Re: [PATCH] FS: timerfd: [Trimmed unreadable long subject line ]

[Thomas Gleixner]

Arul,

On Fri, 16 Aug 2019, arul.jeniston@xxxxxxxxx wrote:

Please write the subject as a short precise sentence which fits into 70
characters and put the long explanation into the body, i.e. here.

See Documentation/process/submitting-patches.rst

> From: ARUL JENISTON MC <arul.jeniston@xxxxxxxxx>

This lacks a Signed-off-by

> ---
>  fs/timerfd.c | 12 ++++++++++--
>  1 file changed, 10 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/timerfd.c b/fs/timerfd.c
> index 6a6fc8aa1de7..f5094e070e9a 100644
> --- a/fs/timerfd.c
> +++ b/fs/timerfd.c
> @@ -284,8 +284,16 @@ static ssize_t timerfd_read(struct file *file, char __user *buf, size_t count,
>  					&ctx->t.alarm, ctx->tintv) - 1;
>  				alarm_restart(&ctx->t.alarm);
>  			} else {
> -				ticks += hrtimer_forward_now(&ctx->t.tmr,
> -							     ctx->tintv) - 1;
> +				u64 nooftimeo = hrtimer_forward_now(&ctx->t.tmr,
> +								 ctx->tintv);
> +				/*
> +				 * ticks shouldn't become zero at this point.
> +				 * Ignore if hrtimer_forward_now returns 0
> +				 * due to larger backward time drift.
> +				 */

What? Backward time drift? Can you please explain how this would happen?

> +				if (likely(nooftimeo)) {
> +					ticks += nooftimeo - 1;
> +				}

Pointless brackets.

Thanks,

	tglx