From: "Nicholas A. Bellinger" <nab@linux-iscsi.org>
To: target-devel <target-devel@vger.kernel.org>
Cc: linux-rdma <linux-rdma@vger.kernel.org>,
Potnuri Bharat Teja <bharat@chelsio.com>,
Steve Wise <swise@opengridcomputing.com>,
Sagi Grimberg <sagi@grimberg.me>,
Nicholas Bellinger <nab@linux-iscsi.org>
Subject: [PATCH 1/3] target: Fix unknown fabric callback queue-full errors
Date: Mon, 31 Oct 2016 05:25:52 +0000 [thread overview]
Message-ID: <1477891554-26222-2-git-send-email-nab@linux-iscsi.org> (raw)
In-Reply-To: <1477891554-26222-1-git-send-email-nab@linux-iscsi.org>
From: Nicholas Bellinger <nab@linux-iscsi.org>
This patch fixes a set of queue-full response handling
bugs, where outgoing responses are leaked when a fabric
driver is propagating non -EAGAIN or -ENOMEM errors
to target-core.
It introduces TRANSPORT_COMPLETE_QF_ERR state used to
signal when CHECK_CONDITION status should be generated,
when fabric driver ->write_pending(), ->queue_data_in(),
or ->queue_status() callbacks fail with non -EAGAIN or
-ENOMEM errors, and data-transfer should not be retried.
Note all fabric driver -EAGAIN and -ENOMEM errors are
still retried indefinately with associated data-transfer
callbacks, following existing queue-full logic.
Also fix two missing ->queue_status() queue-full cases
related to CMD_T_ABORTED w/ TAS status handling.
Reported-by: Potnuri Bharat Teja <bharat@chelsio.com>
Cc: Potnuri Bharat Teja <bharat@chelsio.com>
Reported-by: Steve Wise <swise@opengridcomputing.com>
Cc: Steve Wise <swise@opengridcomputing.com>
Cc: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
---
drivers/target/target_core_transport.c | 102 ++++++++++++++++++++++-----------
include/target/target_core_base.h | 1 +
2 files changed, 69 insertions(+), 34 deletions(-)
diff --git a/drivers/target/target_core_transport.c b/drivers/target/target_core_transport.c
index 7dfefd6..489aa95 100644
--- a/drivers/target/target_core_transport.c
+++ b/drivers/target/target_core_transport.c
@@ -64,8 +64,9 @@
struct kmem_cache *t10_alua_lba_map_mem_cache;
static void transport_complete_task_attr(struct se_cmd *cmd);
+static int translate_sense_reason(struct se_cmd *cmd, sense_reason_t reason);
static void transport_handle_queue_full(struct se_cmd *cmd,
- struct se_device *dev);
+ struct se_device *dev, int err, bool write_pending);
static int transport_put_cmd(struct se_cmd *cmd);
static void target_complete_ok_work(struct work_struct *work);
@@ -811,7 +812,8 @@ void target_qf_do_work(struct work_struct *work)
if (cmd->t_state == TRANSPORT_COMPLETE_QF_WP)
transport_write_pending_qf(cmd);
- else if (cmd->t_state == TRANSPORT_COMPLETE_QF_OK)
+ else if (cmd->t_state == TRANSPORT_COMPLETE_QF_OK ||
+ cmd->t_state == TRANSPORT_COMPLETE_QF_ERR)
transport_complete_qf(cmd);
}
}
@@ -1720,7 +1722,7 @@ void transport_generic_request_failure(struct se_cmd *cmd,
}
trace_target_cmd_complete(cmd);
ret = cmd->se_tfo->queue_status(cmd);
- if (ret == -EAGAIN || ret == -ENOMEM)
+ if (ret)
goto queue_full;
goto check_stop;
default:
@@ -1731,7 +1733,7 @@ void transport_generic_request_failure(struct se_cmd *cmd,
}
ret = transport_send_check_condition_and_sense(cmd, sense_reason, 0);
- if (ret == -EAGAIN || ret == -ENOMEM)
+ if (ret)
goto queue_full;
check_stop:
@@ -1740,8 +1742,7 @@ void transport_generic_request_failure(struct se_cmd *cmd,
return;
queue_full:
- cmd->t_state = TRANSPORT_COMPLETE_QF_OK;
- transport_handle_queue_full(cmd, cmd->se_dev);
+ transport_handle_queue_full(cmd, cmd->se_dev, ret, false);
}
EXPORT_SYMBOL(transport_generic_request_failure);
@@ -1980,13 +1981,29 @@ static void transport_complete_qf(struct se_cmd *cmd)
int ret = 0;
transport_complete_task_attr(cmd);
+ /*
+ * If a fabric driver ->write_pending() or ->queue_data_in() callback
+ * has returned neither -ENOMEM or -EAGAIN, assume it's fatal and
+ * the same callbacks should not be retried. Return CHECK_CONDITION
+ * if a scsi_status is not already set.
