Re: 2.6.0-test9-mm4 (does not boot)

Re: 2.6.0-test9-mm4 (does not boot)

[Frank Dekervel]

hello,

2.6.0-test9-mm4 doesn't boot for me ... oops followed by 
kernel panic - attempted to kill init (2.6.0-test9 works fine). 
it crashes right after initialising PNP  bios. The (undecoded) oops doesn't 
seem to make a lot of  sense (i wrote the oops down and i typed it in), and 
ksymoops doesn't show a lot too. 

Someone has an idea what this could be, or a hint to improve oops 
output ? (i can reproduce it as much as i want to)

------------------- oops output
general protection fault: 0000 [#1]
PREEMPT SMP
CPU: 0
EIP: 0098:[<00002d6c>] Not tainted VLI
EFLAGS: 00010097
EIP is at 0x2d6c
eax: 00003410 ebx: 00000082  ecx: 00020000 edx: 00000002
esi: 00002630 edi: c1a4004d  ebp: c1a40000 esp: c1a47ee2
ds: 0060 es: 0060 ss:0068
Process swapper (PID:1 threadinfo=c1a46000 task=c1a5f980)
Stack: 00000410 341026de 00000000 836d004d 0004cfea 00020002 7f28830c cfeacff2
       64090909 01090109 007b6264 6000007b 00a00246 622000b0 00a861e6 00000086
       000b0000 00010090 00a80000 00b00000 00a00002 bee90000 0060c02b 00820000
Call Trace:

Code:
      bad EIP Value

------------------ ksymoops -V -K -L -o /lib/modules/2.6.0-test9-mm4/ -m /boot/System.map-2.6.0-test9-mm4 < /root/oops.txt

Warning (Oops_read): Code line not seen, dumping what data is available


>>EIP; 00002d6c Before first symbol   <=====

>>edi; c1a4004d <__crc_unregister_chrdev+112bbc/1668b5>
>>ebp; c1a40000 <__crc_unregister_chrdev+112b6f/1668b5>
>>esp; c1a47ee2 <__crc_unregister_chrdev+11aa51/1668b5>

----------------





op Wednesday 19 November 2003 07:52 , schreef Andrew Morton  in <20031118225120.1d213db2.akpm@osdl.org> :

> 
> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test9/2.6.0-test9-mm4/
> 
> 
> . Several fixes against patches which are only in -mm at present.
> 
> . Minor fixes which we'll queue for post-2.6.0.
> 
> . The interactivity problems which the ACPI PM timer patch showed up
>   should be fixed here - please sing out if not.
-- 
Frank Dekervel - frank.dekervel@student.kuleuven.ac.be
Mechelsestraat 88
3000 Leuven (Belgium)

Re: 2.6.0-test9-mm4 (does not boot)

[Andrew Morton]

Frank Dekervel <kervel@drie.kotnet.org> wrote:
>
> 2.6.0-test9-mm4 doesn't boot for me ... oops followed by 
> kernel panic - attempted to kill init (2.6.0-test9 works fine). 
> it crashes right after initialising PNP  bios. 

Please make sure that you have CONFIG_KALLSYMS set.

It would help to add `initcall_debug' to the kernel boot command line. 
That way you will find out the address of the final initcall which was
invoked before the kernel crashed.  Please look that up in System.map.

Thanks.

Re: 2.6.0-test9-mm4 (does not boot)

[Frank Dekervel]

hello,

Op Thursday 20 November 2003 01:59, schreef Andrew Morton:
> > 2.6.0-test9-mm4 doesn't boot for me ... oops followed by
> > kernel panic - attempted to kill init (2.6.0-test9 works fine).
> > it crashes right after initialising PNP  bios.
>
> Please make sure that you have CONFIG_KALLSYMS set.

it is set ... but no decoded output. probably something is messed up
kervel@bakvis:~$ cat /boot/config-2.6.0-test9-mm4 | grep -i kall
CONFIG_KALLSYMS=y

>
> It would help to add `initcall_debug' to the kernel boot command line.
> That way you will find out the address of the final initcall which was
> invoked before the kernel crashed.  Please look that up in System.map.

kervel@bakvis:~$ cat /boot/System.map-2.6.0-test9-mm4 | grep c052ed91
c052ed91 T pnpbios_init

pnpbios says something like this:
 found installation structure 0xc00f5560
 version 1.0 entry 0xf0000:0x6149 dseg 0xf0000

i'm going to try without pnpbios i think.

