Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   Alexandre Courbot <gnurou@gmail.com>
To:     Hans Verkuil <hverkuil-cisco@xs4all.nl>,
        Tomasz Figa <tfiga@chromium.org>,
        Marek Szyprowski <m.szyprowski@samsung.com>,
        Mauro Carvalho Chehab <mchehab@kernel.org>
Cc:     linux-media@vger.kernel.org, linux-kernel@vger.kernel.org,
        Alexandre Courbot <gnurou@gmail.com>
Subject: [PATCH v2 1/2] media: videobuf2: always call poll_wait() on queues
Date:   Tue, 24 Nov 2020 00:18:42 +0900
Message-Id: <20201123151843.798205-2-gnurou@gmail.com>
In-Reply-To: <20201123151843.798205-1-gnurou@gmail.com>
References: <20201123151843.798205-1-gnurou@gmail.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

do_poll()/do_select() seem to set the _qproc member of poll_table to
NULL the first time they are called on a given table, making subsequent
calls of poll_wait() on that table no-ops. This is a problem for vb2
which calls poll_wait() on the V4L2 queues' waitqueues only when a
queue-related event is requested, which may not necessarily be the case
during the first poll.

Fix this by making the call to poll_wait() happen first thing and
unconditionally in vb2_core_poll().

Signed-off-by: Alexandre Courbot <gnurou@gmail.com>
---
 drivers/media/common/videobuf2/videobuf2-core.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/drivers/media/common/videobuf2/videobuf2-core.c b/drivers/media/common/videobuf2/videobuf2-core.c
index 4eab6d81cce1..ef06f90f5c6b 100644
--- a/drivers/media/common/videobuf2/videobuf2-core.c
+++ b/drivers/media/common/videobuf2/videobuf2-core.c
@@ -2363,13 +2363,20 @@ __poll_t vb2_core_poll(struct vb2_queue *q, struct file *file,
 	struct vb2_buffer *vb = NULL;
 	unsigned long flags;
 
+	/*
+	 * poll_wait() MUST be called on the first invocation on all the
+	 * potential queues of interest, even if we are not interested in their
+	 * events during this first call. Failure to do so will result in
+	 * queue's events to be ignored because the poll_table won't be capable
+	 * of adding new wait queues thereafter.
+	 */
+	poll_wait(file, &q->done_wq, wait);
+
 	if (!q->is_output && !(req_events & (EPOLLIN | EPOLLRDNORM)))
 		return 0;
 	if (q->is_output && !(req_events & (EPOLLOUT | EPOLLWRNORM)))
 		return 0;
 
-	poll_wait(file, &q->done_wq, wait);
-
 	/*
 	 * Start file I/O emulator only if streaming API has not been used yet.
 	 */
-- 
2.29.2