Re: [PATCH] do_work_pending: Enable interrupts in do_work_pending()

Re: [PATCH] do_work_pending: Enable interrupts in do_work_pending()

[Russell King - ARM Linux]

On Fri, Dec 05, 2014 at 08:21:06PM +0530, Anand Moon wrote:
> @@ -574,12 +574,14 @@ asmlinkage int
>  do_work_pending(struct pt_regs *regs, unsigned int thread_flags, int syscall)
>  {
>  	do {
> +		if (unlikely(!user_mode(regs)))
> +			return 0;
> +		/* Enable interrupts; they are disabled again on return to
> +		 * caller. */
> +		local_irq_enable();
>  		if (likely(thread_flags & _TIF_NEED_RESCHED)) {
>  			schedule();
>  		} else {
> -			if (unlikely(!user_mode(regs)))
> -				return 0;
> -			local_irq_enable();
>  			if (thread_flags & _TIF_SIGPENDING) {
>  				int restart = do_signal(regs, syscall);
>  				if (unlikely(restart)) {

I'm happy with the hunk above, but:

> @@ -588,6 +590,7 @@ do_work_pending(struct pt_regs *regs, unsigned int thread_flags, int syscall)
>  					 * Deal with it without leaving
>  					 * the kernel space.
>  					 */
> +					local_irq_disable();
>  					return restart;

not this one.  The code expects in the non-zero return case, that
interrupts will be enabled, otherwise we will be restarting the syscall
with IRQs disabled, and calling into the syscall function with IRQs
disabled.

-- 
FTTC broadband for 0.8mile line: currently at 9.5Mbps down 400kbps up
according to speedtest.net.