Re: [PATCH] ata: libata-core: Add NO_NCQ_TRIM quirk for Samsung PM893 (MZ7L3)

[Niklas Cassel <cassel@kernel.org>]

Hello Tobias,

On Tue, Nov 10, 2026 at 09:19:58AM +0100, Tobias Kriebisch wrote:
> On Mon, Nov 09, 2026 at 03:56:41PM +0100, Niklas Cassel wrote:
> > Please boot the latest kernel, 7.4-rc2, and verify that the problem still
> > exists.
> 
> The three drives hold a ZFS pool on a production server. OpenZFS does
> not support 7.4 yet (the latest release supports up to 6.19), so I
> cannot run 7.4-rc2 on that machine, and I don't have a spare machine
> with this drive and controller.
> 
> Would results from the newest kernel that a released OpenZFS supports
> be acceptable? On that kernel I can test noncqtrim and nolpm separately,
> and then boot a kernel with only the needed quirks and no libata.force.

Normally, any problem where OpenZFS is involved is ignored, because the
CDDL license is incompatible with the GPL...


> > If the problem still exist on the latest kernel, apply the minimal set of
> > quirks needed to fix the problem.
> 
> Understood, the v2 will only contain the quirks that are actually needed.
> 
> > And yes, build and boot a kernel with your patch to verify that the
> problem
> > is actually solved.
> 
> If testing on 6.19 is helpful here, I will build and boot 6.19 with the
> patch, verify that the problem is solved, and send a v2 with the results.

... however, since it seems like almost all Samsung drives already have:

ATA_QUIRK_ZERO_AFTER_TRIM and ATA_QUIRK_NO_NCQ_TRIM

(ATA_QUIRK_ZERO_AFTER_TRIM is actually a whitelist, saying that the data
after trim is guaranteed to be zero, so this one tells the upper layers
about the FW behavior, so that it can use the most optimal commands.)

I think it is reasonable to accept quirk(s) for this device if you test
with 6.19 and actually verify that the patch solves your problem.


Kind regards,
Niklas