Crash on BenH's latest rsync (2.4.6-pre3)

Crash on BenH's latest rsync (2.4.6-pre3)

[Michel Lanners]

Hi all,

While booting BanH's latest rsync tree, I get the following crash:
(from memory, with what I noted down as xmon's printout)

Transfer Error Ack,
vector 200 at pc c00067b8 lr=c0005d58 msr 41000 sp c01e1420

Here's where pc and lr point to:

c00067a8 <__sti_end>:
c00067a8:       7c 00 00 a6     mfmsr   r0
c00067ac:       7c 00 18 78     andc    r0,r0,r3
c00067b0:       7c 00 23 78     or      r0,r0,r4
c00067b4:       7c 00 01 24     mtmsr   r0
c00067b8:       4c 00 01 2c     isync
c00067bc:       4e 80 00 20     blr

c0005cf8 <power_save>:
c0005cf8:       94 21 ff f0     stwu    r1,-16(r1)
c0005cfc:       7c 08 02 a6     mflr    r0
c0005d00:       90 01 00 14     stw     r0,20(r1)
c0005d04:       38 60 00 00     li      r3,0
c0005d08:       60 63 80 00     ori     r3,r3,32768
c0005d0c:       38 80 00 00     li      r4,0
c0005d10:       48 00 0a 99     bl      c00067a8 <__sti_end>
c0005d14:       80 02 00 14     lwz     r0,20(r2)
c0005d18:       2c 00 00 00     cmpwi   r0,0
c0005d1c:       40 82 00 3c     bne     c0005d58 <power_save+0x60>
c0005d20:       3d 20 c0 20     lis     r9,-16352
c0005d24:       7d 70 fa a6     mfdbsr  r11
c0005d28:       80 09 a1 f4     lwz     r0,-24076(r9)
c0005d2c:       55 6b 02 ce     rlwinm  r11,r11,0,11,7
c0005d30:       2c 00 00 00     cmpwi   r0,0
c0005d34:       41 82 00 0c     beq     c0005d40 <power_save+0x48>
c0005d38:       65 6b 00 50     oris    r11,r11,80
c0005d3c:       48 00 00 08     b       c0005d44 <power_save+0x4c>
c0005d40:       65 6b 00 90     oris    r11,r11,144
c0005d44:       7d 70 fb a6     mtdbsr  r11
c0005d48:       3c 80 00 04     lis     r4,4
c0005d4c:       38 60 00 00     li      r3,0
c0005d50:       60 84 80 00     ori     r4,r4,32768
c0005d54:       48 00 0a 55     bl      c00067a8 <__sti_end>
c0005d58:       38 80 00 00     li      r4,0
c0005d5c:       38 60 00 00     li      r3,0
c0005d60:       60 84 80 00     ori     r4,r4,32768
c0005d64:       48 00 0a 45     bl      c00067a8 <__sti_end>
c0005d68:       80 01 00 14     lwz     r0,20(r1)
c0005d6c:       7c 08 03 a6     mtlr    r0
c0005d70:       38 21 00 10     addi    r1,r1,16
c0005d74:       4e 80 00 20     blr

The strange part is that it crashes at varying points shortly after
bootup (when starting daemons; later when X is up, ...).

I was running 2.4.5-pre3 before (and now to compose this ;-); which is
rock solid for me.

Anyone seen this? Anyone want me to look further into debugging this?

Ah yes, I forget, this is on a TiBook (running Debian testing; but I
don't think that the distribution makes any difference...)

Cheers

Michel

-------------------------------------------------------------------------
Michel Lanners                 |  " Read Philosophy.  Study Art.
23, Rue Paul Henkes            |    Ask Questions.  Make Mistakes.
L-1710 Luxembourg              |
email   mlan@cpu.lu            |
http://www.cpu.lu/~mlan        |                     Learn Always. "


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Crash on BenH's latest rsync (2.4.6-pre3)

[Tom Rini]

On Mon, Jun 25, 2001 at 10:29:13PM +0200, Michel Lanners wrote:

> Anyone seen this? Anyone want me to look further into debugging this?

