qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] [PATCH 02/11] migration: qemu_savevm_state_cleanup() in


From: Peter Xu
Subject: Re: [Qemu-devel] [PATCH 02/11] migration: qemu_savevm_state_cleanup() in cleanup
Date: Wed, 3 Jan 2018 19:39:45 +0800
User-agent: Mutt/1.9.1 (2017-09-22)

On Wed, Jan 03, 2018 at 12:18:54PM +0100, Juan Quintela wrote:
> Peter Xu <address@hidden> wrote:
> > On Wed, Jan 03, 2018 at 11:21:31AM +0100, Juan Quintela wrote:
> >> Peter Xu <address@hidden> wrote:
> >> > On Wed, Jan 03, 2018 at 10:15:41AM +0100, Juan Quintela wrote:
> >> >> Peter Xu <address@hidden> wrote:
> >> >> > Moving existing callers all into migrate_fd_cleanup().  It simplifies
> >> >> > migration_thread() a bit.
> >> >> >
> >> >> > Signed-off-by: Peter Xu <address@hidden>
> >> >> 
> >> >> Reviewed-by: Juan Quintela <address@hidden>
> >> >
> >> > Thanks.
> >> >
> >> >> 
> >> >> I am trying to see if we can call migrate_fd_cleanup() twice.  As far as
> >> >> I can see, we are not doing it.  But, and it is a big but, we are not
> >> >> checking that we are not calling qemu_savevm_state_cleanup() twice.  If
> >> >> that happens, we can get double frees and similar.
> >> >> 
> >> >> I put the reviewed-by anyways, because I *think* that we are doing it
> >> >> right now, and otherwise, we should make sure that we are not calling it
> >> >> twice, not papering over it.
> >> >> 
> >> >> Once here, I have notice that we call block_cleanup_parameters() in
> >> >> *three* places.  We call notifier_list_notify() on two of this places (I
> >> >> can't see any good reason *why* we don't call the notifier for
> >> >> migrate_fd_cancel).
> >> >
> >> > Indeed.
> >> >
> >> > IMHO we can remove two calls of block_cleanup_parameters(), only keep
> >> > the one in migrate_fd_cleanup(), and remove on notifier_list_notify()
> >> > in migrate_fd_error() (these can be two more patches).  What do you
> >> > think?
> >> 
> >> I think we need to make sure that we have a function that we always
> >> call at the end.  I think that we have that on migration_fd_cleanup(),
> >> so put everything there should be ok, no?
> >
> > IMHO that's exactly what I mean, no? :)
> >
> > For notifier_list_notify(), it's different - I just remove the extra
> > one in migrate_fd_error() because it'll be called in
> > migrate_fd_cleanup() as well, which is a duplicate.
> 
> then what call the one when we do a cancel?  the one in cleanup also?

Yes, IIUC migrate_fd_cancel() tells the thread to cancel, then the
migrate_fd_cleanup() is called there (finally in the bottom half of
main loop, of course).

-- 
Peter Xu



reply via email to

[Prev in Thread] Current Thread [Next in Thread]