[Qemu-devel] kernel-kqemu and linux

[Qemu-devel] kernel-kqemu and linux

[Kevin F. Quinn]

[-- Attachment #1: Type: text/plain, Size: 1430 bytes --]

Hi.

I'm successfully running Windows 2000 guest on qemu (linux host) with
kernel-kqemu, and the speed is excellent.  However I can't get linux to
run as a guest (still linux host); no matter what kernel or kernel
config I create, it always BUGs at the same point:

 hda: hda1 hda2 hda3
------------[ cut here ]------------
kernel BUG at mm/swap.c:215!
invalid operand: 0000 [#1]
Modules linked in:
CPU:    0
EIP:    0060:[<c0141981>]    Not tainted VLI
EFLAGS: 00010256   (2.6.15-gentoo-r1)
EIP is at release_pages+0x131/0x140
eax: 00000000   ebx: c12f98e0   ecx: c0458c94   edx: c12f98e0
esi: 00000000   edi: 00000000   ebp: 00000001   esp: d7fc1da8
ds: 007b   es: 007b   ss: 0068
Process swapper (pid: 1, threadinfo=d7fc0000 task=d7fe4a10)
Stack: 00000000 00000000 00000000 c136eca8 d7fc1e30 00000040 0000000e
c12fb160
       c0458bc0 00000001 c04f2ac0 c0141bca c04f2ac8 00000001 00000000
d7fc1e28
       00000001 d7fc1e28 00000001 00000001 ffffffff c01419b5 d7fc1e30
00000001
Call TRace:
 [<c0141bca>] __pagevec_lru_add_active+0xaa/0xc0
 [<c01419b5>] __pagevec_release+0x25/0x30
 [<c0142119>] invalidate_mapping_pages+0xf9/0x100
 [<c014213e>] invalidate_inode_pages+0x1e/0x30
 [<c015d658>] kill_bdev+0x19/0x40
 [<c0232db9>] add_disk+0x49/0x60
... (during ide probe)


If kernel-kqemu works with linux 2.6 for anyone, could you email
a .config that works?)

Thanks,
-- 
Kevin F. Quinn

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 191 bytes --]

Re: [Qemu-devel] kernel-kqemu and linux

[Fabrice Bellard]

Try the following patch:

diff -u -w -r1.39 helper2.c
--- helper2.c   4 Dec 2005 18:46:06 -0000       1.39
+++ helper2.c   20 Mar 2006 23:38:51 -0000
@@ -110,6 +110,7 @@
          env->pat = 0x0007040600070406ULL;
          env->cpuid_ext_features = 0;
          env->cpuid_features |= CPUID_FXSR | CPUID_MMX | CPUID_SSE | 
CPUID_SSE2
| CPUID_PAE | CPUID_SEP;
+        env->cpuid_features |= CPUID_APIC; /* TEST */
          env->cpuid_xlevel = 0;
          {
              const char *model_id = "QEMU Virtual CPU version " 
QEMU_VERSION;

If it works then APIC usage will become the default on i386...

Fabrice.

