Western Digital SSD freezes on power saving

[Frank Tornack]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512

Please do add me in CC

Am Sonntag, dem 28.01.2024 um 23:17 +0900 schrieb Damien Le Moal:
> On 1/28/24 23:04, Bagas Sanjaya wrote:
> > On Sun, Jan 28, 2024 at 01:26:23PM +0100, Frank Tornack wrote:
> > > Hash: SHA512
> > > 
> > > Hello Linux kernel developers and experts,
> > > 
> > > I have a question that I think is not tied to a specific
> > > distribution
> > > and goes pretty deep into the Linux kernel. I hope you can help
> > > me
> > > without subscribing to the mailing list and remember to include
> > > me in
> > > the answer discussion. 
> > > 
> > > In my old notebook I used an nvme-ssd from WD (wds500g1b0c-
> > > 00s6u0),
> > > which is affected by a well-known bug. I currently use the
> > > parameter
> > > `nvme_core.default_ps_max_latency_us=0` to prevent the system
> > > from
> > > freezing. 
> > > 
> > 
> > What well-known bug?
> 
> Indeed... What bug ? Is there a firmware update for that SSD ?
I'm not sure if it is this bug:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=538e4a8c571efdf131834431e0c14808bcfb1004
but the temp fix and the log is the same.

No, WD does not provide a firmware update for that SSD. I checked with
the Windows Tool of the vendor and there website as well.

> 
> > 
> > > I would like to use a new SSD in my new notebook. However, as it
> > > is
> > > difficult to sell used SSDs, I would like to continue using the
> > > old SSD
> > > affected by the bug. I plan to move the SSD to a Thunderbolt
> > > enclosure
> > > for this purpose. I think since Thunderbolt is based on PCIe, the
> > > same
> > > energy saving measures and the associated bugs will probably
> > > apply.
> > > 
> > > In the hope that my new SSD doesn't have the same bug, I would
> > > prefer
> > > not to disable the Powersafe in the laptop. Is there a way to set
> > > this
> > > flag only for a specific SSD? By identifying it by name or serial
> > > number? 
> 
> echo 0 > /sys/block/nvmeXnY/device/power/pm_qos_latency_tolerance_us
> 
> will have the same effect as nvme_core.default_ps_max_latency_us=0
> but for this
> device only. For reliable device name, see under /dev/disk/by-id and
> find your
> "buggy" nvme device there using its model name+fw-revision.
> 
> If a more permanent fix is needed, then we would need to add a quirk
> for this
> device to the nvme driver.
> 
that sounds like something I could put in a udev rule. I don't know
much about udev, but I think I can find someone who can help me.
Thank you

Sidenote: I don't know how widespread the SSD is and therefore can't
judge whether a quirk is worth.
> > > 
> > > I apologise prophylactically if my question is completely out of
> > > place
> > > and thank you for your help. As a long-time Linux user, I would
> > > also
> > > like to thank you in general. If you have any follow-up
> > > questions,
> > > please do not hesitate to contact me.
> > > 
> > 
> > Confused...
> > 
> 

- -- 
Frank Tornack <f-tornack at t-online.de>
Privat
-----BEGIN PGP SIGNATURE-----

iHUEARYKAB0WIQSKs5ZJ/d3p8HqrcSZVreqZ0wXqzQUCZbaOCQAKCRBVreqZ0wXq
zWRQAP4lxan8LHllVHKHgI48nY2sHZibNE1SnnD2dCvzn1wPYwEAp2kCEopKN+cW
IgsgPsekmyEO3nX9kxdlhdnVw/bBFA0=
=QsNp
-----END PGP SIGNATURE-----