+ *
+ * If a fabric driver ->queue_status() has returned non zero, always
+ * keep retrying no matter what..
+ */
+ if (cmd->t_state == TRANSPORT_COMPLETE_QF_ERR) {
+ if (cmd->scsi_status)
+ goto queue_status;
- if (cmd->se_cmd_flags & SCF_TRANSPORT_TASK_SENSE) {
- trace_target_cmd_complete(cmd);
- ret = cmd->se_tfo->queue_status(cmd);
- goto out;
+ cmd->se_cmd_flags |= SCF_EMULATED_TASK_SENSE;
+ cmd->scsi_status = SAM_STAT_CHECK_CONDITION;
+ cmd->scsi_sense_length = TRANSPORT_SENSE_BUFFER;
+ translate_sense_reason(cmd, TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE);
+ goto queue_status;
}
+ if (cmd->se_cmd_flags & SCF_TRANSPORT_TASK_SENSE)
+ goto queue_status;
+
switch (cmd->data_direction) {
case DMA_FROM_DEVICE:
if (cmd->scsi_status)
@@ -2010,19 +2027,33 @@ static void transport_complete_qf(struct se_cmd *cmd)
break;
}
-out:
if (ret < 0) {
- transport_handle_queue_full(cmd, cmd->se_dev);
+ transport_handle_queue_full(cmd, cmd->se_dev, ret, false);
return;
}
transport_lun_remove_cmd(cmd);
transport_cmd_check_stop_to_fabric(cmd);
}
-static void transport_handle_queue_full(
- struct se_cmd *cmd,
- struct se_device *dev)
+static void transport_handle_queue_full(struct se_cmd *cmd, struct se_device *dev,
+ int err, bool write_pending)
{
+ /*
+ * -EAGAIN or -ENOMEM signals retry of ->write_pending() and/or
+ * ->queue_data_in() callbacks from new process context.
+ *
+ * Otherwise for other errors, transport_complete_qf() will send
+ * CHECK_CONDITION via ->queue_status() instead of attempting to
+ * retry associated fabric driver data-transfer callbacks.
