[re] libata: blacklist NCQ on Seagate Barracuda ST380817AS

[re] libata: blacklist NCQ on Seagate Barracuda ST380817AS

[Alexey Fisher]

Commit da6f0ec2d54acf545e784363aa5bad0dafc5ffa8
libata: blacklist NCQ on Seagate Barracuda ST380817AS

I tested Seagate Barracuda ST380817AS on my system without the blacklist 
  entry on system with ext4 fs and it seems to work. Two days with 
normal load, tar, kvm, torrent... i can't reproduce the error described 
in this commit.

I use board with ICH10R south bridge, AHCI mode.

Alexey

Re: [re] libata: blacklist NCQ on Seagate Barracuda ST380817AS

[Tejun Heo]

Hello,

Alexey Fisher wrote:
> Commit da6f0ec2d54acf545e784363aa5bad0dafc5ffa8
> libata: blacklist NCQ on Seagate Barracuda ST380817AS
> 
> I tested Seagate Barracuda ST380817AS on my system without the blacklist
>  entry on system with ext4 fs and it seems to work. Two days with normal
> load, tar, kvm, torrent... i can't reproduce the error described in this
> commit.
> 
> I use board with ICH10R south bridge, AHCI mode.

Heh.. I don't know.  Things like this are pretty difficult to reliably
reproduce and verify.  As a rule of thumb, cases which are reported
multiple times and in close proximity with existing entries in product
line or firmware revision tend to be included.  Given that earlier NCQ
implementations didn't actually help performance much (or rather often
degraded it) and it has a sibling entry with the same firmware
revision, I'm a bit reluctant to lift the blacklist.  You can
overrided it with kernel parameter "libata.force=P.DD:ncq" where P is
the ata port number and DD is the device number.

Thanks.

-- 
tejun

Re: [re] libata: blacklist NCQ on Seagate Barracuda ST380817AS

[Paolo Ornati]

On Sun, 01 Mar 2009 17:28:06 +0100
Alexey Fisher <bug-track-M18mAb7Tlt0yCq4wW13eYl6hYfS7NtTn@public.gmane.org> wrote:

> Commit da6f0ec2d54acf545e784363aa5bad0dafc5ffa8
> libata: blacklist NCQ on Seagate Barracuda ST380817AS
> 
> I tested Seagate Barracuda ST380817AS on my system without the blacklist 
>   entry on system with ext4 fs and it seems to work. Two days with 
> normal load, tar, kvm, torrent... i can't reproduce the error described 
> in this commit.
> 
> I use board with ICH10R south bridge, AHCI mode.

In my case it was happening _only_ with barriers enabled (like ext3 mounted with
"barrier=1") or with XFS (that enables barriers by default).

Does it survives an untar with barriers?

-- 
	Paolo Ornati
	Linux 2.6.29-rc6-00216-g778ef1e on x86_64

Re: [re] libata: blacklist NCQ on Seagate Barracuda ST380817AS

[Alexey Fisher]

Paolo Ornati schrieb:
> On Sun, 01 Mar 2009 17:28:06 +0100
> Alexey Fisher <bug-track-M18mAb7Tlt0yCq4wW13eYl6hYfS7NtTn@public.gmane.org> wrote:
> 
>> Commit da6f0ec2d54acf545e784363aa5bad0dafc5ffa8
>> libata: blacklist NCQ on Seagate Barracuda ST380817AS
>>
>> I tested Seagate Barracuda ST380817AS on my system without the blacklist 
>>   entry on system with ext4 fs and it seems to work. Two days with 
>> normal load, tar, kvm, torrent... i can't reproduce the error described 
>> in this commit.
>>
>> I use board with ICH10R south bridge, AHCI mode.
> 
> In my case it was happening _only_ with barriers enabled (like ext3 mounted with
> "barrier=1") or with XFS (that enables barriers by default).
> 
> Does it survives an untar with barriers?

ext4 have barriers enabled by default, but it will disable it if hdd is 
not capable to do it. So it will just return:
JBD: barrier-based sync failed on dm-1:8 - disabling barriers

Re: [re] libata: blacklist NCQ on Seagate Barracuda ST380817AS

[Paolo Ornati]

On Tue, 03 Mar 2009 21:39:32 +0100
Alexey Fisher <bug-track-M18mAb7Tlt0yCq4wW13eYl6hYfS7NtTn@public.gmane.org> wrote:

> >> I use board with ICH10R south bridge, AHCI mode.  
> > 
> > In my case it was happening _only_ with barriers enabled (like ext3 mounted with
> > "barrier=1") or with XFS (that enables barriers by default).
> > 
> > Does it survives an untar with barriers?  
> 
> ext4 have barriers enabled by default, but it will disable it if hdd is 
> not capable to do it. So it will just return:
> JBD: barrier-based sync failed on dm-1:8 - disabling barriers

HD in question _CAN_ do barriers.

You are using device mapper, maybe it's that additional layer that
causes barriers to be disabled.

Try with a "raw" partition and see what happens.

-- 
	Paolo Ornati
	Linux 2.6.29-rc6-00305-g2450cf5 on x86_64