.898364] [<ffffffffa3118ad9>] __blk_run_queue+0x39/0x50 [30259.903954] [<ffffffffa3114393>] __elv_add_request+0xd3/0x260 [30259.909805] [<ffffffffa311baf0>] blk_insert_cloned_request+0xf0/0x1b0 [30259.916358] [<ffffffffc010b622>] map_request+0x142/0x220 [dm_mod] [30259.922560] [<ffffffffc010b716>] map_tio_request+0x16/0x40 [dm_mod] [30259.928932] [<ffffffffa2ebb1f5>] kthread_worker_fn+0x85/0x180 [30259.934782] [<ffffffffa2ebb170>] ? kthread_stop+0xf0/0xf0 [30259.940284] [<ffffffffa2ebae31>] kthread+0xd1/0xe0 [30259.945176] [<ffffffffa2ebad60>] ? insert_kthread_work+0x40/0x40 [30259.951290] [<ffffffffa351f61d>] ret_from_fork_nospec_begin+0x7/0x21 [30259.957750] [<ffffffffa2ebad60>] ? insert_kthread_work+0x40/0x40 [30259.963860] Code: fe 41 55 49 89 d5 41 54 53 48 89 f3 48 83 ec 58 4c 8b 67 28 4c 8b 4e 18 65 48 8b 04 25 28 00 00 00 48 89 45 d0 31 c0 4c 8b 7e 58 <49> 8b 84 24 98 01 00 00 48 8b 00 f6 80 31 01 00 00 10 0f 85 0b [30259.983372] RIP [<ffffffffc035b1ed>] qedf_init_task.isra.16+0x3d/0x450 [qedf] [30259.990630] RSP <ffff8c801efdbbb0> [30259.994127] CR2: 0000000000000198 Signed-off-by: Chad Dupuis <cdupuis@marvell.com> Signed-off-by: Saurav Kashyap <skashyap@marvell.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-03-26 16:38:35 +09:00
return -EINVAL;
}
/* Record LUN number for later use if we neeed them */
io_req->lun = (int)sc_cmd->device->lun;
/* Obtain free SQE */
sqe_idx = qedf_get_sqe_idx(fcport);
sqe = &fcport->sq[sqe_idx];
memset(sqe, 0, sizeof(struct fcoe_wqe));
/* Get the task context */
task_ctx = qedf_get_task_mem(&qedf->tasks, xid);
if (!task_ctx) {
QEDF_WARN(&(qedf->dbg_ctx), "task_ctx is NULL, xid=%d.\n",
xid);
/* Release cmd will release io_req, but sc_cmd is assigned */
io_req->sc_cmd = NULL;
kref_put(&io_req->refcount, qedf_release_cmd);
return -EINVAL;
}
qedf_init_task(fcport, lport, io_req, task_ctx, sqe);
/* Ring doorbell */
qedf_ring_doorbell(fcport);
/* Set that command is with the firmware now */
set_bit(QEDF_CMD_OUTSTANDING, &io_req->flags);
if (qedf_io_tracing && io_req->sc_cmd)
qedf_trace_io(fcport, io_req, QEDF_IO_TRACE_REQ);
return false;
}
int
qedf_queuecommand(struct Scsi_Host *host, struct scsi_cmnd *sc_cmd)
{
struct fc_lport *lport = shost_priv(host);
struct qedf_ctx *qedf = lport_priv(lport);
struct fc_rport *rport = starget_to_rport(scsi_target(sc_cmd->device));
struct fc_rport_libfc_priv *rp = rport->dd_data;
struct qedf_rport *fcport;
struct qedf_ioreq *io_req;
int rc = 0;
int rval;
unsigned long flags = 0;
int num_sgs = 0;
num_sgs = scsi_sg_count(sc_cmd);
if (scsi_sg_count(sc_cmd) > QEDF_MAX_BDS_PER_CMD) {
QEDF_ERR(&qedf->dbg_ctx,
"Number of SG elements %d exceeds what hardware limitation of %d.\n",
num_sgs, QEDF_MAX_BDS_PER_CMD);
sc_cmd->result = DID_ERROR;
sc_cmd->scsi_done(sc_cmd);
return 0;
}
if (test_bit(QEDF_UNLOADING, &qedf->flags) ||
test_bit(QEDF_DBG_STOP_IO, &qedf->flags)) {
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO,
"Returning DNC as unloading or stop io, flags 0x%lx.\n",
qedf->flags);
sc_cmd->result = DID_NO_CONNECT << 16;
sc_cmd->scsi_done(sc_cmd);
return 0;
}
if (!qedf->pdev->msix_enabled) {
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_IO,
"Completing sc_cmd=%p DID_NO_CONNECT as MSI-X is not enabled.\n",
sc_cmd);
sc_cmd->result = DID_NO_CONNECT << 16;
sc_cmd->scsi_done(sc_cmd);
return 0;
}
rval = fc_remote_port_chkready(rport);
if (rval) {
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO,
"fc_remote_port_chkready failed=0x%x for port_id=0x%06x.\n",
rval, rport->port_id);
sc_cmd->result = rval;
sc_cmd->scsi_done(sc_cmd);
return 0;
}
/* Retry command if we are doing a qed drain operation */
if (test_bit(QEDF_DRAIN_ACTIVE, &qedf->flags)) {
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO, "Drain active.\n");
rc = SCSI_MLQUEUE_HOST_BUSY;
goto exit_qcmd;
}
if (lport->state != LPORT_ST_READY ||
atomic_read(&qedf->link_state) != QEDF_LINK_UP) {
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO, "Link down.\n");
rc = SCSI_MLQUEUE_HOST_BUSY;
goto exit_qcmd;
}
/* rport and tgt are allocated together, so tgt should be non-NULL */
fcport = (struct qedf_rport *)&rp[1];
if (!test_bit(QEDF_RPORT_SESSION_READY, &fcport->flags) ||
test_bit(QEDF_RPORT_UPLOADING_CONNECTION, &fcport->flags)) {
/*
* Session is not offloaded yet. Let SCSI-ml retry
* the command.
*/
rc = SCSI_MLQUEUE_TARGET_BUSY;
goto exit_qcmd;
}
atomic_inc(&fcport->ios_to_queue);
if (fcport->retry_delay_timestamp) {
if (time_after(jiffies, fcport->retry_delay_timestamp)) {
fcport->retry_delay_timestamp = 0;
} else {
/* If retry_delay timer is active, flow off the ML */
rc = SCSI_MLQUEUE_TARGET_BUSY;
atomic_dec(&fcport->ios_to_queue);
goto exit_qcmd;
}
}
io_req = qedf_alloc_cmd(fcport, QEDF_SCSI_CMD);
if (!io_req) {
rc = SCSI_MLQUEUE_HOST_BUSY;
atomic_dec(&fcport->ios_to_queue);
goto exit_qcmd;
}
io_req->sc_cmd = sc_cmd;
/* Take fcport->rport_lock for posting to fcport send queue */
spin_lock_irqsave(&fcport->rport_lock, flags);
if (qedf_post_io_req(fcport, io_req)) {
QEDF_WARN(&(qedf->dbg_ctx), "Unable to post io_req\n");
/* Return SQE to pool */
atomic_inc(&fcport->free_sqes);
rc = SCSI_MLQUEUE_HOST_BUSY;
}
spin_unlock_irqrestore(&fcport->rport_lock, flags);
atomic_dec(&fcport->ios_to_queue);
exit_qcmd:
return rc;
}
static void qedf_parse_fcp_rsp(struct qedf_ioreq *io_req,
struct fcoe_cqe_rsp_info *fcp_rsp)
{
struct scsi_cmnd *sc_cmd = io_req->sc_cmd;
struct qedf_ctx *qedf = io_req->fcport->qedf;
u8 rsp_flags = fcp_rsp->rsp_flags.flags;
int fcp_sns_len = 0;
int fcp_rsp_len = 0;
uint8_t *rsp_info, *sense_data;
io_req->fcp_status = FC_GOOD;
io_req->fcp_resid = 0;
if (rsp_flags & (FCOE_FCP_RSP_FLAGS_FCP_RESID_OVER |
FCOE_FCP_RSP_FLAGS_FCP_RESID_UNDER))
io_req->fcp_resid = fcp_rsp->fcp_resid;
io_req->scsi_comp_flags = rsp_flags;
CMD_SCSI_STATUS(sc_cmd) = io_req->cdb_status =
fcp_rsp->scsi_status_code;
if (rsp_flags &
FCOE_FCP_RSP_FLAGS_FCP_RSP_LEN_VALID)
fcp_rsp_len = fcp_rsp->fcp_rsp_len;
if (rsp_flags &
FCOE_FCP_RSP_FLAGS_FCP_SNS_LEN_VALID)
fcp_sns_len = fcp_rsp->fcp_sns_len;
io_req->fcp_rsp_len = fcp_rsp_len;
io_req->fcp_sns_len = fcp_sns_len;
rsp_info = sense_data = io_req->sense_buffer;
/* fetch fcp_rsp_code */
if ((fcp_rsp_len == 4) || (fcp_rsp_len == 8)) {
/* Only for task management function */
io_req->fcp_rsp_code = rsp_info[3];
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_IO,
"fcp_rsp_code = %d\n", io_req->fcp_rsp_code);
/* Adjust sense-data location. */
sense_data += fcp_rsp_len;
}
if (fcp_sns_len > SCSI_SENSE_BUFFERSIZE) {
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_IO,
"Truncating sense buffer\n");
fcp_sns_len = SCSI_SENSE_BUFFERSIZE;
}
scsi: qedf: Check if sense buffer has been allocated during completion sc_cmd->sense_buffer is not guaranteed to be allocated so we need to sc_cmd->check if the pointer is NULL before trying to copy anything into it. Fixes the crash: [ 143.793176] [0000:00:00.0]:[qedf_eh_device_reset:626]: LUN RESET Issued... [ 143.802996] BUG: unable to handle kernel NULL pointer dereference at (null) [ 143.803063] IP: qedf_parse_fcp_rsp+0xe2/0x290 [qedf] [ 143.803077] PGD 0 [ 143.803078] P4D 0 [ 143.803103] Oops: 0002 [#1] SMP [ 143.803115] Modules linked in: msr(E) ebtable_filter(E) ebtables(E) ip6table_filter(E) ip6_tables(E) iptable_filter(E) ip_tables(E) x_tables(E) raw(E) scsi_transport_iscsi(E) br_netfilter(E) bridge(E) iscsi_ibft(E) iscsi_boot_sysfs(E) intel_rapl(E) sb_edac(E) x86_pkg_temp_thermal(E) intel_powerclamp(E) coretemp(E) kvm_intel(E) kvm(E) irqbypass(E) crct10dif_pclmul(E) crc32_pclmul(E) xfs(E) ghash_clmulni_intel(E) pcbc(E) aesni_intel(E) aes_x86_64(E) crypto_simd(E) ipmi_ssif(E) glue_helper(E) iTCO_wdt(E) iTCO_vendor_support(E) lpc_ich(E) ipmi_si(E) pcspkr(E) hpilo(E) ioatdma(E) cryptd(E) ipmi_devintf(E) hpwdt(E) mfd_core(E) shpchp(E) dca(E) thermal(E) pcc_cpufreq(E) ipmi_msghandler(E) acpi_cpufreq(E) af_packet(E) btrfs(E) xor(E) raid6_pq(E) sr_mod(E) cdrom(E) ata_generic(E) sd_mod(E) 8021q(E) garp(E) [ 143.803302] stp(E) llc(E) mrp(E) bnx2fc(E) cnic(E) uio(E) mgag200(E) ata_piix(E) i2c_algo_bit(E) drm_kms_helper(E) syscopyarea(E) sysfillrect(E) sysimgblt(E) ahci(E) fb_sys_fops(E) bnx2x(E) qedf(E) serio_raw(E) libahci(E) ttm(E) uhci_hcd(E) ehci_pci(E) qed(E) mdio(E) libcrc32c(E) ehci_hcd(E) crc32c_intel(E) drm(E) libata(E) usbcore(E) tg3(E) ptp(E) hpsa(E) pps_core(E) scsi_transport_sas(E) libphy(E) wmi(E) button(E) fcoe(E) libfcoe(E) libfc(E) scsi_transport_fc(E) sg(E) dm_multipath(E) dm_mod(E) scsi_dh_rdac(E) scsi_dh_emc(E) scsi_dh_alua(E) scsi_mod(E) autofs4(E) [ 143.803438] CPU: 31 PID: 494 Comm: kworker/31:2 Tainted: G E 4.12.0-rc1-69-default+ #1 [ 143.803461] Hardware name: HP ProLiant DL380p Gen8, BIOS P70 08/20/2012 [ 143.803480] Workqueue: qedf_io_wq qedf_fp_io_handler [qedf] [ 143.803496] task: ffff8804181a0000 task.stack: ffffc90003b64000 [ 143.803514] RIP: 0010:qedf_parse_fcp_rsp+0xe2/0x290 [qedf] [ 143.803529] RSP: 0018:ffffc90003b67dc8 EFLAGS: 00010246 [ 143.803544] RAX: 0000000000000000 RBX: ffff880401abdd48 RCX: 000000000000000c [ 143.803563] RDX: 0000000000000060 RSI: ffffffffa039c740 RDI: 0000000000000000 [ 143.803581] RBP: ffffc90003b67df0 R08: ffffffffa039dba8 R09: 0000000000000000 [ 143.803600] R10: 0000000000000000 R11: 0000000000000018 R12: 0000000000000000 [ 143.803619] R13: ffff88040ac80bc8 R14: 0000000000000008 R15: ffff880407c14008 [ 143.803638] FS: 0000000000000000(0000) GS:ffff88043f7c0000(0000) knlGS:0000000000000000 [ 143.804360] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 143.805065] CR2: 0000000000000000 CR3: 0000000001c09000 CR4: 00000000000406e0 [ 143.805753] Call Trace: [ 143.806436] qedf_process_tmf_compl+0x19/0x30 [qedf] [ 143.807124] qedf_process_cqe+0x265/0x280 [qedf] [ 143.807800] qedf_fp_io_handler+0x26/0x60 [qedf] [ 143.808469] process_one_work+0x138/0x370 [ 143.809133] worker_thread+0x4d/0x3b0 [ 143.809797] kthread+0x109/0x140 [ 143.810451] ? rescuer_thread+0x320/0x320 [ 143.811100] ? kthread_park+0x60/0x60 [ 143.811743] ret_from_fork+0x2c/0x40 Signed-off-by: Chad Dupuis <chad.dupuis@cavium.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-06-03 04:02:05 +09:00
