Prev: CFS scheduler, -v8
Next: [patch 06/10] Immediate Value - i386 Optimization
From: Davide Libenzi on 31 May 2007 16:30
On Thu, 31 May 2007, Ulrich Drepper wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Davide Libenzi wrote:
> > What I meant, was that the vast majority of MT+exec
> > apps wants all their fds (but an handfull, maybe)
>
> You never listen when you are told something. Such a statement cannot
> be made. The application doesn't know what the runtime libraries need
> (not just libc, libstdc++, libxml, whatever) and vice versa. Policies
> are always wrong for somebody and changing the standardized behavior is
> not acceptable.

I do always listen when you make a good point. And this is a good one.
Some libraries might not like a global behaviour change, I agree.



- Davide


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Andrew Morton on 31 May 2007 19:30
On Thu, 31 May 2007 14:09:15 -0400
Ulrich Drepper <drepper(a)redhat.com> wrote:

> I've brought this topic up before but didn't provide a patch. Well, here
> we go again, this time with a patch. I even throw in a test program.
>
> The problem is as follows: in multi-threaded code (or more correctly: all
> code using clone() with CLONE_FILES) we have a race when exec'ing.
>
> thread #1 thread #2
>
> fd=open()
>
> fork + exec
>
> fcntl(fd,F_SETFD,FD_CLOEXEC)
>
> In some applications this can happen frequently. Take a web browser. One
> thread opens a file and another thread starts, say, an external PDF viewer.
> The result can even be a security issue if that open file descriptor refers
> to a sensitive file and the external program can somehow be tricked into
> using that descriptor.
>
> Just adding O_CLOEXEC support to open() doesn't solve the whole set of
> problems. There are other ways to create file descriptors (socket,
> epoll_create, Unix domain socket transfer, etc). These can and should
> be addressed separately though. open() is such an easy case that it makes
> not much sense putting the fix off.
>
> ...
>
> diff --git a/include/asm-generic/fcntl.h b/include/asm-generic/fcntl.h
> index c154b9d..b847741 100644
> --- a/include/asm-generic/fcntl.h
> +++ b/include/asm-generic/fcntl.h
> @@ -48,6 +48,9 @@
> #ifndef O_NOATIME
> #define O_NOATIME 01000000
> #endif
> +#ifndef O_CLOEXEC
> +#define O_CLOEXEC 02000000 /* set close_on_exec */
> +#endif
> #ifndef O_NDELAY
> #define O_NDELAY O_NONBLOCK
> #endif

This will break xtensa, because that architecture (and only that
architecture) doesn't include asm-generic/fcntl.h from asm/fcntl.h.

But let's leave this patch as-is: it's xtensa which needs fixing.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Stephen Rothwell on 31 May 2007 21:10
On Thu, 31 May 2007 16:20:21 -0700 Andrew Morton <akpm(a)linux-foundation.org> wrote:
>
> This will break xtensa, because that architecture (and only that
> architecture) doesn't include asm-generic/fcntl.h from asm/fcntl.h.

It used to (for a whole 15 months). Consolidation and cleanup is hard
enough to do once ...

> But let's leave this patch as-is: it's xtensa which needs fixing.

See my other patch ...

--
Cheers,
Stephen Rothwell sfr(a)canb.auug.org.au
http://www.canb.auug.org.au/~sfr/
From: Stephen Rothwell on 31 May 2007 21:40
On Thu, 31 May 2007 16:20:21 -0700 Andrew Morton <akpm(a)linux-foundation.org> wrote:
>
> > diff --git a/include/asm-generic/fcntl.h b/include/asm-generic/fcntl.h
> > index c154b9d..b847741 100644
> > --- a/include/asm-generic/fcntl.h
> > +++ b/include/asm-generic/fcntl.h
> > @@ -48,6 +48,9 @@
> > #ifndef O_NOATIME
> > #define O_NOATIME 01000000
> > #endif
> > +#ifndef O_CLOEXEC
> > +#define O_CLOEXEC 02000000 /* set close_on_exec */
> > +#endif
> > #ifndef O_NDELAY
> > #define O_NDELAY O_NONBLOCK
> > #endif
>
> This will break xtensa, because that architecture (and only that
> architecture) doesn't include asm-generic/fcntl.h from asm/fcntl.h.

This also breaks Alpha (which uses 02000000 for O_DIRECT) and parisc
(which uses 02000000 for O_RSYNC). So you ether need to choose a
different value or define O_CLOEXEC for those two architectures.

--
Cheers,
Stephen Rothwell sfr(a)canb.auug.org.au
http://www.canb.auug.org.au/~sfr/
From: Kyle McMartin on 31 May 2007 23:10
On Fri, Jun 01, 2007 at 11:38:40AM +1000, Stephen Rothwell wrote:
> This also breaks Alpha (which uses 02000000 for O_DIRECT) and parisc
> (which uses 02000000 for O_RSYNC). So you ether need to choose a
> different value or define O_CLOEXEC for those two architectures.
>

That's easy enough to fix...

Signed-off-by: Kyle McMartin <kyle(a)parisc-linux.org>

diff --git a/include/asm-parisc/fcntl.h b/include/asm-parisc/fcntl.h
index 317851f..4ca0fb0 100644
--- a/include/asm-parisc/fcntl.h
+++ b/include/asm-parisc/fcntl.h
@@ -14,6 +14,7 @@
#define O_DSYNC 01000000 /* HPUX only */
#define O_RSYNC 02000000 /* HPUX only */
#define O_NOATIME 04000000
+#define O_CLOEXEC 08000000 /* set close_on_exec */

#define O_DIRECTORY 00010000 /* must be a directory */
#define O_NOFOLLOW 00000200 /* don't follow links */
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
First  |  Prev  |  Next  |  Last
Pages: 1 2 3 4
Prev: CFS scheduler, -v8
Next: [patch 06/10] Immediate Value - i386 Optimization