Re: Design-Question: end_that_request_* and bh->b_end_io hooks

From: Jens Axboe (axboe@suse.de)
Date: Thu Jul 26 2001 - 01:40:04 EST


On Tue, Jul 24 2001, Carsten Otte wrote:
> Hi Folks,
>
> as you are deeper into block-devices & filesystems than me,
> here are my two simple questions in short:
> Is it legal for a filesystem (or whatever) to install a hook into
> bh->b_end_io
> which calls generic_make_request?

No, b_end_io might be called from irq context (IDE for instance) which
will break for __make_request (both the _irq spin locks and the schedule
on request slot empty).

You could do bh stacking and defer stuff like this to a thread, that's
probably the way to go.

> Do block drivers need or are they allowed to hold the io_request_lock or
> other (local) locks when calling end_that_request_*?

Yes they may, in fact they _must_ hold it for end_that_request_last.
Look at blkdev_release_request -- it meddles with the queue free and
pending lists and must be protected against reentrancy.

-- 
Jens Axboe

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Tue Jul 31 2001 - 21:00:25 EST