Re: Kernel SeekCompleteErrors... Different from Re: LibPATA code issues / 2.6.15.4

[Mark Lord <liml@rtr.ca>]

Alan Cox wrote:
> On Sul, 2006-02-26 at 15:36 -0500, Mark Lord wrote:
>> It still is unreliable, as being discussed in another thread.
>>
>> libata wrongly says "medium error" any time it issues a command
>> that the drive rejects (unsupported, invalid parameters, etc..).
> 
> It seems to still get a single case wrong. But it does the report the
> ATA state correctly still.
> 
>> This is biting a few people in 2.6.16-rc*, due to the FUA stuff.
> 
> It is driven by a table in 
> 
> libata-scsi.c:ata_to_sense_error()
> 
> so if you can figure out the wrong entry and tweak the table that would be great

It's the fall-through case, where the table is not used.


         /* No error?  Undecoded? */
         printk(KERN_WARNING "ata%u: no sense translation for op=0x%02x status: 0x%02x\n",
                id, opcode, drv_stat);

         /* For our last chance pick, use medium read error because
          * it's much more common than an ATA drive telling you a write
          * has failed.
          */
         *sk = MEDIUM_ERROR;
         *asc = 0x11; /* "unrecovered read error" */
         *ascq = 0x04; /*  "auto-reallocation failed" */

Cheers