/* The sense buffer can be NULL for TMF commands */
if (sc_cmd->sense_buffer) {
memset(sc_cmd->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
if (fcp_sns_len)
memcpy(sc_cmd->sense_buffer, sense_data,
fcp_sns_len);
}
}
static void qedf_unmap_sg_list(struct qedf_ctx *qedf, struct qedf_ioreq *io_req)
{
struct scsi_cmnd *sc = io_req->sc_cmd;
if (io_req->bd_tbl->bd_valid && sc && scsi_sg_count(sc)) {
dma_unmap_sg(&qedf->pdev->dev, scsi_sglist(sc),
scsi_sg_count(sc), sc->sc_data_direction);
io_req->bd_tbl->bd_valid = 0;
}
}
void qedf_scsi_completion(struct qedf_ctx *qedf, struct fcoe_cqe *cqe,
struct qedf_ioreq *io_req)
{
u16 xid;
struct e4_fcoe_task_context *task_ctx;
struct scsi_cmnd *sc_cmd;
struct fcoe_cqe_rsp_info *fcp_rsp;
struct qedf_rport *fcport;
int refcount;
u16 scope, qualifier = 0;
u8 fw_residual_flag = 0;
if (!io_req)
return;
if (!cqe)
return;
if (!test_bit(QEDF_CMD_OUTSTANDING, &io_req->flags) ||
test_bit(QEDF_CMD_IN_CLEANUP, &io_req->flags) ||
test_bit(QEDF_CMD_IN_ABORT, &io_req->flags)) {
QEDF_ERR(&qedf->dbg_ctx,
"io_req xid=0x%x already in cleanup or abort processing or already completed.\n",
io_req->xid);
return;
}
xid = io_req->xid;
task_ctx = qedf_get_task_mem(&qedf->tasks, xid);
sc_cmd = io_req->sc_cmd;
fcp_rsp = &cqe->cqe_info.rsp_info;
if (!sc_cmd) {
QEDF_WARN(&(qedf->dbg_ctx), "sc_cmd is NULL!\n");
return;
}
if (!sc_cmd->SCp.ptr) {
QEDF_WARN(&(qedf->dbg_ctx), "SCp.ptr is NULL, returned in "
"another context.\n");
return;
}
if (!sc_cmd->device) {
QEDF_ERR(&qedf->dbg_ctx,
"Device for sc_cmd %p is NULL.\n", sc_cmd);
return;
}
if (!sc_cmd->request) {
QEDF_WARN(&(qedf->dbg_ctx), "sc_cmd->request is NULL, "
"sc_cmd=%p.\n", sc_cmd);
return;
}
if (!sc_cmd->request->q) {
QEDF_WARN(&(qedf->dbg_ctx), "request->q is NULL so request "
"is not valid, sc_cmd=%p.\n", sc_cmd);
return;
}
fcport = io_req->fcport;
/*
* When flush is active, let the cmds be completed from the cleanup
* context
*/
if (test_bit(QEDF_RPORT_IN_TARGET_RESET, &fcport->flags) ||
(test_bit(QEDF_RPORT_IN_LUN_RESET, &fcport->flags) &&
sc_cmd->device->lun == (u64)fcport->lun_reset_lun)) {
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO,
"Dropping good completion xid=0x%x as fcport is flushing",
io_req->xid);
return;
}
qedf_parse_fcp_rsp(io_req, fcp_rsp);
qedf_unmap_sg_list(qedf, io_req);
/* Check for FCP transport error */
if (io_req->fcp_rsp_len > 3 && io_req->fcp_rsp_code) {
QEDF_ERR(&(qedf->dbg_ctx),
"FCP I/O protocol failure xid=0x%x fcp_rsp_len=%d "
"fcp_rsp_code=%d.\n", io_req->xid, io_req->fcp_rsp_len,
io_req->fcp_rsp_code);
sc_cmd->result = DID_BUS_BUSY << 16;
goto out;
}
fw_residual_flag = GET_FIELD(cqe->cqe_info.rsp_info.fw_error_flags,
FCOE_CQE_RSP_INFO_FW_UNDERRUN);
if (fw_residual_flag) {
QEDF_ERR(&qedf->dbg_ctx,
"Firmware detected underrun: xid=0x%x fcp_rsp.flags=0x%02x fcp_resid=%d fw_residual=0x%x lba=%02x%02x%02x%02x.\n",
io_req->xid, fcp_rsp->rsp_flags.flags,
io_req->fcp_resid,
cqe->cqe_info.rsp_info.fw_residual, sc_cmd->cmnd[2],
sc_cmd->cmnd[3], sc_cmd->cmnd[4], sc_cmd->cmnd[5]);
if (io_req->cdb_status == 0)
sc_cmd->result = (DID_ERROR << 16) | io_req->cdb_status;
else
sc_cmd->result = (DID_OK << 16) | io_req->cdb_status;
/*
* Set resid to the whole buffer length so we won't try to resue
* any previously data.
*/
scsi_set_resid(sc_cmd, scsi_bufflen(sc_cmd));
goto out;
}
switch (io_req->fcp_status) {
case FC_GOOD:
if (io_req->cdb_status == 0) {
/* Good I/O completion */
sc_cmd->result = DID_OK << 16;
} else {
refcount = kref_read(&io_req->refcount);
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_IO,
"%d:0:%d:%lld xid=0x%0x op=0x%02x "
"lba=%02x%02x%02x%02x cdb_status=%d "
"fcp_resid=0x%x refcount=%d.\n",
qedf->lport->host->host_no, sc_cmd->device->id,
sc_cmd->device->lun, io_req->xid,
sc_cmd->cmnd[0], sc_cmd->cmnd[2], sc_cmd->cmnd[3],
sc_cmd->cmnd[4], sc_cmd->cmnd[5],
io_req->cdb_status, io_req->fcp_resid,
refcount);
sc_cmd->result = (DID_OK << 16) | io_req->cdb_status;
if (io_req->cdb_status == SAM_STAT_TASK_SET_FULL ||
io_req->cdb_status == SAM_STAT_BUSY) {
/*
* Check whether we need to set retry_delay at
* all based on retry_delay module parameter
* and the status qualifier.
*/
/* Upper 2 bits */
scope = fcp_rsp->retry_delay_timer & 0xC000;
/* Lower 14 bits */
qualifier = fcp_rsp->retry_delay_timer & 0x3FFF;
if (qedf_retry_delay &&
scope > 0 && qualifier > 0 &&
qualifier <= 0x3FEF) {
/* Check we don't go over the max */
if (qualifier > QEDF_RETRY_DELAY_MAX)
qualifier =
QEDF_RETRY_DELAY_MAX;
fcport->retry_delay_timestamp =
jiffies + (qualifier * HZ / 10);
}
/* Record stats */
if (io_req->cdb_status ==
SAM_STAT_TASK_SET_FULL)
qedf->task_set_fulls++;
else
qedf->busy++;
}
}
if (io_req->fcp_resid)
scsi_set_resid(sc_cmd, io_req->fcp_resid);
break;
default:
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_IO, "fcp_status=%d.\n",
io_req->fcp_status);
break;
}
out:
if (qedf_io_tracing)
qedf_trace_io(fcport, io_req, QEDF_IO_TRACE_RSP);
/*
* We wait till the end of the function to clear the
* outstanding bit in case we need to send an abort
*/
clear_bit(QEDF_CMD_OUTSTANDING, &io_req->flags);
io_req->sc_cmd = NULL;
sc_cmd->SCp.ptr = NULL;
sc_cmd->scsi_done(sc_cmd);
kref_put(&io_req->refcount, qedf_release_cmd);
}
/* Return a SCSI command in some other context besides a normal completion */
void qedf_scsi_done(struct qedf_ctx *qedf, struct qedf_ioreq *io_req,
int result)
{
u16 xid;
struct scsi_cmnd *sc_cmd;
int refcount;
if (!io_req) {
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO, "io_req is NULL\n");
return;
}
if (test_and_set_bit(QEDF_CMD_ERR_SCSI_DONE, &io_req->flags)) {
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO,
"io_req:%p scsi_done handling already done\n",
io_req);
return;
}
/*
* We will be done with this command after this call so clear the
* outstanding bit.
*/
clear_bit(QEDF_CMD_OUTSTANDING, &io_req->flags);
xid = io_req->xid;
sc_cmd = io_req->sc_cmd;
if (!sc_cmd) {
QEDF_WARN(&(qedf->dbg_ctx), "sc_cmd is NULL!\n");
return;
}
if (!virt_addr_valid(sc_cmd)) {
QEDF_ERR(&qedf->dbg_ctx, "sc_cmd=%p is not valid.", sc_cmd);
goto bad_scsi_ptr;
}
if (!sc_cmd->SCp.ptr) {
QEDF_WARN(&(qedf->dbg_ctx), "SCp.ptr is NULL, returned in "
"another context.\n");
return;
}
if (!sc_cmd->device) {
QEDF_ERR(&qedf->dbg_ctx, "Device for sc_cmd %p is NULL.\n",
sc_cmd);
goto bad_scsi_ptr;
}
if (!virt_addr_valid(sc_cmd->device)) {
QEDF_ERR(&qedf->dbg_ctx,
"Device pointer for sc_cmd %p is bad.\n", sc_cmd);
goto bad_scsi_ptr;
}
if (!sc_cmd->sense_buffer) {
QEDF_ERR(&qedf->dbg_ctx,
"sc_cmd->sense_buffer for sc_cmd %p is NULL.\n",
sc_cmd);
goto bad_scsi_ptr;
}
if (!virt_addr_valid(sc_cmd->sense_buffer)) {
QEDF_ERR(&qedf->dbg_ctx,
"sc_cmd->sense_buffer for sc_cmd %p is bad.\n",
sc_cmd);
goto bad_scsi_ptr;
}
if (!sc_cmd->scsi_done) {
QEDF_ERR(&qedf->dbg_ctx,
"sc_cmd->scsi_done for sc_cmd %p is NULL.\n",
sc_cmd);
goto bad_scsi_ptr;
}
qedf_unmap_sg_list(qedf, io_req);
sc_cmd->result = result << 16;
refcount = kref_read(&io_req->refcount);
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_IO, "%d:0:%d:%lld: Completing "
"sc_cmd=%p result=0x%08x op=0x%02x lba=0x%02x%02x%02x%02x, "
"allowed=%d retries=%d refcount=%d.\n",
qedf->lport->host->host_no, sc_cmd->device->id,
sc_cmd->device->lun, sc_cmd, sc_cmd->result, sc_cmd->cmnd[0],
sc_cmd->cmnd[2], sc_cmd->cmnd[3], sc_cmd->cmnd[4],
sc_cmd->cmnd[5], sc_cmd->allowed, sc_cmd->retries,
refcount);
/*
* Set resid to the whole buffer length so we won't try to resue any
* previously read data
*/
scsi_set_resid(sc_cmd, scsi_bufflen(sc_cmd));
if (qedf_io_tracing)
qedf_trace_io(io_req->fcport, io_req, QEDF_IO_TRACE_RSP);
io_req->sc_cmd = NULL;
sc_cmd->SCp.ptr = NULL;
sc_cmd->scsi_done(sc_cmd);
kref_put(&io_req->refcount, qedf_release_cmd);
return;
bad_scsi_ptr:
/*
* Clear the io_req->sc_cmd backpointer so we don't try to process
* this again
*/
io_req->sc_cmd = NULL;
kref_put(&io_req->refcount, qedf_release_cmd); /* ID: 001 */
}
/*
* Handle warning type CQE completions. This is mainly used for REC timer
* popping.