+ */
+ if (err == -EAGAIN || err == -ENOMEM) {
+ cmd->t_state = (write_pending) ? TRANSPORT_COMPLETE_QF_WP :
+ TRANSPORT_COMPLETE_QF_OK;
+ } else {
+ pr_warn_ratelimited("Got unknown fabric queue status: %d\n", err);
+ cmd->t_state = TRANSPORT_COMPLETE_QF_ERR;
+ }
+
spin_lock_irq(&dev->qf_cmd_lock);
list_add_tail(&cmd->se_qf_node, &cmd->se_dev->qf_cmd_list);
atomic_inc_mb(&dev->dev_qf_count);
@@ -2086,7 +2117,7 @@ static void target_complete_ok_work(struct work_struct *work)
WARN_ON(!cmd->scsi_status);
ret = transport_send_check_condition_and_sense(
cmd, 0, 1);
- if (ret == -EAGAIN || ret == -ENOMEM)
+ if (ret)
goto queue_full;
transport_lun_remove_cmd(cmd);
@@ -2112,7 +2143,7 @@ static void target_complete_ok_work(struct work_struct *work)
} else if (rc) {
ret = transport_send_check_condition_and_sense(cmd,
rc, 0);
- if (ret == -EAGAIN || ret == -ENOMEM)
+ if (ret)
goto queue_full;
transport_lun_remove_cmd(cmd);
@@ -2137,7 +2168,7 @@ static void target_complete_ok_work(struct work_struct *work)
if (target_read_prot_action(cmd)) {
ret = transport_send_check_condition_and_sense(cmd,
cmd->pi_err, 0);
- if (ret == -EAGAIN || ret == -ENOMEM)
+ if (ret)
goto queue_full;
transport_lun_remove_cmd(cmd);
@@ -2147,7 +2178,7 @@ static void target_complete_ok_work(struct work_struct *work)
trace_target_cmd_complete(cmd);
ret = cmd->se_tfo->queue_data_in(cmd);
- if (ret == -EAGAIN || ret == -ENOMEM)
+ if (ret)
goto queue_full;
break;
case DMA_TO_DEVICE:
@@ -2160,7 +2191,7 @@ static void target_complete_ok_work(struct work_struct *work)
atomic_long_add(cmd->data_length,
&cmd->se_lun->lun_stats.tx_data_octets);
ret = cmd->se_tfo->queue_data_in(cmd);
- if (ret == -EAGAIN || ret == -ENOMEM)
+ if (ret)
goto queue_full;
break;
}
@@ -2169,7 +2200,7 @@ static void target_complete_ok_work(struct work_struct *work)
queue_status:
trace_target_cmd_complete(cmd);
ret = cmd->se_tfo->queue_status(cmd);
- if (ret == -EAGAIN || ret == -ENOMEM)
+ if (ret)
goto queue_full;
break;
default:
@@ -2183,8 +2214,8 @@ static void target_complete_ok_work(struct work_struct *work)
queue_full:
pr_debug("Handling complete_ok QUEUE_FULL: se_cmd: %p,"
" data_direction: %d\n", cmd, cmd->data_direction);
- cmd->t_state = TRANSPORT_COMPLETE_QF_OK;
- transport_handle_queue_full(cmd, cmd->se_dev);
+
+ transport_handle_queue_full(cmd, cmd->se_dev, ret, false);
}
void target_free_sgl(struct scatterlist *sgl, int nents)
@@ -2434,18 +2465,14 @@ void transport_kunmap_data_sg(struct se_cmd *cmd)
transport_cmd_check_stop(cmd, false, true);
ret = cmd->se_tfo->write_pending(cmd);
- if (ret == -EAGAIN || ret == -ENOMEM)
+ if (ret)
goto queue_full;
- /* fabric drivers should only return -EAGAIN or -ENOMEM as error */
- WARN_ON(ret);
-
- return (!ret) ? 0 : TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
+ return 0;
queue_full:
pr_debug("Handling write_pending QUEUE__FULL: se_cmd: %p\n", cmd);
- cmd->t_state = TRANSPORT_COMPLETE_QF_WP;
- transport_handle_queue_full(cmd, cmd->se_dev);
+ transport_handle_queue_full(cmd, cmd->se_dev, ret, true);
return 0;
}
EXPORT_SYMBOL(transport_generic_new_cmd);
