Re: [PATCH v7 0/2] fixup handling of timeouts with deferred QCs

[Igor Pylypiv <ipylypiv@google.com>]

On Tue, Jul 21, 2026 at 03:40:25PM +0900, Damien Le Moal wrote:
> This patch series fixes libata and libsas to correctly handles deferred
> queued commands in case of a timeout error, to avoid excessive delays in
> waking up the scsi EH task.
> 
> Igor,
> 
> My apologies for the churn, but please retest!

Test results look good! Thank you for all the time and effort that you
spent on this issue. I learned a lot!

Tested-by: Igor Pylypiv <ipylypiv@google.com>

Best,
Igor

> Also, I added your Signed-off-by on patch 2 since half of it is yours.
> 
> Martin,
> 
> Once reviewed, I or you can take both patches ?
> 
> Changes from v6:
>  - In patch 1, go back to using direct command completion for retrying
>    deferred QCs, but adds libata EH trigger to avoid problems with the
>    block layer immediately re-issuing the retried commands. This same
>    method also allows directly handling timed out deferred QCs, which
>    simplifies libata EH.
> 
> Changes from v5:
>  - Reworked patch 1 to have deferred command retries all go through libata
>    EH to ensure that we do not run into issues with the block layer
>    immediately re-issuing retried deffered commands (which would create
>    again the problem we are trying to solve). This change necessitate the
>    introduction of a new QC flag and changes to the completion path for
>    commands to ATAPI devices.
> 
> Changes from v4:
>  - Simplified sas_eh_timed_out() code in patch 2
> 
> Changes from v3:
>  - Reimplement ata_scsi_requeue_deferred_qc() in patch 1 as
>    ata_eh_retry_deferred_qc() so that all requeue pathes use the same
>    function.
> 
> Changes from v2:
>  - Modified patch 1 to avoid the problem reported by Sashiko that requeued
>    deferred QCs may be re-ssued immediately by the block layer, thus
>    potentially keeping the device busy. The modification now relies on
>    libata-EH to perform the requeue instead of immediately doing it from
>    the eh_timed_out operation.
>  - Modified patch 2 to use the new helper function defined in patch 1.
> 
> Changes from v1:
>  - Modified patch 1 to ignore timed out deferred QCs in
>    ata_scsi_requeue_deferred_qc() to let ata_scsi_cmd_error_handler()
>    correctly handle this case.
> 
> Damien Le Moal (2):
>   ata: libata-scsi: terminate deferred commands on time out
>   scsi: libsas: terminate deferred commands on time out
> 
>  drivers/ata/libata-eh.c             |  29 ++------
>  drivers/ata/libata-scsi.c           | 109 +++++++++++++++++++++++++---
>  drivers/ata/libata.h                |   3 +-
>  drivers/scsi/libsas/sas_scsi_host.c |  15 ++++
>  include/linux/libata.h              |   4 +
>  include/scsi/libsas.h               |   2 +
>  6 files changed, 127 insertions(+), 35 deletions(-)
> 
> -- 
> 2.55.0
>