From: Hans Verkuil <hverkuil+cisco@kernel.org>
To: Fan Wu <fanwu01@zju.edu.cn>, mchehab@kernel.org
Cc: linux-media@vger.kernel.org, linux-kernel@vger.kernel.org,
stable@vger.kernel.org
Subject: Re: [PATCH] media: cx231xx-audio: gate wq_trigger on an audio-local teardown flag
Date: Wed, 15 Jul 2026 13:51:02 +0200 [thread overview]
Message-ID: <8c7e1294-b906-4636-890c-b64d03b0e1d0@kernel.org> (raw)
In-Reply-To: <20260708141622.3936490-1-fanwu01@zju.edu.cn>
On 08/07/2026 16:16, Fan Wu wrote:
> audio_trigger() is deferred work (dev->wq_trigger) armed from
> snd_cx231xx_capture_trigger() on every PCM START/STOP; it dereferences
> dev->adev state and may free the URBs via cx231xx_isoc_audio_deinit().
> cx231xx_audio_fini() tore down that state (snd_card_free_when_closed,
> alt_max_pkt_size) without draining wq_trigger, so work armed before or
> racing fini ran against freed state.
>
> Adding cancel_work_sync() alone is insufficient: in capture_trigger() the
> DEV_DISCONNECTED test and schedule_work() were not atomic, and
> DEV_DISCONNECTED is only set on USB disconnect, but fini also runs on
> cx231xx-alsa module unload (cx231xx_unregister_extension()), which never
> sets it. A trigger that passed the check could still queue work after
> fini's cancel returned an empty queue.
>
> Add an audio-local teardown gate (dev->adev.teardown): fini raises it under
> adev.slock, releases the lock, then calls cancel_work_sync() outside the
> spinlock. Both arm sites perform the teardown check and schedule_work()
> inside one adev.slock section, so once the gate is visible no new work can
> arm after cancel returns. Initialize the lock, work and gate at the top of
> cx231xx_audio_init(), before any fallible allocation, and clear the
> partially-built audio state on its error path, so fini is safe even if a
> later step fails.
>
> This issue was found by an in-house static analysis tool.
>
> Fixes: 61b04cb24a12 ("[media] cx231xx-audio: fix some locking issues")
> Cc: stable@vger.kernel.org
> Assisted-by: Codex:gpt-5.5
> Signed-off-by: Fan Wu <fanwu01@zju.edu.cn>
> ---
> drivers/media/usb/cx231xx/cx231xx-audio.c | 42 +++++++++++++++++++----
> drivers/media/usb/cx231xx/cx231xx.h | 1 +
> 2 files changed, 36 insertions(+), 7 deletions(-)
>
> diff --git a/drivers/media/usb/cx231xx/cx231xx-audio.c b/drivers/media/usb/cx231xx/cx231xx-audio.c
> index 9c71b32552df..44ca75b18a5d 100644
> --- a/drivers/media/usb/cx231xx/cx231xx-audio.c
> +++ b/drivers/media/usb/cx231xx/cx231xx-audio.c
> @@ -441,6 +441,7 @@ static int snd_cx231xx_capture_open(struct snd_pcm_substream *substream)
> static int snd_cx231xx_pcm_close(struct snd_pcm_substream *substream)
> {
> int ret;
> + unsigned long flags;
> struct cx231xx *dev = snd_pcm_substream_chip(substream);
>
> dev_dbg(dev->dev, "closing device\n");
> @@ -470,7 +471,11 @@ static int snd_cx231xx_pcm_close(struct snd_pcm_substream *substream)
> dev_dbg(dev->dev, "released lock\n");
> if (atomic_read(&dev->stream_started) > 0) {
> atomic_set(&dev->stream_started, 0);
> - schedule_work(&dev->wq_trigger);
> +
> + spin_lock_irqsave(&dev->adev.slock, flags);
> + if (!dev->adev.teardown)
> + schedule_work(&dev->wq_trigger);
> + spin_unlock_irqrestore(&dev->adev.slock, flags);
> }
> }
> return 0;
> @@ -509,11 +514,14 @@ static int snd_cx231xx_capture_trigger(struct snd_pcm_substream *substream,
> {
> struct cx231xx *dev = snd_pcm_substream_chip(substream);
> int retval = 0;
> + unsigned long flags;
>
> - if (dev->state & DEV_DISCONNECTED)
> + spin_lock_irqsave(&dev->adev.slock, flags);
> + if (dev->adev.teardown || (dev->state & DEV_DISCONNECTED)) {
> + spin_unlock_irqrestore(&dev->adev.slock, flags);
> return -ENODEV;
> + }
>
> - spin_lock(&dev->adev.slock);
> switch (cmd) {
> case SNDRV_PCM_TRIGGER_START:
> atomic_set(&dev->stream_started, 1);
> @@ -525,10 +533,10 @@ static int snd_cx231xx_capture_trigger(struct snd_pcm_substream *substream,
> retval = -EINVAL;
> break;
> }
> - spin_unlock(&dev->adev.slock);
>
> schedule_work(&dev->wq_trigger);
>
> + spin_unlock_irqrestore(&dev->adev.slock, flags);
> return retval;
> }
>
> @@ -576,12 +584,20 @@ static int cx231xx_audio_init(struct cx231xx *dev)
> dev_dbg(dev->dev,
> "probing for cx231xx non standard usbaudio\n");
>
> + /*
> + * Extension init errors are ignored by the cx231xx core, so fini()
> + * must be safe even if initialization fails part way through.
