<!DOCTYPE html><html><head> <meta http-equiv="Content-Type" content="text/html; charset=utf-8"> </head> <body> Am 30.08.24 um 23:43 schrieb Matthew Brost: <blockquote type="cite" cite="mid:ZtI9EMzHZW3DkHw%2F@DUT025-TGLU.fm.intel.com"> <pre class="moz-quote-pre" wrap="">On Fri, Aug 30, 2024 at 10:14:18AM +0200, Christian König wrote: </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap="">Am 29.08.24 um 19:12 schrieb Boris Brezillon: </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap="">dma_fence objects created by an entity might outlive the drm_gpu_scheduler this entity was bound to if those fences are retained by other other objects, like a dma_buf resv. This means that drm_sched_fence::sched might be invalid when the resv is walked, which in turn leads to a UAF when dma_fence_ops::get_timeline_name() is called. This probably went unnoticed so far, because the drm_gpu_scheduler had the lifetime of the drm_device, so, unless you were removing the device, there were no reasons for the scheduler to be gone before its fences. </pre> </blockquote> <pre class="moz-quote-pre" wrap=""> Nope, that is intentional design. get_timeline_name() is not safe to be called after the fence signaled because that would causes circular dependency problems. </pre> </blockquote> <pre class="moz-quote-pre" wrap=""> I'm quite sure happens, ftrace for example can and will call get_timeline_name in trace_dma_fence_destroy which is certainly after the fence is signaled. There are likely other cases too - this just quickly came to mind.</pre> </blockquote> Good point, completely forgotten about ftrace. <blockquote type="cite" cite="mid:ZtI9EMzHZW3DkHw%2F@DUT025-TGLU.fm.intel.com"> <blockquote type="cite"> <pre class="moz-quote-pre" wrap="">E.g. when you have hardware fences it can happen that fences reference a driver module (for the function printing the name) and the module in turn keeps fences around. </pre> </blockquote> <pre class="moz-quote-pre" wrap=""> I am almost positive without this patch this problematic in Xe or any driver in which schedulers are tied to IOCTLs rather than kernel module. In Xe 'fence->sched' maps to an xe_exec_queue which can be freed once the destroy exec queue IOCTL is called and all jobs are free'd (i.e. 'fence' signals). The fence could be live on after in dma-resv objects, drm syncobjs, etc... I know this issue has been raised before and basically NACK'd but I have a strong opinion this is valid and in fact required.</pre> </blockquote> I've NACK'd automatically signaling pending fences on destruction of the scheduler (that reminds me that I wanted to add a warning for that) and copying the name into every scheduler fence. As long as we don't do any of that I'm perfectly fine fixing this issue. The approach of creating a reference counted object for the name looks rather valid to me. Especially since we then pretty much get the module references correct for free as well. Christian. <blockquote type="cite" cite="mid:ZtI9EMzHZW3DkHw%2F@DUT025-TGLU.fm.intel.com"> <pre class="moz-quote-pre" wrap=""> Matt </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap="">So you easily end up with a module you can never unload. </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap="">With the introduction of a new model where each entity has its own drm_gpu_scheduler instance, this situation is likely to happen every time a GPU context is destroyed and some of its fences remain attached to dma_buf objects still owned by other drivers/processes. In order to make drm_sched_fence_get_timeline_name() safe, we need to copy the scheduler name into our own refcounted object that's only destroyed when both the scheduler and all its fences are gone. The fact drm_sched_fence might have a