A question about ll_10byte_cmd_build

A question about ll_10byte_cmd_build

[peterc]

(Linux 2.5.8)
In ll_rw_blk.c there's a function, ll_10byte_cmd_build() which is
supposed to be used to generate `10-byte commands'.

It appears to generate a SCSI READ_10 or WRITE_10 command (which
happen to be identical in format to the ATAPI GPCMD_{READ,WRITE}_10 commands)

Is this IDE specific, or is it meant to cover all block devices?
If it's IDE specific, why is it in ll_rw_blk.c, which is meant to be
common to all block devices?


As far as I can tell, only ide-cd.c actually uses the function in a
stock 2.5.8 kernel --- so it could theoretically be moved to ide-cd.c.

Peter C

Re: A question about ll_10byte_cmd_build

[Jens Axboe]

On Thu, Apr 18 2002, peterc@gelato.unsw.edu.au wrote:
> 
> 
> (Linux 2.5.8)
> In ll_rw_blk.c there's a function, ll_10byte_cmd_build() which is
> supposed to be used to generate `10-byte commands'.
> 
> It appears to generate a SCSI READ_10 or WRITE_10 command (which
> happen to be identical in format to the ATAPI GPCMD_{READ,WRITE}_10 commands)
> 
> Is this IDE specific, or is it meant to cover all block devices?
> If it's IDE specific, why is it in ll_rw_blk.c, which is meant to be
> common to all block devices?

It's not IDE specific, since when does IDE use packet commands? It's
not ATAPI specific either.

> As far as I can tell, only ide-cd.c actually uses the function in a
> stock 2.5.8 kernel --- so it could theoretically be moved to ide-cd.c.

It's meant to be an example of a generic prep_rq_fn() queue function, a
start the conversion of using struct request as the generic passer of
cdb's. Only ide-cd uses it for now as you see, that's merely because
lots of bits of the infrastructure are still missing.

-- 
Jens Axboe