Messages in this thread | | | Date | Fri, 17 Dec 2010 15:49:47 -0800 | From | Andrew Morton <> | Subject | Re: [PATCH] delay.h: add __must_check to msleep_interruptible |
| |
On Thu, 4 Nov 2010 10:55:41 +0200 Baruch Siach <baruch@tkos.co.il> wrote:
> Code calling msleep_interruptible() must be aware that sleep time might be > shorter than intended as a result of a signal being caught. Code not checking > the return value of msleep_interruptible() is probably buggy, unless it's doing > the signal_pending() check itself, which is redundant. >
True. But there are around 250 callsites which don't check the msleep_interruptible() return value.
I don't think I want to add 250 new warnings to the kernel build - it'll take *years* to get them all weeded out and meanwhile it will cause people to miss other warnings while they're ignoring the msleep_interruptible() warnings.
So. Some lucky duck needs to get down and start fixing all these things first, please. Meanwhile, a checkpatch rule which prevents new occurrences would be good.
Except lots of developers and maintainers can't be assed running checkpatch, so a volunteer who regularly runs linux-next.patch and patch-2.6.Y-rcY through checkpatch and then hassles people would also be good.
| |