| Message ID | 20191222233558.2201901-1-chris@chris-wilson.co.uk (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | drm/i915: Mark the GEM context link as RCU protected | expand |
On 22/12/2019 23:35, Chris Wilson wrote: > The only protection for intel_context.gem_cotext is granted by RCU, so > annotate it as a rcu protected pointer and carefully dereference it in > the few occasions we need to use it. > > Fixes: 9f3ccd40acf4 ("drm/i915: Drop GEM context as a direct link from i915_request") > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> > Cc: Andi Shyti <andi.shyti@intel.com> > --- > drivers/gpu/drm/i915/gem/i915_gem_context.c | 5 ++- > drivers/gpu/drm/i915/gt/intel_context_types.h | 2 +- > drivers/gpu/drm/i915/gt/intel_reset.c | 26 +++++++++--- > .../gpu/drm/i915/gt/intel_ring_submission.c | 2 +- > drivers/gpu/drm/i915/i915_gpu_error.c | 40 ++++++++++++------- > drivers/gpu/drm/i915/i915_request.c | 6 +-- > drivers/gpu/drm/i915/i915_request.h | 8 ++++ > 7 files changed, 62 insertions(+), 27 deletions(-) > [snip] > > static void engine_record_requests(struct intel_engine_cs *engine, > @@ -1298,28 +1304,34 @@ static void error_record_engine_execlists(const struct intel_engine_cs *engine, > static bool record_context(struct drm_i915_error_context *e, > const struct i915_request *rq) > { > - const struct i915_gem_context *ctx = rq->context->gem_context; > + struct i915_gem_context *ctx; > + struct task_struct *task; > + bool capture; > > + rcu_read_lock(); > + ctx = rcu_dereference(rq->context->gem_context); > + if (ctx && !kref_get_unless_zero(&ctx->ref)) > + ctx = NULL; > + rcu_read_unlock(); > if (!ctx) > return false; > > - if (ctx->pid) { > - struct task_struct *task; > - > - rcu_read_lock(); > - task = pid_task(ctx->pid, PIDTYPE_PID); > - if (task) { > - strcpy(e->comm, task->comm); > - e->pid = task->pid; > - } > - rcu_read_unlock(); > + rcu_read_lock(); > + task = pid_task(ctx->pid, PIDTYPE_PID); > + if (task) { > + strcpy(e->comm, task->comm); > + e->pid = task->pid; > } > + rcu_read_unlock(); Why is this rcu_read_lock section needed? The first one obtained the reference to the context so should be good. Regards, Tvrtko
On 07/01/2020 11:15, Tvrtko Ursulin wrote: > > On 22/12/2019 23:35, Chris Wilson wrote: >> The only protection for intel_context.gem_cotext is granted by RCU, so >> annotate it as a rcu protected pointer and carefully dereference it in >> the few occasions we need to use it. >> >> Fixes: 9f3ccd40acf4 ("drm/i915: Drop GEM context as a direct link from >> i915_request") >> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> >> Cc: Andi Shyti <andi.shyti@intel.com> >> --- >> drivers/gpu/drm/i915/gem/i915_gem_context.c | 5 ++- >> drivers/gpu/drm/i915/gt/intel_context_types.h | 2 +- >> drivers/gpu/drm/i915/gt/intel_reset.c | 26 +++++++++--- >> .../gpu/drm/i915/gt/intel_ring_submission.c | 2 +- >> drivers/gpu/drm/i915/i915_gpu_error.c | 40 ++++++++++++------- >> drivers/gpu/drm/i915/i915_request.c | 6 +-- >> drivers/gpu/drm/i915/i915_request.h | 8 ++++ >> 7 files changed, 62 insertions(+), 27 deletions(-) >> > > [snip] > >> static void engine_record_requests(struct intel_engine_cs *engine, >> @@ -1298,28 +1304,34 @@ static void >> error_record_engine_execlists(const struct intel_engine_cs *engine, >> static bool record_context(struct drm_i915_error_context *e, >> const struct i915_request *rq) >> { >> - const struct i915_gem_context *ctx = rq->context->gem_context; >> + struct i915_gem_context *ctx; >> + struct task_struct *task; >> + bool capture; >> + rcu_read_lock(); >> + ctx = rcu_dereference(rq->context->gem_context); >> + if (ctx && !kref_get_unless_zero(&ctx->ref)) >> + ctx = NULL; >> + rcu_read_unlock(); >> if (!ctx) >> return false; >> - if (ctx->pid) { >> - struct task_struct *task; >> - >> - rcu_read_lock(); >> - task = pid_task(ctx->pid, PIDTYPE_PID); >> - if (task) { >> - strcpy(e->comm, task->comm); >> - e->pid = task->pid; >> - } >> - rcu_read_unlock(); >> + rcu_read_lock(); >> + task = pid_task(ctx->pid, PIDTYPE_PID); >> + if (task) { >> + strcpy(e->comm, task->comm); >> + e->pid = task->pid; >> } >> + rcu_read_unlock(); > > Why is this rcu_read_lock section needed? The first one obtained the > reference to the context so should be good. Hmm pid_task() does: ... first = rcu_dereference_check(hlist_first_rcu(&pid->tasks[type]), lockdep_tasklist_lock_is_held()); ... Note, tasklist_lock_is_held. Regards, Tvrtko
Quoting Tvrtko Ursulin (2020-01-07 11:15:39) > > On 22/12/2019 23:35, Chris Wilson wrote: > > The only protection for intel_context.gem_cotext is granted by RCU, so > > annotate it as a rcu protected pointer and carefully dereference it in > > the few occasions we need to use it. > > > > Fixes: 9f3ccd40acf4 ("drm/i915: Drop GEM context as a direct link from i915_request") > > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> > > Cc: Andi Shyti <andi.shyti@intel.com> > > --- > > drivers/gpu/drm/i915/gem/i915_gem_context.c | 5 ++- > > drivers/gpu/drm/i915/gt/intel_context_types.h | 2 +- > > drivers/gpu/drm/i915/gt/intel_reset.c | 26 +++++++++--- > > .../gpu/drm/i915/gt/intel_ring_submission.c | 2 +- > > drivers/gpu/drm/i915/i915_gpu_error.c | 40 ++++++++++++------- > > drivers/gpu/drm/i915/i915_request.c | 6 +-- > > drivers/gpu/drm/i915/i915_request.h | 8 ++++ > > 7 files changed, 62 insertions(+), 27 deletions(-) > > > > [snip] > > > > > static void engine_record_requests(struct intel_engine_cs *engine, > > @@ -1298,28 +1304,34 @@ static void error_record_engine_execlists(const struct intel_engine_cs *engine, > > static bool record_context(struct drm_i915_error_context *e, > > const struct i915_request *rq) > > { > > - const struct i915_gem_context *ctx = rq->context->gem_context; > > + struct i915_gem_context *ctx; > > + struct task_struct *task; > > + bool capture; > > > > + rcu_read_lock(); > > + ctx = rcu_dereference(rq->context->gem_context); > > + if (ctx && !kref_get_unless_zero(&ctx->ref)) > > + ctx = NULL; > > + rcu_read_unlock(); > > if (!ctx) > > return false; > > > > - if (ctx->pid) { > > - struct task_struct *task; > > - > > - rcu_read_lock(); > > - task = pid_task(ctx->pid, PIDTYPE_PID); > > - if (task) { > > - strcpy(e->comm, task->comm); > > - e->pid = task->pid; > > - } > > - rcu_read_unlock(); > > + rcu_read_lock(); > > + task = pid_task(ctx->pid, PIDTYPE_PID); > > + if (task) { > > + strcpy(e->comm, task->comm); > > + e->pid = task->pid; > > } > > + rcu_read_unlock(); > > Why is this rcu_read_lock section needed? The first one obtained the > reference to the context so should be good. The task returned by ctx->pid is not protected by that reference, and pid_task() doesn't increment the reference on the task. That's what I remember of the pid_task() interface, that requires rcu to be held while you look inside, where get_pid_task() does not. -Chris
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c index 6167e68bbb25..dc90b044a217 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_context.