Re: ll_rw_blk.c fails to merge requests. Help!

From: Giuliano Pochini (pochini@denise.shiny.it)
Date: Wed Aug 30 2000 - 21:45:16 EDT

Next message: David Weis: "Re: 2.2.16 deadlocking in schedule()"

Previous message: Eugene Crosser: "2.4.0-test7: umount + mount = same directory (CD)"
In reply to: David Mansfield: "Re: ll_rw_blk.c fails to merge requests. Help!"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

> I have two comments about this patch. (the reqfree-batching-3)
>
> First. Couldn't it end up indefinitely starving a writer?
>
> Assume the free_list is empty [...]

I think that can also happen in 2.2.x and plain 2.4-test7.

I have a question. There is a rq queue for eache device, right ? Well, why
when I flood the queue of the removable also write accesses to my harddisc
get starved ? A simple "cp 100KBfile foobar" from HD to HD can stay
blocked in D state for 10's seconds even with a lot of free mem.
Perhaps because the queue spinlock is global ? m, no I can't see a code
path that holds the lock for a lot of time...
The HD and the removable are attached to two different SCSI controllers (2
Adaptec 2930U).

Bye.
-
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/

Next message: David Weis: "Re: 2.2.16 deadlocking in schedule()"
Previous message: Eugene Crosser: "2.4.0-test7: umount + mount = same directory (CD)"
In reply to: David Mansfield: "Re: ll_rw_blk.c fails to merge requests. Help!"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

This archive was generated by hypermail 2b29 : Wed Aug 30 2000 - 15:45:01 EDT