[PATCH] libata-dev: recognize WRITE_MULTI_FUA_EXT for r/w multiple

[PATCH] libata-dev: recognize WRITE_MULTI_FUA_EXT for r/w multiple

[Albert Lee]

Jeff,

Changes:
Recognize ATA_CMD_WRITE_MULTI_FUA_EXT as r/w multiple commands.

Patch against the irq-pio branch
(1ed82ae3df3882bfcb5bfe5e9f97e4e8e23af299).

For your review, thanks.

Albert

Signed-off-by: Albert Lee <albertcc@tw.ibm.com> 
---

--- irq-pio/include/linux/ata.h	2006-03-02 15:22:58.000000000 +0800
+++ irq-pio-catchup/include/linux/ata.h	2006-03-02 17:05:17.000000000 +0800
@@ -314,7 +314,8 @@ static inline int is_multi_taskfile(stru
 	return (tf->command == ATA_CMD_READ_MULTI) ||
 	       (tf->command == ATA_CMD_WRITE_MULTI) ||
 	       (tf->command == ATA_CMD_READ_MULTI_EXT) ||
-	       (tf->command == ATA_CMD_WRITE_MULTI_EXT);
+	       (tf->command == ATA_CMD_WRITE_MULTI_EXT) ||
+	       (tf->command == ATA_CMD_WRITE_MULTI_FUA_EXT);
 }
 
 static inline int ata_ok(u8 status)

Re: [PATCH] libata-dev: recognize WRITE_MULTI_FUA_EXT for r/w multiple

[Mark Lord]

Albert Lee wrote:
> Jeff,
> 
> Changes:
> Recognize ATA_CMD_WRITE_MULTI_FUA_EXT as r/w multiple commands.

This one will require exhaustive testing on various controllers.
There are quite a few PIO boards with FIFOs that "recognize"
opcodes for READs and WRITEs.  And practically all of them predate
any concept of FUA commands.

Very VERY risky.

Re: [PATCH] libata-dev: recognize WRITE_MULTI_FUA_EXT for r/w multiple

[Jens Axboe]

On Fri, Mar 03 2006, Mark Lord wrote:
> Albert Lee wrote:
> >Jeff,
> >
> >Changes:
> >Recognize ATA_CMD_WRITE_MULTI_FUA_EXT as r/w multiple commands.
> 
> This one will require exhaustive testing on various controllers.
> There are quite a few PIO boards with FIFOs that "recognize"
> opcodes for READs and WRITEs.  And practically all of them predate
> any concept of FUA commands.
> 
> Very VERY risky.

Which is why it's so annoying and seemingly idiotic that t13 keeps
making up new opcodes for the exact same thing instead of allowing
options bits.

-- 
Jens Axboe

Re: [PATCH] libata-dev: recognize WRITE_MULTI_FUA_EXT for r/w multiple

[Jeff Garzik]

Albert Lee wrote:
> Jeff,
> 
> Changes:
> Recognize ATA_CMD_WRITE_MULTI_FUA_EXT as r/w multiple commands.
> 
> Patch against the irq-pio branch
> (1ed82ae3df3882bfcb5bfe5e9f97e4e8e23af299).
> 
> For your review, thanks.
> 
> Albert
> 
> Signed-off-by: Albert Lee <albertcc@tw.ibm.com> 

I would request that you adjust your patch submission slightly.  Stuff 
like "Jeff," "Changes:", "Patch against...", "For your review", "Albert" 
all must be hand-edited out before applying the patch.  Everything 
before the "---" terminator is copied directly into the kernel changelog.

Take a look at Tejun's recent patches for an example.  The message body 
of his emailed submissions look like:

	[change description]
	[signed-off-by]
	---
	[notes, comments, insults, etc.  :)]


> --- irq-pio/include/linux/ata.h	2006-03-02 15:22:58.000000000 +0800
> +++ irq-pio-catchup/include/linux/ata.h	2006-03-02 17:05:17.000000000 +0800
> @@ -314,7 +314,8 @@ static inline int is_multi_taskfile(stru
>  	return (tf->command == ATA_CMD_READ_MULTI) ||
>  	       (tf->command == ATA_CMD_WRITE_MULTI) ||
>  	       (tf->command == ATA_CMD_READ_MULTI_EXT) ||
> -	       (tf->command == ATA_CMD_WRITE_MULTI_EXT);
> +	       (tf->command == ATA_CMD_WRITE_MULTI_EXT) ||
> +	       (tf->command == ATA_CMD_WRITE_MULTI_FUA_EXT);

Applied to irq-pio branch, though Mark Lord makes a correct point.  We 
may have to create a whitelist for PATA devices that support FUA, rather 
than a blacklist.  Something to think about...

Thanks,

	Jeff