Re: [RFC 5/8] drm/i915: Contexts can use struct pid stored in the client

[Tvrtko Ursulin <tvrtko.ursulin@xxxxxxxxxxxxxxx>]

On 10/01/2020 13:42, Chris Wilson wrote:
> Quoting Tvrtko Ursulin (2020-01-10 13:30:46)
> > From: Tvrtko Ursulin <tvrtko.ursulin@xxxxxxxxx>
> >
> > Now that contexts keep their parent client reference counted, we can
> > remove separate struct pid reference owned by contexts in favour of the
> > one already held by the client.
> >
> > Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@xxxxxxxxx>
> > ---
> >   drivers/gpu/drm/i915/gem/i915_gem_context.c       | 13 ++++---------
> >   drivers/gpu/drm/i915/gem/i915_gem_context_types.h | 10 ----------
> >   drivers/gpu/drm/i915/i915_debugfs.c               |  7 ++++---
> >   drivers/gpu/drm/i915/i915_gpu_error.c             |  8 ++++----
> >   4 files changed, 12 insertions(+), 26 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
> > index ba8ccc754f20..758cebb99ba4 100644
> > --- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
> > +++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
> > @@ -323,7 +323,6 @@ static void i915_gem_context_free(struct i915_gem_context *ctx)
> >          if (ctx->timeline)
> >                  intel_timeline_put(ctx->timeline);
> >   
> > -       put_pid(ctx->pid);
> >          mutex_destroy(&ctx->mutex);
> >   
> >          kfree_rcu(ctx, rcu);
> > @@ -794,24 +793,20 @@ static int gem_context_register(struct i915_gem_context *ctx,
> >                  WRITE_ONCE(vm->file, fpriv); /* XXX */
> >          mutex_unlock(&ctx->mutex);
> >   
> > -       ctx->pid = get_task_pid(current, PIDTYPE_PID);
> >          snprintf(ctx->name, sizeof(ctx->name), "%s[%d]",
> > -                current->comm, pid_nr(ctx->pid));
> > +                current->comm, pid_nr(client->pid));
> >   
> >          /* And finally expose ourselves to userspace via the idr */
> >          ret = xa_alloc(&fpriv->context_xa, id, ctx, xa_limit_32b, GFP_KERNEL);
> > -       if (ret) {
> > -               put_pid(fetch_and_zero(&ctx->pid));
> > -               goto out;
> > -       }
> > +       if (ret)
> > +               return ret;
> >   
> >          ctx->client = client = i915_drm_client_get(fpriv->client);
> >          spin_lock(&client->ctx_lock);
> >          list_add_tail_rcu(&ctx->client_link, &client->ctx_list);
> >          spin_unlock(&client->ctx_lock);
> >   
> > -out:
> > -       return ret;
> > +       return 0;
> >   }
> >   
> >   int i915_gem_context_open(struct drm_i915_private *i915,
> > diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
> > index 879824159646..23421377a43d 100644
> > --- a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
> > +++ b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
> > @@ -90,16 +90,6 @@ struct i915_gem_context {
> >           */
> >          struct i915_address_space __rcu *vm;
> >   
> > -       /**
> > -        * @pid: process id of creator
> > -        *
> > -        * Note that who created the context may not be the principle user,
> > -        * as the context may be shared across a local socket. However,
> > -        * that should only affect the default context, all contexts created
> > -        * explicitly by the client are expected to be isolated.
> > -        */
> > -       struct pid *pid;
> > -
> >          /** link: place with &drm_i915_private.context_list */
> >          struct list_head link;
> >          struct llist_node free_link;
> > diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
> > index 8f01c2bc7355..bc533501b4e0 100644
> > --- a/drivers/gpu/drm/i915/i915_debugfs.c
> > +++ b/drivers/gpu/drm/i915/i915_debugfs.c
> > @@ -346,7 +346,8 @@ static void print_context_stats(struct seq_file *m,
> >                          rcu_read_unlock();
> >   
> >                          rcu_read_lock();
> > -                       task = pid_task(ctx->pid ?: file->pid, PIDTYPE_PID);
> > +                       task = pid_task(ctx->client->pid ?: file->pid,
> > +                                       PIDTYPE_PID);
> >                          snprintf(name, sizeof(name), "%s",
> >                                   task ? task->comm : "<unknown>");
> >                          rcu_read_unlock();
> > @@ -1492,10 +1493,10 @@ static int i915_context_status(struct seq_file *m, void *unused)
> >                  spin_unlock(&i915->gem.contexts.lock);
> >   
> >                  seq_puts(m, "HW context ");
> > -               if (ctx->pid) {
> > +               if (ctx->client->pid) {
> >                          struct task_struct *task;
> >   
> > -                       task = get_pid_task(ctx->pid, PIDTYPE_PID);
> > +                       task = get_pid_task(ctx->client->pid, PIDTYPE_PID);
> >                          if (task) {
> >                                  seq_printf(m, "(%s [%d]) ",
> >                                             task->comm, task->pid);
> > diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
> > index fda0977d2059..9240327bdb7d 100644
> > --- a/drivers/gpu/drm/i915/i915_gpu_error.c
> > +++ b/drivers/gpu/drm/i915/i915_gpu_error.c
> > @@ -1235,8 +1235,8 @@ static void record_request(const struct i915_request *request,
> >          erq->pid = 0;
> >          rcu_read_lock();
> >          ctx = rcu_dereference(request->context->gem_context);
> > -       if (ctx)
> > -               erq->pid = pid_nr(ctx->pid);
> > +       if (ctx && ctx->client->pid)
> > +               erq->pid = pid_nr(ctx->client->pid);
> >          rcu_read_unlock();
> >   }
> >   
> > @@ -1313,11 +1313,11 @@ static bool record_context(struct drm_i915_error_context *e,
> >          if (ctx && !kref_get_unless_zero(&ctx->ref))
> >                  ctx = NULL;
> >          rcu_read_unlock();
> > -       if (!ctx)
> > +       if (!ctx || !ctx->client->pid)
> >                  return false;
>
> Can we have a client without a pid?

Yes, when fbdev registers I ignore it as a client because sysfs is not 
there yet. I contemplated stashing it somewhere and registering later 
but did not yet decide to actually do it. This was actually because 
without sysfs I had no use for pid and name. Perhaps now that I have 
moved this patch earlier I can change this to always have those two.. 
okay I'll see.

> Holding a reference to the ctx is sufficent to protect the
> i915_drm_client, ok.
>
> >          rcu_read_lock();
> > -       task = pid_task(ctx->pid, PIDTYPE_PID);
> > +       task = pid_task(ctx->client->pid, PIDTYPE_PID);
> >          if (task) {
> >                  strcpy(e->comm, task->comm);
> >                  e->pid = task->pid;
>
> Ok, if you move this ahead to patch 3, we can use this as justification
> for i915_drm_client all by itself.

Yes mostly done but will see the above, plus a small regression in 
busyness tracking crept in after rebase, somehow. Will post the series 
when I resolve that.

Regards,

Tvrtko

_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx