First off, please send SCSI patches to the SCSI list:
<linux-scsi@vger.kernel.org>

On Tue, 2007-01-30 at 19:47 -0500, Mark Lord wrote:
quoted text
> In ancient kernels, the SCSI disk code used to continue after
> encountering a MEDIUM_ERROR.  It would "complete" the good
> sectors before the error, fail the bad sector/block, and then
> continue with the rest of the request.
> 
> Kernels since about 2.6.16 or so have been broken in this regard.
> They "complete" the good sectors before the error,
> and then fail the entire remaining portions of the request.

What was the commit that introduced the change? ... I have a vague
memory of it being deliberate.

quoted text
> This is very risky behaviour, as a request is often a merge
> of several bios, and just because one application hits a bad sector
> is no reason to pretend that (for example) an adjacent directly lookup also failed.
> 
> This patch fixes the behaviour to be similar to what we had originally.
> 
> When a bad sector is encounted, SCSI will now work around it again,
> failing *only* the bad sector itself.

Erm, but the corollary is that if we get a large read failure because of
a bad track, you're going to try and chunk up it a sector at a time
forcing an individual error for each sector is going to annoy some
people ... particularly removable medium ones which return this error if
the medium isn't present ... Are you sure this is really what we want to
do?

quoted text
> Signed-off-by:  Mark Lord <mlord@pobox.com>
> ---
> diff -u --recursive --new-file --exclude-from=linux_17//Documentation/dontdiff old/drivers/scsi/scsi_lib.c linux/drivers/scsi/scsi_lib.c
> --- old/drivers/scsi/scsi_lib.c	2007-01-30 13:58:05.000000000 -0500
> +++ linux/drivers/scsi/scsi_lib.c	2007-01-30 18:30:01.000000000 -0500
> @@ -865,6 +865,12 @@
>  	 */
>  	if (sense_valid && !sense_deferred) {
>  		switch (sshdr.sense_key) {
> +		case MEDIUM_ERROR:
> +			// Bad sector.  Fail it, and then continue the rest of the request:
> +			if (scsi_end_request(cmd, 0, cmd->device->sector_size, 1) == NULL) {

The sense key may have come with additional information  I think we want
to parse that (if it exists) rather than just blindly failing the first
sector of the request.

quoted text
> +				cmd->retries = 0;	// go around again..
> +				return;
> +			}

This would drop through to the UNIT_ATTENTION case if scsi_end_request()
fails ... I don't think that's correct.

quoted text
>  		case UNIT_ATTENTION:
>  			if (cmd->device->removable) {
>  				/* Detected disc change.  Set a bit

-
unsubscribe notice
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/