Re: [PATCH 3/4] Consolidate sys32_select

[Arnd Bergmann <arnd@arndb.de>]

On Sunday 04 April 2004 13:07, William Lee Irwin III wrote:
> I can second that one; I've seen stuff that "passes LTP" but blatantly
> flunks rather simplistic off-the-cuff a priori sanity checks. One
> example that stands out is a recent patch ppl were rather foolishly
> trying to get my endorsement on that pandered to (ir)Rational crap code.

I just found another example when I tried to open a ppp connection
with my patches :(

This needs to be applied on top of the sys32_select consolidation
patch to get the correct return value from select when called with
nonzero timeout.

===== fs/compat.c 1.25 vs edited =====
--- 1.25/fs/compat.c	Sat Apr  3 23:51:59 2004
+++ edited/fs/compat.c	Mon Apr  5 00:48:52 2004
@@ -1377,8 +1377,9 @@
 			usec = timeout % HZ;
 			usec *= (1000000/HZ);
 		}
-		ret = __put_user(sec, &tvp->tv_sec);
-		ret |= __put_user(usec, &tvp->tv_usec);
+		if (put_user(sec, &tvp->tv_sec) ||
+		    put_user(usec, &tvp->tv_usec))
+			ret = -EFAULT;
 	}
 
 	if (ret < 0)