Not this one in particular, but I'm tracking 2 different oopses w/
the tree on a new iBook right now.  2.4.5-pre3 is rock solid too :)

How did you get that oops?  There seems to be a problem w/ inserting
power to a sleeping 'book.  Or at least w/ this one.

--
Tom Rini (TR1265)
http://gate.crashing.org/~trini/

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Crash on BenH's latest rsync (2.4.6-pre3)

[Michel Lanners]

Hi all,

On  26 Jun, this message from Tom Rini echoed through cyberspace:
> On Mon, Jun 25, 2001 at 10:29:13PM +0200, Michel Lanners wrote:
>
>> Anyone seen this? Anyone want me to look further into debugging this?
>
> Not this one in particular, but I'm tracking 2 different oopses w/
> the tree on a new iBook right now.  2.4.5-pre3 is rock solid too :)

;-)

> How did you get that oops?  There seems to be a problem w/ inserting
> power to a sleeping 'book.  Or at least w/ this one.

I was only booting the book; doing nothing particular. It wasn't stable
enough be to be up long enough to go to sleep....

Ben suggested trying without the USB OHCI driver, but no dice; same
crash.

Michel

-------------------------------------------------------------------------
Michel Lanners                 |  " Read Philosophy.  Study Art.
23, Rue Paul Henkes            |    Ask Questions.  Make Mistakes.
L-1710 Luxembourg              |
email   mlan@cpu.lu            |
http://www.cpu.lu/~mlan        |                     Learn Always. "


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Crash on BenH's latest rsync (2.4.6-pre3)

[Benjamin Herrenschmidt]

>Not this one in particular, but I'm tracking 2 different oopses w/
>the tree on a new iBook right now.  2.4.5-pre3 is rock solid too :)
>
>How did you get that oops?  There seems to be a problem w/ inserting
>power to a sleeping 'book.  Or at least w/ this one.

The biggest difference between 2.4.5-pre3 and my current tree is
(except from linux maintstream changes) the new MM layer and
merging of all the linuxppc_2_4_devel work.

It would be helpful for me if you could both try out the current
_2_4_devel tree and tell me if it exhibit the same symtoms. Well,
I beleive sleep won't work on the iBook, so that cannot be tested,
but at least Michel can try it out and tell me.

I currently fail to see what in my recent changes could explain
such weird/random crashes. As far as the iBook is concerned, it
may well be yet another HW issue with a device playing mad
tricks to memory, but I still doubt it.


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Crash on BenH's latest rsync (2.4.6-pre3)

[Takashi Oe]

Hi,

BootX support on your rsync tree and linuxppc_2_4_devel tree is
broken.  I guess no one with BootX bothered to try
linuxppc_2_4_devel..


Takashi Oe

--- linux-2.4-benh/arch/ppc/kernel/setup.c.ORIG	Wed Jun 20 17:49:30 2001
+++ linux-2.4-benh/arch/ppc/kernel/setup.c	Tue Jun 26 16:24:06 2001
@@ -480,6 +480,11 @@
 	if ( ppc_md.progress ) ppc_md.progress("id mach(): start", 0x100);

 #ifdef CONFIG_ALL_PPC
+	/* prom_init has already been called from __start */
+	/* relocate_nodes() must be called before intuit_machine_type() */
+	if (boot_infos)
+		relocate_nodes();
+
 	/* if we didn't get any bootinfo telling us what we are... */
 	if ( _machine == 0 )
 	{
@@ -494,10 +499,6 @@

 	if ( have_of )
 	{
-		/* prom_init has already been called from __start */
-		if (boot_infos)
-			relocate_nodes();
-
 		/* _machine must be defined by now, so this is safe. */
 		finish_device_tree();
 		/*

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Crash on BenH's latest rsync (2.4.6-pre3)

[Tom Rini]

On Tue, Jun 26, 2001 at 04:29:13PM -0500, Takashi Oe wrote:

> BootX support on your rsync tree and linuxppc_2_4_devel tree is
> broken.  I guess no one with BootX bothered to try
> linuxppc_2_4_devel..

