Commit a5af081d authored by Chris Wilson's avatar Chris Wilson

drm/i915/perf: Mark up the racy use of perf->exclusive_stream

Inside the general i915_oa_init_reg_state() we avoid using the
perf->mutex. However, we rely on perf->exclusive_stream being valid to
access at that point, and for that we have to control the race with
disabling perf. This relies on the disabling being a heavy barrier that
inspects all active contexts, after marking the perf->exclusive_stream
as not available. This should ensure that there are no more concurrent
accesses to the perf->exclusive_stream as we destroy it.

Mark up the races around the perf->exclusive_stream so that they stand
out much more. (And hopefully we will be running kcsan to start
validating that the only races we have are carefully controlled.)
Signed-off-by: default avatarChris Wilson <chris@chris-wilson.co.uk>
Cc: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: default avatarMika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200227085723.1961649-2-chris@chris-wilson.co.uk
parent 6875eb3f
...@@ -1405,8 +1405,10 @@ static void i915_oa_stream_destroy(struct i915_perf_stream *stream) ...@@ -1405,8 +1405,10 @@ static void i915_oa_stream_destroy(struct i915_perf_stream *stream)
/* /*
* Unset exclusive_stream first, it will be checked while disabling * Unset exclusive_stream first, it will be checked while disabling
* the metric set on gen8+. * the metric set on gen8+.
*
* See i915_oa_init_reg_state() and lrc_configure_all_contexts()
*/ */
perf->exclusive_stream = NULL; WRITE_ONCE(perf->exclusive_stream, NULL);
perf->ops.disable_metric_set(stream); perf->ops.disable_metric_set(stream);
free_oa_buffer(stream); free_oa_buffer(stream);
...@@ -2847,7 +2849,7 @@ static int i915_oa_stream_init(struct i915_perf_stream *stream, ...@@ -2847,7 +2849,7 @@ static int i915_oa_stream_init(struct i915_perf_stream *stream,
goto err_oa_buf_alloc; goto err_oa_buf_alloc;
stream->ops = &i915_oa_stream_ops; stream->ops = &i915_oa_stream_ops;
perf->exclusive_stream = stream; WRITE_ONCE(perf->exclusive_stream, stream);
ret = perf->ops.enable_metric_set(stream); ret = perf->ops.enable_metric_set(stream);
if (ret) { if (ret) {
...@@ -2867,7 +2869,7 @@ static int i915_oa_stream_init(struct i915_perf_stream *stream, ...@@ -2867,7 +2869,7 @@ static int i915_oa_stream_init(struct i915_perf_stream *stream,
return 0; return 0;
err_enable: err_enable:
perf->exclusive_stream = NULL; WRITE_ONCE(perf->exclusive_stream, NULL);
perf->ops.disable_metric_set(stream); perf->ops.disable_metric_set(stream);
free_oa_buffer(stream); free_oa_buffer(stream);
...@@ -2893,12 +2895,11 @@ void i915_oa_init_reg_state(const struct intel_context *ce, ...@@ -2893,12 +2895,11 @@ void i915_oa_init_reg_state(const struct intel_context *ce,
{ {
struct i915_perf_stream *stream; struct i915_perf_stream *stream;
/* perf.exclusive_stream serialised by lrc_configure_all_contexts() */
if (engine->class != RENDER_CLASS) if (engine->class != RENDER_CLASS)
return; return;
stream = engine->i915->perf.exclusive_stream; /* perf.exclusive_stream serialised by lrc_configure_all_contexts() */
stream = READ_ONCE(engine->i915->perf.exclusive_stream);
/* /*
* For gen12, only CTX_R_PWR_CLK_STATE needs update, but the caller * For gen12, only CTX_R_PWR_CLK_STATE needs update, but the caller
* is already doing that, so nothing to be done for gen12 here. * is already doing that, so nothing to be done for gen12 here.
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment