Oops 2.6.13-git6

Oops 2.6.13-git6

[Lukas Hejtmanek]

Hello,

I did some experiments with my cdrom and battery switchnig.

I have modular bay for cdrom or battery. 

I have two ide devices: hda - pata disk, hdb - cdrom. I suspended into ram and
removed cdrom replacing it with battery. I resumed, then I did hotswap -c
0 rescan-ide; hotswap -c probe-ide

dd if=/dev/hdb of=/dev/null bs=512 count=1 results in:
hdb: ATAPI reset complete
hdb: status error: status=0x00 { }
ide: failed opcode was: unknown
hdb: status error: status=0x00 { }
ide: failed opcode was: unknown
hdb: status error: status=0x00 { }
ide: failed opcode was: unknown
hdb: status error: status=0x00 { }
ide: failed opcode was: unknown
hdb: ATAPI reset complete
hdb: status error: status=0x00 { }
ide: failed opcode was: unknown
Unable to handle kernel NULL pointer dereference at virtual address 00000000
 printing eip:
c0160f20
*pde = 00000000
Oops: 0002 [#1]
PREEMPT 
Modules linked in: skge i915 asus_acpi rtc snd_hda_intel snd_hda_codec ehci_hcd 
uhci_hcd
CPU:    0
EIP:    0060:[<c0160f20>]    Not tainted VLI
EFLAGS: 00010282   (2.6.13-git6) 
EIP is at create_empty_buffers+0x30/0xb0
eax: 00000000   ebx: c12d6240   ecx: 00000000   edx: 00000000
esi: 00000000   edi: 00000000   ebp: c1450144   esp: cba8aca0
ds: 007b   es: 007b   ss: 0068
Process dd (pid: 2750, threadinfo=cba8a000 task=d5f145e0)
Stack: c12d6240 00010000 00000001 c12d6258 c12d6240 c0161aee c12d6240 00010000 
       00000000 c1450148 c01deddc 00000001 00000020 00010000 c01df051 c1450148 
       c14500a4 00000000 cba8a000 c12d6240 c1450144 c013dba8 c1450148 00000000 
Call Trace:
 [<c0161aee>] block_read_full_page+0x29e/0x2f0
 [<c01deddc>] radix_tree_node_alloc+0x1c/0x60
 [<c01df051>] radix_tree_insert+0x81/0x130
 [<c013dba8>] add_to_page_cache+0x68/0xc0
 [<c0145811>] read_pages+0xe1/0x160
 [<c0166130>] blkdev_get_block+0x0/0x70
 [<c01430c0>] __alloc_pages+0x300/0x420
 [<c01459e0>] __do_page_cache_readahead+0x150/0x180
 [<c0145b75>] blockable_page_cache_readahead+0x55/0xd0
 [<c0145dd8>] page_cache_readahead+0x118/0x1c0
 [<c013e7b7>] do_generic_mapping_read+0x4d7/0x630
 [<c01771d3>] dput+0x93/0x240
 [<c013ebb4>] __generic_file_aio_read+0x1a4/0x210
 [<c013e910>] file_read_actor+0x0/0x100
 [<c013ed54>] generic_file_read+0xb4/0xe0
 [<c014ecb4>] __handle_mm_fault+0x1b4/0x210
 [<c0131580>] autoremove_wake_function+0x0/0x60
 [<c0116919>] do_page_fault+0x1b9/0x5b1
 [<c015e03c>] vfs_read+0x14c/0x160
 [<c015e331>] sys_read+0x51/0x80
 [<c01031b5>] syscall_call+0x7/0xb
Code: 00 53 83 ec 0c 8b 5c 24 18 89 44 24 08 8b 44 24 1c 8b 74 24 20 89 1c 24 89
 44 24 04 e8 1a f6 ff ff 89 c1 89 c2 8d b6 00 00 00 00 <09> 32 89 d0 8b 52 04 85
 d2 75 f5 89 48 04 b8 00 f0 ff ff 21 e0 


Is there some way how to change battery and cdrom without reboot?

-- 
Lukáš Hejtmánek

Re: Oops 2.6.13-git6

[Alan Cox]

On Llu, 2005-09-12 at 12:20 +0200, Lukas Hejtmanek wrote:
> Is there some way how to change battery and cdrom without reboot?

The 2.6 IDE layer doesn't support IDE drive hot or warm plug. 2.4-ac did
but that was the only one that ever supported it.

Nevertheless it should not have crashed only errored so the bug report
is useful 8)

Re: Oops 2.6.13-git6

[Lukas Hejtmanek]

On Mon, Sep 12, 2005 at 02:11:10PM +0100, Alan Cox wrote:
> The 2.6 IDE layer doesn't support IDE drive hot or warm plug. 2.4-ac did
> but that was the only one that ever supported it.
> 
> Nevertheless it should not have crashed only errored so the bug report
> is useful 8)

However, in the case of 2 IDE channels hotplug program works with unregister and
rescan commands. I think there is still problem with DMA, but it works partly.

Another issue, when cdrecord finalizes DVD or CD, then whole IDE channel is
blocked. Is that hardware problem or some preemption problem in kernel?

-- 
Lukáš Hejtmánek

Re: Oops 2.6.13-git6

[Alan Cox]

On Llu, 2005-09-12 at 20:00 +0200, Lukas Hejtmanek wrote:
> However, in the case of 2 IDE channels hotplug program works with unregister and
> rescan commands. I think there is still problem with DMA, but it works partly.

In a tiny subset of cases and for PIO modes only for devices that match
IDE generic PIO interfaces, providing the interface is not boot
detected, you don't remove it when in use and the moon is in the right
phase.


> Another issue, when cdrecord finalizes DVD or CD, then whole IDE channel is
> blocked. Is that hardware problem or some preemption problem in kernel?

This is an IDE protocol limitation. See the cdrecord manual for other
options if you want to avoid that occurring.

Re: Oops 2.6.13-git6

[Lukas Hejtmanek]

On Mon, Sep 12, 2005 at 11:01:15PM +0100, Alan Cox wrote:
> In a tiny subset of cases and for PIO modes only for devices that match
> IDE generic PIO interfaces, providing the interface is not boot
> detected, you don't remove it when in use and the moon is in the right
> phase.

I'm not sure, if I understand correctly. However, I meant, in the case of hdd as
/dev/hda and cdrom as /dev/hdc, it works:
/usr/bin/hotswap --ide-controller 1 -n probe-ide
/usr/bin/hotswap --ide-controller 1 -n unregister-ide
remove cdrom here

after pluging cdrom back:
/usr/bin/hotswap --ide-controller 1 -n probe-ide
/usr/bin/hotswap --ide-controller 1 -n rescan-ide

I think, DMA cannot be turned on if cdrom was not plugged in at boot time for
some reason.

But things above do not work if hdd is /dev/hda and cdrom is /dev/hdb, i.e. on
the same channel.

Again, is it an IDE conception bug?

-- 
Lukáš Hejtmánek