On Wed, Jul 30, 2025 at 09:24:40AM +0900, Damien Le Moal wrote:
> Commit 8ae720449fca ("libata: whitespace fixes in ata_to_sense_error()")
> inadvertantly added the entry 0x40 (ATA_DRDY) to the stat_table array in
> the function ata_to_sense_error(). This entry ties a failed qc which has
> a status filed equal to ATA_DRDY to the sense key ILLEGAL REQUEST with
> the additional sense code UNALIGNED WRITE COMMAND. This entry will be
> used to generate a failed qc sense key and sense code when the qc is
> missing sense data and there is no match for the qc error field in the
> sense_table array of ata_to_sense_error().
>
> As a result, for a failed qc for which we failed to get sense data (e.g.
> read log 10h failed if qc is an NCQ command, or REQUEST SENSE EXT
> command failed for the non-ncq case, the user very often end up seeing
> the completely misleading "unaligned write command" error, even if qc
> was not a write command. E.g.:
>
> sd 0:0:0:0: [sda] tag#12 FAILED Result: hostbyte=DID_OK driverbyte=DRIVER_OK cmd_age=0s
> sd 0:0:0:0: [sda] tag#12 Sense Key : Illegal Request [current]
> sd 0:0:0:0: [sda] tag#12 Add. Sense: Unaligned write command
> sd 0:0:0:0: [sda] tag#12 CDB: Read(10) 28 00 00 00 10 00 00 00 08 00
> I/O error, dev sda, sector 4096 op 0x0:(READ) flags 0x80700 phys_seg 1 prio class 0
>
> Fix this by removing the ATA_DRDY entry from the stat_table array so
> that we default to always returning ABORTED COMMAND without any
> additional sense code, since we do not know any better. The entry 0x08
> (ATA_DRQ) is also removed since signaling ABORTED COMMAND with a parity
> error is also misleading (as a parity error would likely be signaled
> through a bus error). So for this case, also default to returning
> ABORTED COMMAND without any additional sense code. With this, the
> previous example error case becomes:
>
> sd 0:0:0:0: [sda] tag#17 FAILED Result: hostbyte=DID_OK driverbyte=DRIVER_OK cmd_age=0s
> sd 0:0:0:0: [sda] tag#17 Sense Key : Aborted Command [current]
> sd 0:0:0:0: [sda] tag#17 Add. Sense: No additional sense information
> sd 0:0:0:0: [sda] tag#17 CDB: Read(10) 28 00 00 00 10 00 00 00 08 00
> I/O error, dev sda, sector 4096 op 0x0:(READ) flags 0x80700 phys_seg 1 prio class 0
>
> Together with these fixes, refactor stat_table to make it more readable
> by putting the entries comments in front of the entries and using the
> defined status bits macros instead of hardcoded values.
>
> Reported-by: Lorenz Brun <lorenz@xxxxxxxx>
> Reported-by: Brandon Schwartz <Brandon.Schwartz@xxxxxxx>
> Fixes: 8ae720449fca ("libata: whitespace fixes in ata_to_sense_error()")
> Cc: stable@xxxxxxxxxxxxxxx
> Signed-off-by: Damien Le Moal <dlemoal@xxxxxxxxxx>
> ---
> drivers/ata/libata-scsi.c | 20 ++++++++------------
> 1 file changed, 8 insertions(+), 12 deletions(-)
>
> diff --git a/drivers/ata/libata-scsi.c b/drivers/ata/libata-scsi.c
> index 27b15176db56..9b16c0f553e0 100644
> --- a/drivers/ata/libata-scsi.c
> +++ b/drivers/ata/libata-scsi.c
> @@ -859,18 +859,14 @@ static void ata_to_sense_error(u8 drv_stat, u8 drv_err, u8 *sk, u8 *asc,
> {0xFF, 0xFF, 0xFF, 0xFF}, // END mark
> };
> static const unsigned char stat_table[][4] = {
> - /* Must be first because BUSY means no other bits valid */
> - {0x80, ABORTED_COMMAND, 0x47, 0x00},
> - // Busy, fake parity for now
> - {0x40, ILLEGAL_REQUEST, 0x21, 0x04},
> - // Device ready, unaligned write command
> - {0x20, HARDWARE_ERROR, 0x44, 0x00},
> - // Device fault, internal target failure
> - {0x08, ABORTED_COMMAND, 0x47, 0x00},
> - // Timed out in xfer, fake parity for now
> - {0x04, RECOVERED_ERROR, 0x11, 0x00},
> - // Recovered ECC error Medium error, recovered
> - {0xFF, 0xFF, 0xFF, 0xFF}, // END mark
> + /* Busy: must be first because BUSY means no other bits valid */
> + { ATA_BUSY, ABORTED_COMMAND, 0x00, 0x00 },
Hi Damien,
ata_to_sense_error() already has a check for ATA_BUSY. Perhaps we could add
a goto statement and fill ABORTED_COMMAND without looking up the same data in
the stat_table?
> + /* Device fault: INTERNAL TARGET FAILURE */
> + { ATA_DF, HARDWARE_ERROR, 0x44, 0x00 },
> + /* Corrected data error */
> + { ATA_CORR, RECOVERED_ERROR, 0x00, 0x00 },
I'm trying to understand what this "Corrected data error" is. ACS-6 does not
seem to have any references to such corrected errors. BIT(2) of STATUS field
is defined as "N/A or ALIGNMENT ERROR bit – See 6.2.2". Does it make sense
to translate this bit to "unaligned write command" instead?
Thanks,
Igor
> +
> + { 0xFF, 0xFF, 0xFF, 0xFF }, /* END mark */
> };
>
> /*
> --
> 2.50.1
>
>