my working 2.6.0test9 also has pnpbios setup:
kervel@bakvis:~$ cat /boot/config-2.6.0-test9 | grep -i pnpbios
CONFIG_PNPBIOS=y


greetings,
frank

-- 
Frank Dekervel - frank.dekervel@student.kuleuven.ac.be
Mechelsestraat 88
3000 Leuven (Belgium)

Re: 2.6.0-test9-mm4 (does not boot)

[Frank Dekervel]

Op Thursday 20 November 2003 11:34, schreef Frank Dekervel:
> pnpbios says something like this:
>  found installation structure 0xc00f5560
>  version 1.0 entry 0xf0000:0x6149 dseg 0xf0000
>
> i'm going to try without pnpbios i think.
>
> my working 2.6.0test9 also has pnpbios setup:
> kervel@bakvis:~$ cat /boot/config-2.6.0-test9 | grep -i pnpbios
> CONFIG_PNPBIOS=y

ok, replying to myself to be more specific:

working pnpbios gives this:
ACPI: PCI Interrupt Link [LNKH] (IRQs 3 4 5 6 7 10 *11 12 14 15)
Linux Plug and Play Support v0.97 (c) Adam Belay
PnPBIOS: Scanning system for PnP BIOS support...
PnPBIOS: Found PnP BIOS installation structure at 0xc00f5560
PnPBIOS: PnP BIOS version 1.0, entry 0xf0000:0x614a, dseg 0xf0000
PnPBIOS: 15 nodes reported by PnP BIOS; 15 recorded by driver
SCSI subsystem initialized

mm4 pnpbios gives the same numbers, but never says 
PnPBIOS: 15 nodes reported by PnP BIOS; 15 recorded by driver
instead it says general protection fault

sorry,
greetings,
frank


-- 
Frank Dekervel - frank.dekervel@student.kuleuven.ac.be
Mechelsestraat 88
3000 Leuven (Belgium)

Re: 2.6.0-test9-mm4 (does not boot)

[Andrew Morton]

Frank Dekervel <kervel@drie.kotnet.org> wrote:
>
> Op Thursday 20 November 2003 11:34, schreef Frank Dekervel:
> > pnpbios says something like this:
> >  found installation structure 0xc00f5560
> >  version 1.0 entry 0xf0000:0x6149 dseg 0xf0000
> >
> > i'm going to try without pnpbios i think.
> >
> > my working 2.6.0test9 also has pnpbios setup:
> > kervel@bakvis:~$ cat /boot/config-2.6.0-test9 | grep -i pnpbios
> > CONFIG_PNPBIOS=y
> 
> ok, replying to myself to be more specific:
> 
> working pnpbios gives this:
> ACPI: PCI Interrupt Link [LNKH] (IRQs 3 4 5 6 7 10 *11 12 14 15)
> Linux Plug and Play Support v0.97 (c) Adam Belay
> PnPBIOS: Scanning system for PnP BIOS support...
> PnPBIOS: Found PnP BIOS installation structure at 0xc00f5560
> PnPBIOS: PnP BIOS version 1.0, entry 0xf0000:0x614a, dseg 0xf0000
> PnPBIOS: 15 nodes reported by PnP BIOS; 15 recorded by driver
> SCSI subsystem initialized
> 
> mm4 pnpbios gives the same numbers, but never says 
> PnPBIOS: 15 nodes reported by PnP BIOS; 15 recorded by driver
> instead it says general protection fault
> 

There are three pnpbios patches in -mm:

ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test9/2.6.0-test9-mm4/broken-out/pnp-fix-1.patch
ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test9/2.6.0-test9-mm4/broken-out/pnp-fix-2.patch
ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test9/2.6.0-test9-mm4/broken-out/pnp-fix-3.patch

It would help if you could determine which (if any) of these are causing
the problem.  You can remove the patches with

	cd /usr/src/linux
	patch -p1 -R < ~/pnp-fix-3.patch

etcetera.


Thanks.

Re: 2.6.0-test9-mm4 (does not boot)

[Frank Dekervel]

hello,

> mm4 pnpbios gives the same numbers, but never says 
> PnPBIOS: 15 nodes reported by PnP BIOS; 15 recorded by driver
> instead it says general protection fault
> 

> There are three pnpbios patches in -mm:

ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test9/2.6.0-test9-mm4/broken-out/pnp-fix-1.patch

"With this patch, the PnPBIOS driver will read static resources 
 initially and then switch to dynamic mode when allocating 
 resources for specific nodes."

-> this one causes the trouble (general protection fault -> kernel panic) for me... without this patch
    my system boots (and recognises pnpbios) fine.

ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test9/2.6.0-test9-mm4/broken-out/pnp-fix-2.patch

"this patch provides an option for PnPBIOS calls to be 
 managed by the PnPBIOS driver exclusively" (no /proc support) 

-> cannot be this one, since i enabled /proc support

ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test9/2.6.0-test9-mm4/broken-out/pnp-fix-3.patch

-> not pnpbios related (isapnp)

greetings,
frank


-- 
Frank Dekervel - frank.dekervel@student.kuleuven.ac.be
Mechelsestraat 88
3000 Leuven (Belgium)

Re: 2.6.0-test9-mm4 (does not boot)

[Frank Dekervel]

hello

something similar:

catting /proc/bus/pnp/devices makes my system oops, doing it twice makes my 
system crash :p

the oops looks very much like the oops (also bad EIP value, also no stack 
trace) i get on boot with the first patch (below) applied. As i already 
mailed, i need to revert that patch to make my system boot.

this oops happens with all 3 patches below reverted, so i guess it'll happen 
too with stock test9.

would the -mm5 pnp-fix-4.patch be worth a try ? it seems related

thanks,
greetings,
frank


Op Thursday 20 November 2003 16:22, schreef Andrew Morton:
> There are three pnpbios patches in -mm:
>
> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test9/2
>.6.0-test9-mm4/broken-out/pnp-fix-1.patch
> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test9/2
>.6.0-test9-mm4/broken-out/pnp-fix-2.patch
> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test9/2
>.6.0-test9-mm4/broken-out/pnp-fix-3.patch
>
> It would help if you could determine which (if any) of these are causing
> the problem.  You can remove the patches with
>
>         cd /usr/src/linux
>         patch -p1 -R < ~/pnp-fix-3.patch

-- 
Frank Dekervel - frank.dekervel@student.kuleuven.ac.be
Mechelsestraat 88
3000 Leuven (Belgium)

Re: 2.6.0-test9-mm4 (does not boot)

[Adam Belay]

On Sat, Nov 22, 2003 at 06:50:36PM +0100, Frank Dekervel wrote:
> hello
> 
> something similar:
> 
> catting /proc/bus/pnp/devices makes my system oops, doing it twice makes my 
> system crash :p
> 
> the oops looks very much like the oops (also bad EIP value, also no stack 
> trace) i get on boot with the first patch (below) applied. As i already 
> mailed, i need to revert that patch to make my system boot.
> 
> this oops happens with all 3 patches below reverted, so i guess it'll happen 
> too with stock test9.
> 
> would the -mm5 pnp-fix-4.patch be worth a try ? it seems related
>
> thanks,
> greetings,
> frank

Hi,

Thanks for the testing.  I don't think pnp-fix-4.patch should affect this problem.
I will probably be creating a blacklist for PnPBIOS systems that have this bug.
Currently, I'm waiting to see if reading static resources has any positive affects
on some additional systems.  If not then I may switch back to the the original
behavior.  DMI information for your system would be helpful.

Thanks,
Adam

Re: 2.6.0-test9-mm4 (does not boot)

[Frank Dekervel]

Op Monday 24 November 2003 00:05, schreef u:
> > - will the original behaviour really solve the problem  (not only a
> > symptom) ? as i wrote, i can trigger almost the same oops (general
> > protection fault #0000 , invalid EIP value ), probably non-fatal because
> > another process is killed instead of the pid=1 process, and i can trigger
> > it on a mm4 with all pnpbios fixes backed out ...
>
> Yes but through the /proc/bus/pnp/devices file.  Correct?  It is
> independent from this change and would also need to be corrected.  Does the
> escd interface in /proc/bus/pnp also trigger an oops?

yup it seems so, but this time with valid backtrace
bakvis:/proc/bus/pnp# cat escd
Segmentation fault

