Re: [BUG] 2.5.23 mpage io vs queue_max_sectors

[William Lee Irwin III <wli@holomorphy.com>]

William Lee Irwin III wrote:
>> This has been haunting me for a while on a 4 cpu Sequent S-quad with
>> 4GB of RAM, and a tray of 12 9GB DCHS09X's attached to a QLogicISP1020.
>> People keep saying "hands off" ... so here I am passing the buck.
>> Kernel BUG at ll_rw_blk.c:1639

On Wed, Jun 19, 2002 at 05:38:25PM -0700, Andrew Morton wrote:
> This is happening because we have BIO_MAX_SIZE = 64k,
> but that particular driver doesn't like requests which are
> that big.

Well, I'm not sure what the driver's doing (or not doing) but
this check is on q->max_sectors...

On Wed, Jun 19, 2002 at 05:38:25PM -0700, Andrew Morton wrote:
> This is the issuewhich Adam, Jens and I have been discussing.
> Looks like the preferred solution is an add_page_to_bio() API
> in the block layer.
> But it's not there yet, so in the short-term, please just do
> -#define MPAGE_BIO_MAX_SIZE BIO_MAX_SIZE
> +#define MPAGE_BIO_MAX_SIZE 16384
> in fs/mpage.c

Hmm, q->max_sectors is 64 in my case (and 96 was fed to the queue) so
I'm not convinced I have to cripple it entirely, I'll go with 32K.


Cheers,
Bill