reference to the drm_gpu_scheduler even after it's been released is worrisome though, but I'd rather discuss that with everyone than come up with a solution that's likely to end up being rejected. Note that the bug was found while repeatedly reading dma_buf's debugfs file, which, at some point, calls dma_resv_describe() on a resv that contains signalled fences coming from a destroyed GPU context. AFAIK, there's nothing invalid there. </pre> </blockquote> <pre class="moz-quote-pre" wrap=""> Yeah but reading debugfs is not guaranteed to crash the kernel. On the other hand the approach with a kref'ed string looks rather sane to me. One comment on this below. </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap=""> Cc: Luben Tuikov <a class="moz-txt-link-rfc2396E" href="mailto:ltuikov89@gmail.com"><ltuikov89@gmail.com></a> Cc: Matthew Brost <a class="moz-txt-link-rfc2396E" href="mailto:matthew.brost@intel.com"><matthew.brost@intel.com></a> Cc: "Christian König" <a class="moz-txt-link-rfc2396E" href="mailto:christian.koenig@amd.com"><christian.koenig@amd.com></a> Cc: Danilo Krummrich <a class="moz-txt-link-rfc2396E" href="mailto:dakr@redhat.com"><dakr@redhat.com></a> Signed-off-by: Boris Brezillon <a class="moz-txt-link-rfc2396E" href="mailto:boris.brezillon@collabora.com"><boris.brezillon@collabora.com></a> --- drivers/gpu/drm/scheduler/sched_fence.c | 8 +++- drivers/gpu/drm/scheduler/sched_main.c | 18 ++++++++- include/drm/gpu_scheduler.h | 52 +++++++++++++++++++++++++ 3 files changed, 75 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/scheduler/sched_fence.c b/drivers/gpu/drm/scheduler/sched_fence.c index 0f35f009b9d3..efa2a71d98eb 100644 --- a/drivers/gpu/drm/scheduler/sched_fence.c +++ b/drivers/gpu/drm/scheduler/sched_fence.c @@ -90,7 +90,7 @@ static const char *drm_sched_fence_get_driver_name(struct dma_fence *fence) static const char *drm_sched_fence_get_timeline_name(struct dma_fence *f) { struct drm_sched_fence *fence = to_drm_sched_fence(f); - return (const char *)fence->sched->name; + return (const char *)fence->timeline->name; } static void drm_sched_fence_free_rcu(struct rcu_head *rcu) @@ -112,8 +112,10 @@ static void drm_sched_fence_free_rcu(struct rcu_head *rcu) */ void drm_sched_fence_free(struct drm_sched_fence *fence) { + drm_sched_fence_timeline_put(fence->timeline); + /* This function should not be called if the fence has been initialized. */ - if (!WARN_ON_ONCE(fence->sched)) + if (!WARN_ON_ONCE(fence->timeline)) kmem_cache_free(sched_fence_slab, fence); } @@ -224,6 +226,8 @@ void drm_sched_fence_init(struct drm_sched_fence *fence, unsigned seq; fence->sched = entity->rq->sched; + fence->timeline = fence->sched->fence_timeline; + drm_sched_fence_timeline_get(fence->timeline); seq = atomic_inc_return(&entity->fence_seq); dma_fence_init(&fence->scheduled, &drm_sched_fence_ops_scheduled, &fence->lock, entity->fence_context, seq); diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c index 7e90c9f95611..1e31a9c8ce15 100644 --- a/drivers/gpu/drm/scheduler/sched_main.c +++ b/drivers/gpu/drm/scheduler/sched_main.c @@ -1288,10 +1288,21 @@ int drm_sched_init(struct drm_gpu_scheduler *sched, sched->own_submit_wq = true; } + sched->fence_timeline = kzalloc(sizeof(*sched->fence_timeline), GFP_KERNEL); + if (!sched->fence_timeline) + goto Out_check_own; + + sched->fence_timeline->name = kasprintf(GFP_KERNEL, "%s", sched->name); + if (!sched->fence_timeline->name) + goto Out_free_fence_timeline; + + kref_init(&sched->fence_timeline->kref); + sched->sched_rq = kmalloc_array(num_rqs, sizeof(*sched->sched_rq), GFP_KERNEL | __GFP_ZERO); if (!sched->sched_rq) - goto Out_check_own; + goto Out_free_fence_timeline; + sched->num_rqs = num_rqs; for (i = DRM_SCHED_PRIORITY_KERNEL; i < sched->num_rqs; i++) { sched->sched_rq[i] = kzalloc(sizeof(*sched->sched_rq[i]), GFP_KERNEL); @@ -1319,6 +1330,10 @@ int drm_sched_init(struct drm_gpu_scheduler *sched, kfree(sched->sched_rq); sched->sched_rq = NULL; +Out_free_fence_timeline: + if (sched->fence_timeline) + kfree(sched->fence_timeline->name); + kfree(sched->fence_timeline); Out_check_own: if (sched->own_submit_wq) destroy_workqueue(sched->submit_wq); @@ -1367,6 +1382,7 @@ void drm_sched_fini(struct drm_gpu_scheduler *sched) sched->ready = false; kfree(sched->sched_rq); sched->sched_rq = NULL; + drm_sched_fence_timeline_put(sched->fence_timeline); } EXPORT_SYMBOL(drm_sched_fini); diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h index 5acc64954a88..615ca89f77dc 100644 --- a/include/drm/gpu_scheduler.h +++ b/include/drm/gpu_scheduler.h @@ -261,6 +261,52 @@ struct drm_sched_rq { struct rb_root_cached rb_tree_root; }; +/** + * struct drm_sched_fence_timeline - Wrapped around the timeline name + * + * This is needed to cope with the fact dma_fence objects created by + * an entity might outlive the drm_gpu_scheduler this entity was bound + * to, making drm_sched_fence::sched invalid and leading to a UAF when + * dma_fence_ops::get_timeline_name() is called. + */ +struct drm_sched_fence_timeline { + /** @kref: Reference count of this timeline object. */ + struct kref kref; + + /** + * @name: Name of the timeline. + * + * This is currently a copy of drm_gpu_scheduler::name. + */ + const char *name; </pre> </blockquote> <pre class="moz-quote-pre" wrap=""> Make that a char name[] and embed the name into the structure. The macro struct_size() can be used to calculate the size. </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap="">+}; + +static inline void +drm_sched_fence_timeline_release(struct kref *kref) +{ + struct drm_sched_fence_timeline *tl = + container_of(kref, struct drm_sched_fence_timeline, kref); + + kfree(tl->name); + kfree(tl); </pre> </blockquote> <pre class="moz-quote-pre" wrap=""> This avoids having two allocations for the timeline name. Regards, Christian. </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap="">+} + +static inline void +drm_sched_fence_timeline_put(struct drm_sched_fence_timeline *tl) +{ + if (tl) + kref_put(&tl->kref, drm_sched_fence_timeline_release); +} + +static inline struct drm_sched_fence_timeline * +drm_sched_fence_timeline_get(struct drm_sched_fence_timeline *tl) +{ + if (tl) + kref_get(&tl->kref); + + return tl; +} + /** * struct drm_sched_fence - fences corresponding to the scheduling of a job. */ @@ -289,6 +335,9 @@ struct drm_sched_fence { */ ktime_t deadline; + /** @timeline: the timeline this fence belongs to. */ + struct drm_sched_fence_timeline *timeline; + /** * @parent: the fence returned by &drm_sched_backend_ops.run_job * when scheduling the job on hardware. We signal the @@ -488,6 +537,8 @@ struct drm_sched_backend_ops { * @credit_count: the current credit count of this scheduler * @timeout: the time after which a job is removed from the scheduler. * @name: name of the ring for which this scheduler is being used. + * @fence_timeline: fence timeline that will be passed to fences created by + * and entity that's bound to this scheduler. * @num_rqs: Number of run-queues. This is at most DRM_SCHED_PRIORITY_COUNT, * as there's usually one run-queue per priority, but could be less. * @sched_rq: An allocated array of run-queues of size @num_rqs; @@ -521,6 +572,7 @@ struct drm_gpu_scheduler { atomic_t credit_count; long timeout; const char *name; + struct drm_sched_fence_timeline *fence_timeline; u32 num_rqs; struct drm_sched_rq **sched_rq; wait_queue_head_t job_scheduled; </pre> </blockquote> <pre class="moz-quote-pre" wrap=""> </pre> </blockquote> </blockquote> </body> </html>