From: Mike Christie <michael.christie@oracle.com>
To: Chengfeng Ye <dg573847474@gmail.com>,
njavali@marvell.com, mrangankar@marvell.com,
GR-QLogic-Storage-Upstream@marvell.com, jejb@linux.ibm.com,
martin.petersen@oracle.com, lduncan@suse.com, cleech@redhat.com
Cc: linux-scsi@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH] scsi: qla4xxx: fix potential deadlock on frwd/back lock
Date: Fri, 29 Sep 2023 11:03:47 -0500 [thread overview]
Message-ID: <b3a5e717-90ab-42f8-8fe1-4a24fff404c3@oracle.com> (raw)
In-Reply-To: <20230928145401.32918-1-dg573847474@gmail.com>
On 9/28/23 9:54 AM, Chengfeng Ye wrote:
> iscsi_complete_pdu() is called by several rx callback, under
> softirq context. Thus the callsite of it inside qla4xxx_task_work()
> should better disable bottom half, as work queue is executed under
> process context, it needs to protect race with softirq context
> locking to avoid deadlock.
>
> <potential deadlock #1>
> qla4xxx_task_work()
> --> iscsi_complete_pdu()
> --> spin_lock(&conn->session->back_lock);
> <interrupt>
> --> iscsi_tcp_data_recv_done()
> --> iscsi_complete_pdu()
> --> spin_lock(&conn->session->back_lock) (deadlock)
>
This will not happen, because qla4xxx does not use the libiscsi_tcp
code. If you had a iscsi session with qla4xxx and another
iscsi session with iscsi_tcp, then you could get an interrupt like
this, but the "conn->session->back_lock" that's taken above would
be for the iscsi_tcp session.
So you would need to make sure that the above can't happen for
iscsi_tcp an iscsi_tcp session. Same with the below analysis.
> <potential deadlock #2>
> qla4xxx_task_work()
> --> iscsi_complete_pdu()
> --> __iscsi_complete_pdu()
> --> spin_lock(&conn->session->frwd_lock)
> <interrupt>
> --> iscsi_tcp_data_recv_done()
> --> iscsi_complete_pdu()
> --> __iscsi_complete_pdu()
> --> spin_lock(&conn->session->frwd_lock) (deadlock)
>
> This flaw was found by an experimental static analysis tool I am
> developing for irq-related deadlock.
>
> To avoid the problem, disable bh inside qla4xxx_task_work() before
> calling iscsi_complete_pdu().
>
> Signed-off-by: Chengfeng Ye <dg573847474@gmail.com>
> ---
> drivers/scsi/qla4xxx/ql4_os.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/drivers/scsi/qla4xxx/ql4_os.c b/drivers/scsi/qla4xxx/ql4_os.c
> index 675332e49a7b..c60781148e6c 100644
> --- a/drivers/scsi/qla4xxx/ql4_os.c
> +++ b/drivers/scsi/qla4xxx/ql4_os.c
> @@ -3382,7 +3382,9 @@ static void qla4xxx_task_work(struct work_struct *wdata)
> hdr->itt = itt;
> data = task_data->resp_buffer + hdr_len;
> data_len = task_data->resp_len - hdr_len;
> + local_bh_disable();
> iscsi_complete_pdu(conn, hdr, data, data_len);
> + local_bh_enable();
> break;
> default:
> ql4_printk(KERN_ERR, ha, "Passthru failed status = 0x%x\n",
next prev parent reply other threads:[~2023-09-29 16:04 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-09-28 14:54 Chengfeng Ye
2023-09-29 16:03 ` Mike Christie [this message]
2023-09-29 16:11 ` Chengfeng Ye
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=b3a5e717-90ab-42f8-8fe1-4a24fff404c3@oracle.com \
--to=michael.christie@oracle.com \
--cc=GR-QLogic-Storage-Upstream@marvell.com \
--cc=cleech@redhat.com \
--cc=dg573847474@gmail.com \
--cc=jejb@linux.ibm.com \
--cc=lduncan@suse.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-scsi@vger.kernel.org \
--cc=martin.petersen@oracle.com \
--cc=mrangankar@marvell.com \
--cc=njavali@marvell.com \
/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
all inboxes | Powered by JetHome®