SATA disk detect problem with silicon image 3132

SATA disk detect problem with silicon image 3132

[KueiHuan Chen]

Hello,

  I have some problem in detect sata disk with silicon image 3132 in
Marvell platform (arm).

  I am using kernel 2.6.24 with silicon image 3132 in Marvell 5281 /
6281 platform(Driver is sata_sil24). both would has same situation
like the messages below.

  The problem is when the 3132 has disk on it, it would know it. But
the identify command would timeout. After this, all ports would not
responce for any hotplug event.

  The interesting thing is when i change the 3132 with marvell 6145
sata chip(both using PCI-Express), and the driver using AHCI in
Marvell 6281 platform.  It just work.

  And when i use 3132 in x86 or MPC85xx platform(same kernel), it also
work fine. So i don't know where might has problem.  Is there any
hint/idea for me ?

  Thanks.

messages during Boot
Intergrated Sata device found
scsi0 : Marvell SCSI to SATA adapter
scsi1 : Marvell SCSI to SATA adapter
Driver 'sd' needs updating - please use bus_type methods
scsi2 : sata_sil24
scsi3 : sata_sil24
ata1: SATA max UDMA/100 host m128@0xe8084000 port 0xe8080000 irq 9
ata2: SATA max UDMA/100 host m128@0xe8084000 port 0xe8082000 irq 9
ata1: SATA link down (SStatus 0 SControl 0)
ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 0)
ata2.00: qc timeout (cmd 0xec)
ata2.00: failed to IDENTIFY (I/O error, err_mask=0x4)
ata2: failed to recover some devices, retrying in 5 secs
ata2: controller in dubious state, performing PORT_RST
ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 0)
ata2.00: qc timeout (cmd 0xec)
ata2.00: failed to IDENTIFY (I/O error, err_mask=0x4)
ata2: failed to recover some devices, retrying in 5 secs
ata2: controller in dubious state, performing PORT_RST
ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 0)
ata2.00: qc timeout (cmd 0xec)
ata2.00: failed to IDENTIFY (I/O error, err_mask=0x4)
ata2: failed to recover some devices, retrying in 5 secs
ata2: controller in dubious state, performing PORT_RST
ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 0)
physmap platform flash device: 00400000 at f8000000
physmap-flash physmap-flash.0: map_probe failed
md: linear personality registered for level -1
md: raid0 personality registered for level 0
md: raid1 personality registered for level 1
...
After this, hotplug would not detect by system.


------------------------------------
After boot, hotplug into port 2
ata2: exception Emask 0x10 SAct 0x0 SErr 0x0 action 0xa frozen
ata2: irq_stat 0x00a40080, device exchanged
ata2: hard resetting link
ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 0)
ata2.00: qc timeout (cmd 0xec)
ata2.00: failed to IDENTIFY (I/O error, err_mask=0x4)
ata2: failed to recover some devices, retrying in 5 secs
ata2: hard resetting link
ata2: controller in dubious state, performing PORT_RST
ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 0)
ata2.00: qc timeout (cmd 0xec)
ata2.00: failed to IDENTIFY (I/O error, err_mask=0x4)
ata2: failed to recover some devices, retrying in 5 secs
ata2: hard resetting link
ata2: controller in dubious state, performing PORT_RST
ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 0)
ata2.00: qc timeout (cmd 0xec)
ata2.00: failed to IDENTIFY (I/O error, err_mask=0x4)
ata2: failed to recover some devices, retrying in 5 secs
ata2: hard resetting link
ata2: controller in dubious state, performing PORT_RST
ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 0)
ata2: EH complete

After this, hotplug would not detect by system.
-- 
Best Regards,
KueiHuan-Chen

Re: SATA disk detect problem with silicon image 3132

[Tejun Heo]

KueiHuan Chen wrote:
> Hello,
> 
>   I have some problem in detect sata disk with silicon image 3132 in
> Marvell platform (arm).
> 
>   I am using kernel 2.6.24 with silicon image 3132 in Marvell 5281 /
> 6281 platform(Driver is sata_sil24). both would has same situation
> like the messages below.
> 
>   The problem is when the 3132 has disk on it, it would know it. But
> the identify command would timeout. After this, all ports would not
> responce for any hotplug event.
> 
>   The interesting thing is when i change the 3132 with marvell 6145
> sata chip(both using PCI-Express), and the driver using AHCI in
> Marvell 6281 platform.  It just work.
> 
>   And when i use 3132 in x86 or MPC85xx platform(same kernel), it also
> work fine. So i don't know where might has problem.  Is there any
> hint/idea for me ?

