On Thu, Apr 26, 2012 at 5:29 PM, Gwendal Grignou [off-list ref] wrote:

Lin,

In the patch d9027470b88631d0956ac37cdadfdeb9cdcf2c99, I did limit the
amount of data cleaning in some of the ata objects.

If setting err_mask to 0 masks the regression I introduced in the
patch, I may have altered too much how ata_device object is
reinitialized when a device is found. I am digging deeper, I may have
change the code to try to preserve the ering as much as possible.

Concerning your patch, isn't adding a test (ent->eflags &
ATA_EFLAG_OLD_ER) in ata_count_probe_trials_cb() more in line with
speed_down_verdict_cb() code?

Hi,

I have updated the patch.

[PATCH] libata: skip old error history when counting probe trials
http://marc.info/?l=linux-ide&m=133600832126645&w=2

Regards,
Lin Ming