*/
void qedf_process_warning_compl(struct qedf_ctx *qedf, struct fcoe_cqe *cqe,
struct qedf_ioreq *io_req)
{
int rval, i;
struct qedf_rport *fcport = io_req->fcport;
u64 err_warn_bit_map;
u8 err_warn = 0xff;
if (!cqe) {
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO,
"cqe is NULL for io_req %p xid=0x%x\n",
io_req, io_req->xid);
return;
}
QEDF_ERR(&(io_req->fcport->qedf->dbg_ctx), "Warning CQE, "
"xid=0x%x\n", io_req->xid);
QEDF_ERR(&(io_req->fcport->qedf->dbg_ctx),
"err_warn_bitmap=%08x:%08x\n",
le32_to_cpu(cqe->cqe_info.err_info.err_warn_bitmap_hi),
le32_to_cpu(cqe->cqe_info.err_info.err_warn_bitmap_lo));
QEDF_ERR(&(io_req->fcport->qedf->dbg_ctx), "tx_buff_off=%08x, "
"rx_buff_off=%08x, rx_id=%04x\n",
le32_to_cpu(cqe->cqe_info.err_info.tx_buf_off),
le32_to_cpu(cqe->cqe_info.err_info.rx_buf_off),
le32_to_cpu(cqe->cqe_info.err_info.rx_id));
/* Normalize the error bitmap value to an just an unsigned int */
err_warn_bit_map = (u64)
((u64)cqe->cqe_info.err_info.err_warn_bitmap_hi << 32) |
(u64)cqe->cqe_info.err_info.err_warn_bitmap_lo;
for (i = 0; i < 64; i++) {
if (err_warn_bit_map & (u64)((u64)1 << i)) {
err_warn = i;
break;
}
}
/* Check if REC TOV expired if this is a tape device */
if (fcport->dev_type == QEDF_RPORT_TYPE_TAPE) {
if (err_warn ==
FCOE_WARNING_CODE_REC_TOV_TIMER_EXPIRATION) {
QEDF_ERR(&(qedf->dbg_ctx), "REC timer expired.\n");
if (!test_bit(QEDF_CMD_SRR_SENT, &io_req->flags)) {
io_req->rx_buf_off =
cqe->cqe_info.err_info.rx_buf_off;
io_req->tx_buf_off =
cqe->cqe_info.err_info.tx_buf_off;
io_req->rx_id = cqe->cqe_info.err_info.rx_id;
rval = qedf_send_rec(io_req);
/*
* We only want to abort the io_req if we
* can't queue the REC command as we want to
* keep the exchange open for recovery.
*/
if (rval)
goto send_abort;
}
return;
}
}
send_abort:
init_completion(&io_req->abts_done);
rval = qedf_initiate_abts(io_req, true);
if (rval)
QEDF_ERR(&(qedf->dbg_ctx), "Failed to queue ABTS.\n");
}
/* Cleanup a command when we receive an error detection completion */
void qedf_process_error_detect(struct qedf_ctx *qedf, struct fcoe_cqe *cqe,
struct qedf_ioreq *io_req)
{
int rval;
scsi: qedf: Add check to synchronize abort and flush [ Upstream commit df99446d5c2a63dc6e6920c8090da0e9da6539d5 ] A race condition was observed between qedf_cleanup_fcport() and qedf_process_error_detect()->qedf_initiate_abts(): [2069091.203145] BUG: unable to handle kernel NULL pointer dereference at 0000000000000030 [2069091.213100] IP: [<ffffffffc0666cc6>] qedf_process_error_detect+0x96/0x130 [qedf] [2069091.223391] PGD 1943049067 PUD 194304e067 PMD 0 [2069091.233420] Oops: 0000 [#1] SMP [2069091.361820] CPU: 1 PID: 14751 Comm: kworker/1:46 Kdump: loaded Tainted: P OE ------------ 3.10.0-1160.25.1.el7.x86_64 #1 [2069091.388474] Hardware name: HPE Synergy 480 Gen10/Synergy 480 Gen10 Compute Module, BIOS I42 04/08/2020 [2069091.402148] Workqueue: qedf_io_wq qedf_fp_io_handler [qedf] [2069091.415780] task: ffff9bb9f5190000 ti: ffff9bacaef9c000 task.ti: ffff9bacaef9c000 [2069091.429590] RIP: 0010:[<ffffffffc0666cc6>] [<ffffffffc0666cc6>] qedf_process_error_detect+0x96/0x130 [qedf] [2069091.443666] RSP: 0018:ffff9bacaef9fdb8 EFLAGS: 00010246 [2069091.457692] RAX: 0000000000000000 RBX: ffff9bbbbbfb18a0 RCX: ffffffffc0672310 [2069091.471997] RDX: 00000000000005de RSI: ffffffffc066e7f0 RDI: ffff9beb3f4538d8 [2069091.486130] RBP: ffff9bacaef9fdd8 R08: 0000000000006000 R09: 0000000000006000 [2069091.500321] R10: 0000000000001551 R11: ffffb582996ffff8 R12: ffffb5829b39cc18 [2069091.514779] R13: ffff9badab380c28 R14: ffffd5827f643900 R15: 0000000000000040 [2069091.529472] FS: 0000000000000000(0000) GS:ffff9beb3f440000(0000) knlGS:0000000000000000 [2069091.543926] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [2069091.558942] CR2: 0000000000000030 CR3: 000000193b9a2000 CR4: 00000000007607e0 [2069091.573424] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [2069091.587876] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 [2069091.602007] PKRU: 00000000 [2069091.616010] Call Trace: [2069091.629902] [<ffffffffc0663969>] qedf_process_cqe+0x109/0x2e0 [qedf] [2069091.643941] [<ffffffffc0663b66>] qedf_fp_io_handler+0x26/0x60 [qedf] [2069091.657948] [<ffffffff85ebddcf>] process_one_work+0x17f/0x440 [2069091.672111] [<ffffffff85ebeee6>] worker_thread+0x126/0x3c0 [2069091.686057] [<ffffffff85ebedc0>] ? manage_workers.isra.26+0x2a0/0x2a0 [2069091.700033] [<ffffffff85ec5da1>] kthread+0xd1/0xe0 [2069091.713891] [<ffffffff85ec5cd0>] ? insert_kthread_work+0x40/0x40 Add check in qedf_process_error_detect(). When flush is active, let the cmds be completed from the cleanup contex. Link: https://lore.kernel.org/r/20210624171802.598-1-jhasan@marvell.com Signed-off-by: Javed Hasan <jhasan@marvell.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
2021-06-25 02:18:02 +09:00
if (io_req == NULL) {
QEDF_INFO(NULL, QEDF_LOG_IO, "io_req is NULL.\n");
return;
}
if (io_req->fcport == NULL) {
QEDF_INFO(NULL, QEDF_LOG_IO, "fcport is NULL.\n");
return;
}
if (!cqe) {
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO,
scsi: qedf: Add check to synchronize abort and flush [ Upstream commit df99446d5c2a63dc6e6920c8090da0e9da6539d5 ] A race condition was observed between qedf_cleanup_fcport() and qedf_process_error_detect()->qedf_initiate_abts(): [2069091.203145] BUG: unable to handle kernel NULL pointer dereference at 0000000000000030 [2069091.213100] IP: [<ffffffffc0666cc6>] qedf_process_error_detect+0x96/0x130 [qedf] [2069091.223391] PGD 1943049067 PUD 194304e067 PMD 0 [2069091.233420] Oops: 0000 [#1] SMP [2069091.361820] CPU: 1 PID: 14751 Comm: kworker/1:46 Kdump: loaded Tainted: P OE ------------ 3.10.0-1160.25.1.el7.x86_64 #1 [2069091.388474] Hardware name: HPE Synergy 480 Gen10/Synergy 480 Gen10 Compute Module, BIOS I42 04/08/2020 [2069091.402148] Workqueue: qedf_io_wq qedf_fp_io_handler [qedf] [2069091.415780] task: ffff9bb9f5190000 ti: ffff9bacaef9c000 task.ti: ffff9bacaef9c000 [2069091.429590] RIP: 0010:[<ffffffffc0666cc6>] [<ffffffffc0666cc6>] qedf_process_error_detect+0x96/0x130 [qedf] [2069091.443666] RSP: 0018:ffff9bacaef9fdb8 EFLAGS: 00010246 [2069091.457692] RAX: 0000000000000000 RBX: ffff9bbbbbfb18a0 RCX: ffffffffc0672310 [2069091.471997] RDX: 00000000000005de RSI: ffffffffc066e7f0 RDI: ffff9beb3f4538d8 [2069091.486130] RBP: ffff9bacaef9fdd8 R08: 0000000000006000 R09: 0000000000006000 [2069091.500321] R10: 0000000000001551 R11: ffffb582996ffff8 R12: ffffb5829b39cc18 [2069091.514779] R13: ffff9badab380c28 R14: ffffd5827f643900 R15: 0000000000000040 [2069091.529472] FS: 0000000000000000(0000) GS:ffff9beb3f440000(0000) knlGS:0000000000000000 [2069091.543926] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [2069091.558942] CR2: 0000000000000030 CR3: 000000193b9a2000 CR4: 00000000007607e0 [2069091.573424] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [2069091.587876] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 [2069091.602007] PKRU: 00000000 [2069091.616010] Call Trace: [2069091.629902] [<ffffffffc0663969>] qedf_process_cqe+0x109/0x2e0 [qedf] [2069091.643941] [<ffffffffc0663b66>] qedf_fp_io_handler+0x26/0x60 [qedf] [2069091.657948] [<ffffffff85ebddcf>] process_one_work+0x17f/0x440 [2069091.672111] [<ffffffff85ebeee6>] worker_thread+0x126/0x3c0 [2069091.686057] [<ffffffff85ebedc0>] ? manage_workers.isra.26+0x2a0/0x2a0 [2069091.700033] [<ffffffff85ec5da1>] kthread+0xd1/0xe0 [2069091.713891] [<ffffffff85ec5cd0>] ? insert_kthread_work+0x40/0x40 Add check in qedf_process_error_detect(). When flush is active, let the cmds be completed from the cleanup contex. Link: https://lore.kernel.org/r/20210624171802.598-1-jhasan@marvell.com Signed-off-by: Javed Hasan <jhasan@marvell.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
2021-06-25 02:18:02 +09:00
"cqe is NULL for io_req %p\n", io_req);
return;
}
QEDF_ERR(&(io_req->fcport->qedf->dbg_ctx), "Error detection CQE, "
"xid=0x%x\n", io_req->xid);
QEDF_ERR(&(io_req->fcport->qedf->dbg_ctx),
"err_warn_bitmap=%08x:%08x\n",
le32_to_cpu(cqe->cqe_info.err_info.err_warn_bitmap_hi),
le32_to_cpu(cqe->cqe_info.err_info.err_warn_bitmap_lo));
QEDF_ERR(&(io_req->fcport->qedf->dbg_ctx), "tx_buff_off=%08x, "
"rx_buff_off=%08x, rx_id=%04x\n",
le32_to_cpu(cqe->cqe_info.err_info.tx_buf_off),
le32_to_cpu(cqe->cqe_info.err_info.rx_buf_off),
le32_to_cpu(cqe->cqe_info.err_info.rx_id));
