Re: [PATCH v3 3/3] PCI: dwc: Enable MSI affinity support

[Brian Norris]

(CC a few more)

Hi Niklas,

On Mon, Jun 01, 2026 at 10:20:07AM +0200, Niklas Cassel wrote:
> On Fri, May 29, 2026 at 05:35:13PM -0700, Brian Norris wrote:
> > 
> > This seems to tell me the irq_set_affinity()/migration process isn't
> > really what's killing things, but something else.
> 
> Hello Brian,
> 
> Perhaps a long-shot, but have you tried to apply:
> https://lore.kernel.org/linux-pci/20260519-l1ss-fix-v2-0-b2c3a4bdeb15@xxxxxxxxxxxxxxxx/

For the record, that's:

[PATCH v2 0/4] PCI: Introduce pci_suspend_retains_context() API

> to see if it helps your problem?

It also seemed like a long-shot, but I tried it and it does resolve my
problems. Maybe I can provide a Tested-by there.

> https://lore.kernel.org/linux-pci/r247uuc7ilkhn65vcbsaoxhgypif37lj3r3mvb7ooqqdg55vnm@vx3ip6klg6ha/
> also seems to mention a recent PSCI fix:
> https://lore.kernel.org/all/20251231162126.7728-1-manivannan.sadhasivam@xxxxxxxxxxxxxxxx/

That's:
[PATCH] firmware: psci: Set pm_set_resume/suspend_via_firmware() for SYSTEM_SUSPEND

> Which also sounds somewhat related.

Yes, that looks somewhat related. I tried it, and it seems to resolve my
issue as well.

It looks like Mani suggests both patch sets are valid proposals,
although solving slightly different variations depending on the Qualcomm
firmware in question? But the PSCI patch looks like it's stuck in
/dev/null. Dunno if a Tested-by would help there too?

I'm also perplexed at why:

1) those patches claim that NVMe wasn't really working on Qcom systems
   before, and yet the laptop in question was working fine before v7.0;
   and

2) this bisected to commit eaf290c404f7 ("PCI: dwc: Enable MSI affinity
   support").

Anyway, thanks for the pointers.

Brian