[PATCH] sound/pci/riptide or drivers/base/firmware: Fix a possible sleep-in-atomic bug

[PATCH] sound/pci/riptide or drivers/base/firmware: Fix a possible sleep-in-atomic bug

[Jia-Ju Bai]

The riptide driver may sleep under a spinlock, and the function call path is:
snd_riptide_prepare (acquire the spinlock)
  setsampleformat
    sendcmd
      riptide_reset
        try_to_load_firmware
          request_firmware
            _request_firmware (drivers/base/firmware_class.c)
              _request_firmware_prepare
                kzalloc(GFP_KERNEL) --> may sleep

To fix it, GFP_KERNEL is replaced with GFP_ATOMIC in _request_firmware_prepare.
This bug is found by my static analysis tool and my code review.

Signed-off-by: Jia-Ju Bai <baijiaju1990@163.com>
---
 drivers/base/firmware_class.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/base/firmware_class.c b/drivers/base/firmware_class.c
index 4b57cf5..d89f676 100644
--- a/drivers/base/firmware_class.c
+++ b/drivers/base/firmware_class.c
@@ -1138,7 +1138,7 @@ static inline void kill_pending_fw_fallback_reqs(bool only_kill_custom) { }
 	struct firmware_buf *buf;
 	int ret;
 
-	*firmware_p = firmware = kzalloc(sizeof(*firmware), GFP_KERNEL);
+	*firmware_p = firmware = kzalloc(sizeof(*firmware), GFP_ATOMIC);
 	if (!firmware) {
 		dev_err(device, "%s: kmalloc(struct firmware) failed\n",
 			__func__);
-- 
1.7.9.5

Re: [alsa-devel] [PATCH] sound/pci/riptide or drivers/base/firmware: Fix a possible sleep-in-atomic bug

[Takashi Iwai]

On Mon, 09 Oct 2017 04:13:16 +0200,
Jia-Ju Bai wrote:
> 
> The riptide driver may sleep under a spinlock, and the function call path is:
> snd_riptide_prepare (acquire the spinlock)
>   setsampleformat
>     sendcmd
>       riptide_reset
>         try_to_load_firmware
>           request_firmware
>             _request_firmware (drivers/base/firmware_class.c)
>               _request_firmware_prepare
>                 kzalloc(GFP_KERNEL) --> may sleep
> 
> To fix it, GFP_KERNEL is replaced with GFP_ATOMIC in _request_firmware_prepare.
> This bug is found by my static analysis tool and my code review.
> 
> Signed-off-by: Jia-Ju Bai <baijiaju1990@163.com>

This doesn't happen.  try_to_load_firmware() aborts before the
request_firmware() call when chip == NULL, which is the case from
sendcmd().

And, moreover, the call of request_firmware() itself inside the
spinlock would be already buggy.


thanks,

Takashi

> ---
>  drivers/base/firmware_class.c |    2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/base/firmware_class.c b/drivers/base/firmware_class.c
> index 4b57cf5..d89f676 100644
> --- a/drivers/base/firmware_class.c
> +++ b/drivers/base/firmware_class.c
> @@ -1138,7 +1138,7 @@ static inline void kill_pending_fw_fallback_reqs(bool only_kill_custom) { }
>  	struct firmware_buf *buf;
>  	int ret;
>  
> -	*firmware_p = firmware = kzalloc(sizeof(*firmware), GFP_KERNEL);
> +	*firmware_p = firmware = kzalloc(sizeof(*firmware), GFP_ATOMIC);
>  	if (!firmware) {
>  		dev_err(device, "%s: kmalloc(struct firmware) failed\n",
>  			__func__);
> -- 
> 1.7.9.5
> 
> 
> _______________________________________________
> Alsa-devel mailing list
> Alsa-devel@alsa-project.org
> http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
>