[PATCH] libata-eh: Add a simple mechanism for silencing error reporting

[PATCH] libata-eh: Add a simple mechanism for silencing error reporting

[Alan]

We want to be able to issue commands that fail silently some of the time
(set_features/xfer rate to CF 1.4 devices, perhaps some others such as
user SG_IO commands ought to be silent too as the error is for the app)

This is a minimal implementation, we can extend it so the QUIET flag
isn't quiet about errors that are not command errors but indicate
infrastructre problems (CRC errors, HSM violation, DeviceFault) if need
be.

Signed-off-by: Alan Cox <alan@redhat.com>

diff -u --new-file --recursive --exclude-from /usr/src/exclude linux.vanilla-2.6.20-rc6-mm3/drivers/ata/libata-eh.c linux-2.6.20-rc6-mm3/drivers/ata/libata-eh.c
--- linux.vanilla-2.6.20-rc6-mm3/drivers/ata/libata-eh.c	2007-01-31 14:20:39.000000000 +0000
+++ linux-2.6.20-rc6-mm3/drivers/ata/libata-eh.c	2007-01-31 14:27:25.000000000 +0000
@@ -1407,6 +1407,8 @@
 			continue;
 		if (qc->flags & ATA_QCFLAG_SENSE_VALID && !qc->err_mask)
 			continue;
+		if (qc->tf.flags & ATA_TFLAG_QUIET)
+			continue;
 
 		nr_failed++;
 	}
@@ -1446,6 +1448,8 @@
 
 		if (!(qc->flags & ATA_QCFLAG_FAILED) || !qc->err_mask)
 			continue;
+		if (qc->tf.flags & ATA_TFLAG_QUIET)
+			continue;
 
 		ata_dev_printk(qc->dev, KERN_ERR,
 			"cmd %02x/%02x:%02x:%02x:%02x:%02x/%02x:%02x:%02x:%02x:%02x/%02x "
diff -u --new-file --recursive --exclude-from /usr/src/exclude linux.vanilla-2.6.20-rc6-mm3/include/linux/ata.h linux-2.6.20-rc6-mm3/include/linux/ata.h
--- linux.vanilla-2.6.20-rc6-mm3/include/linux/ata.h	2007-01-31 14:20:43.000000000 +0000
+++ linux-2.6.20-rc6-mm3/include/linux/ata.h	2007-02-05 10:53:45.000000000 +0000
@@ -232,6 +232,7 @@
 	ATA_TFLAG_LBA		= (1 << 4), /* enable LBA */
 	ATA_TFLAG_FUA		= (1 << 5), /* enable FUA */
 	ATA_TFLAG_POLLING	= (1 << 6), /* set nIEN to 1 and use polling */
+	ATA_TFLAG_QUIET		= (1 << 7), /* don't log rejection */
 };
 
 enum ata_tf_protocols {

Re: [PATCH] libata-eh: Add a simple mechanism for silencing error reporting

[Tejun Heo]

Alan wrote:
> We want to be able to issue commands that fail silently some of the time
> (set_features/xfer rate to CF 1.4 devices, perhaps some others such as
> user SG_IO commands ought to be silent too as the error is for the app)
> 
> This is a minimal implementation, we can extend it so the QUIET flag
> isn't quiet about errors that are not command errors but indicate
> infrastructre problems (CRC errors, HSM violation, DeviceFault) if need
> be.
> 
> Signed-off-by: Alan Cox <alan@redhat.com>
> 
> diff -u --new-file --recursive --exclude-from /usr/src/exclude linux.vanilla-2.6.20-rc6-mm3/drivers/ata/libata-eh.c linux-2.6.20-rc6-mm3/drivers/ata/libata-eh.c
> --- linux.vanilla-2.6.20-rc6-mm3/drivers/ata/libata-eh.c	2007-01-31 14:20:39.000000000 +0000
> +++ linux-2.6.20-rc6-mm3/drivers/ata/libata-eh.c	2007-01-31 14:27:25.000000000 +0000
> @@ -1407,6 +1407,8 @@
>  			continue;
>  		if (qc->flags & ATA_QCFLAG_SENSE_VALID && !qc->err_mask)
>  			continue;
> +		if (qc->tf.flags & ATA_TFLAG_QUIET)
> +			continue;
>  
>  		nr_failed++;
>  	}
> @@ -1446,6 +1448,8 @@
>  
>  		if (!(qc->flags & ATA_QCFLAG_FAILED) || !qc->err_mask)
>  			continue;
> +		if (qc->tf.flags & ATA_TFLAG_QUIET)
> +			continue;
>  
>  		ata_dev_printk(qc->dev, KERN_ERR,
>  			"cmd %02x/%02x:%02x:%02x:%02x:%02x/%02x:%02x:%02x:%02x:%02x/%02x "

I think the result can be a bit confusing if ehc->i.err_mask is set
(port/dev exception message without failed command report).  How about
counting nr_quiet and doing if (nr_failed == nr_quiet) return right
after the counting block?

-- 
tejun

Re: [PATCH] libata-eh: Add a simple mechanism for silencing error reporting

[Jeff Garzik]

Alan wrote:
> We want to be able to issue commands that fail silently some of the time
> (set_features/xfer rate to CF 1.4 devices, perhaps some others such as
> user SG_IO commands ought to be silent too as the error is for the app)
> 
> This is a minimal implementation, we can extend it so the QUIET flag
> isn't quiet about errors that are not command errors but indicate
> infrastructre problems (CRC errors, HSM violation, DeviceFault) if need
> be.
> 
> Signed-off-by: Alan Cox <alan@redhat.com>
> 
> diff -u --new-file --recursive --exclude-from /usr/src/exclude linux.vanilla-2.6.20-rc6-mm3/drivers/ata/libata-eh.c linux-2.6.20-rc6-mm3/drivers/ata/libata-eh.c
> --- linux.vanilla-2.6.20-rc6-mm3/drivers/ata/libata-eh.c	2007-01-31 14:20:39.000000000 +0000
> +++ linux-2.6.20-rc6-mm3/drivers/ata/libata-eh.c	2007-01-31 14:27:25.000000000 +0000
> @@ -1407,6 +1407,8 @@
>  			continue;
>  		if (qc->flags & ATA_QCFLAG_SENSE_VALID && !qc->err_mask)
>  			continue;
> +		if (qc->tf.flags & ATA_TFLAG_QUIET)
> +			continue;
>  
>  		nr_failed++;
>  	}

I would rather pass an ok-to-fail-without-recovery type flag to EH. 
This TFLAG_QUIET approach is poorly defined, probably temporary, and not 
really interesting for upstream

	Jeff

Re: [PATCH] libata-eh: Add a simple mechanism for silencing error reporting

[Alan]

> I would rather pass an ok-to-fail-without-recovery type flag to EH. 
> This TFLAG_QUIET approach is poorly defined, probably temporary, and not 
> really interesting for upstream

Fail without recovery flags make sense, but then we also need to ensure
it is a fail without recovery messages. We could add qc->ignore_mask to
mirror qc->err_mask ?

Alan