Re: [PATCH v5 0/4] ata: fix deferred QC handling for port multipliers

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Wed, May 27, 2026 at 10:26:02AM +0200, Niklas Cassel wrote:
> On Tue, May 26, 2026 at 08:47:53PM +0200, David Dreschner wrote:
> > Hello Niklas,
> > 
> > thank you very much for fixing the issue!
> > 
> > I tested the first three patches manually on my Debian 13 machine with
> > kernel 6.12.90 and can indeed confirm that your patches fixes the issue. The
> > fourth patch had some merge-conflicts and were therefore skipped - I'm
> > unable to test the changes anyway (my HBA only supports CBS).
> > 
> > Machine: HP MicroServer Gen8
> > HBA: HP Dynamic Smart Array B120i
> > Port multiplier: StarTech S322M225R (SATA to 2x M.2 SATA using ASM1092R
> > chipset in JBOD mode)
> > 
> > Could you please make sure that the fixes are being backported to all
> > versions that contain the faulty commit 0ea84089dbf6? Looks like that is
> > necessary for v7.0, v6.18, v6.12 and v6.6.
> > 
> > If not, it would be great to get the fixes at least merged into v6.12, as
> > I'm otherwise forced to manually build any new stable kernel that's being
> > pushed by the Debian team.
> > 
> > Thanks again and greetings from Germany!
> > David
> 
> Dear stable maintainers,
> 
> Sorry for the extra work put on you.
> 
> Could your please help us backport:
> 
> $ git log --reverse --oneline v7.1-rc4..v7.1-rc5 drivers/ata/
> 360190bd965f ata: libata-scsi: improve readability of ata_scsi_qc_issue()
> ce4548807d2e ata: libata-scsi: do not use the deferred QC feature for ATA_DEFER_PORT
> f233124fb36c ata: libata-scsi: do not use the deferred QC feature on PMPs with CBS
> 759e8756da00 ata: libata-scsi: do not needlessly defer commands when using PMP with FBS
> 
> 
> The first three patches apply cleanly to at least v6.12.
> 
> The final commit does not apply cleanly, attaching a backported version.
> 
> Again, sorry for the trouble.
> Hopefully our AI overlords / Sashiko will manage to find some of these
> problems that was missed during the review of the faulty commit.

Now queued up, thanks.

greg k-h