Hi,
Here is the whole story for those who prefer just the summary instead of
the whole thread.
1. Patrick Mau noticed a problem caused by the bug in the patch I sumitted
for the poll() change in 2.3.41
2. Manfred Spraul noticed that this problem could be caused by userspace
passing nfds=0.
3. I now submit the fix to this problem. Instead of naively (like I have
done in the 0th iteration of the patch) calling schedule_timeout() (or
sys_nanosleep()) in a special case of nfds=0, we just let the code fall
through (like it happens for select and like it used to happen for poll
before my change) only checking that we don't leak memory and don't pass
garbage to kmalloc. This way we correctly get -EINTR if interrupted and
0 if not interrupted, like on other UNIXen (e.g. UnixWare7)
A copy of the patch is on:
http://www.ocston.org/~tigran/patches/pollfix4.patch
Patrick, please let me know if your squid is now happy.
Regards,
------
Tigran A. Aivazian | http://www.sco.com
Escalations Research Group | tel: +44-(0)1923-813796
Santa Cruz Operation Ltd | http://www.ocston.org/~tigran
--- select.c.0 Mon Jan 31 08:48:52 2000
+++ select.c Mon Jan 31 09:39:04 2000
@@ -414,11 +414,13 @@
wait = wait_table;
}
- fds = (struct pollfd **)kmalloc(
- (1 + (nfds - 1) / POLLFD_PER_PAGE) * sizeof(struct pollfd *),
- GFP_KERNEL);
- if (fds == NULL)
- goto out;
+ if (nfds != 0) {
+ fds = (struct pollfd **)kmalloc(
+ (1 + (nfds - 1) / POLLFD_PER_PAGE) * sizeof(struct pollfd *),
+ GFP_KERNEL);
+ if (fds == NULL)
+ goto out;
+ }
nchunks = 0;
nleft = nfds;
@@ -467,7 +469,8 @@
out_fds:
for (i=0; i < nchunks; i++)
free_page((unsigned long)(fds[i]));
- kfree(fds);
+ if (nfds != 0)
+ kfree(fds);
out:
if (wait)
free_wait(wait_table);
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Mon Jan 31 2000 - 21:00:27 EST