scsi: qedf: Add check to synchronize abort and flush [ Upstream commit df99446d5c2a63dc6e6920c8090da0e9da6539d5 ] A race condition was observed between qedf_cleanup_fcport() and qedf_process_error_detect()->qedf_initiate_abts(): [2069091.203145] BUG: unable to handle kernel NULL pointer dereference at 0000000000000030 [2069091.213100] IP: [<ffffffffc0666cc6>] qedf_process_error_detect+0x96/0x130 [qedf] [2069091.223391] PGD 1943049067 PUD 194304e067 PMD 0 [2069091.233420] Oops: 0000 [#1] SMP [2069091.361820] CPU: 1 PID: 14751 Comm: kworker/1:46 Kdump: loaded Tainted: P OE ------------ 3.10.0-1160.25.1.el7.x86_64 #1 [2069091.388474] Hardware name: HPE Synergy 480 Gen10/Synergy 480 Gen10 Compute Module, BIOS I42 04/08/2020 [2069091.402148] Workqueue: qedf_io_wq qedf_fp_io_handler [qedf] [2069091.415780] task: ffff9bb9f5190000 ti: ffff9bacaef9c000 task.ti: ffff9bacaef9c000 [2069091.429590] RIP: 0010:[<ffffffffc0666cc6>] [<ffffffffc0666cc6>] qedf_process_error_detect+0x96/0x130 [qedf] [2069091.443666] RSP: 0018:ffff9bacaef9fdb8 EFLAGS: 00010246 [2069091.457692] RAX: 0000000000000000 RBX: ffff9bbbbbfb18a0 RCX: ffffffffc0672310 [2069091.471997] RDX: 00000000000005de RSI: ffffffffc066e7f0 RDI: ffff9beb3f4538d8 [2069091.486130] RBP: ffff9bacaef9fdd8 R08: 0000000000006000 R09: 0000000000006000 [2069091.500321] R10: 0000000000001551 R11: ffffb582996ffff8 R12: ffffb5829b39cc18 [2069091.514779] R13: ffff9badab380c28 R14: ffffd5827f643900 R15: 0000000000000040 [2069091.529472] FS: 0000000000000000(0000) GS:ffff9beb3f440000(0000) knlGS:0000000000000000 [2069091.543926] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [2069091.558942] CR2: 0000000000000030 CR3: 000000193b9a2000 CR4: 00000000007607e0 [2069091.573424] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [2069091.587876] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 [2069091.602007] PKRU: 00000000 [2069091.616010] Call Trace: [2069091.629902] [<ffffffffc0663969>] qedf_process_cqe+0x109/0x2e0 [qedf] [2069091.643941] [<ffffffffc0663b66>] qedf_fp_io_handler+0x26/0x60 [qedf] [2069091.657948] [<ffffffff85ebddcf>] process_one_work+0x17f/0x440 [2069091.672111] [<ffffffff85ebeee6>] worker_thread+0x126/0x3c0 [2069091.686057] [<ffffffff85ebedc0>] ? manage_workers.isra.26+0x2a0/0x2a0 [2069091.700033] [<ffffffff85ec5da1>] kthread+0xd1/0xe0 [2069091.713891] [<ffffffff85ec5cd0>] ? insert_kthread_work+0x40/0x40 Add check in qedf_process_error_detect(). When flush is active, let the cmds be completed from the cleanup contex. Link: https://lore.kernel.org/r/20210624171802.598-1-jhasan@marvell.com Signed-off-by: Javed Hasan <jhasan@marvell.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
2021-06-25 02:18:02 +09:00
/* When flush is active, let the cmds be flushed out from the cleanup context */
if (test_bit(QEDF_RPORT_IN_TARGET_RESET, &io_req->fcport->flags) ||
(test_bit(QEDF_RPORT_IN_LUN_RESET, &io_req->fcport->flags) &&
io_req->sc_cmd->device->lun == (u64)io_req->fcport->lun_reset_lun)) {
QEDF_ERR(&qedf->dbg_ctx,
"Dropping EQE for xid=0x%x as fcport is flushing",
io_req->xid);
return;
}
if (qedf->stop_io_on_error) {
qedf_stop_all_io(qedf);
return;
}
init_completion(&io_req->abts_done);
rval = qedf_initiate_abts(io_req, true);
if (rval)
QEDF_ERR(&(qedf->dbg_ctx), "Failed to queue ABTS.\n");
}
static void qedf_flush_els_req(struct qedf_ctx *qedf,
struct qedf_ioreq *els_req)
{
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_IO,
"Flushing ELS request xid=0x%x refcount=%d.\n", els_req->xid,
kref_read(&els_req->refcount));
/*
* Need to distinguish this from a timeout when calling the
* els_req->cb_func.
*/
els_req->event = QEDF_IOREQ_EV_ELS_FLUSH;
/* Cancel the timer */
cancel_delayed_work_sync(&els_req->timeout_work);
/* Call callback function to complete command */
if (els_req->cb_func && els_req->cb_arg) {
els_req->cb_func(els_req->cb_arg);
els_req->cb_arg = NULL;
}
/* Release kref for original initiate_els */
kref_put(&els_req->refcount, qedf_release_cmd);
}
/* A value of -1 for lun is a wild card that means flush all
* active SCSI I/Os for the target.
*/
void qedf_flush_active_ios(struct qedf_rport *fcport, int lun)
{
struct qedf_ioreq *io_req;
struct qedf_ctx *qedf;
struct qedf_cmd_mgr *cmd_mgr;
int i, rc;
unsigned long flags;
int flush_cnt = 0;
int wait_cnt = 100;
int refcount = 0;
if (!fcport) {
QEDF_ERR(NULL, "fcport is NULL\n");
return;
}
scsi: qedf: Add check for offload before flushing I/Os for target We need to check that a fcport is offloaded before we try to flush any requests. No doing so could lead to undefined results and most likely a crash. Fixes the oops: [ 343.971886] [0000:42:00.3]:[qedf_execute_tmf:2070]:8: wait for tm_cmpl timeout! [ 343.971933] BUG: unable to handle kernel paging request at 00000000000024a8 [ 343.971949] IP: [<ffffffffa06b8cc6>] qedf_flush_active_ios+0x46/0x260 [qedf] [ 343.971952] PGD 42c569067 PUD 4160fe067 PMD 0 [ 343.971954] Oops: 0000 [#1] SMP [ 343.972008] Modules linked in: qedf(OEX) qed(OEX) bnx2i cnic fuse af_packet iscsi_ibft msr xfs intel_rapl sb_edac edac_core x86_pkg_temp_thermal bnx2x geneve intel_powerclamp vxlan coretemp ipmi_ssif ipmi_devintf kvm_intel kvm libiscsi joydev irqbypass crct10dif_pclmul crc32_pclmul ghash_clmulni_intel tg3 ip6_udp_tunnel udp_tunnel mdio libcrc32c iTCO_wdt scsi_transport_iscsi uio drbg iTCO_vendor_support iscsi_boot_sysfs dcdbas(X) ipmi_si ansi_cprng aesni_intel aes_x86_64 lrw gf128mul glue_helper ablk_helper ptp pps_core pcspkr libphy lpc_ich mfd_core cryptd fjes wmi ipmi_msghandler button crc8 libfcoe libfc scsi_transport_fc mei_me mei shpchp processor acpi_pad btrfs xor hid_generic usbhid raid6_pq sd_mod sr_mod cdrom mgag200 crc32c_intel i2c_algo_bit drm_kms_helper syscopyarea sysfillrect sysimgblt [ 343.972020] fb_sys_fops ttm ahci ehci_pci libahci ehci_hcd drm libata usbcore megaraid_sas usb_common sg dm_multipath dm_mod scsi_dh_rdac scsi_dh_emc scsi_dh_alua scsi_mod autofs4 [last unloaded: qedf] [ 343.972022] Supported: Yes, External [ 343.972026] CPU: 30 PID: 12777 Comm: sg_reset Tainted: G W OE X 4.4.73-5-default #1 [ 343.972027] Hardware name: Dell Inc. PowerEdge R720/0X3D66, BIOS 2.1.3 11/20/2013 [ 343.972029] task: ffff88018dfc0e80 ti: ffff88042bd7c000 task.ti: ffff88042bd7c000 [ 343.972036] RIP: 0010:[<ffffffffa06b8cc6>] [<ffffffffa06b8cc6>] qedf_flush_active_ios+0x46/0x260 [qedf] [ 343.972038] RSP: 0018:ffff88042bd7fbe0 EFLAGS: 00010286 [ 343.972039] RAX: 0000000000000000 RBX: ffff88042ce37800 RCX: 0000000000000400 [ 343.972040] RDX: 000000000000060e RSI: ffffffffa06be830 RDI: ffff8807e5072cc0 [ 343.972041] RBP: 0000000000001000 R08: ffffffffa06bff4d R09: ffff88018dd84580 [ 343.972042] R10: 000000000000018b R11: 0000000000000002 R12: 0000000000002003 [ 343.972043] R13: 0000000000000000 R14: 0000000000000000 R15: ffff8807e5072cc0 [ 343.972046] FS: 00007fc1c8809700(0000) GS:ffff88042fbc0000(0000) knlGS:0000000000000000 [ 343.972048] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 343.972049] CR2: 00000000000024a8 CR3: 00000004236ec000 CR4: 00000000001406e0 [ 343.972050] Stack: [ 343.972053] 504c78750607e154 ffffffff810a7d10 ffff88042ce37800 0000000000000010 [ 343.972055] 0000000000002003 ffff8807ff480c48 ffff8807e5072cc0 ffffc90004ec4ff8 [ 343.972057] ffffffffa06b9b86 ffff880800000010 0000000000000282 ffff88042ce37800 [ 343.972058] Call Trace: [ 343.972094] [<ffffffffa06b9b86>] qedf_initiate_tmf+0x346/0x3e0 [qedf] [ 343.972120] [<ffffffffa000fa06>] scsi_try_bus_device_reset+0x26/0x40 [scsi_mod] [ 343.972133] [<ffffffffa001038e>] scsi_ioctl_reset+0x13e/0x260 [scsi_mod] [ 343.972145] [<ffffffffa000f416>] scsi_ioctl+0x136/0x3d0 [scsi_mod] [ 343.972154] [<ffffffff812ff6eb>] blkdev_ioctl+0x6bb/0x950 [ 343.972164] [<ffffffff8123cfed>] block_ioctl+0x3d/0x40 [ 343.972170] [<ffffffff81217e2d>] do_vfs_ioctl+0x2cd/0x4a0 [ 343.972186] [<ffffffff81218074>] SyS_ioctl+0x74/0x80 [ 343.972193] [<ffffffff8160916e>] entry_SYSCALL_64_fastpath+0x12/0x6d [ 343.975285] DWARF2 unwinder stuck at entry_SYSCALL_64_fastpath+0x12/0x6d Signed-off-by: Chad Dupuis <chad.dupuis@cavium.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2018-04-25 22:08:49 +09:00
/* Check that fcport is still offloaded */
if (!test_bit(QEDF_RPORT_SESSION_READY, &fcport->flags)) {
QEDF_ERR(NULL, "fcport is no longer offloaded.\n");
return;
}
qedf = fcport->qedf;
if (!qedf) {
QEDF_ERR(NULL, "qedf is NULL.\n");
return;
}
/* Only wait for all commands to be queued in the Upload context */
if (test_bit(QEDF_RPORT_UPLOADING_CONNECTION, &fcport->flags) &&
(lun == -1)) {
while (atomic_read(&fcport->ios_to_queue)) {
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO,
"Waiting for %d I/Os to be queued\n",
atomic_read(&fcport->ios_to_queue));
if (wait_cnt == 0) {
QEDF_ERR(NULL,
"%d IOs request could not be queued\n",
atomic_read(&fcport->ios_to_queue));
}
msleep(20);
wait_cnt--;
}
}
cmd_mgr = qedf->cmd_mgr;
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO,
"Flush active i/o's num=0x%x fcport=0x%p port_id=0x%06x scsi_id=%d.\n",
atomic_read(&fcport->num_active_ios), fcport,
fcport->rdata->ids.port_id, fcport->rport->scsi_target_id);
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO, "Locking flush mutex.\n");
mutex_lock(&qedf->flush_mutex);
if (lun == -1) {
set_bit(QEDF_RPORT_IN_TARGET_RESET, &fcport->flags);
} else {
set_bit(QEDF_RPORT_IN_LUN_RESET, &fcport->flags);
fcport->lun_reset_lun = lun;
}
for (i = 0; i < FCOE_PARAMS_NUM_TASKS; i++) {
io_req = &cmd_mgr->cmds[i];
if (!io_req)
continue;
if (!io_req->fcport)
continue;
spin_lock_irqsave(&cmd_mgr->lock, flags);
if (io_req->alloc) {
if (!test_bit(QEDF_CMD_OUTSTANDING, &io_req->flags)) {
if (io_req->cmd_type == QEDF_SCSI_CMD)
QEDF_ERR(&qedf->dbg_ctx,
"Allocated but not queued, xid=0x%x\n",
io_req->xid);
}
spin_unlock_irqrestore(&cmd_mgr->lock, flags);
} else {
spin_unlock_irqrestore(&cmd_mgr->lock, flags);
continue;
}
if (io_req->fcport != fcport)
continue;
/* In case of ABTS, CMD_OUTSTANDING is cleared on ABTS response,
* but RRQ is still pending.
