In the case when a SCSI_CMD times out, bnx2fc will initiate the sending of the
ABTS. However, if the SCSI layer's SCSI command timer also times out, it'll
instantiate a task abort of the same xid.
The race condition this patch tries to fix is as follows:
SCSI_CMD timeout (20s)
thread 1 thread 2
send ABTS
rx ABTS cmpl
task abort_eh
explicit LOGO since ABTS was engaged
CLEANUP cmpl
SCSI_CMD cmpl (ABTS cmpl)
instantiate RRQ
wait 10s
attempt to send RRQ (because of LOGO, it wouldn't continue)
Note that there is no call to scsi_done for this SCSI_CMD cmpletion
in this path.
The patch changes the path of execution to call scsi_done immediately
instead of instantiating the RRQ.
Signed-off-by: Eddie Wai <eddie.wai@xxxxxxxxxxxx>
---
drivers/scsi/bnx2fc/bnx2fc_io.c | 6 ++++++
1 files changed, 6 insertions(+), 0 deletions(-)
diff --git a/drivers/scsi/bnx2fc/bnx2fc_io.c b/drivers/scsi/bnx2fc/bnx2fc_io.c
index 575142e..ed88089 100644
--- a/drivers/scsi/bnx2fc/bnx2fc_io.c
+++ b/drivers/scsi/bnx2fc/bnx2fc_io.c
@@ -1246,6 +1246,12 @@ int bnx2fc_eh_abort(struct scsi_cmnd *sc_cmd)
kref_put(&io_req->refcount,
bnx2fc_cmd_release); /* drop timer hold */
rc = bnx2fc_expl_logo(lport, io_req);
+ /* This only occurs when an task abort was requested while ABTS
+ is in progress. Setting the IO_CLEANUP flag will skip the
+ RRQ process in the case when the fw generated SCSI_CMD cmpl
+ was a result from the ABTS request rather than the CLEANUP
+ request */
+ set_bit(BNX2FC_FLAG_IO_CLEANUP, &io_req->req_flags);
goto out;
}
--
1.7.1
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
