On 19/06/2021 04:09, Chunyou Tang wrote: > Hi Steve, > 1, > from > https://lore.kernel.org/lkml/31644881-134a-2d6e-dddf-e658a3a8176b@xxxxxxx/ > I can see what your sent,I used a wrong email address,Now it correct. > 2, >>> Unless I'm mistaken the situation where some mappings may be NULL is >>> caused by the loop in panfrost_lookup_bos() not completing >>> successfully >>> (panfrost_gem_mapping_get() returning NULL). In this case if >>> mappings[i] >>> is NULL then all following mappings must also be NULL. So 'break' >>> allows >>> us to skip the later ones. Admittedly the performance here isn't >>> important so I'm not sure it's worth the optimisation, but AIUI this >>> code isn't actually wrong. > > from panfrost_lookup_bos(),you can see: > for (i = 0; i < job->bo_count; i++) { > struct panfrost_gem_mapping *mapping; > > bo = to_panfrost_bo(job->bos[i]); > ICUBE_DEBUG_PRINTK("panfrost bo gem handle=0x%x > is_dumb=%d\n", bo->gem_handle, bo->is_dumb); > if (!bo->is_dumb) { > mapping = panfrost_gem_mapping_get(bo, priv); > if (!mapping) { > ret = -EINVAL; > break; > } > > atomic_inc(&bo->gpu_usecount); > job->mappings[i] = mapping; > } else { > atomic_inc(&bo->gpu_usecount); > job->mappings[i] = NULL; > } > } This code isn't upstream - in drm-misc/drm-misc-next (and all mainline kernels from what I can tell) this doesn't have any "is_dumb" test. Which branch are you using? > if bo->is_dumb is TRUE,the job->mappings[i] will set to NULL,and the > while will be continue,so if job->mappings[i] is NULL,the following > can not be NULL. I agree that with the above code the panfrost_job_cleanup() would need changing. But we don't (currently) have this code upstream, so this change doesn't make sense upstream. Thanks, Steve > 3, > I've had this problem in our project,the value of is_dumb like these: > 0 > 0 > 0 > 1 > 0 > 0 > 0 > so,when job->mappings[i] is NULL,we can not break the while in > panfrost_job_cleanup(). > > thanks > Chunyou > > ÓÚ Fri, 18 Jun 2021 13:43:25 +0100 > Steven Price <steven.price@xxxxxxx> дµÀ: > >> On 17/06/2021 09:04, ChunyouTang wrote: >>> From: ChunyouTang <tangchunyou@xxxxxxxxxxxx> >>> >>> The 'break' can cause 'Memory manager not clean during takedown' >>> >>> It cannot use break to finish the circulation,it should use >>> >>> continue to traverse the circulation.it should put every mapping >>> >>> which is not NULL. >> >> You don't appear to have answered my question about whether you've >> actually seen this happen (and ideally what circumstances). In my >> previous email[1] I explained why I don't think this is needed. You >> need to convince me that I've overlooked something. >> >> Thanks, >> >> Steve >> >> [1] >> https://lore.kernel.org/r/31644881-134a-2d6e-dddf-e658a3a8176b%40arm.com >> >>> Signed-off-by: ChunyouTang <tangchunyou@xxxxxxxxxxxx> >>> --- >>> drivers/gpu/drm/panfrost/panfrost_job.c | 2 +- >>> 1 file changed, 1 insertion(+), 1 deletion(-) >>> >>> diff --git a/drivers/gpu/drm/panfrost/panfrost_job.c >>> b/drivers/gpu/drm/panfrost/panfrost_job.c index >>> 6003cfeb1322..52bccc1d2d42 100644 --- >>> a/drivers/gpu/drm/panfrost/panfrost_job.c +++ >>> b/drivers/gpu/drm/panfrost/panfrost_job.c @@ -281,7 +281,7 @@ >>> static void panfrost_job_cleanup(struct kref *ref) if >>> (job->mappings) { for (i = 0; i < job->bo_count; i++) { >>> if (!job->mappings[i]) >>> - break; >>> + continue; >>> >>> atomic_dec(&job->mappings[i]->obj->gpu_usecount); >>> panfrost_gem_mapping_put(job->mappings[i]); >>> > >
