Al,
This is going to be a continuing problem for non-Unix file systems like
NTFS and NWFS that rely on the ability to read and write variable length
sector runs. At some point, the AIO subsystem needs to get fixed. I
submitted a patch based on Linus' suggestion that the check in
ll_rw_block() be removed to allow folks to chain variable length sector
requests. After I did, half the disk drivers seemed to croak and
break. I doubt 2.4 could be changed, but 2.5 needs this or it's going
to keep biting. The whole buffer head chaining crap is too memory
intensive.
How about
<disk, starting sector, sector count, callback> for the request?
Forcing us to use a 512 byte block size sucks up tons of memory and
creates too much overhead to get high performance numbers. I notice
very significant performance differences when using 512 blocks in NWFS
and EXT2 with 1024 and 4K block sizes.
Jeff
Alexander Viro wrote:
>
> On Sat, 30 Sep 2000, Anton Altaparmakov wrote:
>
> > All those problems disappear as soon as you change BLOCK_SIZE to 512. And
>
> Have you actually tried that? Go ahead, just do full backup before the
> experiment...
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> Please read the FAQ at http://www.tux.org/lkml/
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sat Sep 30 2000 - 21:00:25 EST