Kevin F. Quinn wrote:
> Hi.
> 
> I'm successfully running Windows 2000 guest on qemu (linux host) with
> kernel-kqemu, and the speed is excellent.  However I can't get linux to
> run as a guest (still linux host); no matter what kernel or kernel
> config I create, it always BUGs at the same point:
> 
>  hda: hda1 hda2 hda3
> ------------[ cut here ]------------
> kernel BUG at mm/swap.c:215!
> invalid operand: 0000 [#1]
> Modules linked in:
> CPU:    0
> EIP:    0060:[<c0141981>]    Not tainted VLI
> EFLAGS: 00010256   (2.6.15-gentoo-r1)
> EIP is at release_pages+0x131/0x140
> eax: 00000000   ebx: c12f98e0   ecx: c0458c94   edx: c12f98e0
> esi: 00000000   edi: 00000000   ebp: 00000001   esp: d7fc1da8
> ds: 007b   es: 007b   ss: 0068
> Process swapper (pid: 1, threadinfo=d7fc0000 task=d7fe4a10)
> Stack: 00000000 00000000 00000000 c136eca8 d7fc1e30 00000040 0000000e
> c12fb160
>        c0458bc0 00000001 c04f2ac0 c0141bca c04f2ac8 00000001 00000000
> d7fc1e28
>        00000001 d7fc1e28 00000001 00000001 ffffffff c01419b5 d7fc1e30
> 00000001
> Call TRace:
>  [<c0141bca>] __pagevec_lru_add_active+0xaa/0xc0
>  [<c01419b5>] __pagevec_release+0x25/0x30
>  [<c0142119>] invalidate_mapping_pages+0xf9/0x100
>  [<c014213e>] invalidate_inode_pages+0x1e/0x30
>  [<c015d658>] kill_bdev+0x19/0x40
>  [<c0232db9>] add_disk+0x49/0x60
> ... (during ide probe)
> 
> 
> If kernel-kqemu works with linux 2.6 for anyone, could you email
> a .config that works?)
> 
> Thanks,
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://lists.nongnu.org/mailman/listinfo/qemu-devel

Re: [Qemu-devel] kernel-kqemu and linux

[Brad Campbell]

Fabrice Bellard wrote:
> Try the following patch:
> 
> diff -u -w -r1.39 helper2.c
> --- helper2.c   4 Dec 2005 18:46:06 -0000       1.39
> +++ helper2.c   20 Mar 2006 23:38:51 -0000
> @@ -110,6 +110,7 @@
>          env->pat = 0x0007040600070406ULL;
>          env->cpuid_ext_features = 0;
>          env->cpuid_features |= CPUID_FXSR | CPUID_MMX | CPUID_SSE | 
> CPUID_SSE2
> | CPUID_PAE | CPUID_SEP;
> +        env->cpuid_features |= CPUID_APIC; /* TEST */
>          env->cpuid_xlevel = 0;
>          {
>              const char *model_id = "QEMU Virtual CPU version " 
> QEMU_VERSION;
> 
> If it works then APIC usage will become the default on i386...
> 

I'm afraid that makes no difference at all here..
PIII-M linux host and both 2.4 compiled for i386 and 2.6 compiled for PII linux guests

Regards,
Brad
-- 
"Human beings, who are almost unique in having the ability
to learn from the experience of others, are also remarkable
for their apparent disinclination to do so." -- Douglas Adams

Re: [Qemu-devel] kernel-kqemu and linux

[Kevin F. Quinn]

[-- Attachment #1: Type: text/plain, Size: 4077 bytes --]

Sorry, still fails at the same place.  It recognises the APIC:

...
Found and enabled local APIC!
mapped APIC to ffffd000 (fee00000)
...

I noticed that before the 'kernel BUG' message I got a warning that
scrolled off the screen; so I halted qemu and captured it piece by
piece:

...
hda: cache flushes not supported
 hda:Badness in blk_remove_plug at block/ll_rw_blk.c:1436
 xx blk_remove_plug+0x69/0x70
 xx ide_do_request+0x3c2/0x3f0
 xx do_ide_request+0x24/0x30
 xx generic_unplug_device+0x10/0x20
 xx block_sync_page+0x3a/0x50
...
 xx kernel_thread_helper+0x5/0xc
 hda1 hda2 hda3
------------[ cut here ]------------
kernel BUG at mm/swap.c:215!
... then as before (I can transcribe the whole trace if you want).


The warning is the following code:

int blk_remove_plug(request_queue_t *q)
{
    WARN_ON(!irqs_disabled())

"Badness in" comes from the WARN_ON macro.  So it appears that linux
expects the irqs to have been disabled, which they are in the normal
emulation but not in the virtualised kernel mode.