Looks like IRQ goes offline after the event.  sata_sil24 uses pretty
heavy hammer to reset the controller.  Maybe it somehow mysteriously
messes with IRQ delivery?

-- 
tejun

Re: SATA disk detect problem with silicon image 3132

[KueiHuan Chen]

Hello,

  Thanks for the reply. But i am curiosity about that is it possible
for two PCI-Epress chip in the same environment, one could work fine,
and another would has IRQ delivery problem ?

Thanks .

2008/9/29 Tejun Heo <htejun@gmail.com>:
> KueiHuan Chen wrote:
>> Hello,
>>
>>   I have some problem in detect sata disk with silicon image 3132 in
>> Marvell platform (arm).
>>
>>   I am using kernel 2.6.24 with silicon image 3132 in Marvell 5281 /
>> 6281 platform(Driver is sata_sil24). both would has same situation
>> like the messages below.
>>
>>   The problem is when the 3132 has disk on it, it would know it. But
>> the identify command would timeout. After this, all ports would not
>> responce for any hotplug event.
>>
>>   The interesting thing is when i change the 3132 with marvell 6145
>> sata chip(both using PCI-Express), and the driver using AHCI in
>> Marvell 6281 platform.  It just work.
>>
>>   And when i use 3132 in x86 or MPC85xx platform(same kernel), it also
>> work fine. So i don't know where might has problem.  Is there any
>> hint/idea for me ?
>
> Looks like IRQ goes offline after the event.  sata_sil24 uses pretty
> heavy hammer to reset the controller.  Maybe it somehow mysteriously
> messes with IRQ delivery?
>
> --
> tejun
>



-- 
Best Regards,
KueiHuan-Chen

Re: SATA disk detect problem with silicon image 3132

[Tejun Heo]

KueiHuan Chen wrote:
>   Thanks for the reply. But i am curiosity about that is it possible
> for two PCI-Epress chip in the same environment, one could work fine,
> and another would has IRQ delivery problem ?

Maybe sata_mv might was in MSI mode?

-- 
tejun

Re: SATA disk detect problem with silicon image 3132

[KueiHuan Chen]

Eh,  my 6145 using AHCI, not sata_mv. And it has the AHCI_HFLAG_NO_MSI
flag. Is there any possible ?

Thanks .

2008/9/29 Tejun Heo <htejun@gmail.com>:
> KueiHuan Chen wrote:
>>   Thanks for the reply. But i am curiosity about that is it possible
>> for two PCI-Epress chip in the same environment, one could work fine,
>> and another would has IRQ delivery problem ?
>
> Maybe sata_mv might was in MSI mode?
>
> --
> tejun
>



-- 
Best Regards,
KueiHuan-Chen

Re: SATA disk detect problem with silicon image 3132

[Tejun Heo]

KueiHuan Chen wrote:
> Eh,  my 6145 using AHCI, not sata_mv. And it has the AHCI_HFLAG_NO_MSI
> flag. Is there any possible ?

Eh... right, 6145 is ahci.  Anyways, yeah, given that pci-e line
interrupts are emulated too, it's quite unlikely that IRQ delivery
gets broken only for sil24 but well I've seen weirder problems and the
failing log definitely looks like IRQ delivery problem.  Does irqpoll
help?

-- 
tejun

Re: SATA disk detect problem with silicon image 3132

[KueiHuan Chen]

Hi,

   irqpoll is not work. And due to some reason, the evaluation board
is taken back. So i can't do more test on it : (   Still thanks a lot.

   If there is any idea or hint, please let me know. I would borrow it again.

Thanks.

2008/9/29 Tejun Heo <htejun@gmail.com>:
> KueiHuan Chen wrote:
>> Eh,  my 6145 using AHCI, not sata_mv. And it has the AHCI_HFLAG_NO_MSI
>> flag. Is there any possible ?
>
> Eh... right, 6145 is ahci.  Anyways, yeah, given that pci-e line
> interrupts are emulated too, it's quite unlikely that IRQ delivery
> gets broken only for sil24 but well I've seen weirder problems and the
> failing log definitely looks like IRQ delivery problem.  Does irqpoll
> help?
>
> --
> tejun
>



-- 
Best Regards,
KueiHuan-Chen