Your scenario is correct, and yes, there is a lot more going on than just
those calls. I'll see if I can get you a more detailed calling sequence.
In the mean time, I'm running with your patch on an otherwise unmodified
XFS 1.1 system. Perhaps I am missing some more recent modifications? I
could grab something more recent from CVS if you think that would help.
IBM Global Services - Federal
Phone: (281) 336 2578
Fax: (281) 335 4231
<roehrich@xxxxxxx To: James A
> cc: "Francis Qu"
Subject: Re: DMAPI bug in
dm_file_getattr() / dm_set_region
>From: "James A Goodwin" <jagoodwi@xxxxxxxxxx>
>Thanks for the patch. It gets us past the problem with dm_get_fileattr().
>However, later in the same function (called when the write event was
>received), my process hangs on dm_set_region() for the same file. Is this
>perhaps a similar problem with a similar fix?
So a user process opens a file for append, and then does a write. This
generates a write event.
The HSM calls dm_get_events to get the write event. Then it calls
dm_get_fileattr on that handle, then dm_set_region on that handle, then
finally responds to the write event with dm_respond_event--is that the
scenario? That scenario doesn't hang for me, so is there anything else
on in there?