Without -kernel-kqemu I don't get the warning (or the BUG):

...
hda: cache flushes not supported
 hda1 hda2 hda3
hdc: ATAPI 4X CD-ROM drive, 512kB Cache

and it boots up & works fine.

Kev.



On Tue, 21 Mar 2006 00:39:15 +0100
Fabrice Bellard <fabrice@bellard.org> wrote:

> Try the following patch:
> 
> diff -u -w -r1.39 helper2.c
> --- helper2.c   4 Dec 2005 18:46:06 -0000       1.39
> +++ helper2.c   20 Mar 2006 23:38:51 -0000
> @@ -110,6 +110,7 @@
>           env->pat = 0x0007040600070406ULL;
>           env->cpuid_ext_features = 0;
>           env->cpuid_features |= CPUID_FXSR | CPUID_MMX | CPUID_SSE | 
> CPUID_SSE2
> | CPUID_PAE | CPUID_SEP;
> +        env->cpuid_features |= CPUID_APIC; /* TEST */
>           env->cpuid_xlevel = 0;
>           {
>               const char *model_id = "QEMU Virtual CPU version " 
> QEMU_VERSION;
> 
> If it works then APIC usage will become the default on i386...
> 
> Fabrice.
> 
> Kevin F. Quinn wrote:
> > Hi.
> > 
> > I'm successfully running Windows 2000 guest on qemu (linux host)
> > with kernel-kqemu, and the speed is excellent.  However I can't get
> > linux to run as a guest (still linux host); no matter what kernel
> > or kernel config I create, it always BUGs at the same point:
> > 
> >  hda: hda1 hda2 hda3
> > ------------[ cut here ]------------
> > kernel BUG at mm/swap.c:215!
> > invalid operand: 0000 [#1]
> > Modules linked in:
> > CPU:    0
> > EIP:    0060:[<c0141981>]    Not tainted VLI
> > EFLAGS: 00010256   (2.6.15-gentoo-r1)
> > EIP is at release_pages+0x131/0x140
> > eax: 00000000   ebx: c12f98e0   ecx: c0458c94   edx: c12f98e0
> > esi: 00000000   edi: 00000000   ebp: 00000001   esp: d7fc1da8
> > ds: 007b   es: 007b   ss: 0068
> > Process swapper (pid: 1, threadinfo=d7fc0000 task=d7fe4a10)
> > Stack: 00000000 00000000 00000000 c136eca8 d7fc1e30 00000040
> > 0000000e c12fb160
> >        c0458bc0 00000001 c04f2ac0 c0141bca c04f2ac8 00000001
> > 00000000 d7fc1e28
> >        00000001 d7fc1e28 00000001 00000001 ffffffff c01419b5
> > d7fc1e30 00000001
> > Call TRace:
> >  [<c0141bca>] __pagevec_lru_add_active+0xaa/0xc0
> >  [<c01419b5>] __pagevec_release+0x25/0x30
> >  [<c0142119>] invalidate_mapping_pages+0xf9/0x100
> >  [<c014213e>] invalidate_inode_pages+0x1e/0x30
> >  [<c015d658>] kill_bdev+0x19/0x40
> >  [<c0232db9>] add_disk+0x49/0x60
> > ... (during ide probe)
> > 
> > 
> > If kernel-kqemu works with linux 2.6 for anyone, could you email
> > a .config that works?)
> > 
> > Thanks,
> > 
> > 
> > ------------------------------------------------------------------------
> > 
> > _______________________________________________
> > Qemu-devel mailing list
> > Qemu-devel@nongnu.org
> > http://lists.nongnu.org/mailman/listinfo/qemu-devel
> 
> 
> 
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://lists.nongnu.org/mailman/listinfo/qemu-devel


-- 
Kevin F. Quinn

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 191 bytes --]

Re: [Qemu-devel] kernel-kqemu and linux

[Brad Campbell]

