On Mon, 19 May 2003, Bob Arendt wrote:
> In the latest *RedHat* kernel, I've noticed that there's a "workaround"
> implemented in:
> /usr/src/linux-2.4.20-13.9/kernel/exit.c, routine sys_wait4():
>
> if (current->sighand->action[SIGCHLD-1].sa.sa_handler == SIG_IGN) {
> static unsigned long last_timestamp;
>
> // rate-limit it to 1 per minute:
> if (jiffies - last_timestamp > 60*HZ) {
> last_timestamp = jiffies;
> printk(KERN_INFO "application bug: %s(%d) has SIGCHLD set to SIG_IGN but calls wait().\n", current->comm, current->pid);
> printk(KERN_INFO "(see the NOTES section of 'man 2 wait'). Workaround activated.\n");
> }
> current->sighand->action[SIGCHLD-1].sa.sa_handler = SIG_DFL;
> workaround = 1;
> }
> add_wait_queue(¤t->wait_chldexit,&wait);
>
> If the parent SIGCHLD is set to SIG_IGN, I get this printed in the
> system log. Nifty. I understand why this is a problem, and I see this in
> the RedHat 9 release notes.
>
> But I don't see this change in Linus's 2.4 or 2.5 bitkeeper trees or
> in the latest kernel.org patch sets (thought it might be in one of
> the "ac" patch sets). Is this becoming part of the official kernel?
>
> Is there some way I can browse RedHat kernel changes to see the history
> of code changes? Something like the BitKeeper or SourceForge browsers would
> be helpfull. The RPM changelog for the kernel-source package didn't
> give any hint of this change either.
It'd really be nice if the rpm changelog were kept more in sync with real
changes...
- Panu -
