Re: [Linux Kernel Bugs] KASAN: slab-use-after-free Read in cec_queue_msg_fh and 4 other crashes in the cec device (`cec_ioctl`)

From: Takashi Iwai
Date: Fri Apr 19 2024 - 10:51:39 EST


On Mon, 26 Feb 2024 13:27:16 +0100,
Yang, Chenyuan wrote:
>
> Hi Hans,
>
> Thank you for your continued efforts in investigating this bug and implementing the new patch!
>
> Regarding the two warnings, they have been addressed by this new patch and are no longer reproducible. Additionally, I conducted a 48-hour fuzzing test on the CEC driver, which has successfully eliminated the previous hanging issue.
>
> One thing to note that the system will now log timeout events:
> ```
> [ 2281.265385][ T2034] cec-vivid-001-vid-out0: transmit timed out
> [ 2282.994510][ T2017] cec-vivid-000-vid-cap0: transmit timed out
> [ 2283.063484][ T2050] cec-vivid-002-vid-out0: transmit timed out
> [ 2283.073468][ T2065] cec-vivid-003-vid-cap0: transmit timed out
> [ 2283.373518][ T2033] cec-vivid-001-vid-cap0: transmit timed out
> [ 2285.113544][ T2018] cec-vivid-000-vid-out0: transmit timed out
> [ 2285.193502][ T2050] cec-vivid-002-vid-out0: transmit timed out
> [ 2285.193570][ T2065] cec-vivid-003-vid-cap0: transmit timed out
> [ 2285.513570][ T2033] cec-vivid-001-vid-cap0: transmit timed out
> ```
>
> Best,
> Chenyuan

Hi Hans,

how is the current status of this bug fix? It seems that the thread
stalled, and I wonder how we can go further.

I'm asking it because CVE-2024-23848 was assigned and we've been asked
about the bug fix.


Thanks!

Takashi

>
> From: Hans Verkuil <hverkuil-cisco@xxxxxxxxx>
> Date: Friday, February 23, 2024 at 8:44 AM
> To: Yang, Chenyuan <cy54@xxxxxxxxxxxx>, linux-media@xxxxxxxxxxxxxxx <linux-media@xxxxxxxxxxxxxxx>, linux-kernel@xxxxxxxxxxxxxxx <linux-kernel@xxxxxxxxxxxxxxx>
> Cc: jani.nikula@xxxxxxxxx <jani.nikula@xxxxxxxxx>, syzkaller@xxxxxxxxxxxxxxxx <syzkaller@xxxxxxxxxxxxxxxx>, mchehab@xxxxxxxxxx <mchehab@xxxxxxxxxx>, Zhao, Zijie <zijie4@xxxxxxxxxxxx>, Zhang, Lingming <lingming@xxxxxxxxxxxx>
> Subject: Re: [Linux Kernel Bugs] KASAN: slab-use-after-free Read in cec_queue_msg_fh and 4 other crashes in the cec device (`cec_ioctl`)
> Hi Chenyuan,
>
> Here is another patch for you to try. I think it is good for blocking CEC_ADAP_S_LOG_ADDRS
> ioctl calls, but if the filehandle is in non-blocking mode, I'm still not certain it
> is correct. But one issue at a time :-)
>
> Regards,
>
>         Hans
>
> diff --git a/drivers/media/cec/core/cec-adap.c b/drivers/media/cec/core/cec-adap.c
> index 559a172ebc6c..a493cbce2456 100644
> --- a/drivers/media/cec/core/cec-adap.c
> +++ b/drivers/media/cec/core/cec-adap.c
> @@ -936,8 +936,7 @@ int cec_transmit_msg_fh(struct cec_adapter *adap, struct cec_msg *msg,
>           */
>          mutex_unlock(&adap->lock);
>          wait_for_completion_killable(&data->c);
> -       if (!data->completed)
> -               cancel_delayed_work_sync(&data->work);
> +       cancel_delayed_work_sync(&data->work);
>          mutex_lock(&adap->lock);
>
>          /* Cancel the transmit if it was interrupted */
> @@ -1575,9 +1574,12 @@ static int cec_config_thread_func(void *arg)
>   */
>  static void cec_claim_log_addrs(struct cec_adapter *adap, bool block)
>  {
> -       if (WARN_ON(adap->is_configuring || adap->is_configured))
> +       if (WARN_ON(adap->is_claiming_log_addrs ||
> +                   adap->is_configuring || adap->is_configured))
>                  return;
>
> +       adap->is_claiming_log_addrs = true;
> +
>          init_completion(&adap->config_completion);
>
>          /* Ready to kick off the thread */
> @@ -1592,6 +1594,7 @@ static void cec_claim_log_addrs(struct cec_adapter *adap, bool block)
>                  wait_for_completion(&adap->config_completion);
>                  mutex_lock(&adap->lock);
>          }
> +       adap->is_claiming_log_addrs = false;
>  }
>
>  /*
> diff --git a/drivers/media/cec/core/cec-api.c b/drivers/media/cec/core/cec-api.c
> index 67dc79ef1705..3ef915344304 100644
> --- a/drivers/media/cec/core/cec-api.c
> +++ b/drivers/media/cec/core/cec-api.c
> @@ -178,7 +178,7 @@ static long cec_adap_s_log_addrs(struct cec_adapter *adap, struct cec_fh *fh,
>                             CEC_LOG_ADDRS_FL_ALLOW_RC_PASSTHRU |
>                             CEC_LOG_ADDRS_FL_CDC_ONLY;
>          mutex_lock(&adap->lock);
> -       if (!adap->is_configuring &&
> +       if (!adap->is_claiming_log_addrs && !adap->is_configuring &&
>              (!log_addrs.num_log_addrs || !adap->is_configured) &&
>              !cec_is_busy(adap, fh)) {
>                  err = __cec_s_log_addrs(adap, &log_addrs, block);
> @@ -664,6 +664,8 @@ static int cec_release(struct inode *inode, struct file *filp)
>                  list_del_init(&data->xfer_list);
>          }
>          mutex_unlock(&adap->lock);
> +
> +       mutex_lock(&fh->lock);
>          while (!list_empty(&fh->msgs)) {
>                  struct cec_msg_entry *entry =
>                          list_first_entry(&fh->msgs, struct cec_msg_entry, list);
> @@ -681,6 +683,7 @@ static int cec_release(struct inode *inode, struct file *filp)
>                          kfree(entry);
>                  }
>          }
> +       mutex_unlock(&fh->lock);
>          kfree(fh);
>
>          cec_put_device(devnode);
> diff --git a/include/media/cec.h b/include/media/cec.h
> index 10c9cf6058b7..cc3fcd0496c3 100644
> --- a/include/media/cec.h
> +++ b/include/media/cec.h
> @@ -258,6 +258,7 @@ struct cec_adapter {
>          u16 phys_addr;
>          bool needs_hpd;
>          bool is_enabled;
> +       bool is_claiming_log_addrs;
>          bool is_configuring;
>          bool must_reconfigure;
>          bool is_configured;
>