Brad Campbell wrote:
> Fabrice Bellard wrote:
>> Try the following patch:
>>
>> diff -u -w -r1.39 helper2.c
>> --- helper2.c   4 Dec 2005 18:46:06 -0000       1.39
>> +++ helper2.c   20 Mar 2006 23:38:51 -0000
>> @@ -110,6 +110,7 @@
>>          env->pat = 0x0007040600070406ULL;
>>          env->cpuid_ext_features = 0;
>>          env->cpuid_features |= CPUID_FXSR | CPUID_MMX | CPUID_SSE | 
>> CPUID_SSE2
>> | CPUID_PAE | CPUID_SEP;
>> +        env->cpuid_features |= CPUID_APIC; /* TEST */
>>          env->cpuid_xlevel = 0;
>>          {
>>              const char *model_id = "QEMU Virtual CPU version " 
>> QEMU_VERSION;
>>
>> If it works then APIC usage will become the default on i386...
>>
> 
> I'm afraid that makes no difference at all here..
> PIII-M linux host and both 2.4 compiled for i386 and 2.6 compiled for 
> PII linux guests

More information here..

http://fnarfbargle.dyndns.org:81/qemu-oops/qemu.oops.log
http://fnarfbargle.dyndns.org:81/qemu-oops/qemu.oops.log.2

Neat trick for capturing such groovy stuff..
In one console :-

cat /dev/ptyzf | tee qemu.oops.log

In another console :-

qemu -hda vm-1.img -kernel /tracks/linux-2.6.15.6/arch/i386/boot/bzImage -serial /dev/ttyzf -append 
"console=ttyS0" -m 64 -kernel-kqemu

Regards,
Brad (who just learned about pseudo terminal devices)
-- 
"Human beings, who are almost unique in having the ability
to learn from the experience of others, are also remarkable
for their apparent disinclination to do so." -- Douglas Adams

Re: [Qemu-devel] kernel-kqemu and linux

[Pascal Terjan]

On 3/21/06, Brad Campbell <brad@wasp.net.au> wrote:
> Neat trick for capturing such groovy stuff..
> In one console :-
>
> cat /dev/ptyzf | tee qemu.oops.log
>
> In another console :-
>
> qemu -hda vm-1.img -kernel /tracks/linux-2.6.15.6/arch/i386/boot/bzImage -serial /dev/ttyzf -append
> "console=ttyS0" -m 64 -kernel-kqemu

More simple and does not need any right on /dev/tty*  :
qemu -hda vm-1.img -kernel theimage -append console=ttyS0 -serial
stdio -kernel-kqemu | tee qemu.log

Here is the output with various Mandriva kernels (last one is vanilla
2.6.16 rc4). The failure is different each time but always occurs in
the same area (cache init). Running it several times with the same
kernel will give the same error.

=====

Linux version 2.6.12-12mdk (apatard@n1.mandriva.com) (gcc version
4.0.1 (4.0.1-5mdk for Mandriva Linux release 2006.0)) #1 Fri Sep 9
18:15:22 CEST 2005
BIOS-provided physical RAM map:
 BIOS-e820: 0000000000000000 - 000000000009fc00 (usable)
 BIOS-e820: 0000000000100000 - 0000000010000000 (usable)
0MB HIGHMEM available.
256MB LOWMEM available.
DMI not present.
ACPI: Unable to locate RSDP
Allocating PCI resources starting at 10000000 (gap: 10000000:f0000000)
Built 1 zonelists
Local APIC disabled by BIOS -- you can enable it with "lapic"
Initializing CPU#0
Kernel command line: console=ttyS0
PID hash table entries: 2048 (order: 11, 32768 bytes)
Detected 407.111 MHz processor.
Using tsc for high-res timesource
Console: colour VGA+ 80x25
Dentry cache hash table entries: 65536 (order: 6, 262144 bytes)
Inode-cache hash table entries: 32768 (order: 5, 131072 bytes)
Memory: 255420k/262144k available (2348k kernel code, 6232k reserved,
717k data, 268k init, 0k highmem, 0k BadRAM)
Checking if this processor honours the WP bit even in supervisor mode... Ok.
Kernel panic - not syncing: pgtable_cache_init(): Cannot create pgd cache
 <4>Losing too many ticks!
TSC cannot be used as a timesource.
Possible reasons for this are:
  You're running with Speedstep,
  You don't have DMA enabled for your hard disk (see hdparm),
  Incorrect TSC synchronization on an SMP system (see dmesg).
Falling back to a sane timesource now.

