media::V4L2VideoDecoder: introduce kMaxNumOfInstances

MS Teams on ARM-based ChromeOS with the new VideoDecoder (not launched)
tries to create many direct video decoders (I see e.g. 42 on kukui);
some succeed but eventually they don't and the tab ends up hitting an
OOM. The process is not well understood and needs further debugging.

This CL replicates the kMaxNumOfInstances [1] logic from the legacy
VideoDecodeAccelerator to the new direct VideoDecoder, so the new
behaves like the old, allowing us to unblock the linked bug.

[1] https://source.chromium.org/search?q=kMaxNumOfInstances&sq=&ss=chromium%2Fchromium%2Fsrc:media%2Fgpu%2Fv4l2%2F

Landing this provision on Kukui was already discussed on
.

https: //chromium-review.googlesource.com/c/chromium/src/+/2372926/10#message-ca9039b5fb3eb9e1fbcd9701ddc476b257f088d6
Bug: b/170870476
Change-Id: Ice0d9f16e7ca0063c03cccdafe54e29ad5602cd0
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2628069
Commit-Queue: Miguel Casas <mcasas@chromium.org>
Reviewed-by: Andres Calderon Jaramillo <andrescj@chromium.org>
Reviewed-by: Fritz Koenig <frkoenig@chromium.org>
Cr-Commit-Position: refs/heads/master@{#843377}

media/gpu/v4l2/v4l2_video_decoder.cc

@@ -46,6 +46,9 @@ constexpr size_t kDpbOutputBufferExtraCount = limits::kMaxVideoFrames + 1;
 
 }  // namespace
 
+// static
+base::AtomicRefCount V4L2VideoDecoder::num_instances_(0);
+
 // static
 std::unique_ptr<DecoderInterface> V4L2VideoDecoder::Create(
     scoped_refptr<base::SequencedTaskRunner> decoder_task_runner,
@@ -80,6 +83,7 @@ V4L2VideoDecoder::V4L2VideoDecoder(
     base::WeakPtr<DecoderInterface::Client> client,
     scoped_refptr<V4L2Device> device)
     : DecoderInterface(std::move(decoder_task_runner), std::move(client)),
+      can_use_decoder_(num_instances_.Increment() < kMaxNumOfInstances),
       device_(std::move(device)),
       weak_this_factory_(this) {
   DCHECK_CALLED_ON_VALID_SEQUENCE(decoder_sequence_checker_);
@@ -110,6 +114,7 @@ V4L2VideoDecoder::~V4L2VideoDecoder() {
   }
 
   weak_this_factory_.InvalidateWeakPtrs();
+  num_instances_.Decrement();
 }
 
 void V4L2VideoDecoder::Initialize(const VideoDecoderConfig& config,
@@ -121,6 +126,13 @@ void V4L2VideoDecoder::Initialize(const VideoDecoderConfig& config,
   DCHECK(state_ == State::kUninitialized || state_ == State::kDecoding);
   DVLOGF(3);
 
+  if (!can_use_decoder_) {
+    VLOGF(1) << "Reached maximum number of decoder instances ("
+             << kMaxNumOfInstances << ")";
+    std::move(init_cb).Run(StatusCode::kDecoderCreationFailed);
+    return;
+  }
+
   if (cdm_context || config.is_encrypted()) {
     VLOGF(1) << "V4L2 decoder does not support encrypted stream";
     std::move(init_cb).Run(StatusCode::kEncryptedContentUnsupported);

media/gpu/v4l2/v4l2_video_decoder.h

@@ -135,6 +135,16 @@ class MEDIA_GPU_EXPORT V4L2VideoDecoder
   // Change the state and check the state transition is valid.
   void SetState(State new_state);
 
+  // Pages with multiple V4L2VideoDecoder instances might run out of memory
+  // (e.g. b/170870476) or crash (e.g. crbug.com/1109312). To avoid that and
+  // while the investigation goes on, limit the maximum number of simultaneous
+  // decoder instances for now. |num_instances_| tracks the number of
+  // simultaneous decoders. |can_use_decoder_| is true iff we haven't reached
+  // the maximum number of instances at the time this decoder is created.
+  static constexpr int kMaxNumOfInstances = 10;
+  static base::AtomicRefCount num_instances_;
+  const bool can_use_decoder_;
+
   // The V4L2 backend, i.e. the part of the decoder that sends
   // decoding jobs to the kernel.
   std::unique_ptr<V4L2VideoDecoderBackend> backend_;