Regression on linux-next (next-20250321)

[Borah, Chaitanya Kumar]

> -----Original Message-----
> From: Nicolin Chen <nicolinc at nvidia.com>
> Sent: Thursday, March 27, 2025 2:14 AM
> To: Borah, Chaitanya Kumar <chaitanya.kumar.borah at intel.com>
> Cc: iommu at lists.linux.dev; intel-gfx at lists.freedesktop.org; intel-
> xe at lists.freedesktop.org; Kurmi, Suresh Kumar
> <suresh.kumar.kurmi at intel.com>; Saarinen, Jani <jani.saarinen at intel.com>;
> jgg at nvidia.com
> Subject: Re: Regression on linux-next (next-20250321)
> 
> Hi Chaitanya,
> 
> On Wed, Mar 26, 2025 at 08:31:15AM +0000, Borah, Chaitanya Kumar
> wrote:
> > > > `````````````````````````````````````````````````````````````````````````````````
> > > > <4>[    0.226495] Unpatched return thunk in use. This should not
> happen!
> > > > <4>[    0.226502] WARNING: CPU: 0 PID: 1 at
> > > arch/x86/kernel/cpu/bugs.c:3107 __warn_thunk+0x62/0x70
> > >
> > > Hmm....I wonder why x86 can be affected...
> > >
> > > The only four callers of iommu_dma_prepare_msi() are ARM platforms.
> [...]
> > > > Details log can be found in [3].
> > >
> > > And I can't see something obvious from the log..
> > >
> > > Would you please give the git-diff a try (drivers/iommu/iommu.c)?
> > > https://lore.kernel.org/linux-iommu/Z+Itnw4ys6dmDsc+@nvidia.com/
> > >
> > > If this doesn't help, would you please give this a try?
> > > https://lore.kernel.org/linux-
> iommu/20250324170743.GA1339275 at ax162/
> > >
> >
> > Thank you, Nicolin, for your reply. Unfortunately, these changes does
> > not solve the issue. (applied individually and together)
> 
> Would you please try the latest linux-next (next-20250326) and see if the
> issue still occurs?
> 
> If it does occur with next-20250326, would you please try reverting
> 06d54f00f3f5a on top of the tree rather than bisect?
>  "06d54f00f3f5 iommu: Drop sw_msi from iommu_domain"
> 

We have tried both of these but the error persists.

> It still feels odd to me and Jason that this change would break x86.
> So, we want to confirm that this is really the culprit.
> 

I think as discussed in the other thread [1]. It makes sense that objtool is injecting the error.
That would explain it.

[1] https://lore.kernel.org/intel-gfx/Z+Rm9LweNAtQBrmD@nvidia.com/T/#t

Regards

Chaitanya

> Also, I think your platform doesn't set CONFIG_IRQ_MSI_IOMMU, would you
> please check you .config file and confirm?
> 
> Thanks
> Nicolin