Re: [linuxsh-dev] Re: bad alignment fixup

[Philipp R. <pr...@pa...>]

On Thu, Nov 23, 2000 at 01:11:20PM +0000, David Woodhouse wrote:
> 
> pr...@pa... said:
> >  Write portable code and it won't.
> 
> insmod(8) was what triggered us to do it.
> 
> 
> pr...@pa... said:
> >  If you don't see your process's exit status your test environment is
> > broken. if the process doesn't die with SEGV there's another kernel
> > bug.
> 
> You don't see the exit status of insmod when request_module() fails, do you?

exec_usermodehelper is broken.  not broken as in "has bugs", but broken as in
"has a fundamentlly broken design".  this isn't exactly news.  nevertheless,
there's an obvious fix (not sure whether WIF* are defined in kernelspace, but
this shouldn't _be_ in kernelspace in the first place)

@@ -167,6 +167,7 @@
 {
 	int pid;
 	int waitpid_result;
+	int waitpid_status;
 	sigset_t tmpsig;
 	int i;
 	static atomic_t kmod_concurrent = ATOMIC_INIT(0);
@@ -215,7 +216,7 @@
 	recalc_sigpending(current);
 	spin_unlock_irq(&current->sigmask_lock);
 
-	waitpid_result = waitpid(pid, NULL, __WCLONE);
+	waitpid_result = waitpid(pid, &waitpid_status, __WCLONE);
 	atomic_dec(&kmod_concurrent);
 
 	/* Allow signals again.. */
@@ -228,6 +229,12 @@
 		printk(KERN_ERR "request_module[%s]: waitpid(%d,...) failed, errno %d\n",
 		       module_name, pid, -waitpid_result);
 	}
+
+	if (!WIFSIGNALED(waitpid_status)) {
+		printk(KERN_ERR "request_module[%s]: modprobe exited with signal %d\n",
+		       module_name,  WTERMSIG(waitpid_status));
+	}
+
 	return 0;
 }
 #endif /* CONFIG_KMOD */