> + */
> + spin_lock_init(&adev->slock);
> + INIT_WORK(&dev->wq_trigger, audio_trigger);
> + adev->teardown = false;
> + atomic_set(&dev->stream_started, 0);
> +
> err = snd_card_new(dev->dev, index[devnr], "Cx231xx Audio",
> THIS_MODULE, 0, &card);
> if (err < 0)
> return err;
>
> - spin_lock_init(&adev->slock);
> err = snd_pcm_new(card, "Cx231xx Audio", 0, 0, 1, &pcm);
> if (err < 0)
> goto err_free_card;
> @@ -596,8 +612,6 @@ static int cx231xx_audio_init(struct cx231xx *dev)
> strscpy(card->shortname, "Cx231xx Audio", sizeof(card->shortname));
> strscpy(card->longname, "Conexant cx231xx Audio", sizeof(card->longname));
>
> - INIT_WORK(&dev->wq_trigger, audio_trigger);
> -
> err = snd_card_register(card);
> if (err < 0)
> goto err_free_card;
> @@ -651,14 +665,18 @@ static int cx231xx_audio_init(struct cx231xx *dev)
>
> err_free_pkt_size:
> kfree(adev->alt_max_pkt_size);
> + adev->alt_max_pkt_size = NULL;
> err_free_card:
> snd_card_free(card);
> + adev->sndcard = NULL;
>
> return err;
> }
>
> static int cx231xx_audio_fini(struct cx231xx *dev)
> {
> + unsigned long flags;
> +
> if (dev == NULL)
> return 0;
>
> @@ -669,6 +687,16 @@ static int cx231xx_audio_fini(struct cx231xx *dev)
> return 0;
> }
>
> + /*
> + * Block new trigger work before draining already queued work.
> + * cancel_work_sync() may sleep, so it must run after dropping slock.
> + */
> + spin_lock_irqsave(&dev->adev.slock, flags);
> + dev->adev.teardown = true;
> + spin_unlock_irqrestore(&dev->adev.slock, flags);
> +
> + cancel_work_sync(&dev->wq_trigger);
If I am not mistaken, it should be enough to use disable_work_sync() instead of
cancel_work_sync(): this disables the workqueue as well, preventing new work from
being queued. That avoids all the fiddly 'teardown' bits.
Regards,
Hans
> +
> if (dev->adev.sndcard) {
> snd_card_free_when_closed(dev->adev.sndcard);
> kfree(dev->adev.alt_max_pkt_size);
> diff --git a/drivers/media/usb/cx231xx/cx231xx.h b/drivers/media/usb/cx231xx/cx231xx.h
> index 19f5036a78d7..99c6a279944f 100644
> --- a/drivers/media/usb/cx231xx/cx231xx.h
> +++ b/drivers/media/usb/cx231xx/cx231xx.h
> @@ -418,6 +418,7 @@ struct cx231xx_audio {
> struct snd_card *sndcard;
>
> int users, shutdown;
> + bool teardown; /* audio fini in progress; set under slock */
> /* locks */
> spinlock_t slock;
>
next prev parent reply other threads:[~2026-07-15 11:51 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-07-08 14:16 Fan Wu
2026-07-15 11:51 ` Hans Verkuil [this message]
2026-07-16 13:48 ` [PATCH v2] media: cx231xx-audio: drain wq_trigger before audio fini Fan Wu
2026-07-28 13:21 ` hverkuil+cisco
2026-07-29 7:26 ` [PATCH v3] " Fan Wu
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=8c7e1294-b906-4636-890c-b64d03b0e1d0@kernel.org \
--to=hverkuil+cisco@kernel.org \
--cc=fanwu01@zju.edu.cn \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-media@vger.kernel.org \
--cc=mchehab@kernel.org \
--cc=stable@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®