* Workaround: Within qedf_send_rrq, we check if the fcport is
* NULL, and we drop the ref on the io_req to clean it up.
*/
if (!test_bit(QEDF_CMD_OUTSTANDING, &io_req->flags)) {
refcount = kref_read(&io_req->refcount);
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO,
"Not outstanding, xid=0x%x, cmd_type=%d refcount=%d.\n",
io_req->xid, io_req->cmd_type, refcount);
scsi: qedf: Cleanup rrq_work after QEDF_CMD_OUTSTANDING is cleared Here is the relevant logs for the problem we are solving: qedf_flush_active_ios:1707]:3: Flush active i/o's num=0x17 fcport=0xffff948168fbcc80 port_id=0x550200 scsi_id=0. qedf_flush_active_ios:1708]:3: Locking flush mutex. qedf_flush_active_ios:1758]:3: Not outstanding, xid=0xaaf, cmd_type=3 refcount=1. qedf_flush_active_ios:1896]:3: Flushed 0x16 I/Os, active=0x1. qedf_flush_active_ios:1901]:3: Flushed 0x16 I/Os, active=0x1 cnt=60. qedf_send_rrq:295]:3: Sending RRQ orig io = ffffb48b8f7d7158, orig_xid = 0xaaf qedf_initiate_els:37]:3: Sending ELS qedf_initiate_els:68]:3: initiate_els els_req = 0xffffb48b8f6d3098 cb_arg = ffff948fd5e4de80 xid = 4c6 qedf_init_mp_req:2172]:3: Entered. qedf_init_mp_task:727]:3: Initializing MP task for cmd_type=4 qedf_initiate_els:134]:3: Ringing doorbell for ELS req qedf_flush_active_ios:1901]:3: Flushed 0x16 I/Os, active=0x2 cnt=20. qedf_cmd_timeout:96]:3: ELS timeout, xid=0x4c6. qedf_rrq_compl:186]:3: Entered. qedf_rrq_compl:204]:3: rrq_compl: orig io = ffffb48b8f7d7158, orig xid = 0xaaf, rrq_xid = 0x4c6, refcount=1 qedf_flush_active_ios:1935]:3: Unlocking flush mutex. qedf_upload_connection:1579]:3: Uploading connection port_id=550200. We found an ABTS command for which CMD_OUTSTANDING was cleared (line 3). For this command, delayed send_rrq was queued, but would take 10 secs to execute. Adding capability to detect that (based on io_req->state that is being introduced), and attempt to cancel rrq_work. If we succeed, we drop the reference and free the io_req. If we cannot, then the els will get sent out and we will wait for 10 secs for it to complete. Signed-off-by: Shyam Sundar <ssundar@marvell.com> Signed-off-by: Saurav Kashyap <skashyap@marvell.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-03-26 16:38:55 +09:00
/* If RRQ work has been queue, try to cancel it and
* free the io_req
*/
if (atomic_read(&io_req->state) ==
QEDFC_CMD_ST_RRQ_WAIT) {
if (cancel_delayed_work_sync
(&io_req->rrq_work)) {
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO,
"Putting reference for pending RRQ work xid=0x%x.\n",
io_req->xid);
/* ID: 003 */
kref_put(&io_req->refcount,
qedf_release_cmd);
}
}
continue;
}
/* Only consider flushing ELS during target reset */
if (io_req->cmd_type == QEDF_ELS &&
lun == -1) {
rc = kref_get_unless_zero(&io_req->refcount);
if (!rc) {
QEDF_ERR(&(qedf->dbg_ctx),
"Could not get kref for ELS io_req=0x%p xid=0x%x.\n",
io_req, io_req->xid);
continue;
}
flush_cnt++;
qedf_flush_els_req(qedf, io_req);
/*
* Release the kref and go back to the top of the
* loop.
*/
goto free_cmd;
}
if (io_req->cmd_type == QEDF_ABTS) {
scsi: qedf: Cleanup rrq_work after QEDF_CMD_OUTSTANDING is cleared Here is the relevant logs for the problem we are solving: qedf_flush_active_ios:1707]:3: Flush active i/o's num=0x17 fcport=0xffff948168fbcc80 port_id=0x550200 scsi_id=0. qedf_flush_active_ios:1708]:3: Locking flush mutex. qedf_flush_active_ios:1758]:3: Not outstanding, xid=0xaaf, cmd_type=3 refcount=1. qedf_flush_active_ios:1896]:3: Flushed 0x16 I/Os, active=0x1. qedf_flush_active_ios:1901]:3: Flushed 0x16 I/Os, active=0x1 cnt=60. qedf_send_rrq:295]:3: Sending RRQ orig io = ffffb48b8f7d7158, orig_xid = 0xaaf qedf_initiate_els:37]:3: Sending ELS qedf_initiate_els:68]:3: initiate_els els_req = 0xffffb48b8f6d3098 cb_arg = ffff948fd5e4de80 xid = 4c6 qedf_init_mp_req:2172]:3: Entered. qedf_init_mp_task:727]:3: Initializing MP task for cmd_type=4 qedf_initiate_els:134]:3: Ringing doorbell for ELS req qedf_flush_active_ios:1901]:3: Flushed 0x16 I/Os, active=0x2 cnt=20. qedf_cmd_timeout:96]:3: ELS timeout, xid=0x4c6. qedf_rrq_compl:186]:3: Entered. qedf_rrq_compl:204]:3: rrq_compl: orig io = ffffb48b8f7d7158, orig xid = 0xaaf, rrq_xid = 0x4c6, refcount=1 qedf_flush_active_ios:1935]:3: Unlocking flush mutex. qedf_upload_connection:1579]:3: Uploading connection port_id=550200. We found an ABTS command for which CMD_OUTSTANDING was cleared (line 3). For this command, delayed send_rrq was queued, but would take 10 secs to execute. Adding capability to detect that (based on io_req->state that is being introduced), and attempt to cancel rrq_work. If we succeed, we drop the reference and free the io_req. If we cannot, then the els will get sent out and we will wait for 10 secs for it to complete. Signed-off-by: Shyam Sundar <ssundar@marvell.com> Signed-off-by: Saurav Kashyap <skashyap@marvell.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-03-26 16:38:55 +09:00
/* ID: 004 */
rc = kref_get_unless_zero(&io_req->refcount);
if (!rc) {
QEDF_ERR(&(qedf->dbg_ctx),
"Could not get kref for abort io_req=0x%p xid=0x%x.\n",
io_req, io_req->xid);
continue;
}
if (lun != -1 && io_req->lun != lun)
goto free_cmd;
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO,
"Flushing abort xid=0x%x.\n", io_req->xid);
if (cancel_delayed_work_sync(&io_req->rrq_work)) {
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO,
scsi: qedf: Cleanup rrq_work after QEDF_CMD_OUTSTANDING is cleared Here is the relevant logs for the problem we are solving: qedf_flush_active_ios:1707]:3: Flush active i/o's num=0x17 fcport=0xffff948168fbcc80 port_id=0x550200 scsi_id=0. qedf_flush_active_ios:1708]:3: Locking flush mutex. qedf_flush_active_ios:1758]:3: Not outstanding, xid=0xaaf, cmd_type=3 refcount=1. qedf_flush_active_ios:1896]:3: Flushed 0x16 I/Os, active=0x1. qedf_flush_active_ios:1901]:3: Flushed 0x16 I/Os, active=0x1 cnt=60. qedf_send_rrq:295]:3: Sending RRQ orig io = ffffb48b8f7d7158, orig_xid = 0xaaf qedf_initiate_els:37]:3: Sending ELS qedf_initiate_els:68]:3: initiate_els els_req = 0xffffb48b8f6d3098 cb_arg = ffff948fd5e4de80 xid = 4c6 qedf_init_mp_req:2172]:3: Entered. qedf_init_mp_task:727]:3: Initializing MP task for cmd_type=4 qedf_initiate_els:134]:3: Ringing doorbell for ELS req qedf_flush_active_ios:1901]:3: Flushed 0x16 I/Os, active=0x2 cnt=20. qedf_cmd_timeout:96]:3: ELS timeout, xid=0x4c6. qedf_rrq_compl:186]:3: Entered. qedf_rrq_compl:204]:3: rrq_compl: orig io = ffffb48b8f7d7158, orig xid = 0xaaf, rrq_xid = 0x4c6, refcount=1 qedf_flush_active_ios:1935]:3: Unlocking flush mutex. qedf_upload_connection:1579]:3: Uploading connection port_id=550200. We found an ABTS command for which CMD_OUTSTANDING was cleared (line 3). For this command, delayed send_rrq was queued, but would take 10 secs to execute. Adding capability to detect that (based on io_req->state that is being introduced), and attempt to cancel rrq_work. If we succeed, we drop the reference and free the io_req. If we cannot, then the els will get sent out and we will wait for 10 secs for it to complete. Signed-off-by: Shyam Sundar <ssundar@marvell.com> Signed-off-by: Saurav Kashyap <skashyap@marvell.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-03-26 16:38:55 +09:00
"Putting ref for cancelled RRQ work xid=0x%x.\n",
io_req->xid);
kref_put(&io_req->refcount, qedf_release_cmd);
}
scsi: qedf: Cleanup rrq_work after QEDF_CMD_OUTSTANDING is cleared Here is the relevant logs for the problem we are solving: qedf_flush_active_ios:1707]:3: Flush active i/o's num=0x17 fcport=0xffff948168fbcc80 port_id=0x550200 scsi_id=0. qedf_flush_active_ios:1708]:3: Locking flush mutex. qedf_flush_active_ios:1758]:3: Not outstanding, xid=0xaaf, cmd_type=3 refcount=1. qedf_flush_active_ios:1896]:3: Flushed 0x16 I/Os, active=0x1. qedf_flush_active_ios:1901]:3: Flushed 0x16 I/Os, active=0x1 cnt=60. qedf_send_rrq:295]:3: Sending RRQ orig io = ffffb48b8f7d7158, orig_xid = 0xaaf qedf_initiate_els:37]:3: Sending ELS qedf_initiate_els:68]:3: initiate_els els_req = 0xffffb48b8f6d3098 cb_arg = ffff948fd5e4de80 xid = 4c6 qedf_init_mp_req:2172]:3: Entered. qedf_init_mp_task:727]:3: Initializing MP task for cmd_type=4 qedf_initiate_els:134]:3: Ringing doorbell for ELS req qedf_flush_active_ios:1901]:3: Flushed 0x16 I/Os, active=0x2 cnt=20. qedf_cmd_timeout:96]:3: ELS timeout, xid=0x4c6. qedf_rrq_compl:186]:3: Entered. qedf_rrq_compl:204]:3: rrq_compl: orig io = ffffb48b8f7d7158, orig xid = 0xaaf, rrq_xid = 0x4c6, refcount=1 qedf_flush_active_ios:1935]:3: Unlocking flush mutex. qedf_upload_connection:1579]:3: Uploading connection port_id=550200. We found an ABTS command for which CMD_OUTSTANDING was cleared (line 3). For this command, delayed send_rrq was queued, but would take 10 secs to execute. Adding capability to detect that (based on io_req->state that is being introduced), and attempt to cancel rrq_work. If we succeed, we drop the reference and free the io_req. If we cannot, then the els will get sent out and we will wait for 10 secs for it to complete. Signed-off-by: Shyam Sundar <ssundar@marvell.com> Signed-off-by: Saurav Kashyap <skashyap@marvell.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-03-26 16:38:55 +09:00
if (cancel_delayed_work_sync(&io_req->timeout_work)) {
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO,
"Putting ref for cancelled tmo work xid=0x%x.\n",
io_req->xid);
qedf_initiate_cleanup(io_req, true);
/* Notify eh_abort handler that ABTS is
* complete
*/
complete(&io_req->abts_done);
clear_bit(QEDF_CMD_IN_ABORT, &io_req->flags);
/* ID: 002 */
kref_put(&io_req->refcount, qedf_release_cmd);
}
flush_cnt++;
goto free_cmd;
}
if (!io_req->sc_cmd)
continue;
if (!io_req->sc_cmd->device) {
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO,
"Device backpointer NULL for sc_cmd=%p.\n",
io_req->sc_cmd);
/* Put reference for non-existent scsi_cmnd */
io_req->sc_cmd = NULL;
qedf_initiate_cleanup(io_req, false);
kref_put(&io_req->refcount, qedf_release_cmd);
continue;
}
if (lun > -1) {
if (io_req->lun != lun)
continue;
}
/*
* Use kref_get_unless_zero in the unlikely case the command
* we're about to flush was completed in the normal SCSI path
*/
rc = kref_get_unless_zero(&io_req->refcount);
if (!rc) {
QEDF_ERR(&(qedf->dbg_ctx), "Could not get kref for "
"io_req=0x%p xid=0x%x\n", io_req, io_req->xid);
continue;
}
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_IO,
"Cleanup xid=0x%x.\n", io_req->xid);
flush_cnt++;
/* Cleanup task and return I/O mid-layer */
qedf_initiate_cleanup(io_req, true);
free_cmd:
scsi: qedf: Cleanup rrq_work after QEDF_CMD_OUTSTANDING is cleared Here is the relevant logs for the problem we are solving: qedf_flush_active_ios:1707]:3: Flush active i/o's num=0x17 fcport=0xffff948168fbcc80 port_id=0x550200 scsi_id=0. qedf_flush_active_ios:1708]:3: Locking flush mutex. qedf_flush_active_ios:1758]:3: Not outstanding, xid=0xaaf, cmd_type=3 refcount=1. qedf_flush_active_ios:1896]:3: Flushed 0x16 I/Os, active=0x1. qedf_flush_active_ios:1901]:3: Flushed 0x16 I/Os, active=0x1 cnt=60. qedf_send_rrq:295]:3: Sending RRQ orig io = ffffb48b8f7d7158, orig_xid = 0xaaf qedf_initiate_els:37]:3: Sending ELS qedf_initiate_els:68]:3: initiate_els els_req = 0xffffb48b8f6d3098 cb_arg = ffff948fd5e4de80 xid = 4c6 qedf_init_mp_req:2172]:3: Entered. qedf_init_mp_task:727]:3: Initializing MP task for cmd_type=4 qedf_initiate_els:134]:3: Ringing doorbell for ELS req qedf_flush_active_ios:1901]:3: Flushed 0x16 I/Os, active=0x2 cnt=20. qedf_cmd_timeout:96]:3: ELS timeout, xid=0x4c6. qedf_rrq_compl:186]:3: Entered. qedf_rrq_compl:204]:3: rrq_compl: orig io = ffffb48b8f7d7158, orig xid = 0xaaf, rrq_xid = 0x4c6, refcount=1 qedf_flush_active_ios:1935]:3: Unlocking flush mutex. qedf_upload_connection:1579]:3: Uploading connection port_id=550200. We found an ABTS command for which CMD_OUTSTANDING was cleared (line 3). For this command, delayed send_rrq was queued, but would take 10 secs to execute. Adding capability to detect that (based on io_req->state that is being introduced), and attempt to cancel rrq_work. If we succeed, we drop the reference and free the io_req. If we cannot, then the els will get sent out and we will wait for 10 secs for it to complete. Signed-off-by: Shyam Sundar <ssundar@marvell.com> Signed-off-by: Saurav Kashyap <skashyap@marvell.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-03-26 16:38:55 +09:00
kref_put(&io_req->refcount, qedf_release_cmd); /* ID: 004 */
}
wait_cnt = 60;
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO,
"Flushed 0x%x I/Os, active=0x%x.\n",
flush_cnt, atomic_read(&fcport->num_active_ios));
/* Only wait for all commands to complete in the Upload context */
if (test_bit(QEDF_RPORT_UPLOADING_CONNECTION, &fcport->flags) &&
(lun == -1)) {
while (atomic_read(&fcport->num_active_ios)) {
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO,
"Flushed 0x%x I/Os, active=0x%x cnt=%d.\n",
flush_cnt,
atomic_read(&fcport->num_active_ios),
wait_cnt);
if (wait_cnt == 0) {
QEDF_ERR(&qedf->dbg_ctx,
"Flushed %d I/Os, active=%d.\n",
flush_cnt,
atomic_read(&fcport->num_active_ios));
for (i = 0; i < FCOE_PARAMS_NUM_TASKS; i++) {
io_req = &cmd_mgr->cmds[i];
if (io_req->fcport &&
io_req->fcport == fcport) {
refcount =
kref_read(&io_req->refcount);
set_bit(QEDF_CMD_DIRTY,
&io_req->flags);
QEDF_ERR(&qedf->dbg_ctx,
"Outstanding io_req =%p xid=0x%x flags=0x%lx, sc_cmd=%p refcount=%d cmd_type=%d.\n",
io_req, io_req->xid,
io_req->flags,
io_req->sc_cmd,
refcount,
io_req->cmd_type);
}
}
WARN_ON(1);
break;
}
msleep(500);
wait_cnt--;
}
}
clear_bit(QEDF_RPORT_IN_LUN_RESET, &fcport->flags);
clear_bit(QEDF_RPORT_IN_TARGET_RESET, &fcport->flags);
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO, "Unlocking flush mutex.\n");
mutex_unlock(&qedf->flush_mutex);
}
/*
* Initiate a ABTS middle path command. Note that we don't have to initialize
* the task context for an ABTS task.
*/
int qedf_initiate_abts(struct qedf_ioreq *io_req, bool return_scsi_cmd_on_abts)
{
struct fc_lport *lport;
struct qedf_rport *fcport = io_req->fcport;
struct fc_rport_priv *rdata;
struct qedf_ctx *qedf;
u16 xid;
u32 r_a_tov = 0;
int rc = 0;
unsigned long flags;
struct fcoe_wqe *sqe;
u16 sqe_idx;
int refcount = 0;
/* Sanity check qedf_rport before dereferencing any pointers */
if (!test_bit(QEDF_RPORT_SESSION_READY, &fcport->flags)) {
QEDF_ERR(NULL, "tgt not offloaded\n");
rc = 1;
goto out;
}
qedf = fcport->qedf;
rdata = fcport->rdata;
if (!rdata || !kref_get_unless_zero(&rdata->kref)) {
QEDF_ERR(&qedf->dbg_ctx, "stale rport\n");
rc = 1;
goto out;
}
r_a_tov = rdata->r_a_tov;
lport = qedf->lport;
if (lport->state != LPORT_ST_READY || !(lport->link_up)) {
QEDF_ERR(&(qedf->dbg_ctx), "link is not ready\n");
rc = 1;
goto drop_rdata_kref;
}
if (atomic_read(&qedf->link_down_tmo_valid) > 0) {
QEDF_ERR(&(qedf->dbg_ctx), "link_down_tmo active.\n");
rc = 1;
goto drop_rdata_kref;
}
/* Ensure room on SQ */
if (!atomic_read(&fcport->free_sqes)) {
QEDF_ERR(&(qedf->dbg_ctx), "No SQ entries available\n");
rc = 1;
goto drop_rdata_kref;
}
if (test_bit(QEDF_RPORT_UPLOADING_CONNECTION, &fcport->flags)) {
QEDF_ERR(&qedf->dbg_ctx, "fcport is uploading.\n");
rc = 1;
goto drop_rdata_kref;
}
if (!test_bit(QEDF_CMD_OUTSTANDING, &io_req->flags) ||
test_bit(QEDF_CMD_IN_CLEANUP, &io_req->flags) ||
test_bit(QEDF_CMD_IN_ABORT, &io_req->flags)) {
QEDF_ERR(&qedf->dbg_ctx,
"io_req xid=0x%x sc_cmd=%p already in cleanup or abort processing or already completed.\n",
io_req->xid, io_req->sc_cmd);
rc = 1;
goto drop_rdata_kref;
}
kref_get(&io_req->refcount);
xid = io_req->xid;
qedf->control_requests++;
qedf->packet_aborts++;
/* Set the command type to abort */
io_req->cmd_type = QEDF_ABTS;
io_req->return_scsi_cmd_on_abts = return_scsi_cmd_on_abts;
set_bit(QEDF_CMD_IN_ABORT, &io_req->flags);
refcount = kref_read(&io_req->refcount);
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_SCSI_TM,
"ABTS io_req xid = 0x%x refcount=%d\n",
xid, refcount);
qedf_cmd_timer_set(qedf, io_req, QEDF_ABORT_TIMEOUT);
spin_lock_irqsave(&fcport->rport_lock, flags);
sqe_idx = qedf_get_sqe_idx(fcport);
sqe = &fcport->sq[sqe_idx];
memset(sqe, 0, sizeof(struct fcoe_wqe));
io_req->task_params->sqe = sqe;
init_initiator_abort_fcoe_task(io_req->task_params);
qedf_ring_doorbell(fcport);
spin_unlock_irqrestore(&fcport->rport_lock, flags);
drop_rdata_kref:
kref_put(&rdata->kref, fc_rport_destroy);
out:
return rc;
}
void qedf_process_abts_compl(struct qedf_ctx *qedf, struct fcoe_cqe *cqe,
struct qedf_ioreq *io_req)
{
uint32_t r_ctl;
uint16_t xid;
int rc;
struct qedf_rport *fcport = io_req->fcport;
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_SCSI_TM, "Entered with xid = "
"0x%x cmd_type = %d\n", io_req->xid, io_req->cmd_type);
xid = io_req->xid;
r_ctl = cqe->cqe_info.abts_info.r_ctl;
/* This was added at a point when we were scheduling abts_compl &
* cleanup_compl on different CPUs and there was a possibility of
* the io_req to be freed from the other context before we got here.
