Re: [PATCH] drm/xe: Make p2pdma distance check verbose via dynamic debug

["Thomas Hellström" <thomas.hellstrom@linux.intel.com>]

On Mon, 2026-09-14 at 12:34 +0100, Matthew Auld wrote:
> On 14/09/2026 10:46, Thomas Hellström wrote:
> > pci_p2pdma_distance() can optionally print details about how it
> > computed the ACS/PCIe topology distance between the two devices,
> > but
> > xe_dma_buf_attach() always called it with verbose set to false,
> > silently hiding useful diagnostics whenever a dma-buf attach unset
> > attach->peer2peer.
> > 
> > Tie the verbose argument to a dedicated dynamic-debug callsite
> > tagged
> > "xe dma-buf p2p debug" (in addition to keeping the existing
> > DRM_UT_DRIVER category as a default), so the extra p2pdma distance
> > logging can be switched on without a recompile in any of the
> > following
> > ways:
> > 
> >   - Enable the whole DRIVER category for xe via drm.debug, which is
> >     already used elsewhere in the driver:
> > 
> >       # echo 0x10 > /sys/module/drm/parameters/debug
> > 
> >   - Enable dynamic debug for the whole file:
> > 
> >       # echo 'file xe_dma_buf.c +p' > \
> >               /sys/kernel/debug/dynamic_debug/control
> > 
> >   - Enable dynamic debug for just this callsite by function name:
> > 
> >       # echo 'func xe_dma_buf_attach +p' > \
> >               /sys/kernel/debug/dynamic_debug/control
> > 
> >   - Enable dynamic debug by matching the format string used to
> >     register the callsite:
> > 
> >       # echo 'format "xe dma-buf p2p debug" +p' > \
> >               /sys/kernel/debug/dynamic_debug/control
> > 
> >   - Enable it at boot via the dyndbg= kernel command line
> > parameter:
> > 
> >       dyndbg="func xe_dma_buf_attach +p"
> > 
> > Any of the above flips verbose_p2p_dist to true for that attach
> > call,
> > causing pci_p2pdma_distance() to log its distance calculation
> > without
> > requiring the whole DRM_UT_DRIVER category to be enabled.
> 
> Is drm.debug not enough here for typical usescases?  If you are 
> interested in this you would maybe toggle that anyway, in case there
> is 
> other interesting stuff? Or do you think the dynamic debug will be
> useful?

There are cases where we just want to check just if p2p is enabled. And
normally IIRC you can do this by targeting a separate pr_debug or
drm_debug using dynamic debug. But since this one triggers printouts in
the p2pdma code, it seems like this construct is neecessary.

> 
> > 
> > Cc: Matthew Brost <matthew.brost@intel.com>
> > Cc: Matthew Auld <matthew.auld@intel.com>
> > Assisted-by: LLM
> > Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
> 
> Either way,
> Reviewed-by: Matthew Auld <matthew.auld@intel.com>

Thanks,
Thomas

> 
> (With the ifdef guard or so)
> 
> > ---
> >   drivers/gpu/drm/xe/xe_dma_buf.c | 10 +++++++++-
> >   1 file changed, 9 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/gpu/drm/xe/xe_dma_buf.c
> > b/drivers/gpu/drm/xe/xe_dma_buf.c
> > index 5d9f1cd24b7f..b63fd70fb4de 100644
> > --- a/drivers/gpu/drm/xe/xe_dma_buf.c
> > +++ b/drivers/gpu/drm/xe/xe_dma_buf.c
> > @@ -22,13 +22,21 @@
> >   
> >   MODULE_IMPORT_NS("DMA_BUF");
> >   
> > +static void xe_ddebug_flip_bool(const struct _ddebug *desc, bool
> > *flag)
> > +{
> > +	*flag = true;
> > +}
> > +
> >   static int xe_dma_buf_attach(struct dma_buf *dmabuf,
> >   			     struct dma_buf_attachment *attach)
> >   {
> >   	struct drm_gem_object *obj = attach->dmabuf->priv;
> > +	bool verbose_p2p_dist = drm_debug_enabled(DRM_UT_DRIVER);
> > +
> > +	_dynamic_func_call("xe dma-buf p2p debug",
> > xe_ddebug_flip_bool, &verbose_p2p_dist);
> >   
> >   	if (attach->peer2peer &&
> > -	    pci_p2pdma_distance(to_pci_dev(obj->dev->dev), attach-
> > >dev, false) < 0)
> > +	    pci_p2pdma_distance(to_pci_dev(obj->dev->dev), attach-
> > >dev, verbose_p2p_dist) < 0)
> >   		attach->peer2peer = false;
> >   
> >   	if (!attach->peer2peer &&
> > !xe_bo_can_migrate(gem_to_xe_bo(obj), XE_PL_TT))