Re: [example PATCH - not for applying] exclude certain commands

[James Bottomley <James.Bottomley@steeleye.com>]

On Thu, 2003-04-24 at 10:05, Alan Stern wrote:
> You probably already know this, but I'll point it out anyway.  The 
> protocol.c file contains 4 different translation routines:
> 
> 	usb_stor_qic157_command(),
> 	usb_stor_ATAPI_command(),
> 	usb_stor_ufi_command(),
> 	usb_stor_transparent_scsi_command().
> 
> The third one already contains the 4->8 modification and your patch adds 
> it to the fourth.  But it's still missing from the first two.
> 
> I agree with James Bottomley's comment that it's not really safe to be 
> doing this at all, even if we do feel sure that sufficient buffer space 
> exists.  After all, some other data may be using that space.
> 
> Better solutions would be not to issue the 4-byte call at all, 
> or to recognize that it failed and re-issue with at least 8 bytes.  That 
> way no questionable buffer-size conversion is needed.  There's also the 
> advantage of not having to do the same thing in 4 different places within 
> usb-storage.

Actually, the better solution might be to enforce a minimum transfer
length for non-scatter gather transfers.  Although we could enforce this
in the mid-layer allocations, we'd be in difficulty for user issued
commands, so the scheme that copes with everything would be a bounce
buffer type of thing (you supply a transfer request too short for the
device, I provide the bigger buffer and then give you back what you
asked for).

Better suggestions?

James