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c @@ -212,8 +212,8 @@ context_get_vm_rcu(struct i915_gem_context *ctx) static void intel_context_set_gem(struct intel_context *ce, struct i915_gem_context *ctx) { - GEM_BUG_ON(ce->gem_context); - ce->gem_context = ctx; + GEM_BUG_ON(rcu_access_pointer(ce->gem_context)); + RCU_INIT_POINTER(ce->gem_context, ctx); if (!test_bit(CONTEXT_ALLOC_BIT, &ce->flags)) ce->ring = __intel_context_ring_size(SZ_16K); @@ -244,6 +244,7 @@ static void __free_engines(struct i915_gem_engines *e, unsigned int count) if (!e->engines[count]) continue; + RCU_INIT_POINTER(e->engines[count]->gem_context, NULL); intel_context_put(e->engines[count]); } kfree(e); diff --git a/drivers/gpu/drm/i915/gt/intel_context_types.h b/drivers/gpu/drm/i915/gt/intel_context_types.h index 7dd03ad9826c..9527a659546c 100644 --- a/drivers/gpu/drm/i915/gt/intel_context_types.h +++ b/drivers/gpu/drm/i915/gt/intel_context_types.h @@ -44,7 +44,7 @@ struct intel_context { #define intel_context_inflight_count(ce) ptr_unmask_bits((ce)->inflight, 2) struct i915_address_space *vm; - struct i915_gem_context *gem_context; + struct i915_gem_context __rcu *gem_context; struct list_head signal_link; struct list_head signals; diff --git a/drivers/gpu/drm/i915/gt/intel_reset.c b/drivers/gpu/drm/i915/gt/intel_reset.c index ab8213b90517..1c51296646e0 100644 --- a/drivers/gpu/drm/i915/gt/intel_reset.c +++ b/drivers/gpu/drm/i915/gt/intel_reset.c @@ -85,20 +85,27 @@ static bool mark_guilty(struct i915_request *rq) bool banned; int i; - ctx = rq->context->gem_context; + rcu_read_lock(); + ctx = rcu_dereference(rq->context->gem_context); + if (ctx && !kref_get_unless_zero(&ctx->ref)) + ctx = NULL; + rcu_read_unlock(); if (!ctx) return false; if (i915_gem_context_is_closed(ctx)) { intel_context_set_banned(rq->context); - return true; + banned = true; + goto out; } atomic_inc(&ctx->guilty_count); /* Cool contexts are too cool to be banned! (Used for reset testing.) */ - if (!i915_gem_context_is_bannable(ctx)) - return false; + if (!i915_gem_context_is_bannable(ctx)) { + banned = false; + goto out; + } dev_notice(ctx->i915->drm.dev, "%s context reset due to GPU hang\n", @@ -122,13 +129,20 @@ static bool mark_guilty(struct i915_request *rq) client_mark_guilty(ctx, banned); +out: + i915_gem_context_put(ctx); return banned; } static void mark_innocent(struct i915_request *rq) { - if (rq->context->gem_context) - atomic_inc(&rq->context->gem_context->active_count); + struct i915_gem_context *ctx; + + rcu_read_lock(); + ctx = rcu_dereference(rq->context->gem_context); + if (ctx) + atomic_inc(&ctx->active_count); + rcu_read_unlock(); } void __i915_request_reset(struct i915_request *rq, bool guilty) diff --git a/drivers/gpu/drm/i915/gt/intel_ring_submission.c b/drivers/gpu/drm/i915/gt/intel_ring_submission.c index 118170eb51b4..81f872f9ef03 100644 --- a/drivers/gpu/drm/i915/gt/intel_ring_submission.c +++ b/drivers/gpu/drm/i915/gt/intel_ring_submission.c @@ -1550,7 +1550,7 @@ static int remap_l3_slice(struct i915_request *rq, int slice) static int remap_l3(struct i915_request *rq) { - struct i915_gem_context *ctx = rq->context->gem_context; + struct i915_gem_context *ctx = i915_request_gem_context(rq); int i, err; if (!ctx || !ctx->remap_slice) diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c index 7e2cb063110c..fda0977d2059 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.c +++ b/drivers/gpu/drm/i915/i915_gpu_error.c @@ -1221,7 +1221,7 @@ static void error_record_engine_registers(struct i915_gpu_state *error, static void record_request(const struct i915_request *request, struct drm_i915_error_request *erq) { - const struct i915_gem_context *ctx = request->context->gem_context; + const struct i915_gem_context *ctx; erq->flags = request->fence.flags; erq->context = request->fence.context; @@ -1231,7 +1231,13 @@ static void record_request(const struct i915_request *request, erq->start = i915_ggtt_offset(request->ring->vma); erq->head = request->head; erq->tail = request->tail; - erq->pid = ctx && ctx->pid ? pid_nr(ctx->pid) : 0; + + erq->pid = 0; + rcu_read_lock(); + ctx = rcu_dereference(request->context->gem_context); + if (ctx) + erq->pid = pid_nr(ctx->pid); + rcu_read_unlock(); } static void engine_record_requests(struct intel_engine_cs *engine, @@ -1298,28 +1304,34 @@ static void error_record_engine_execlists(const struct intel_engine_cs *engine, static bool record_context(struct drm_i915_error_context *e, const struct i915_request *rq) { - const struct i915_gem_context *ctx = rq->context->gem_context; + struct i915_gem_context *ctx; + struct task_struct *task; + bool capture; + rcu_read_lock(); + ctx = rcu_dereference(rq->context->gem_context); + if (ctx && !kref_get_unless_zero(&ctx->ref)) + ctx = NULL; + rcu_read_unlock(); if (!ctx) return false; - if (ctx->pid) { - struct task_struct *task; - - rcu_read_lock(); - task = pid_task(ctx->pid, PIDTYPE_PID); - if (task) { - strcpy(e->comm, task->comm); - e->pid = task->pid; - } - rcu_read_unlock(); + rcu_read_lock(); + task = pid_task(ctx->pid, PIDTYPE_PID); + if (task) { + strcpy(e->comm, task->comm); + e->pid = task->pid; } + rcu_read_unlock(); e->sched_attr = ctx->sched; e->guilty = atomic_read(&ctx->guilty_count); e->active = atomic_read(&ctx->active_count); - return i915_gem_context_no_error_capture(ctx); + capture = i915_gem_context_no_error_capture(ctx); + + i915_gem_context_put(ctx); + return capture; } struct capture_vma { diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c index 14a5a99284fa..44a0d1a950c5 100644 --- a/drivers/gpu/drm/i915/i915_request.c +++ b/drivers/gpu/drm/i915/i915_request.c @@ -76,7 +76,7 @@ static const char *i915_fence_get_timeline_name(struct dma_fence *fence) if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags)) return "signaled"; - ctx = to_request(fence)->context->gem_context; + ctx = i915_request_gem_context(to_request(fence)); if (!ctx) return "[" DRIVER_NAME "]"; @@ -1312,8 +1312,8 @@ void i915_request_add(struct i915_request *rq) prev = __i915_request_commit(rq); - if (rq->context->gem_context) - attr = rq->context->gem_context->sched; + if (rcu_access_pointer(rq->context->gem_context)) + attr = i915_request_gem_context(rq)->sched; /* * Boost actual workloads past semaphores! diff --git a/drivers/gpu/drm/i915/i915_request.h b/drivers/gpu/drm/i915/i915_request.h index 0e4fe3205ce7..565322640378 100644 --- a/drivers/gpu/drm/i915/i915_request.h +++ b/drivers/gpu/drm/i915/i915_request.h @@ -28,6 +28,7 @@ #include <linux/dma-fence.h> #include <linux/lockdep.h> +#include "gem/i915_gem_context_types.h" #include "gt/intel_context_types.h" #include "gt/intel_engine_types.h" #include "gt/intel_timeline_types.h" @@ -463,6 +464,13 @@ i915_request_timeline(struct i915_request *rq) lockdep_is_held(&rcu_access_pointer(rq->timeline)->mutex)); } +static inline struct i915_gem_context * +i915_request_gem_context(struct i915_request *rq) +{ + /* Valid only while the request is being constructed (or retired). */ + return rcu_dereference_protected(rq->context->gem_context, true); +} + static inline struct intel_timeline * i915_request_active_timeline(struct i915_request *rq) {
The only protection for intel_context.gem_cotext is granted by RCU, so annotate it as a rcu protected pointer and carefully dereference it in the few occasions we need to use it. Fixes: 9f3ccd40acf4 ("drm/i915: Drop GEM context as a direct link from i915_request") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Andi Shyti <andi.shyti@intel.com> --- drivers/gpu/drm/i915/gem/i915_gem_context.c | 5 ++- drivers/gpu/drm/i915/gt/intel_context_types.h | 2 +- drivers/gpu/drm/i915/gt/intel_reset.c | 26 +++++++++--- .../gpu/drm/i915/gt/intel_ring_submission.c | 2 +- drivers/gpu/drm/i915/i915_gpu_error.c | 40 ++++++++++++------- drivers/gpu/drm/i915/i915_request.c | 6 +-- drivers/gpu/drm/i915/i915_request.h | 8 ++++ 7 files changed, 62 insertions(+), 27 deletions(-)