sys_poll with timeout -1 bug fix

sys_poll with timeout -1 bug fix

[frode isaksen]

If you do a poll() call with timeout -1, the wait will be a big  
number (depending on HZ)  instead of infinite wait, since -1 is  
passed to the msecs_to_jiffies function.
Tested on i386 and MIPS with latest (2.6.16) kernel.

Frode


--- linux-2.6.16.20/fs/select.c.orig.c	2006-06-05 19:18:23.000000000  
+0200
+++ linux-2.6.16.20/fs/select.c	2006-06-15 14:20:47.000000000 +0200
@@ -699,9 +699,9 @@ out_fds:
asmlinkage long sys_poll(struct pollfd __user *ufds, unsigned int nfds,
			long timeout_msecs)
{
-	s64 timeout_jiffies = 0;
+	s64 timeout_jiffies;
-	if (timeout_msecs) {
+	if (timeout_msecs > 0) {
#if HZ > 1000
		/* We can only overflow if HZ > 1000 */
		if (timeout_msecs / 1000 > (s64)0x7fffffffffffffffULL / (s64)HZ)
@@ -709,6 +709,9 @@ asmlinkage long sys_poll(struct pollfd _
		else
#endif
			timeout_jiffies = msecs_to_jiffies(timeout_msecs);
+	} else {
+		/* Infinite (-1) or no (0) timeout */
+		timeout_jiffies = timeout_msecs;
	}
	return do_sys_poll(ufds, nfds, &timeout_jiffies);

Re: sys_poll with timeout -1 bug fix

[Nishanth Aravamudan]

On 15.06.2006 [15:05:00 +0200], frode isaksen wrote:
> If you do a poll() call with timeout -1, the wait will be a big  
> number (depending on HZ)  instead of infinite wait, since -1 is  
> passed to the msecs_to_jiffies function.
> Tested on i386 and MIPS with latest (2.6.16) kernel.

This looks very close to a fix I have on my machine, as well, after
looking at sys_poll() a bit. I suppose the only times you would notice
are if an application actually wants as long a wait as possible, as
opposed to a very long one.

Your patch was line-wrapped and does not have a Signed-off-by line,
however. Care to try again?

Thanks,
Nish

-- 
Nishanth Aravamudan <nacc@us.ibm.com>
IBM Linux Technology Center