*/
if (!fcport) {
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO,
"Dropping ABTS completion xid=0x%x as fcport is NULL",
io_req->xid);
return;
}
/*
* When flush is active, let the cmds be completed from the cleanup
* context
*/
if (test_bit(QEDF_RPORT_IN_TARGET_RESET, &fcport->flags) ||
test_bit(QEDF_RPORT_IN_LUN_RESET, &fcport->flags)) {
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO,
"Dropping ABTS completion xid=0x%x as fcport is flushing",
io_req->xid);
return;
}
if (!cancel_delayed_work(&io_req->timeout_work)) {
QEDF_ERR(&qedf->dbg_ctx,
"Wasn't able to cancel abts timeout work.\n");
}
switch (r_ctl) {
case FC_RCTL_BA_ACC:
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_SCSI_TM,
"ABTS response - ACC Send RRQ after R_A_TOV\n");
io_req->event = QEDF_IOREQ_EV_ABORT_SUCCESS;
scsi: qedf: Cleanup rrq_work after QEDF_CMD_OUTSTANDING is cleared Here is the relevant logs for the problem we are solving: qedf_flush_active_ios:1707]:3: Flush active i/o's num=0x17 fcport=0xffff948168fbcc80 port_id=0x550200 scsi_id=0. qedf_flush_active_ios:1708]:3: Locking flush mutex. qedf_flush_active_ios:1758]:3: Not outstanding, xid=0xaaf, cmd_type=3 refcount=1. qedf_flush_active_ios:1896]:3: Flushed 0x16 I/Os, active=0x1. qedf_flush_active_ios:1901]:3: Flushed 0x16 I/Os, active=0x1 cnt=60. qedf_send_rrq:295]:3: Sending RRQ orig io = ffffb48b8f7d7158, orig_xid = 0xaaf qedf_initiate_els:37]:3: Sending ELS qedf_initiate_els:68]:3: initiate_els els_req = 0xffffb48b8f6d3098 cb_arg = ffff948fd5e4de80 xid = 4c6 qedf_init_mp_req:2172]:3: Entered. qedf_init_mp_task:727]:3: Initializing MP task for cmd_type=4 qedf_initiate_els:134]:3: Ringing doorbell for ELS req qedf_flush_active_ios:1901]:3: Flushed 0x16 I/Os, active=0x2 cnt=20. qedf_cmd_timeout:96]:3: ELS timeout, xid=0x4c6. qedf_rrq_compl:186]:3: Entered. qedf_rrq_compl:204]:3: rrq_compl: orig io = ffffb48b8f7d7158, orig xid = 0xaaf, rrq_xid = 0x4c6, refcount=1 qedf_flush_active_ios:1935]:3: Unlocking flush mutex. qedf_upload_connection:1579]:3: Uploading connection port_id=550200. We found an ABTS command for which CMD_OUTSTANDING was cleared (line 3). For this command, delayed send_rrq was queued, but would take 10 secs to execute. Adding capability to detect that (based on io_req->state that is being introduced), and attempt to cancel rrq_work. If we succeed, we drop the reference and free the io_req. If we cannot, then the els will get sent out and we will wait for 10 secs for it to complete. Signed-off-by: Shyam Sundar <ssundar@marvell.com> Signed-off-by: Saurav Kashyap <skashyap@marvell.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-03-26 16:38:55 +09:00
rc = kref_get_unless_zero(&io_req->refcount); /* ID: 003 */
if (!rc) {
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_SCSI_TM,
"kref is already zero so ABTS was already completed or flushed xid=0x%x.\n",
io_req->xid);
return;
}
/*
* Dont release this cmd yet. It will be relesed
* after we get RRQ response
*/
queue_delayed_work(qedf->dpc_wq, &io_req->rrq_work,
msecs_to_jiffies(qedf->lport->r_a_tov));
scsi: qedf: Cleanup rrq_work after QEDF_CMD_OUTSTANDING is cleared Here is the relevant logs for the problem we are solving: qedf_flush_active_ios:1707]:3: Flush active i/o's num=0x17 fcport=0xffff948168fbcc80 port_id=0x550200 scsi_id=0. qedf_flush_active_ios:1708]:3: Locking flush mutex. qedf_flush_active_ios:1758]:3: Not outstanding, xid=0xaaf, cmd_type=3 refcount=1. qedf_flush_active_ios:1896]:3: Flushed 0x16 I/Os, active=0x1. qedf_flush_active_ios:1901]:3: Flushed 0x16 I/Os, active=0x1 cnt=60. qedf_send_rrq:295]:3: Sending RRQ orig io = ffffb48b8f7d7158, orig_xid = 0xaaf qedf_initiate_els:37]:3: Sending ELS qedf_initiate_els:68]:3: initiate_els els_req = 0xffffb48b8f6d3098 cb_arg = ffff948fd5e4de80 xid = 4c6 qedf_init_mp_req:2172]:3: Entered. qedf_init_mp_task:727]:3: Initializing MP task for cmd_type=4 qedf_initiate_els:134]:3: Ringing doorbell for ELS req qedf_flush_active_ios:1901]:3: Flushed 0x16 I/Os, active=0x2 cnt=20. qedf_cmd_timeout:96]:3: ELS timeout, xid=0x4c6. qedf_rrq_compl:186]:3: Entered. qedf_rrq_compl:204]:3: rrq_compl: orig io = ffffb48b8f7d7158, orig xid = 0xaaf, rrq_xid = 0x4c6, refcount=1 qedf_flush_active_ios:1935]:3: Unlocking flush mutex. qedf_upload_connection:1579]:3: Uploading connection port_id=550200. We found an ABTS command for which CMD_OUTSTANDING was cleared (line 3). For this command, delayed send_rrq was queued, but would take 10 secs to execute. Adding capability to detect that (based on io_req->state that is being introduced), and attempt to cancel rrq_work. If we succeed, we drop the reference and free the io_req. If we cannot, then the els will get sent out and we will wait for 10 secs for it to complete. Signed-off-by: Shyam Sundar <ssundar@marvell.com> Signed-off-by: Saurav Kashyap <skashyap@marvell.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-03-26 16:38:55 +09:00
atomic_set(&io_req->state, QEDFC_CMD_ST_RRQ_WAIT);
break;
/* For error cases let the cleanup return the command */
case FC_RCTL_BA_RJT:
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_SCSI_TM,
"ABTS response - RJT\n");
io_req->event = QEDF_IOREQ_EV_ABORT_FAILED;
break;
default:
QEDF_ERR(&(qedf->dbg_ctx), "Unknown ABTS response\n");
break;
}
clear_bit(QEDF_CMD_IN_ABORT, &io_req->flags);
if (io_req->sc_cmd) {
if (!io_req->return_scsi_cmd_on_abts)
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_SCSI_TM,
"Not call scsi_done for xid=0x%x.\n",
io_req->xid);
if (io_req->return_scsi_cmd_on_abts)
qedf_scsi_done(qedf, io_req, DID_ERROR);
}
/* Notify eh_abort handler that ABTS is complete */
complete(&io_req->abts_done);
kref_put(&io_req->refcount, qedf_release_cmd);
}
int qedf_init_mp_req(struct qedf_ioreq *io_req)
{
struct qedf_mp_req *mp_req;
struct scsi_sge *mp_req_bd;
struct scsi_sge *mp_resp_bd;
struct qedf_ctx *qedf = io_req->fcport->qedf;
dma_addr_t addr;
uint64_t sz;
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_MP_REQ, "Entered.\n");
mp_req = (struct qedf_mp_req *)&(io_req->mp_req);
memset(mp_req, 0, sizeof(struct qedf_mp_req));
if (io_req->cmd_type != QEDF_ELS) {
mp_req->req_len = sizeof(struct fcp_cmnd);
io_req->data_xfer_len = mp_req->req_len;
} else
mp_req->req_len = io_req->data_xfer_len;
mp_req->req_buf = dma_alloc_coherent(&qedf->pdev->dev, QEDF_PAGE_SIZE,
&mp_req->req_buf_dma, GFP_KERNEL);
if (!mp_req->req_buf) {
QEDF_ERR(&(qedf->dbg_ctx), "Unable to alloc MP req buffer\n");
qedf_free_mp_resc(io_req);
return -ENOMEM;
}
mp_req->resp_buf = dma_alloc_coherent(&qedf->pdev->dev,
QEDF_PAGE_SIZE, &mp_req->resp_buf_dma, GFP_KERNEL);
if (!mp_req->resp_buf) {
QEDF_ERR(&(qedf->dbg_ctx), "Unable to alloc TM resp "
"buffer\n");
qedf_free_mp_resc(io_req);
return -ENOMEM;
}
/* Allocate and map mp_req_bd and mp_resp_bd */
sz = sizeof(struct scsi_sge);
mp_req->mp_req_bd = dma_alloc_coherent(&qedf->pdev->dev, sz,
&mp_req->mp_req_bd_dma, GFP_KERNEL);
if (!mp_req->mp_req_bd) {
QEDF_ERR(&(qedf->dbg_ctx), "Unable to alloc MP req bd\n");
qedf_free_mp_resc(io_req);
return -ENOMEM;
}
mp_req->mp_resp_bd = dma_alloc_coherent(&qedf->pdev->dev, sz,
&mp_req->mp_resp_bd_dma, GFP_KERNEL);
if (!mp_req->mp_resp_bd) {
QEDF_ERR(&(qedf->dbg_ctx), "Unable to alloc MP resp bd\n");
qedf_free_mp_resc(io_req);
return -ENOMEM;
}
/* Fill bd table */
addr = mp_req->req_buf_dma;
mp_req_bd = mp_req->mp_req_bd;
mp_req_bd->sge_addr.lo = U64_LO(addr);
mp_req_bd->sge_addr.hi = U64_HI(addr);
mp_req_bd->sge_len = QEDF_PAGE_SIZE;
/*
* MP buffer is either a task mgmt command or an ELS.
* So the assumption is that it consumes a single bd
* entry in the bd table
*/
mp_resp_bd = mp_req->mp_resp_bd;
addr = mp_req->resp_buf_dma;
mp_resp_bd->sge_addr.lo = U64_LO(addr);
mp_resp_bd->sge_addr.hi = U64_HI(addr);
mp_resp_bd->sge_len = QEDF_PAGE_SIZE;
return 0;
}
/*
* Last ditch effort to clear the port if it's stuck. Used only after a
* cleanup task times out.
*/
static void qedf_drain_request(struct qedf_ctx *qedf)
{
if (test_bit(QEDF_DRAIN_ACTIVE, &qedf->flags)) {
QEDF_ERR(&(qedf->dbg_ctx), "MCP drain already active.\n");
return;
}
/* Set bit to return all queuecommand requests as busy */
set_bit(QEDF_DRAIN_ACTIVE, &qedf->flags);
/* Call qed drain request for function. Should be synchronous */
qed_ops->common->drain(qedf->cdev);
/* Settle time for CQEs to be returned */
msleep(100);
/* Unplug and continue */
clear_bit(QEDF_DRAIN_ACTIVE, &qedf->flags);
}
/*
* Returns SUCCESS if the cleanup task does not timeout, otherwise return
* FAILURE.
*/
int qedf_initiate_cleanup(struct qedf_ioreq *io_req,
bool return_scsi_cmd_on_abts)
{
struct qedf_rport *fcport;
struct qedf_ctx *qedf;
uint16_t xid;
struct e4_fcoe_task_context *task;
int tmo = 0;
int rc = SUCCESS;
unsigned long flags;
struct fcoe_wqe *sqe;
u16 sqe_idx;
int refcount = 0;
fcport = io_req->fcport;
if (!fcport) {
QEDF_ERR(NULL, "fcport is NULL.\n");
return SUCCESS;
}
/* Sanity check qedf_rport before dereferencing any pointers */
if (!test_bit(QEDF_RPORT_SESSION_READY, &fcport->flags)) {
QEDF_ERR(NULL, "tgt not offloaded\n");
rc = 1;
return SUCCESS;
}
qedf = fcport->qedf;
if (!qedf) {
QEDF_ERR(NULL, "qedf is NULL.\n");
return SUCCESS;
}
if (!test_bit(QEDF_CMD_OUTSTANDING, &io_req->flags) ||
test_and_set_bit(QEDF_CMD_IN_CLEANUP, &io_req->flags)) {
QEDF_ERR(&(qedf->dbg_ctx), "io_req xid=0x%x already in "
"cleanup processing or already completed.\n",
io_req->xid);
return SUCCESS;
}
set_bit(QEDF_CMD_IN_CLEANUP, &io_req->flags);
/* Ensure room on SQ */
if (!atomic_read(&fcport->free_sqes)) {
QEDF_ERR(&(qedf->dbg_ctx), "No SQ entries available\n");
/* Need to make sure we clear the flag since it was set */
clear_bit(QEDF_CMD_IN_CLEANUP, &io_req->flags);
return FAILED;
}
if (io_req->cmd_type == QEDF_CLEANUP) {
QEDF_ERR(&qedf->dbg_ctx,
"io_req=0x%x is already a cleanup command cmd_type=%d.\n",
io_req->xid, io_req->cmd_type);
clear_bit(QEDF_CMD_IN_CLEANUP, &io_req->flags);
return SUCCESS;
}
refcount = kref_read(&io_req->refcount);
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_IO,
"Entered xid=0x%x sc_cmd=%p cmd_type=%d flags=0x%lx refcount=%d fcport=%p port_id=0x%06x\n",
io_req->xid, io_req->sc_cmd, io_req->cmd_type, io_req->flags,
refcount, fcport, fcport->rdata->ids.port_id);
/* Cleanup cmds re-use the same TID as the original I/O */
xid = io_req->xid;
io_req->cmd_type = QEDF_CLEANUP;
io_req->return_scsi_cmd_on_abts = return_scsi_cmd_on_abts;
task = qedf_get_task_mem(&qedf->tasks, xid);
init_completion(&io_req->cleanup_done);
spin_lock_irqsave(&fcport->rport_lock, flags);
sqe_idx = qedf_get_sqe_idx(fcport);
sqe = &fcport->sq[sqe_idx];
memset(sqe, 0, sizeof(struct fcoe_wqe));
io_req->task_params->sqe = sqe;
init_initiator_cleanup_fcoe_task(io_req->task_params);
qedf_ring_doorbell(fcport);
spin_unlock_irqrestore(&fcport->rport_lock, flags);
tmo = wait_for_completion_timeout(&io_req->cleanup_done,
QEDF_CLEANUP_TIMEOUT * HZ);
if (!tmo) {
rc = FAILED;
/* Timeout case */
QEDF_ERR(&(qedf->dbg_ctx), "Cleanup command timeout, "
"xid=%x.\n", io_req->xid);
clear_bit(QEDF_CMD_IN_CLEANUP, &io_req->flags);
/* Issue a drain request if cleanup task times out */
QEDF_ERR(&(qedf->dbg_ctx), "Issuing MCP drain request.\n");
qedf_drain_request(qedf);
}
/* If it TASK MGMT handle it, reference will be decreased
* in qedf_execute_tmf
*/
if (io_req->tm_flags == FCP_TMF_LUN_RESET ||
io_req->tm_flags == FCP_TMF_TGT_RESET) {
clear_bit(QEDF_CMD_OUTSTANDING, &io_req->flags);
io_req->sc_cmd = NULL;
complete(&io_req->tm_done);
}
if (io_req->sc_cmd) {
if (!io_req->return_scsi_cmd_on_abts)
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_SCSI_TM,
"Not call scsi_done for xid=0x%x.\n",
io_req->xid);
if (io_req->return_scsi_cmd_on_abts)
qedf_scsi_done(qedf, io_req, DID_ERROR);
}
if (rc == SUCCESS)
io_req->event = QEDF_IOREQ_EV_CLEANUP_SUCCESS;
else
io_req->event = QEDF_IOREQ_EV_CLEANUP_FAILED;
return rc;
}
void qedf_process_cleanup_compl(struct qedf_ctx *qedf, struct fcoe_cqe *cqe,
struct qedf_ioreq *io_req)
{
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_IO, "Entered xid = 0x%x\n",
io_req->xid);
clear_bit(QEDF_CMD_IN_CLEANUP, &io_req->flags);
/* Complete so we can finish cleaning up the I/O */
complete(&io_req->cleanup_done);
}
static int qedf_execute_tmf(struct qedf_rport *fcport, struct scsi_cmnd *sc_cmd,
uint8_t tm_flags)
{
struct qedf_ioreq *io_req;
struct e4_fcoe_task_context *task;
struct qedf_ctx *qedf = fcport->qedf;
struct fc_lport *lport = qedf->lport;
int rc = 0;
uint16_t xid;
int tmo = 0;
int lun = 0;
unsigned long flags;
struct fcoe_wqe *sqe;
u16 sqe_idx;
if (!sc_cmd) {
QEDF_ERR(&qedf->dbg_ctx, "sc_cmd is NULL\n");
return FAILED;
}
lun = (int)sc_cmd->device->lun;
if (!test_bit(QEDF_RPORT_SESSION_READY, &fcport->flags)) {
QEDF_ERR(&(qedf->dbg_ctx), "fcport not offloaded\n");
rc = FAILED;
goto no_flush;
}
io_req = qedf_alloc_cmd(fcport, QEDF_TASK_MGMT_CMD);
if (!io_req) {
QEDF_ERR(&(qedf->dbg_ctx), "Failed TMF");
rc = -EAGAIN;
goto no_flush;
}
if (tm_flags == FCP_TMF_LUN_RESET)
qedf->lun_resets++;
else if (tm_flags == FCP_TMF_TGT_RESET)
qedf->target_resets++;
/* Initialize rest of io_req fields */
io_req->sc_cmd = sc_cmd;
io_req->fcport = fcport;
io_req->cmd_type = QEDF_TASK_MGMT_CMD;
/* Record which cpu this request is associated with */
io_req->cpu = smp_processor_id();
/* Set TM flags */
io_req->io_req_flags = QEDF_READ;
io_req->data_xfer_len = 0;
io_req->tm_flags = tm_flags;
/* Default is to return a SCSI command when an error occurs */
io_req->return_scsi_cmd_on_abts = false;
/* Obtain exchange id */
xid = io_req->xid;
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_SCSI_TM, "TMF io_req xid = "
"0x%x\n", xid);
/* Initialize task context for this IO request */
task = qedf_get_task_mem(&qedf->tasks, xid);
init_completion(&io_req->tm_done);
spin_lock_irqsave(&fcport->rport_lock, flags);
sqe_idx = qedf_get_sqe_idx(fcport);
sqe = &fcport->sq[sqe_idx];
memset(sqe, 0, sizeof(struct fcoe_wqe));
qedf_init_task(fcport, lport, io_req, task, sqe);
qedf_ring_doorbell(fcport);
spin_unlock_irqrestore(&fcport->rport_lock, flags);
set_bit(QEDF_CMD_OUTSTANDING, &io_req->flags);
tmo = wait_for_completion_timeout(&io_req->tm_done,
QEDF_TM_TIMEOUT * HZ);
if (!tmo) {
rc = FAILED;
QEDF_ERR(&(qedf->dbg_ctx), "wait for tm_cmpl timeout!\n");
/* Clear outstanding bit since command timed out */
clear_bit(QEDF_CMD_OUTSTANDING, &io_req->flags);
io_req->sc_cmd = NULL;
} else {
/* Check TMF response code */
if (io_req->fcp_rsp_code == 0)
rc = SUCCESS;
else
rc = FAILED;
}
/*
* Double check that fcport has not gone into an uploading state before
* executing the command flush for the LUN/target.