in dmesg:
Unable to handle kernel paging request at virtual address fffec01a
 printing eip:
000055bf
*pde = 00004067
*pte = 00000000
Oops: 0000 [#1]
PREEMPT SMP
CPU:    1
EIP:    0098:[<000055bf>]    Tainted: PF  VLI
EFLAGS: 00210086
EIP is at 0x55bf
eax: 000001ff   ebx: 00b06341   ecx: 000000a0   edx: 00000000
esi: 0000001a   edi: 00000000   ebp: ce47de9c   esp: ce47de68
ds: 00b0   es: 00a8   ss: 0068
Process cat (pid: 9496, threadinfo=ce47c000 task=e87fe080)
Stack: 000a0002 00b00000 000600a8 5b995598 00000000 635f00a0 007b0033 c000007b
       0206ce47 61e60020 008600a8 00000100 0090000b 00000042 00b000a8 000000a0
       00000000 c02bc9d2 00000060 00200082 00200033 c1210000 0000007b c017007b
Call Trace:
 [<c02bc9d2>] __pnp_bios_read_escd+0x130/0x1bb
 [<c017007b>] flush_old_exec+0x2cd/0xab2
 [<c02bca78>] pnp_bios_read_escd+0x1b/0x40
 [<c02bdcea>] proc_read_escd+0x65/0xf4
 [<c02bdc85>] proc_read_escd+0x0/0xf4
 [<c019d321>] proc_file_read+0xc4/0x26c
 [<c01641b2>] vfs_read+0xb0/0x119
 [<c0164459>] sys_read+0x42/0x63
 [<c03f353b>] syscall_call+0x7/0xb

Code:  Bad EIP value.
 <6>note: cat[9496] exited with preempt_count 2

greetings,
frank

-- 
Frank Dekervel - frank.dekervel@student.kuleuven.ac.be
Mechelsestraat 88
3000 Leuven (Belgium)

Re: 2.6.0-test9-mm4 (does not boot)

[Adam Belay]

On Mon, Nov 24, 2003 at 09:10:29AM +0100, Frank Dekervel wrote:
> Op Monday 24 November 2003 00:05, schreef u:
> > > - will the original behaviour really solve the problem ?(not only a
> > > symptom) ? as i wrote, i can trigger almost the same oops (general
> > > protection fault #0000 , invalid EIP value ), probably non-fatal because
> > > another process is killed instead of the pid=1 process, and i can trigger
> > > it on a mm4 with all pnpbios fixes backed out ...

Is the oops triggered by reading /proc/bus/pnp/devices a general protection fault?
Once in a while it will produce a different error in userspace.  If so, I'd like
to see the output.

> >
> > Yes but through the /proc/bus/pnp/devices file. ?Correct? ?It is
> > independent from this change and would also need to be corrected. ?Does the
> > escd interface in /proc/bus/pnp also trigger an oops?
>
> yup it seems so, but this time with valid backtrace
> bakvis:/proc/bus/pnp# cat escd
> Segmentation fault

Could you please try this patch (without pnp-patch-1)... It may fix the ESCD
reading problem.  If it doesn't oops when catted, you may want to test it out
with lsescd by Gunther Mayer.
(http://home.t-online.de/home/gunther.mayer/lsescd-0.10.tar.bz2)

--- a/drivers/pnp/pnpbios/bioscalls.c	2003-11-26 20:44:47.000000000 +0000
+++ b/drivers/pnp/pnpbios/bioscalls.c	2003-12-02 21:17:42.000000000 +0000
@@ -493,7 +493,7 @@
 	if (!pnp_bios_present())
 		return ESCD_FUNCTION_NOT_SUPPORTED;
 	status = call_pnp_bios(PNP_READ_ESCD, 0, PNP_TS1, PNP_TS2, PNP_DS, 0, 0, 0,
-			       data, 65536, (void *)nvram_base, 65536);
+			       data, 65536, __va((void *)nvram_base), 65536);
 	return status;
 }

@@ -516,7 +516,7 @@
 	if (!pnp_bios_present())
 		return ESCD_FUNCTION_NOT_SUPPORTED;
 	status = call_pnp_bios(PNP_WRITE_ESCD, 0, PNP_TS1, PNP_TS2, PNP_DS, 0, 0, 0,
-			       data, 65536, nvram_base, 65536);
+			       data, 65536, __va((void *)nvram_base), 65536);
 	return status;
 }
 #endif


Also just out of curiosity, could I see what the real mode CS and DS are?  I'm
looking for patterns...

So far all of the offenders have a datasegement at 0xf0000.

--- a/drivers/pnp/pnpbios/core.c	2003-11-26 20:42:52.000000000 +0000
+++ b/drivers/pnp/pnpbios/core.c	2003-12-02 21:59:18.000000000 +0000
@@ -460,6 +460,9 @@
                        check->fields.version >> 4, check->fields.version & 15,
 		       check->fields.pm16cseg, check->fields.pm16offset,
 		       check->fields.pm16dseg);
+		printk(KERN_INFO "PnPBIOS: realmode entry 0x%x:0x%x, dseg 0x%x\n",
+		       check->fields.rmcseg, check->fields.rmoffset,
+		       check->fields.rmdseg);
 		pnp_bios_install = check;
 		return 1;
 	}

Thanks,
Adam