@@ -2455,10 +2482,10 @@ static void transport_write_pending_qf(struct se_cmd *cmd)
int ret;
ret = cmd->se_tfo->write_pending(cmd);
- if (ret == -EAGAIN || ret == -ENOMEM) {
+ if (ret) {
pr_debug("Handling write_pending QUEUE__FULL: se_cmd: %p\n",
cmd);
- transport_handle_queue_full(cmd, cmd->se_dev);
+ transport_handle_queue_full(cmd, cmd->se_dev, ret, true);
}
}
@@ -2951,6 +2978,8 @@ static int __transport_check_aborted_status(struct se_cmd *cmd, int send_status)
__releases(&cmd->t_state_lock)
__acquires(&cmd->t_state_lock)
{
+ int ret;
+
assert_spin_locked(&cmd->t_state_lock);
WARN_ON_ONCE(!irqs_disabled());
@@ -2974,7 +3003,9 @@ static int __transport_check_aborted_status(struct se_cmd *cmd, int send_status)
trace_target_cmd_complete(cmd);
spin_unlock_irq(&cmd->t_state_lock);
- cmd->se_tfo->queue_status(cmd);
+ ret = cmd->se_tfo->queue_status(cmd);
+ if (ret)
+ transport_handle_queue_full(cmd, cmd->se_dev, ret, false);
spin_lock_irq(&cmd->t_state_lock);
return 1;
@@ -2995,6 +3026,7 @@ int transport_check_aborted_status(struct se_cmd *cmd, int send_status)
void transport_send_task_abort(struct se_cmd *cmd)
{
unsigned long flags;
+ int ret;
spin_lock_irqsave(&cmd->t_state_lock, flags);
if (cmd->se_cmd_flags & (SCF_SENT_CHECK_CONDITION)) {
@@ -3030,7 +3062,9 @@ void transport_send_task_abort(struct se_cmd *cmd)
cmd->t_task_cdb[0], cmd->tag);
trace_target_cmd_complete(cmd);
- cmd->se_tfo->queue_status(cmd);
+ ret = cmd->se_tfo->queue_status(cmd);
+ if (ret)
+ transport_handle_queue_full(cmd, cmd->se_dev, ret, false);
}
static void target_tmr_work(struct work_struct *work)
diff --git a/include/target/target_core_base.h b/include/target/target_core_base.h
index c211900..6a5e439 100644
--- a/include/target/target_core_base.h
+++ b/include/target/target_core_base.h
@@ -119,6 +119,7 @@ enum transport_state_table {
TRANSPORT_ISTATE_PROCESSING = 11,
TRANSPORT_COMPLETE_QF_WP = 18,
TRANSPORT_COMPLETE_QF_OK = 19,
+ TRANSPORT_COMPLETE_QF_ERR = 20,
};
/* Used for struct se_cmd->se_cmd_flags */
--
1.9.1
next prev parent reply other threads:[~2016-10-31 5:25 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-10-31 5:25 [PATCH 0/3] target: Fix queue-full callback error signaling Nicholas A. Bellinger
2016-10-31 5:25 ` Nicholas A. Bellinger [this message]
2016-10-31 5:25 ` [PATCH 2/3] iscsi-target: Propigate queue_data_in + queue_status errors Nicholas A. Bellinger
[not found] ` <1477891554-26222-1-git-send-email-nab-IzHhD5pYlfBP7FQvKIMDCQ@public.gmane.org>
2016-10-31 5:25 ` [PATCH 3/3] iser-target: Fix queue-full response handling Nicholas A. Bellinger
2017-03-23 5:36 ` [PATCH 0/3] target: Fix queue-full callback error signaling Potnuri Bharat Teja
2017-03-31 4:32 ` Nicholas A. Bellinger
2017-04-03 6:23 ` Potnuri Bharat Teja
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1477891554-26222-2-git-send-email-nab@linux-iscsi.org \
--to=nab@linux-iscsi.org \
--cc=bharat@chelsio.com \
--cc=linux-rdma@vger.kernel.org \
--cc=sagi@grimberg.me \
--cc=swise@opengridcomputing.com \
--cc=target-devel@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox