From: Hannes Reinecke <hare@suse.de>
To: Himanshu Madhani <himanshu.madhani@qlogic.com>,
target-devel@vger.kernel.org, nab@linux-iscsi.org
Cc: giridhar.malavali@qlogic.com, linux-scsi@vger.kernel.org
Subject: Re: [PATCH 12/20] target/tmr: LUN reset cause cmd premature free.
Date: Wed, 9 Dec 2015 08:03:13 +0100 [thread overview]
Message-ID: <5667D231.9080702@suse.de> (raw)
In-Reply-To: <1449535747-2850-13-git-send-email-himanshu.madhani@qlogic.com>
On 12/08/2015 01:48 AM, Himanshu Madhani wrote:
> From: Quinn Tran <quinn.tran@qlogic.com>
>
> During LUN/Target reset, the TMR code attempt to intercept
> cmds and try to aborted them. Current code assume cmds are
> always intercepted at the back end device. The cleanup code
> would issue a "queue_status() & check_stop_free()" to terminate
> the command. However, when a cmd is intercepted at the front
> end/Fabric layer, current code introduce premature free or
> cause Fabric to double free.
>
> When command is intercepted at Fabric layer, it means a
> check_stop_free(cmd_kref--) has been called. The extra
> check_stop_free in the Lun Reset cleanup code causes early
> free. When a cmd in the Fabric layer is completed, the normal
> free code adds another another free which introduce a double free.
>
> To fix the issue:
> - add a new flag/CMD_T_SENT_STATUS to track command that have
> made it down to fabric layer after back end good/bad completion.
> - if cmd reach Fabric Layer at Lun Reset time, add an extra
> cmd_kref count to prevent premature free.
>
> Signed-off-by: Quinn Tran <quinn.tran@qlogic.com>
> Signed-off-by: Himanshu Madhani <himanshu.madhani@qlogic.com>
> ---
> drivers/target/target_core_tmr.c | 33 +++++++++++++++++++++++++++++++-
> drivers/target/target_core_transport.c | 30 +++++++++++++++++++++++++++++
> include/target/target_core_base.h | 1 +
> 3 files changed, 63 insertions(+), 1 deletions(-)
>
> diff --git a/drivers/target/target_core_tmr.c b/drivers/target/target_core_tmr.c
> index 28fb301..41f8b57 100644
> --- a/drivers/target/target_core_tmr.c
> +++ b/drivers/target/target_core_tmr.c
> @@ -243,7 +243,9 @@ static void core_tmr_drain_state_list(
> {
> LIST_HEAD(drain_task_list);
> struct se_cmd *cmd, *next;
> - unsigned long flags;
> + unsigned long flags, flags2;
> + int rmkref;
> + struct se_session *se_sess;
>
> /*
> * Complete outstanding commands with TASK_ABORTED SAM status.
> @@ -282,6 +284,16 @@ static void core_tmr_drain_state_list(
> if (prout_cmd == cmd)
> continue;
>
> + se_sess = cmd->se_sess;
> + /* take an extra kref to prevent cmd free race condition. */
> + spin_lock_irqsave(&se_sess->sess_cmd_lock, flags2);
> + if (!kref_get_unless_zero(&cmd->cmd_kref)) {
> + /* cmd is already in the free process */
> + spin_unlock_irqrestore(&se_sess->sess_cmd_lock, flags2);
> + continue;
> + }
> + spin_unlock_irqrestore(&se_sess->sess_cmd_lock, flags2);
> +
> list_move_tail(&cmd->state_list, &drain_task_list);
> cmd->state_active = false;
> }
> @@ -320,9 +332,28 @@ static void core_tmr_drain_state_list(
> target_stop_cmd(cmd, &flags);
>
> cmd->transport_state |= CMD_T_ABORTED;
> +
> + /* CMD_T_SENT_STATUS: cmd is down in fabric layer.
> + * A check stop has been called. Keep the extra kref
> + * from above because core_tmr_handle_tas_abort will
> + * generate another check_stop.
> + *
> + * !CMD_T_SENT_STATUS: cmd intercepted at back end.
> + * Remove the extra kref from above because only
> + * 1 check_stop is required or generated by
> + * core_tmr_handle_tas_abort()
> + */
> + rmkref = 0;
> + if (!((cmd->t_state == TRANSPORT_COMPLETE) &&
> + (cmd->transport_state & CMD_T_SENT_STATUS)))
> + rmkref = 1;
> +
> spin_unlock_irqrestore(&cmd->t_state_lock, flags);
>
> core_tmr_handle_tas_abort(tmr_nacl, cmd, tas);
> +
> + if (rmkref)
> + target_put_sess_cmd(cmd);
> }
> }
>
> diff --git a/drivers/target/target_core_transport.c b/drivers/target/target_core_transport.c
> index 4fdcee2..cdd18bf 100644
> --- a/drivers/target/target_core_transport.c
> +++ b/drivers/target/target_core_transport.c
> @@ -639,9 +639,14 @@ void transport_cmd_finish_abort(struct se_cmd *cmd, int remove)
> static void target_complete_failure_work(struct work_struct *work)
> {
> struct se_cmd *cmd = container_of(work, struct se_cmd, work);
> + unsigned long flags;
>
> transport_generic_request_failure(cmd,
> TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE);
> +
> + spin_lock_irqsave(&cmd->t_state_lock, flags);
> + cmd->transport_state |= CMD_T_SENT_STATUS;
> + spin_unlock_irqrestore(&cmd->t_state_lock, flags);
> }
>
> /*
> @@ -1659,6 +1664,7 @@ void transport_generic_request_failure(struct se_cmd *cmd,
> sense_reason_t sense_reason)
> {
> int ret = 0, post_ret = 0;
> + unsigned long flags;
>
> pr_debug("-----[ Storage Engine Exception for cmd: %p ITT: 0x%08llx"
> " CDB: 0x%02x\n", cmd, cmd->tag, cmd->t_task_cdb[0]);
> @@ -1670,6 +1676,10 @@ void transport_generic_request_failure(struct se_cmd *cmd,
> (cmd->transport_state & CMD_T_STOP) != 0,
> (cmd->transport_state & CMD_T_SENT) != 0);
>
> + spin_lock_irqsave(&cmd->t_state_lock, flags);
> + cmd->transport_state |= CMD_T_SENT_STATUS;
> + spin_unlock_irqrestore(&cmd->t_state_lock, flags);
> +
> /*
> * For SAM Task Attribute emulation for failed struct se_cmd
> */
> @@ -1951,6 +1961,7 @@ static void transport_complete_task_attr(struct se_cmd *cmd)
> static void transport_complete_qf(struct se_cmd *cmd)
> {
> int ret = 0;
> + unsigned long flags;
>
> transport_complete_task_attr(cmd);
>
> @@ -1986,6 +1997,10 @@ out:
> }
> transport_lun_remove_cmd(cmd);
> transport_cmd_check_stop_to_fabric(cmd);
> +
> + spin_lock_irqsave(&cmd->t_state_lock, flags);
> + cmd->transport_state |= CMD_T_SENT_STATUS;
> + spin_unlock_irqrestore(&cmd->t_state_lock, flags);
> }
>
> static void transport_handle_queue_full(
> @@ -2032,6 +2047,7 @@ static void target_complete_ok_work(struct work_struct *work)
> {
> struct se_cmd *cmd = container_of(work, struct se_cmd, work);
> int ret;
> + unsigned long flags;
>
> /*
> * Check if we need to move delayed/dormant tasks from cmds on the
> @@ -2060,6 +2076,10 @@ static void target_complete_ok_work(struct work_struct *work)
>
> transport_lun_remove_cmd(cmd);
> transport_cmd_check_stop_to_fabric(cmd);
> +
> + spin_lock_irqsave(&cmd->t_state_lock, flags);
> + cmd->transport_state |= CMD_T_SENT_STATUS;
> + spin_unlock_irqrestore(&cmd->t_state_lock, flags);
> return;
> }
> /*
> @@ -2086,6 +2106,11 @@ static void target_complete_ok_work(struct work_struct *work)
>
> transport_lun_remove_cmd(cmd);
> transport_cmd_check_stop_to_fabric(cmd);
> +
> + spin_lock_irqsave(&cmd->t_state_lock, flags);
> + cmd->transport_state |= CMD_T_SENT_STATUS;
> + spin_unlock_irqrestore(&cmd->t_state_lock, flags);
> +
> return;
> }
> }
> @@ -2136,6 +2161,7 @@ queue_rsp:
> ret = cmd->se_tfo->queue_status(cmd);
> if (ret == -EAGAIN || ret == -ENOMEM)
> goto queue_full;
> +
> break;
> default:
> break;
> @@ -2143,6 +2169,10 @@ queue_rsp:
>
> transport_lun_remove_cmd(cmd);
> transport_cmd_check_stop_to_fabric(cmd);
> +
> + spin_lock_irqsave(&cmd->t_state_lock, flags);
> + cmd->transport_state |= CMD_T_SENT_STATUS;
> + spin_unlock_irqrestore(&cmd->t_state_lock, flags);
> return;
>
> queue_full:
> diff --git a/include/target/target_core_base.h b/include/target/target_core_base.h
> index aabf0ac..efccd71 100644
> --- a/include/target/target_core_base.h
> +++ b/include/target/target_core_base.h
> @@ -490,6 +490,7 @@ struct se_cmd {
> #define CMD_T_DEV_ACTIVE (1 << 7)
> #define CMD_T_REQUEST_STOP (1 << 8)
> #define CMD_T_BUSY (1 << 9)
> +#define CMD_T_SENT_STATUS (1 << 10)
> spinlock_t t_state_lock;
> struct kref cmd_kref;
> struct completion t_transport_stop_comp;
>
Same here: using bitops would save you taking the spinlock when
modifying flags.
Cheers,
Hannes
--
Dr. Hannes Reinecke zSeries & Storage
hare@suse.de +49 911 74053 688
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: J. Hawn, J. Guild, F. Imendörffer, HRB 16746 (AG Nürnberg)
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
next prev parent reply other threads:[~2015-12-09 7:03 UTC|newest]
Thread overview: 51+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-12-08 0:48 [PATCH 00/20] qla2xxx: Patches for target-pending branch Himanshu Madhani
2015-12-08 0:48 ` [PATCH 01/20] qla2xxx: Enable Extended Login support Himanshu Madhani
2015-12-08 15:51 ` Hannes Reinecke
2015-12-08 19:35 ` Himanshu Madhani
2015-12-08 0:48 ` [PATCH 02/20] qla2xxx: Enable Exchange offload support Himanshu Madhani
2015-12-08 15:52 ` Hannes Reinecke
2015-12-08 0:48 ` [PATCH 03/20] qla2xxx: Enable Target counters in DebugFS Himanshu Madhani
2015-12-08 15:52 ` Hannes Reinecke
2015-12-08 0:48 ` [PATCH 04/20] qla2xxx: Add FW resource count " Himanshu Madhani
2015-12-08 15:53 ` Hannes Reinecke
2015-12-08 0:48 ` [PATCH 05/20] qla2xxx: Added interface to send ELS commands from driver Himanshu Madhani
2015-12-08 2:10 ` kbuild test robot
2015-12-08 2:10 ` [PATCH] qla2xxx: fix ifnullfree.cocci warnings kbuild test robot
2015-12-08 15:54 ` [PATCH 05/20] qla2xxx: Added interface to send ELS commands from driver Hannes Reinecke
2015-12-08 0:48 ` [PATCH 06/20] qla2xxx: Delete session if initiator is gone from FW Himanshu Madhani
2015-12-08 1:41 ` kbuild test robot
2015-12-08 15:58 ` Hannes Reinecke
2015-12-08 0:48 ` [PATCH 07/20] qla2xxx: Wait for all conflicts before ack'ing PLOGI Himanshu Madhani
2015-12-08 16:00 ` Hannes Reinecke
2015-12-08 0:48 ` [PATCH 08/20] qla2xxx: Replace QLA_TGT_STATE_ABORTED with a bit Himanshu Madhani
2015-12-08 16:01 ` Hannes Reinecke
2015-12-08 0:48 ` [PATCH 09/20] qla2xxx: Change check_stop_free to always return 1 Himanshu Madhani
2015-12-08 2:33 ` Christoph Hellwig
2015-12-09 6:56 ` Hannes Reinecke
2015-12-10 1:06 ` Quinn Tran
2015-12-08 0:48 ` [PATCH 10/20] qla2xxx: Fix interaction issue between qla2xxx and Target Core Module Himanshu Madhani
2015-12-08 2:37 ` Christoph Hellwig
2015-12-09 22:07 ` Quinn Tran
2015-12-14 10:34 ` Christoph Hellwig
2015-12-14 21:59 ` Quinn Tran
2015-12-09 7:01 ` Hannes Reinecke
2015-12-09 22:41 ` Quinn Tran
2015-12-08 0:48 ` [PATCH 11/20] qla2xxx: Add TAS detection for kernel 3.15 n newer Himanshu Madhani
2015-12-08 2:48 ` Christoph Hellwig
2015-12-09 20:24 ` Quinn Tran
2015-12-14 10:37 ` Christoph Hellwig
2015-12-14 22:00 ` Quinn Tran
2015-12-09 7:02 ` Hannes Reinecke
2015-12-08 0:48 ` [PATCH 12/20] target/tmr: LUN reset cause cmd premature free Himanshu Madhani
2015-12-08 2:48 ` Christoph Hellwig
2015-12-09 20:11 ` Quinn Tran
2016-01-04 7:44 ` Bart Van Assche
2015-12-09 7:03 ` Hannes Reinecke [this message]
2015-12-08 0:49 ` [PATCH 13/20] qla2xxx: Remove dependency on hardware_lock to reduce lock contention Himanshu Madhani
2015-12-08 0:49 ` [PATCH 14/20] qla2xxx: Add irq affinity notification Himanshu Madhani
2015-12-08 0:49 ` [PATCH 15/20] qla2xxx: Add selective command queuing Himanshu Madhani
2015-12-08 0:49 ` [PATCH 16/20] qla2xxx: Move atioq to a different lock to reduce lock contention Himanshu Madhani
2015-12-08 0:49 ` [PATCH 17/20] qla2xxx: Disable ZIO at start time Himanshu Madhani
2015-12-08 0:49 ` [PATCH 18/20] qla2xxx: Set all queues to 4k Himanshu Madhani
2015-12-08 0:49 ` [PATCH 19/20] qla2xxx: Add bulk send for atio & ctio completion paths Himanshu Madhani
2015-12-08 0:49 ` [PATCH 20/20] qla2xxx: Check for online flag instead of active reset when transmitting responses Himanshu Madhani
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=5667D231.9080702@suse.de \
--to=hare@suse.de \
--cc=giridhar.malavali@qlogic.com \
--cc=himanshu.madhani@qlogic.com \
--cc=linux-scsi@vger.kernel.org \
--cc=nab@linux-iscsi.org \
--cc=target-devel@vger.kernel.org \
--subject='Re: [PATCH 12/20] target/tmr: LUN reset cause cmd premature free.' \
/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
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.