Does this work as well?
        /* if we didn't get any bootinfo telling us what we are... */
        if ( _machine == 0 )
        {
                /* prep boot loader tells us if we're prep or not */
                if ( *(unsigned long *)(KERNELBASE) == (0xdeadc0de) )
                        _machine = _MACH_prep;
                else {
                        /* This must be called before intuit_machine_type */
                        if (boot_infos)
                                relocate_nodes();
                        /* If we aren't PReP, we can find out if we're Pmac
                         * or CHRP with this. */
                        intuit_machine_type();
                }
        }

?

--
Tom Rini (TR1265)
http://gate.crashing.org/~trini/

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Crash on BenH's latest rsync (2.4.6-pre3)

[Takashi Oe]

On Tue, 26 Jun 2001 15:13:46 -0700, Tom Rini wrote:

> On Tue, Jun 26, 2001 at 04:29:13PM -0500, Takashi Oe wrote:
>
> > BootX support on your rsync tree and linuxppc_2_4_devel tree is
> > broken.  I guess no one with BootX bothered to try
> > linuxppc_2_4_devel..
>
> Does this work as well?

It sure does, though we might as well do:

	if (boot_infos) {
		_machine = _MACH_Pmac;
		relocate_nodes();
	}

or similar as calling intuit_machine_type() is silly on BootX ;)


Takashi Oe

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Crash on BenH's latest rsync (2.4.6-pre3)

[Tom Rini]

On Tue, Jun 26, 2001 at 05:31:23PM -0500, Takashi Oe wrote:
> On Tue, 26 Jun 2001 15:13:46 -0700, Tom Rini wrote:
>
> > On Tue, Jun 26, 2001 at 04:29:13PM -0500, Takashi Oe wrote:
> >
> > > BootX support on your rsync tree and linuxppc_2_4_devel tree is
> > > broken.  I guess no one with BootX bothered to try
> > > linuxppc_2_4_devel..
> >
> > Does this work as well?
>
> It sure does, though we might as well do:
[snip]
> or similar as calling intuit_machine_type() is silly on BootX ;)

True enough.  I just pushed that.

--
Tom Rini (TR1265)
http://gate.crashing.org/~trini/

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Crash on BenH's latest rsync (2.4.6-pre3)

[Stephen R Marenka]

[-- Attachment #1: Type: text/plain, Size: 978 bytes --]

On Tue, Jun 26, 2001 at 05:31:23PM -0500, Takashi Oe wrote:
> 
> On Tue, 26 Jun 2001 15:13:46 -0700, Tom Rini wrote:
> 
> > On Tue, Jun 26, 2001 at 04:29:13PM -0500, Takashi Oe wrote:
> >
> > > BootX support on your rsync tree and linuxppc_2_4_devel tree is
> > > broken.  I guess no one with BootX bothered to try
> > > linuxppc_2_4_devel..
> >
> > Does this work as well?
> 
> It sure does, though we might as well do:
> 
> 	if (boot_infos) {
> 		_machine = _MACH_Pmac;
> 		relocate_nodes();
> 	}
> 
> or similar as calling intuit_machine_type() is silly on BootX ;)

If I understand correctly, this failed for me because this causes 
relocate_nodes to get called again if have_of.

The following does work for me.

 	if (boot_infos) {
 		_machine = _MACH_Pmac;
                if ( !have_of )
 		    relocate_nodes();
 	}

Thanks,

Stephen

-- 
Stephen R. Marenka     If life's not fun, you're not doing it right!
<stephen@marenka.net>

[-- Attachment #2: Type: application/pgp-signature, Size: 0 bytes --]