linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
* mix-block size for raw_io ??
@ 2000-10-12 16:18 BenHanokh Gabriel
  2000-10-12 18:06 ` Stephen Tweedie
  0 siblings, 1 reply; 2+ messages in thread
From: BenHanokh Gabriel @ 2000-10-12 16:18 UTC (permalink / raw)
  To: linux-mm

hi

i saw that the raw interface is using 1/2K blocks for io.
i understand that raw_io needs to support disk accesses in sector 
resolution, but why is it not possible to mix block sizes for the same 
device -> giving a much better performance using something like:
---------------------
left_bytes    = io_size % 4K
bytes            = io_size - left_bytes

call brw_kiovec(bytes)         // using_4K_blocks
call brw_kiovec(left_bytes) // using_1/2K_blocks
----------------------

the brw_kiovec() doesn't seems to remember block_size between calls  as 
it calculates everything based on the block_size passed as parameter.

is there any reason why we don't use mix-block-size ?

please CC me for answers

THX
/Gabriel BenHanokh

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux.eu.org/Linux-MM/

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: mix-block size for raw_io ??
  2000-10-12 16:18 mix-block size for raw_io ?? BenHanokh Gabriel
@ 2000-10-12 18:06 ` Stephen Tweedie
  0 siblings, 0 replies; 2+ messages in thread
From: Stephen Tweedie @ 2000-10-12 18:06 UTC (permalink / raw)
  To: BenHanokh Gabriel; +Cc: linux-mm

Hi,

On Thu, Oct 12, 2000 at 07:18:49PM +0300, BenHanokh Gabriel wrote:
> 
> i saw that the raw interface is using 1/2K blocks for io.
> i understand that raw_io needs to support disk accesses in sector 
> resolution, but why is it not possible to mix block sizes for the same 
> device

It is possible in principle, but I can't guarantee that every device
driver will work properly if you do this.

> is there any reason why we don't use mix-block-size ?

Because it would be a short-term hack --- the real solution to this
problem is to push kiobufs right down into the block device request
layer.  SGI have already got patches to do this, with support in the
scsi and ide drivers.

Cheers, 
 Stephen
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux.eu.org/Linux-MM/

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2000-10-12 18:06 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2000-10-12 16:18 mix-block size for raw_io ?? BenHanokh Gabriel
2000-10-12 18:06 ` Stephen Tweedie

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox