2.4.19-pre4-ac4 kills my gdm

2.4.19-pre4-ac4 kills my gdm

[Patrick R. McManus]

This is kind of odd - I have an Athlon 850 with a sis 730 chipset. X
detects the "sis 630" as my video chipset.

under 2.4.19-pre4-ac3 all is good - reboot to 2.4.19-pre4-ac4 and gdm
won't start.

X does start successfully.. but not gdm. I can go to runlevel 3 and
run startx without a problem (i.e. get a window manager, etc..)

If I boot back to 2.4.19-pre4-ac3 all is well again.

/var/log/gdm/:0.log complains of a lock

-----
Fatal server error:
Server is already active for display 0
        If this server is no longer running, remove /tmp/.X0-lock
        and start again.


When reporting a problem related to a server crash, please send
the full server output, not just the last messages.
Please report problems to xfree86@xfree86.org.
-----

However clearing the lock (and making sure its gone when the sever
starts) is of no help. 

-Patrick

Re: 2.4.19-pre4-ac4 kills my gdm

[Alan Cox]

> X does start successfully.. but not gdm. I can go to runlevel 3 and
> run startx without a problem (i.e. get a window manager, etc..)
> 
> If I boot back to 2.4.19-pre4-ac3 all is well again.
> 

Can you do a clean build with pre4-ac4 for non athlon cpu, try that, then
a clean built back to with athlon cpu just to verify that is the actual
issue ?

Re: 2.4.19-pre4-ac4 kills my gdm

[Patrick R. McManus]

Alan,

mea culpa on this one - my problem appears to be with the only other
patch I was running - I believed I was also running it on ac3, but it
now appears that I was thinking of another box.

The other patch was corey minyard's "allow signal handler to not call
handler" patch that I was interested in seeing its impact on a
userspace project of mine. It kills gdm (at least with ac4.. maybe
others?)

thanks,
-Pat

[Alan Cox: Thu, Apr 04, 2002 at 04:16:24PM +0100]
> > X does start successfully.. but not gdm. I can go to runlevel 3 and
> > run startx without a problem (i.e. get a window manager, etc..)
> > 
> > If I boot back to 2.4.19-pre4-ac3 all is well again.
> > 
> 
> Can you do a clean build with pre4-ac4 for non athlon cpu, try that, then
> a clean built back to with athlon cpu just to verify that is the actual
> issue ?

Re: 2.4.19-pre4-ac4 kills my gdm

[Corey Minyard]

[-- Attachment #1: Type: text/plain, Size: 813 bytes --]

  Patrick R. McManus wrote:

>Alan,
>
>mea culpa on this one - my problem appears to be with the only other
>patch I was running - I believed I was also running it on ac3, but it
>now appears that I was thinking of another box.
>
>The other patch was corey minyard's "allow signal handler to not call
>handler" patch that I was interested in seeing its impact on a
>userspace project of mine. It kills gdm (at least with ac4.. maybe
>others?)
>
>thanks,
>-Pat
>
Yet, it does seem to kill gdm. xdm and kdm seem to work fine.

Ok, I see the problem. I've fixed it, booted and tried gdm, and it works 
fine. The SA_ONESHOT was checked in the wrong place, the handler is set 
in the frame_setup routines, so I have to check and set the handler due 
to the SA_ONESHOT flag after this. The patch is attached.

-Corey



[-- Attachment #2: linux-nocallhndlr.patch --]
[-- Type: text/plain, Size: 1572 bytes --]

--- ./arch/i386/kernel/signal.c.nocallhndlr	Wed Mar 27 10:56:29 2002
+++ ./arch/i386/kernel/signal.c	Thu Apr  4 13:10:30 2002
@@ -558,22 +558,25 @@
 		}
 	}
 
-	/* Set up the stack frame */
-	if (ka->sa.sa_flags & SA_SIGINFO)
-		setup_rt_frame(sig, ka, info, oldset, regs);
-	else
-		setup_frame(sig, ka, oldset, regs);
+	/* Set up the stack frame if we are calling the handler. */
+	if (! (ka->sa.sa_flags & SA_NOCALLHNDLR)) {
+		if (ka->sa.sa_flags & SA_SIGINFO)
+			setup_rt_frame(sig, ka, info, oldset, regs);
+		else
+			setup_frame(sig, ka, oldset, regs);
+
+		if (!(ka->sa.sa_flags & SA_NODEFER)) {
+			spin_lock_irq(&current->sigmask_lock);
+			sigorsets(&current->blocked,&current->blocked,
+				  &ka->sa.sa_mask);
+			sigaddset(&current->blocked,sig);
+			recalc_sigpending(current);
+			spin_unlock_irq(&current->sigmask_lock);
+		}
+	}
 
 	if (ka->sa.sa_flags & SA_ONESHOT)
 		ka->sa.sa_handler = SIG_DFL;
-
-	if (!(ka->sa.sa_flags & SA_NODEFER)) {
-		spin_lock_irq(&current->sigmask_lock);
-		sigorsets(&current->blocked,&current->blocked,&ka->sa.sa_mask);
-		sigaddset(&current->blocked,sig);
-		recalc_sigpending(current);
-		spin_unlock_irq(&current->sigmask_lock);
-	}
 }
 
 /*
--- ./include/asm-i386/signal.h.nocallhndlr	Wed Mar 27 10:56:12 2002
+++ ./include/asm-i386/signal.h	Thu Apr  4 13:02:07 2002
@@ -91,6 +91,7 @@
 #define SA_RESTART	0x10000000
 #define SA_NODEFER	0x40000000
 #define SA_RESETHAND	0x80000000
+#define SA_NOCALLHNDLR	0x00800000 /* Don't really call the handler. */
 
 #define SA_NOMASK	SA_NODEFER
 #define SA_ONESHOT	SA_RESETHAND