| Message ID | 20230601144450.792228-1-andrzej.hajda@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | drm/i915/gt: limit lmem allocation size to succeed on SmallBars | expand |
Hi Andrzej, On Thu, Jun 01, 2023 at 04:44:50PM +0200, Andrzej Hajda wrote: > In case system is short on mappable memory (256MB on SmallBar) allocation > of two 1GB buffers will fail. > > Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/8300 > Signed-off-by: Andrzej Hajda <andrzej.hajda@intel.com> > --- > drivers/gpu/drm/i915/gt/selftest_tlb.c | 11 +++++++++-- > 1 file changed, 9 insertions(+), 2 deletions(-) > > diff --git a/drivers/gpu/drm/i915/gt/selftest_tlb.c b/drivers/gpu/drm/i915/gt/selftest_tlb.c > index 4493c8518e91b2..3bd6b540257b46 100644 > --- a/drivers/gpu/drm/i915/gt/selftest_tlb.c > +++ b/drivers/gpu/drm/i915/gt/selftest_tlb.c > @@ -190,11 +190,18 @@ pte_tlbinv(struct intel_context *ce, > > static struct drm_i915_gem_object *create_lmem(struct intel_gt *gt) > { > + struct intel_memory_region *mr = gt->i915->mm.regions[INTEL_REGION_LMEM_0]; > + resource_size_t size = SZ_1G; > + > /* > * Allocation of largest possible page size allows to test all types > - * of pages. > + * of pages. To succeed with both allocations, especially in case of Small > + * BAR, try to allocate no more than quarter of mappable memory. > */ > - return i915_gem_object_create_lmem(gt->i915, SZ_1G, I915_BO_ALLOC_CONTIGUOUS); > + if (mr && size > mr->io_size / 4) > + size = mr->io_size / 4; > + > + return i915_gem_object_create_lmem(gt->i915, size, I915_BO_ALLOC_CONTIGUOUS); makes sense to me. Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com> Thanks, Andi
On 02.06.2023 11:07, Andi Shyti wrote: > Hi Andrzej, > > On Thu, Jun 01, 2023 at 04:44:50PM +0200, Andrzej Hajda wrote: >> In case system is short on mappable memory (256MB on SmallBar) allocation >> of two 1GB buffers will fail. >> >> Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/8300 >> Signed-off-by: Andrzej Hajda <andrzej.hajda@intel.com> >> --- >> drivers/gpu/drm/i915/gt/selftest_tlb.c | 11 +++++++++-- >> 1 file changed, 9 insertions(+), 2 deletions(-) >> >> diff --git a/drivers/gpu/drm/i915/gt/selftest_tlb.c b/drivers/gpu/drm/i915/gt/selftest_tlb.c >> index 4493c8518e91b2..3bd6b540257b46 100644 >> --- a/drivers/gpu/drm/i915/gt/selftest_tlb.c >> +++ b/drivers/gpu/drm/i915/gt/selftest_tlb.c >> @@ -190,11 +190,18 @@ pte_tlbinv(struct intel_context *ce, >> >> static struct drm_i915_gem_object *create_lmem(struct intel_gt *gt) >> { >> + struct intel_memory_region *mr = gt->i915->mm.regions[INTEL_REGION_LMEM_0]; >> + resource_size_t size = SZ_1G; >> + >> /* >> * Allocation of largest possible page size allows to test all types >> - * of pages. >> + * of pages. To succeed with both allocations, especially in case of Small >> + * BAR, try to allocate no more than quarter of mappable memory. >> */ >> - return i915_gem_object_create_lmem(gt->i915, SZ_1G, I915_BO_ALLOC_CONTIGUOUS); >> + if (mr && size > mr->io_size / 4) >> + size = mr->io_size / 4; >> + >> + return i915_gem_object_create_lmem(gt->i915, size, I915_BO_ALLOC_CONTIGUOUS); > > makes sense to me. > > Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com> Thx, patch merged. Regards Andrzej > > Thanks, > Andi
diff --git a/drivers/gpu/drm/i915/gt/selftest_tlb.c b/drivers/gpu/drm/i915/gt/selftest_tlb.c index 4493c8518e91b2..3bd6b540257b46 100644 --- a/drivers/gpu/drm/i915/gt/selftest_tlb.c +++ b/drivers/gpu/drm/i915/gt/selftest_tlb.c @@ -190,11 +190,18 @@ pte_tlbinv(struct intel_context *ce, static struct drm_i915_gem_object *create_lmem(struct intel_gt *gt) { + struct intel_memory_region *mr = gt->i915->mm.regions[INTEL_REGION_LMEM_0]; + resource_size_t size = SZ_1G; + /* * Allocation of largest possible page size allows to test all types - * of pages. + * of pages. To succeed with both allocations, especially in case of Small + * BAR, try to allocate no more than quarter of mappable memory. */ - return i915_gem_object_create_lmem(gt->i915, SZ_1G, I915_BO_ALLOC_CONTIGUOUS); + if (mr && size > mr->io_size / 4) + size = mr->io_size / 4; + + return i915_gem_object_create_lmem(gt->i915, size, I915_BO_ALLOC_CONTIGUOUS); } static struct drm_i915_gem_object *create_smem(struct intel_gt *gt)
In case system is short on mappable memory (256MB on SmallBar) allocation of two 1GB buffers will fail. Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/8300 Signed-off-by: Andrzej Hajda <andrzej.hajda@intel.com> --- drivers/gpu/drm/i915/gt/selftest_tlb.c | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-)