Re: Odd behaviour of device in response to idleimmediate with unload

[Tejun Heo <tj@kernel.org>]

Evgeni Golov wrote:
> On Mon, 17 Nov 2008 16:19:59 +0900 Tejun Heo wrote:
> 
>> Evgeni Golov wrote:
>>> On Sun, 16 Nov 2008 18:39:08 +0900 Tejun Heo wrote:
>>>
>>>> So, this comes before actually issuing the park.  It sounds like it has
>>>> nothing to do with park command itself.  If you do "echo - - - >
>>>> /sys/class/scsi_host/host0/scan" right after boot, what does the kernel say?
>>> # echo - - - > /sys/class/scsi_host/host0/scan
>>> echo: write error: invalid argument
>> Eh... strange.  It works perfectly fine here.  Can you play with quoting
>> and -n?
> 
> Heh, did try -n but no quoting, damn - :)
> Get the same reset too:
> 
> shinkupaddo# dmesg
> ata1: exception Emask 0x10 SAct 0x0 SErr 0x50000 action 0xf
> ata1: SError: { PHYRdyChg CommWake }

Heh... you're not supposed to see these events here, so the PHY event
and resetting don't have anything to do with head unloading per se.
It's just discovering previously set SError values.  The question is
when did they get set and why didn't ahci catch it.  libata EH enables
enable reporting and then clear SError before finishing up so there
shouldn't be any window where events can get lost.

Can you please attach lspci -nn result and full boot log?

Thanks.

-- 
tejun