S.M.A.R.T. Monitoring Tools

On 13/10/10 08:00, fortaisur wrote:
> I noticed a problem I didn't notice before: running smartd seems
> to take an Intel SSD out of DMA mode.  Here is a demonstration of
>    

That's arguably a Linux bug, I'd say.  Can you use an libata driver 
instead, e.g. force the use of the libata PIIX driver, instead of the 
old-IDE piix driver, or put the controller into AHCI mode, or use an 
sil3132 PCIe card or similar?  Note that the libata drivers use the 
/dev/sd? namespace, so you'll need to make sure you don't have /dev/hda 
hard-coded if you use this.

On the plus side, if you use AHCI, or the Sil card then you'll get NCQ 
(which you won't with PIIX), and thus better performance (hdparm -t just 
gave 250M per second on a similar Intel 80G SSD using ahci on this 
particular machine).

Tim.

-- 
South East Open Source Solutions Limited
Registered in England and Wales with company number 06134732.
Registered Office: 2 Powell Gardens, Redhill, Surrey, RH1 1TQ
VAT number: 900 6633 53  http://seoss.co.uk/ +44-(0)1273-808309

Tim Small <tim@...> wrote:

>On 13/10/10 08:00, fortaisur wrote:
>> I noticed a problem I didn't notice before: running smartd seems
>> to take an Intel SSD out of DMA mode.  Here is a demonstration of
>>    
>
>That's arguably a Linux bug, I'd say.  Can you use an libata driver 
>instead, e.g. force the use of the libata PIIX driver, instead of the 
>old-IDE piix driver, or put the controller into AHCI mode, or use an 
>sil3132 PCIe card or similar?  Note that the libata drivers use the 
>/dev/sd? namespace, so you'll need to make sure you don't have /dev/hda 
>hard-coded if you use this.
>
>On the plus side, if you use AHCI, or the Sil card then you'll get NCQ 
>(which you won't with PIIX), and thus better performance (hdparm -t just 
>gave 250M per second on a similar Intel 80G SSD using ahci on this 
>particular machine).

Thanks for the response.

It's a VIA M800 Mini-ITX unit, it has a single PCI slot but there
isn't much room so I doubt I could use a card.

I don't know how to try different drivers.

Wikipedia says that AHCI is supported in kernel 2.6.19, the
current RHEL5/CentOS is 2.6.18-194.17.1 so maybe I can't try
that; don't even know if the PC supports it.

I poked around in the AwardBIOS, I see nothing to tell it use
AHCI or an alternate SATA non-emulation mode or anything like
that.

In a web search I found mention to try "hda=noprobe hda=none" but
that failed early in booting, it can't find various /dev items,
goes into kernel panic, so won't work on this existing built
system.

On 10-10-14 03:12 AM, fortaisur wrote:
> Tim Small<tim@...>  wrote:
>
> Thanks for the response.
>
> It's a VIA M800 Mini-ITX unit, it has a single PCI slot but there
> isn't much room so I doubt I could use a card.
>
> I don't know how to try different drivers.
>
> Wikipedia says that AHCI is supported in kernel 2.6.19, the
> current RHEL5/CentOS is 2.6.18-194.17.1 so maybe I can't try
> that; don't even know if the PC supports it.

You might want to try a recent LiveCD which will have a more recent 
kernel. This has the advantage of not making the situation worse as you 
don't have to install anything to your existing system.

Since I use Ubuntu I'll give Ubuntu centric hints. There  are lots of 
RedHat Live CDs as well, but I don't use them.

The Ubuntu 10.10 install CD is a live CD as well as an installer and has 
a far more recent kernel than the one you are using.

Just boot the LiveCD and select "try ubuntu". Then install the 
smartmontools by: sudo apt-get install smartmontools. You shouldn't even 
have to mount your existing SDD, just exercise it with dd based reads 
(dd if=/dev/sdx of=/dev/null bs=1M count=16 sort of thing) and run 
smartctl on the same device in another window.

None of this will actually change anything on your SSD, so if it doesn't 
work you haven't lost anything.

Andrew Reid <AndrewReid@...> wrote:

>On 10-10-14 03:12 AM, fortaisur wrote:
>> Tim Small<tim@...>  wrote:
>>
>> Thanks for the response.
>>
>> It's a VIA M800 Mini-ITX unit, it has a single PCI slot but there
>> isn't much room so I doubt I could use a card.
>>
>> I don't know how to try different drivers.
>>
>> Wikipedia says that AHCI is supported in kernel 2.6.19, the
>> current RHEL5/CentOS is 2.6.18-194.17.1 so maybe I can't try
>> that; don't even know if the PC supports it.
>
>You might want to try a recent LiveCD which will have a more recent 
>kernel. This has the advantage of not making the situation worse as you 
>don't have to install anything to your existing system.
>
>Since I use Ubuntu I'll give Ubuntu centric hints. There  are lots of 
>RedHat Live CDs as well, but I don't use them.
>
>The Ubuntu 10.10 install CD is a live CD as well as an installer and has 
>a far more recent kernel than the one you are using.
>
>Just boot the LiveCD and select "try ubuntu". Then install the 
>smartmontools by: sudo apt-get install smartmontools. You shouldn't even 
>have to mount your existing SDD, just exercise it with dd based reads 
>(dd if=/dev/sdx of=/dev/null bs=1M count=16 sort of thing) and run 
>smartctl on the same device in another window.
>
>None of this will actually change anything on your SSD, so if it doesn't 
>work you haven't lost anything.

Thanks, I tried Ubuntu 10.10 as a Live CD.  Couldn't get its
networking going so transferred smartmontools via USB stick. I
can't tell if DMA is on or not, doing hdparm -d /dev/sda gives:

HDIO_GET_DMA: Inappropriate ioctl for device

But based on hdparm -t speed of about 200MB/s before running the
smartd daemon, and the same speed after, and no disk errors in
syslog or daemon.log running smartd, I don't think that smartd
causes the falling-out-of-DMA problem on this kernel (2.6.35-22).

Unfortunately that doesn't help me much, I'm long-settled on the
CentOS platform.  I guess I just won't be able to use the Intel
drive in the Mini-ITX, so will put it in a Windows machine, and
use another SSD here, probably OCZ.  Either that, or just not use
smartmontools at all for now.

On 10-10-24 12:48 AM, fortaisur wrote:


> But based on hdparm -t speed of about 200MB/s before running the
> smartd daemon, and the same speed after, and no disk errors in
> syslog or daemon.log running smartd, I don't think that smartd
> causes the falling-out-of-DMA problem on this kernel (2.6.35-22).
>

You are almost there, I wouldn't give up yet.

Centos 5.5 Has a live CD (https://projects.centos.org/trac/livecd/), try 
that to make sure it fixes your issue.

I would use something more rigorous than hdparm to exercise the disk.

Try:
dd if=/dev/sda of=/dev/null bs=1M count=16


which will read 16GB from the disk, enough to remove caching effects and 
long enough to expose any issues.