Re: [patch] Re: vfork failing - posix compliance fix

Andrea Arcangeli (andrea@e-mind.com)
Thu, 18 Mar 1999 03:25:12 +0100 (CET)

On Wed, 17 Mar 1999, Pavel Machek wrote:

>It seems to me that this patch is required for correct return code
>(according to posix). (It is much better idea than returning
>ENOMEM, because we return ENOMEM for almost anything else.)

Agreed.

>--- clean/kernel/fork.c Wed Mar 10 22:37:42 1999
>+++ linux/kernel/fork.c Tue Mar 16 00:03:14 1999
>@@ -511,8 +533,10 @@
> lock_kernel();
>
> if (p->user) {
>- if (atomic_read(&p->user->count) >= p->rlim[RLIMIT_NPROC].rlim_cur)
>+ if (atomic_read(&p->user->count) >= p->rlim[RLIMIT_NPROC].rlim_cur) {
>+ retval = -EAGAIN;
> goto bad_fork_free;
>+ }
> }

I have a _minor_ remark only about the implementation. I think this my
approch will perform a bit better.

In general I think that the goto has to be the only thing to be done in
the if(true)path to make sure that the fast path will be preserved.

Index: kernel/fork.c
===================================================================
RCS file: /var/cvs/linux/kernel/fork.c,v
retrieving revision 1.1.1.4
retrieving revision 1.1.2.14
diff -u -r1.1.1.4 -r1.1.2.14
--- fork.c 1999/03/09 00:55:51 1.1.1.4
+++ linux/kernel/fork.c 1999/03/17 18:38:29 1.1.2.14
@@ -506,6 +509,7 @@
goto fork_out;

*p = *current;
+ retval = -EAGAIN;

down(&current->mm->mmap_sem);
lock_kernel();
@@ -518,7 +522,6 @@
{
struct task_struct **tslot;
tslot = find_empty_process();
- retval = -EAGAIN;
if (!tslot)
goto bad_fork_free;
p->tarray_ptr = tslot;

Andrea Arcangeli

-
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/