Re: [PATCH] drm/xe/pf: Keep VF LMEM BAR size low if no VFs enabled

[Rodrigo Vivi <rodrigo.vivi@intel.com>]

On Mon, Sep 21, 2026 at 10:28:01AM +0200, Bernatowicz, Marcin wrote:
> 
> On 9/18/2026 4:50 PM, Rodrigo Vivi wrote:
> > On Fri, Sep 18, 2026 at 01:01:30PM +0200, Marcin Bernatowicz wrote:
> > > When VFs are enabled on dGFX the driver resizes the PF VF_LMEM_BAR to
> > > fit the requested layout. After VFs are disabled the PF VF BAR
> > > size is left as-is. On platforms with tight MMIO apertures a
> > > subsequent unplug/rescan followed by another enable may fail with:
> > > 
> > >    "VF BAR …: can't assign; no space"
> > > 
> > > because the PCI core reserves address space based on the (now large) VF
> > > template, often multiplied by totalvfs.
> > > 
> > > Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/5937
> > > Fixes: 94eae6ee4c2d ("drm/xe/pf: Set VF LMEM BAR size")
> > > Signed-off-by: Marcin Bernatowicz <marcin.bernatowicz@linux.intel.com>
> > > Cc: Michał Wajdeczko <michal.wajdeczko@intel.com>
> > > Cc: Michał Winiarski <michal.winiarski@intel.com>
> > > ---
> > > v4:
> > >   - use directly pf_resize_vf_vram_bar helper with device_total_vfs value.
> > > v3:
> > >   - rebased to resolve conflicts, no other functional changes were done
> > > v2:
> > >   - Rename helper to restore_vf_vram_bar_size() (Michal)
> > >   - Use xe->sriov.pf.device_total_vfs instead of pci_sriov_get_totalvfs(),
> > >     which may be capped (Michal)
> > >   - Switch logging to %pe (Michal)
> > >   - Call restore unconditionally on enable-fail path (Michal)
> > >     (drop vf_vram_bar_resized flag)
> > > ---
> > > ---
> > >   drivers/gpu/drm/xe/xe_pci_sriov.c | 3 +++
> > >   1 file changed, 3 insertions(+)
> > > 
> > > diff --git a/drivers/gpu/drm/xe/xe_pci_sriov.c b/drivers/gpu/drm/xe/xe_pci_sriov.c
> > > index c023a2a1fc17..31abcee55211 100644
> > > --- a/drivers/gpu/drm/xe/xe_pci_sriov.c
> > > +++ b/drivers/gpu/drm/xe/xe_pci_sriov.c
> > > @@ -177,6 +177,7 @@ static int pf_enable_vfs(struct xe_device *xe, int num_vfs)
> > >   	return num_vfs;
> > >   failed:
> > > +	pf_resize_vf_vram_bar(xe, xe->sriov.pf.device_total_vfs);
> > should we resize before unprovisioning?
> > or it doesn't even matter at this point?!
> 
> The BAR size is independent of Xe/GuC provisioning, so resizing before
> unprovisioning is safe. It only requires SR-IOV memory decoding to be
> disabled.
> 
> > 
> > >   	xe_sriov_pf_unprovision_vfs(xe, num_vfs);
> > >   	xe_pm_runtime_put(xe);
> > >   	pf_finish_vfs_enabling(xe);
> > > @@ -204,6 +205,8 @@ static int pf_disable_vfs(struct xe_device *xe)
> > >   	pci_disable_sriov(pdev);
> > > +	pf_resize_vf_vram_bar(xe, xe->sriov.pf.device_total_vfs);
> > I just noticed that no users of this function handles the error.
> > Should we convert it to void? (I see it already prints some messages,
> > so not needed to duplicate anywhere)
> > 
> > this is not a blocker and if the order above is okay:
> The helper’s return type is unrelated to this fix, so I prefer to keep its
> interface unchanged here; it can be converted to void separately.

fair enough. pushed to drm-xe-next. Thanks for the confirmations.

> > 
> > Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
> 
> Thanks for the review.
> 
> marcin
> 
> > > +
> > >   	xe_sriov_pf_reprovision_default(xe);
> > >   	pf_reset_vfs(xe, num_vfs);
> > > -- 
> > > 2.43.0
> > >