*/
if (test_bit(QEDF_RPORT_UPLOADING_CONNECTION, &fcport->flags)) {
QEDF_ERR(&qedf->dbg_ctx,
"fcport is uploading, not executing flush.\n");
goto no_flush;
}
/* We do not need this io_req any more */
kref_put(&io_req->refcount, qedf_release_cmd);
if (tm_flags == FCP_TMF_LUN_RESET)
qedf_flush_active_ios(fcport, lun);
else
qedf_flush_active_ios(fcport, -1);
no_flush:
if (rc != SUCCESS) {
QEDF_ERR(&(qedf->dbg_ctx), "task mgmt command failed...\n");
rc = FAILED;
} else {
QEDF_ERR(&(qedf->dbg_ctx), "task mgmt command success...\n");
rc = SUCCESS;
}
return rc;
}
int qedf_initiate_tmf(struct scsi_cmnd *sc_cmd, u8 tm_flags)
{
struct fc_rport *rport = starget_to_rport(scsi_target(sc_cmd->device));
struct fc_rport_libfc_priv *rp = rport->dd_data;
struct qedf_rport *fcport = (struct qedf_rport *)&rp[1];
struct qedf_ctx *qedf;
struct fc_lport *lport = shost_priv(sc_cmd->device->host);
int rc = SUCCESS;
int rval;
struct qedf_ioreq *io_req = NULL;
int ref_cnt = 0;
struct fc_rport_priv *rdata = fcport->rdata;
QEDF_ERR(NULL,
"tm_flags 0x%x sc_cmd %p op = 0x%02x target_id = 0x%x lun=%d\n",
tm_flags, sc_cmd, sc_cmd->cmd_len ? sc_cmd->cmnd[0] : 0xff,
rport->scsi_target_id, (int)sc_cmd->device->lun);
if (!rdata || !kref_get_unless_zero(&rdata->kref)) {
QEDF_ERR(NULL, "stale rport\n");
return FAILED;
}
QEDF_ERR(NULL, "portid=%06x tm_flags =%s\n", rdata->ids.port_id,
(tm_flags == FCP_TMF_TGT_RESET) ? "TARGET RESET" :
"LUN RESET");
if (sc_cmd->SCp.ptr) {
io_req = (struct qedf_ioreq *)sc_cmd->SCp.ptr;
ref_cnt = kref_read(&io_req->refcount);
QEDF_ERR(NULL,
"orig io_req = %p xid = 0x%x ref_cnt = %d.\n",
io_req, io_req->xid, ref_cnt);
}
rval = fc_remote_port_chkready(rport);
if (rval) {
QEDF_ERR(NULL, "device_reset rport not ready\n");
rc = FAILED;
goto tmf_err;
}
rc = fc_block_scsi_eh(sc_cmd);
if (rc)
goto tmf_err;
if (!fcport) {
QEDF_ERR(NULL, "device_reset: rport is NULL\n");
rc = FAILED;
goto tmf_err;
}
qedf = fcport->qedf;
if (!qedf) {
QEDF_ERR(NULL, "qedf is NULL.\n");
rc = FAILED;
goto tmf_err;
}
if (test_bit(QEDF_RPORT_UPLOADING_CONNECTION, &fcport->flags)) {
QEDF_ERR(&qedf->dbg_ctx, "Connection is getting uploaded.\n");
rc = SUCCESS;
goto tmf_err;
}
if (test_bit(QEDF_UNLOADING, &qedf->flags) ||
test_bit(QEDF_DBG_STOP_IO, &qedf->flags)) {
rc = SUCCESS;
goto tmf_err;
}
if (lport->state != LPORT_ST_READY || !(lport->link_up)) {
QEDF_ERR(&(qedf->dbg_ctx), "link is not ready\n");
rc = FAILED;
goto tmf_err;
}
if (test_bit(QEDF_RPORT_UPLOADING_CONNECTION, &fcport->flags)) {
if (!fcport->rdata)
QEDF_ERR(&qedf->dbg_ctx, "fcport %p is uploading.\n",
fcport);
else
QEDF_ERR(&qedf->dbg_ctx,
"fcport %p port_id=%06x is uploading.\n",
fcport, fcport->rdata->ids.port_id);
rc = FAILED;
goto tmf_err;
}
rc = qedf_execute_tmf(fcport, sc_cmd, tm_flags);
tmf_err:
kref_put(&rdata->kref, fc_rport_destroy);
return rc;
}
void qedf_process_tmf_compl(struct qedf_ctx *qedf, struct fcoe_cqe *cqe,
struct qedf_ioreq *io_req)
{
struct fcoe_cqe_rsp_info *fcp_rsp;
clear_bit(QEDF_CMD_OUTSTANDING, &io_req->flags);
fcp_rsp = &cqe->cqe_info.rsp_info;
qedf_parse_fcp_rsp(io_req, fcp_rsp);
io_req->sc_cmd = NULL;
complete(&io_req->tm_done);
}
void qedf_process_unsol_compl(struct qedf_ctx *qedf, uint16_t que_idx,
struct fcoe_cqe *cqe)
{
unsigned long flags;
uint16_t tmp;
uint16_t pktlen = cqe->cqe_info.unsolic_info.pkt_len;
u32 payload_len, crc;
struct fc_frame_header *fh;
struct fc_frame *fp;
struct qedf_io_work *io_work;
u32 bdq_idx;
void *bdq_addr;
struct scsi_bd *p_bd_info;
p_bd_info = &cqe->cqe_info.unsolic_info.bd_info;
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_UNSOL,
"address.hi=%x, address.lo=%x, opaque_data.hi=%x, opaque_data.lo=%x, bdq_prod_idx=%u, len=%u\n",
le32_to_cpu(p_bd_info->address.hi),
le32_to_cpu(p_bd_info->address.lo),
le32_to_cpu(p_bd_info->opaque.fcoe_opaque.hi),
le32_to_cpu(p_bd_info->opaque.fcoe_opaque.lo),
qedf->bdq_prod_idx, pktlen);
bdq_idx = le32_to_cpu(p_bd_info->opaque.fcoe_opaque.lo);
if (bdq_idx >= QEDF_BDQ_SIZE) {
QEDF_ERR(&(qedf->dbg_ctx), "bdq_idx is out of range %d.\n",
bdq_idx);
goto increment_prod;
}
bdq_addr = qedf->bdq[bdq_idx].buf_addr;
if (!bdq_addr) {
QEDF_ERR(&(qedf->dbg_ctx), "bdq_addr is NULL, dropping "
"unsolicited packet.\n");
goto increment_prod;
}
if (qedf_dump_frames) {
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_UNSOL,
"BDQ frame is at addr=%p.\n", bdq_addr);
print_hex_dump(KERN_WARNING, "bdq ", DUMP_PREFIX_OFFSET, 16, 1,
(void *)bdq_addr, pktlen, false);
}
/* Allocate frame */
payload_len = pktlen - sizeof(struct fc_frame_header);
fp = fc_frame_alloc(qedf->lport, payload_len);
if (!fp) {
QEDF_ERR(&(qedf->dbg_ctx), "Could not allocate fp.\n");
goto increment_prod;
}
/* Copy data from BDQ buffer into fc_frame struct */
fh = (struct fc_frame_header *)fc_frame_header_get(fp);
memcpy(fh, (void *)bdq_addr, pktlen);
QEDF_WARN(&qedf->dbg_ctx,
"Processing Unsolicated frame, src=%06x dest=%06x r_ctl=0x%x type=0x%x cmd=%02x\n",
ntoh24(fh->fh_s_id), ntoh24(fh->fh_d_id), fh->fh_r_ctl,
fh->fh_type, fc_frame_payload_op(fp));
/* Initialize the frame so libfc sees it as a valid frame */
crc = fcoe_fc_crc(fp);
fc_frame_init(fp);
fr_dev(fp) = qedf->lport;
fr_sof(fp) = FC_SOF_I3;
fr_eof(fp) = FC_EOF_T;
fr_crc(fp) = cpu_to_le32(~crc);
/*
* We need to return the frame back up to libfc in a non-atomic
* context
*/
io_work = mempool_alloc(qedf->io_mempool, GFP_ATOMIC);
if (!io_work) {
QEDF_WARN(&(qedf->dbg_ctx), "Could not allocate "
"work for I/O completion.\n");
fc_frame_free(fp);
goto increment_prod;
}
memset(io_work, 0, sizeof(struct qedf_io_work));
INIT_WORK(&io_work->work, qedf_fp_io_handler);
/* Copy contents of CQE for deferred processing */
memcpy(&io_work->cqe, cqe, sizeof(struct fcoe_cqe));
io_work->qedf = qedf;
io_work->fp = fp;
queue_work_on(smp_processor_id(), qedf_io_wq, &io_work->work);
increment_prod:
spin_lock_irqsave(&qedf->hba_lock, flags);
/* Increment producer to let f/w know we've handled the frame */
qedf->bdq_prod_idx++;
/* Producer index wraps at uint16_t boundary */
if (qedf->bdq_prod_idx == 0xffff)
qedf->bdq_prod_idx = 0;
writew(qedf->bdq_prod_idx, qedf->bdq_primary_prod);
tmp = readw(qedf->bdq_primary_prod);
writew(qedf->bdq_prod_idx, qedf->bdq_secondary_prod);
tmp = readw(qedf->bdq_secondary_prod);
spin_unlock_irqrestore(&qedf->hba_lock, flags);
}