Skip to contentSkip to conversation

libata sata_sil error handling

DISCUSSION Arnd Hannemann 2 participants 2 messages Apr 22, 2007 – Apr 23, 2007 libata

lore
Share & tools
lore

Arnd Hannemann

author
#0 lore raw

Hi,

tried to use my SiI 3512 card in the 2nd slot of a riser card on my mobo. Probably the BIOS did not assign a correct IRQ to the card, so I got this error message when loading the sata_sil module on my 2.6.20.7 kernel:

Quoted 6 lines

libata version 2.00 loaded.
sata_sil 0000:00:0b.0: version 2.0
sata_sil 0000:00:0b.0: Applying R_ERR on DMA activate FIS errata fix
ata1: SATA max UDMA/100 cmd 0xEF980080 ctl 0xEF98008A bmdma 0xEF980000 irq 255
ata2: SATA max UDMA/100 cmd 0xEF9800C0 ctl 0xEF9800CA bmdma 0xEF980008 irq 255 sata_sil 0000:00:0b.0: irq 255 request failed: -22

Of course I don't blame libata for above, because the error seems to be with my BIOS or riser card. But now when I tried to rmmod sata_sil, I got these:

Quoted 24 lines
BUG: unable to handle kernel NULL pointer dereference at virtual address 00000010
 printing eip:
ef9c8b05
*pde = 00000000
Oops: 0000 [#1]
Modules linked in: af_packet tun xt_state ipt_MASQUERADE ipt_REJECT ipt_recent ipt_LOG iptable_mangle iptax
CPU:    0
EIP:    0060:[<ef9c8b05>]    Not tainted VLI
EFLAGS: 00010282   (2.6.20.7 #2)

EIP is at ata_host_remove+0x5/0xd0 [libata]
eax: 00000000 ebx: b18d9400 ecx: b18d9448 edx: ef9cce20
esi: 00000000 edi: ef97bbe4 ebp: e3a51000 esp: e3a51ef4
ds: 007b es: 007b ss: 0068
Process rmmod (pid: 4110, ti=e3a51000 task=e7444550 task.ti=e3a51000)

Stack: b18d9400 b18d9448 ef97bbe4 ef9cce3a b18d9400 ef97bb94 b01dc036 b18d9448
       b02262b4 b18d9448 ef97bb94 b02267e5 ef97bb94 ef97bb94 b033d180 b0225e89
       ef97bb94 ef97bd00 00000000 b0226818 ef97bb60 b01dc10d 00000880 ef97bd00
Call Trace:
 [<ef9cce3a>] ata_pci_remove_one+0x1a/0x40 [libata]
 [<b01dc036>] pci_device_remove+0x16/0x40
 [<b02262b4>] __device_release_driver+0x64/0x90
 [<b02267e5>] driver_detach+0x95/0xa0
 [<b0225e89>] bus_remove_driver+0x69/0x90
 [<b0226818>] driver_unregister+0x8/0x20
Quoted 8 lines
 [<b01dc10d>] pci_unregister_driver+0xd/0x60
 [<b01305d0>] sys_delete_module+0x140/0x1b0
 [<b0142499>] remove_vma+0x39/0x50
 [<b0142e37>] do_munmap+0x197/0x1f0
 [<b0102cb0>] syscall_call+0x7/0xb
 =======================
Code: 74 24 04 8b 7c 24 08 8b 6c 24 0c 83 c4 10 e9 f3 cb fd ff 8d 76 00 0f ba b7 f0 1f 00 00 00 eb d0 8d b
EIP: [<ef9c8b05>] ata_host_remove+0x5/0xd0 [libata] SS:ESP 0068:e3a51ef4

I don't know if there exists a clean way to bail out at this type of error (i.e. is it possible refusing to load the module at all if IRQ Request failed?) but just wanted to give you some input.

Best regards,
Arnd Hannemann

#1 lore raw
Quoted 7 lines · Arnd Hannemann
Quoted 7 lines
BUG: unable to handle kernel NULL pointer dereference at virtual address 00000010
 printing eip:
ef9c8b05
EIP is at ata_host_remove+0x5/0xd0 [libata]
 [<ef9cce3a>] ata_pci_remove_one+0x1a/0x40 [libata]
 [<b01dc036>] pci_device_remove+0x16/0x40
 [<b02262b4>] __device_release_driver+0x64/0x90
Quoted 8 lines
Quoted 8 lines
 [<b02267e5>] driver_detach+0x95/0xa0
 [<b0225e89>] bus_remove_driver+0x69/0x90
 [<b0226818>] driver_unregister+0x8/0x20
 [<b01dc10d>] pci_unregister_driver+0xd/0x60
 [<b01305d0>] sys_delete_module+0x140/0x1b0
 [<b0142499>] remove_vma+0x39/0x50
 [<b0142e37>] do_munmap+0x197/0x1f0
 [<b0102cb0>] syscall_call+0x7/0xb

This should be fixed in the current devel kernel. Can you give a shot at 2.6.21-rc7?

Thanks.

Signature · 1 line
tejun