=====

Linux version 2.6.12-13mdk-i686-up-4GB (apatard@n1.mandriva.com) (gcc
version 4.0.2 (4.0.2-1mdk for Mandriva Linux release 2006.1)) #1 Mon
Nov 21 18:31:00 CET 2005
BIOS-provided physical RAM map:
 BIOS-e820: 0000000000000000 - 000000000009fc00 (usable)
 BIOS-e820: 0000000000100000 - 0000000010000000 (usable)
0MB HIGHMEM available.
256MB LOWMEM available.
DMI not present.
ACPI: Unable to locate RSDP
Allocating PCI resources starting at 10000000 (gap: 10000000:f0000000)
Built 1 zonelists
Local APIC disabled by BIOS -- you can enable it with "lapic"
Initializing CPU#0
Kernel command line: console=ttyS0
PID hash table entries: 2048 (order: 11, 32768 bytes)
Detected 433.098 MHz processor.
Using tsc for high-res timesource
Console: colour VGA+ 80x25
Dentry cache hash table entries: 65536 (order: 6, 262144 bytes)
Inode-cache hash table entries: 32768 (order: 5, 131072 bytes)
Memory: 255432k/262144k available (2335k kernel code, 6220k reserved,
714k data, 264k init, 0k highmem, 0k BadRAM)
Checking if this processor honours the WP bit even in supervisor mode... Ok.
Kernel panic - not syncing: kmem_cache_create(): failed to create slab
`anon_vma'

 <4>Losing too many ticks!
TSC cannot be used as a timesource.
Possible reasons for this are:
  You're running with Speedstep,
  You don't have DMA enabled for your hard disk (see hdparm),
  Incorrect TSC synchronization on an SMP system (see dmesg).
Falling back to a sane timesource now.

=====

Linux version 2.6.14-1mdk-i686-up-4GB (sbellabes@n3.mandriva.com) (gcc
version 4.0.2 (4.0.2-1mdk for Mandriva Linux release 2006.1)) #1 Wed
Mar 8 20:09:39 CET 2006
BIOS-provided physical RAM map:
 BIOS-e820: 0000000000000000 - 000000000009fc00 (usable)
 BIOS-e820: 0000000000100000 - 0000000010000000 (usable)
0MB HIGHMEM available.
256MB LOWMEM available.
DMI not present.
ACPI: Unable to locate RSDP
Allocating PCI resources starting at 20000000 (gap: 10000000:f0000000)
Built 1 zonelists
Local APIC disabled by BIOS -- you can enable it with "lapic"
Initializing CPU#0
Kernel command line: console=ttyS0
PID hash table entries: 2048 (order: 11, 32768 bytes)
Detected 372.477 MHz processor.
Using tsc for high-res timesource
Console: colour VGA+ 80x25
Dentry cache hash table entries: 65536 (order: 6, 262144 bytes)
Inode-cache hash table entries: 32768 (order: 5, 131072 bytes)
Memory: 254792k/262144k available (2305k kernel code, 6208k reserved,
713k data, 244k init, 0k highmem, 0k BadRAM)
Checking if this processor honours the WP bit even in supervisor mode... Ok.
Kernel panic - not syncing: kmem_cache_create(): failed to create slab `size-32'

