Re: [PATCH] swiotlb: suppress warning when __GFP_NOWARN is set v2

[Christian KÃnig]

Am 02.01.2018 um 16:39 schrieb Chris Wilson:
> Quoting Christian KÃnig (2018-01-02 12:13:58)
> > TTM tries to allocate coherent memory in chunks of 2MB first to improve
> > TLB efficiency and falls back to allocating 4K pages if that fails.
> >
> > Suppress the warning when the 2MB allocations fails since there is a
> > valid fall back path.
> >
> > v2: suppress warnings from swiotlb_tbl_map_single as well
> >
> > Signed-off-by: Christian KÃnig <christian.koenig@xxxxxxx>
> > Reported-by: Mike Galbraith <efault@xxxxxx>
> > Bug: https://bugs.freedesktop.org/show_bug.cgi?id=104082
> > CC: stable@xxxxxxxxxxxxxxx
> > ---
> > @@ -989,7 +1002,8 @@ swiotlb_map_sg_attrs(struct device *hwdev, struct scatterlist *sgl, int nelems,
> >                  if (swiotlb_force == SWIOTLB_FORCE ||
> >                      !dma_capable(hwdev, dev_addr, sg->length)) {
> >                          phys_addr_t map = map_single(hwdev, sg_phys(sg),
> > -                                                    sg->length, dir, attrs);
> > +                                                    sg->length, dir, attrs,
> > +                                                    true);
> This keeps the warning for dma_map_sg(), so seems quite ineffective.

That is the desired result, dma_map_sg() should still print the warning.

We only want to disable it for swiotlb_alloc_coherent().

Regards,
Christian.

> >                          if (map == SWIOTLB_MAP_ERROR) {
> >                                  /* Don't panic here, we expect map_sg users
> >                                     to do proper error handling. */
> And counter to the suggestion here?

We don't use swiotlb_map_sg_attrs() in TTM, so that is irrelevant for my 
use case and I can't test any change regarding it.

Regards,
Christian.

> -Chris