From: Lee Duncan <lduncan@suse.com>

Commit 1a1975551943 "scsi: fcoe: Fix potential deadlock on &fip->ctlr_lock"
changed locking for fnic/FCoE, but it did so by disabling interrupts
where they weren't disabled before, and this caused FCoE targets
to go offline. Reverting that patch fixed the issue.

But to handle the problem originally addressed by the commit,
instead of modifying the locking, move the work to be done
into a work queue.

Differences in v4:
  - Corrected "Fixes" attributes in both patches
  - Added identifier name in fnic_flush_tx() prototype, for checkpatch

Differences in v3:
  - Added "fixes" clause to the fnic patch, as requested by Hannes

Differences in V2:
  - Fix kerneldoc comments in fnic_flush_tx()

Lee Duncan (1):
  Revert "scsi: fcoe: Fix potential deadlock on &fip->ctlr_lock"

Hannes Reinecke (1):
  fnic: move fnic_fnic_flush_tx() to a work queue

 drivers/scsi/fcoe/fcoe_ctlr.c | 20 ++++++++------------
 drivers/scsi/fnic/fnic.h      |  3 ++-
 drivers/scsi/fnic/fnic_fcs.c  |  5 +++--
 drivers/scsi/fnic/fnic_main.c |  1 +
 drivers/scsi/fnic/fnic_scsi.c |  4 ++--
 5 files changed, 16 insertions(+), 17 deletions(-)

-- 
2.35.3

On Fri, 09 Feb 2024 10:07:33 -0800, Lee Duncan wrote:

> Commit 1a1975551943 "scsi: fcoe: Fix potential deadlock on &fip->ctlr_lock"
> changed locking for fnic/FCoE, but it did so by disabling interrupts
> where they weren't disabled before, and this caused FCoE targets
> to go offline. Reverting that patch fixed the issue.
> 
> But to handle the problem originally addressed by the commit,
> instead of modifying the locking, move the work to be done
> into a work queue.
> 
> [...]

Applied to 6.8/scsi-fixes, thanks!

[1/2] Revert "scsi: fcoe: Fix potential deadlock on &fip->ctlr_lock"
      https://git.kernel.org/mkp/scsi/c/977fe773dcc7
[2/2] fnic: move fnic_fnic_flush_tx() to a work queue
      https://git.kernel.org/mkp/scsi/c/379a58caa199

-- 
Martin K. Petersen	Oracle Linux Engineering