Re: [PATCH v4 22/36] scsi: Have scsi-ml retry read_capacity_16 errors

[Mike Christie <michael.christie@oracle.com>]

On 10/19/22 5:15 PM, Bart Van Assche wrote:
> On 10/16/22 12:59, Mike Christie wrote:
>> +    struct scsi_failure failures[] = {
>> +        {
>> +            .sense = UNIT_ATTENTION,
>> +            .asc = 0x29,
>> +            .ascq = 0,
>> +            /* Device reset might occur several times */
>> +            .allowed = READ_CAPACITY_RETRIES_ON_RESET,
>> +            .result = SAM_STAT_CHECK_CONDITION,
>> +        },
>> +        {
>> +            .result = SCMD_FAILURE_ANY,
>> +            .allowed = 3,
>> +        },
>> +        {},
>> +    };
> 
> Is the implication of the above that any failure reported in the .result field will result in a retry? I don't think that matches the current behavior of read_capacity_16().

I think I made 2 mistakes:

1. Retrying on media_not_present
2. Retrying on the specific ILLEGAL_REQUEST error.

I made a similar mistakes on some other patches so will fix
them as well.