Re: POWER MODE and sat,12 again

Alfred Ganz wrote:
>>   Does the SMART STATUS (smartctl -H) work with this kernel?
> Yes,

Sorry no, because ...


>    # smartctl -H -d sat,12 /dev/sdb
>    smartctl 5.43 2012-06-30 r3573 [i686-linux-2.6.32-573.1.1.el6.i686] (local build)
>    Copyright (C) 2002-12 by Bruce Allen, http://smartmontools.sourceforge.net
>    
>    === START OF READ SMART DATA SECTION ===
>    SMART overall-health self-assessment test result: PASSED
>    Warning: This result is based on an Attribute check.

... the "Warning" means that SMART status command itself failed, likely 
due to same broken SAT behavior.
Try "smartctl -r ioctl,2 -H ..." and see the same broken sense key 4.


>    Please note the following marginal Attributes:
>    ID# ATTRIBUTE_NAME          FLAG     VALUE WORST THRESH TYPE      UPDATED  WHEN_FAILED RAW_VALUE
>    190 Airflow_Temperature_Cel 0x0022   067   041   045    Old_age   Always   In_the_past 33 (0 13 33 33 0)
>
>> Did you check -d sat ?
> Yes, It doesn't work.
>
> These are relatively up to date kernels, maintained in the RedHat
> fashion (i.e changes are retrofitted to the base kernel release).
> smart{d,ctl} have worked with sat,12 on this disk+enclosure without
> problems in a sequence of kernels that ended with kernel 2.6.32-504.30.3.el6.
> But now suddenly with the new kernel 2.6.32-573.1.1.el6 there are
> problems (which eerely look like the problems we addressed in 2009).

Then RH possibly broke something in the SAT layer.
IIRC the SAT layer of more recent 2.6.* upstream kernels worked as 
expected and also supported -d sat[,16].


> The strange thing is that only some SAT ATA PASS-THROUGH(12) commands
> seem to fail, hdparm fails with "-S 180 --prefer-ata12" but works with
> "-I --prefer-ata12" and smart{d,ctl} fail with "-n" but otherwise
> work. In both failed cases POWER MODE is accessed. I haven't discovered
> any other sat,12 accesses that caused problems. Is there something
> special with accessing POWER MODE?

Yes. The problem only occurs in the rare cases when the result is 
returned in ATA output registers:
smartctl -H, -n standby, -l scterc

In SAT ATA PASS-THROUGH commands, the output registers are requested by 
CK_COND bit and returned in ATA RETURN DESCRIPTOR sense data. The SAT 
layer MUST NOT return a sense key indicating an error in this case 
(except when the ATA command has actually failed).


> Any advice on how to proceed? Thanks, AG

I probably will add some flag, e.g. "-d sat,12,i" to indicate that the 
error sense key should be ignored in this case.

Thanks,
Christian


------------------------------------------------------------------------------