On 25 Nov 2000, at 2:58, Adrian Chadd <adrian@creative.net.au> wrote:
> Here is a patch against HEAD. If you assert() on the close_request with
> this patch inside storeDiskdReadDone() to die if close_request is 1,
> I can reliably triugger it by ctrl-C'ing a polyclt during a polygraph
> run.
>
> Can someone check my logic here? We only have a single op on an FS
> (read or write) scheduled on one sio at a time, so we can implement
> a cancel based on close_request ..
I'm not familiar with diskd, (aufs also not too well), but it seems
to me that we cannot safely issue close request until we have dequeued
pending read. And as this happens from callback checks, we can't really
do much other than flag pending read for cancel. We can do that at
close_request, but we shouldn't callback handlers if any disk op is
pending. We should relay on checkcallbacks to call done handlers, who
check for flags.close_request and do what we need.
------------------------------------
Andres Kroonmaa <andre@online.ee>
Delfi Online
Tel: 6501 731, Fax: 6501 708
Pärnu mnt. 158, Tallinn,
11317 Estonia
Received on Fri Nov 24 2000 - 14:54:11 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:13:00 MST