NCR5380: Fix a possible sleep-in-atomic bug in NCR5380_poll_politely2

Commit Message

Jia-Ju Bai Dec. 12, 2017, 2:50 a.m. UTC

From: Jia-Ju Bai <[e-mail]>

The kernel module may sleep under a spinlock.
The function call paths are:
NCR5380_select (acquire the spinlock)
  NCR5380_reselect
    NCR5380_poll_politely
      NCR5380_poll_politely2
        schedule_timeout_uninterruptible --> may sleep

NCR5380_abort (acquire the spinlock)
  do_abort
    NCR5380_poll_politely
      NCR5380_poll_politely2
        schedule_timeout_uninterruptible --> may sleep

To fix it, schedule_timeout_uninterruptible is replaced with mdelay.

This bug is found by my static analysis tool(DSAC) and checked by my code review.


Signed-off-by: Jia-Ju Bai <[e-mail]>
---
 drivers/scsi/NCR5380.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Comments

Finn Thain Dec. 12, 2017, 3:38 a.m. UTC | #1

On Tue, 12 Dec 2017, Jia-Ju Bai wrote:

> From: Jia-Ju Bai <[e-mail]>
> 
> The kernel module may sleep under a spinlock.

The spinlock is always taken in irq mode, and the 
schedule_timeout_uninterruptible() is conditional on !irqs_disabled().

> The function call paths are:
> NCR5380_select (acquire the spinlock)
>   NCR5380_reselect
>     NCR5380_poll_politely
>       NCR5380_poll_politely2
>         schedule_timeout_uninterruptible --> may sleep
> 
> NCR5380_abort (acquire the spinlock)
>   do_abort
>     NCR5380_poll_politely
>       NCR5380_poll_politely2
>         schedule_timeout_uninterruptible --> may sleep
> 

Well, it's expected to sleep here, hence the "sleep for 1ms" comment.

(I notice that you left the comment unchanged in your "fix", was that an 
oversight?)

> To fix it, schedule_timeout_uninterruptible is replaced with mdelay.
> 
> This bug is found by my static analysis tool(DSAC) and checked by my 
> code review.
> 
> 
> Signed-off-by: Jia-Ju Bai <[e-mail]>

NAK.

> ---
>  drivers/scsi/NCR5380.c |    2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/scsi/NCR5380.c b/drivers/scsi/NCR5380.c
> index 90ea0f5..4176aca 100644
> --- a/drivers/scsi/NCR5380.c
> +++ b/drivers/scsi/NCR5380.c
> @@ -202,7 +202,7 @@ static int NCR5380_poll_politely2(struct NCR5380_hostdata *hostdata,
>  

Here's a little more context:

        if (irqs_disabled() || in_interrupt())
                return -ETIMEDOUT;

>  	/* Repeatedly sleep for 1 ms until deadline */
>  	while (time_is_after_jiffies(deadline)) {
> -		schedule_timeout_uninterruptible(1);
> +		mdelay(1);
>  		if ((NCR5380_read(reg1) & bit1) == val1)
>  			return 0;
>  		if ((NCR5380_read(reg2) & bit2) == val2)
> 

--

Jia-Ju Bai Dec. 12, 2017, 3:49 a.m. UTC | #2

Thanks for your reply :)

On 2017/12/12 11:38, Finn Thain wrote:
> On Tue, 12 Dec 2017, Jia-Ju Bai wrote:
>
>> From: Jia-Ju Bai <[e-mail]>
>>
>> The kernel module may sleep under a spinlock.
> The spinlock is always taken in irq mode, and the
> schedule_timeout_uninterruptible() is conditional on !irqs_disabled().
>
I think I ignore this check, which causes a false bug report, sorry.


Thanks,
Jia-Ju Bai

Patch

diff --git a/drivers/scsi/NCR5380.c b/drivers/scsi/NCR5380.c
index 90ea0f5..4176aca 100644
--- a/drivers/scsi/NCR5380.c
+++ b/drivers/scsi/NCR5380.c
@@ -202,7 +202,7 @@  static int NCR5380_poll_politely2(struct NCR5380_hostdata *hostdata,
 
 	/* Repeatedly sleep for 1 ms until deadline */
 	while (time_is_after_jiffies(deadline)) {
-		schedule_timeout_uninterruptible(1);
+		mdelay(1);
 		if ((NCR5380_read(reg1) & bit1) == val1)
 			return 0;
 		if ((NCR5380_read(reg2) & bit2) == val2)