should drivers avoid the use of in_atomic()?

[Greg KH]

On Sun, May 15, 2011 at 03:22:10PM -0400, Robert P. J. Day wrote:
> 
>   perusing some timer-related kernel stuff and i ran across this in
> include/linux/hardirq.h:
> 
> /*
>  * Are we running in atomic context?  WARNING: this macro cannot
>  * always detect atomic context; in particular, it cannot know about
>  * held spinlocks in non-preemptible kernels.  Thus it should not be
>  * used in the general case to determine whether sleeping is possible.
>  * Do not use in_atomic() in driver code.
>  */
> #define in_atomic()     ((preempt_count() & ~PREEMPT_ACTIVE) != PREEMPT_INATOMIC_BASE)
> 
>   but a quick check shows a sprinkling of in_atomic() checks in the
> drivers/ directory.  is that admonition overly strict?  or what?

No, it's right, using it is wrong, but unfortunatly, it sometimes is the
best that we have to use.  Fixing this in the drivers would be great to
do, feel free to add it to the kernel janitor's TODO list.

greg k-h