=====

Linux version 2.6.16-rc4.11mdk (lcapitulino@n1.mandriva.com) (gcc
version 4.0.3 (4.0.3-0.20060215.1mdk for Mandriva Linux release
2006.1)) #1 Sat Feb 18 23:35:38 CET 2006
BIOS-provided physical RAM map:
 BIOS-e820: 0000000000000000 - 000000000009fc00 (usable)
 BIOS-e820: 0000000000100000 - 0000000010000000 (usable)
0MB HIGHMEM available.
256MB LOWMEM available.
DMI not present or invalid.
ACPI: Unable to locate RSDP
Allocating PCI resources starting at 20000000 (gap: 10000000:f0000000)
Built 1 zonelists
Kernel command line: console=ttyS0
Local APIC disabled by BIOS -- you can enable it with "lapic"
Enabling fast FPU save and restore... done.
Enabling unmasked SIMD FPU exception support... done.
Initializing CPU#0
CPU 0 irqstacks, hard=c0386000 soft=c0387000
PID hash table entries: 2048 (order: 11, 32768 bytes)
Detected 429.499 MHz processor.
Using tsc for high-res timesource
Console: colour VGA+ 80x25
Dentry cache hash table entries: 65536 (order: 6, 262144 bytes)
Inode-cache hash table entries: 32768 (order: 5, 131072 bytes)
Memory: 256228k/262144k available (1585k kernel code, 5448k reserved,
775k data, 196k init, 0k highmem)
Checking if this processor honours the WP bit even in supervisor mode... Ok.
------------[ cut here ]------------
kernel BUG at mm/slab.c:1998!
invalid opcode: 0000 [#1]
Modules linked in:
CPU:    0
EIP:    0060:[<c01468b4>]    Not tainted VLI
EFLAGS: 00010002   (2.6.16-rc4.11mdk #1)
EIP is at check_irq_off+0xa/0x14
eax: 00000046   ebx: c127f120   ecx: c0147da2   edx: 00000000
esi: 000000d0   edi: 00000246   ebp: c034ff70   esp: c034ff70
ds: 007b   es: 007b   ss: 0068
Process swapper (pid: 0, threadinfo=c034f000 task=c02d9200)
Stack: <0>c034ff88 c0147949 c0147da2 00000008 00000004 00000008
c034ff98 c01479ab
       00000004 c127c140 c034ffc0 c0147da2 c034ffcc c0289f25 c036018e 00000000
       22222222 00000008 c127c140 c127e5fc c034ffd4 c0148071 00000000 c127c140
Call Trace:
 [<c0103efb>] show_stack_log_lvl+0xaa/0xb5
 [<c010401c>] show_registers+0x116/0x17c
 [<c01042d6>] die+0x129/0x1a8
 [<c028a72d>] do_trap+0x7c/0x96
 [<c0104685>] do_invalid_op+0x89/0x93
 [<c0103a4f>] error_code+0x4f/0x60
 [<c0147949>] __kmalloc_track_caller+0x54/0x99
 [<c01479ab>] alloc_arraycache+0x1d/0x38
 [<c0147da2>] do_tune_cpucache+0x1d/0x18a
 [<c0148071>] enable_cpucache+0x46/0x65
 [<c036019d>] kmem_cache_init+0x25e/0x28f
 [<c03502bf>] start_kernel+0x10a/0x258
 [<c0100199>] 0xc0100199
Code: c0 e8 1f 09 fd ff 46 58 43 5a 39 fe 7c e9 68 f5 d2 2b c0 e8 0d
09 fd ff 5e 8d 65 f4 5b 5e 5f 5d c3 55 89 e5 9c 58 f6 c4 02 74 08 <0f>
0b ce 07 ad 53 2a c0 5d c3 55 89 e5 9c 58 f6 c4 02 75 08 0f
 <0>Kernel panic - not syncing: Attempted to kill the idle task!

=====