Hi,

Here is a small series of fixes for the monitor, mostly related to
threading issues.

v2: after Peter review
 - patch 2: fix resuming with oob=off
 - patch 4: keep MUX case explicit, improve commit message

Marc-André Lureau (6):
  monitor: inline ambiguous helper functions
  monitor: accept chardev input from iothread
  char: add a QEMU_CHAR_FEATURE_GCONTEXT flag
  monitor: check if chardev can switch gcontext for OOB
  monitor: prevent inserting new monitors after cleanup
  monitor: avoid potential dead-lock when cleaning up

 include/chardev/char.h |  3 +++
 chardev/char.c         | 11 +++++++++
 monitor.c              | 51 +++++++++++++++++++++++++++---------------
 3 files changed, 47 insertions(+), 18 deletions(-)

-- 
2.19.0.271.gfe8321ec05

Marc-André Lureau <marcandre.lureau@redhat.com> writes:

> Hi,
>
> Here is a small series of fixes for the monitor, mostly related to
> threading issues.

Series looks good on first glance, but thread safety is subtle business,
so I'd like to give it a bit more time for review.  Bug fixes like these
can go in after soft freeze.

On Mon, Oct 29, 2018 at 04:57:27PM +0400, Marc-André Lureau wrote:
> Hi,
> 
> Here is a small series of fixes for the monitor, mostly related to
> threading issues.

Hi, Marc-André, Markus,

I'd be glad to know how you think about these monitor series,
considering that I just noticed it's exactly softfreeze for QEMU 3.1
today...

I gave a quick look on this series, IMHO patch 1,2,6 could be
candidate for 3.1 (though I'm not sure for patch 6; it seems even fine
to me with some comments there but I'd like to see how other people
think about it too, and whether it's really possible to insert a new
monitor at that time) and I'm not sure whether the rest can be
postponed (say, patch 3,4,5).  Meanwhile, I think we need Wolfgang's
patches to fix the daemonize chaos, and hopefully I also like to know
your thoughts on my oob enablement series (especially Marc-André,
since I talked to Markus face to face during the forum after all :).
I would like it to be in 3.1 as well but I'll see my luck.

Thoughts?

Regards,

-- 
Peter Xu

Hi

On Tue, Oct 30, 2018 at 9:48 AM Peter Xu <peterx@redhat.com> wrote:
>
> On Mon, Oct 29, 2018 at 04:57:27PM +0400, Marc-André Lureau wrote:
> > Hi,
> >
> > Here is a small series of fixes for the monitor, mostly related to
> > threading issues.
>
> Hi, Marc-André, Markus,
>
> I'd be glad to know how you think about these monitor series,
> considering that I just noticed it's exactly softfreeze for QEMU 3.1
> today...
>
> I gave a quick look on this series, IMHO patch 1,2,6 could be
> candidate for 3.1 (though I'm not sure for patch 6; it seems even fine
> to me with some comments there but I'd like to see how other people
> think about it too, and whether it's really possible to insert a new
> monitor at that time) and I'm not sure whether the rest can be

Patch 5 & 6 are the result of the discussion of the previous proposed fixed:
https://lists.nongnu.org/archive/html/qemu-devel/2018-08/msg00018.html

They fix a dead-lock on cleanup, so it should be considered for 3.1.

> postponed (say, patch 3,4,5).  Meanwhile, I think we need Wolfgang's

Patch 3 & 4 should also be considered, as OOB won't have the expected
behaviour with all chardevs. With non-gcontext feature chardev, they
will behave more or less like non-oob.

> patches to fix the daemonize chaos, and hopefully I also like to know
> your thoughts on my oob enablement series (especially Marc-André,
> since I talked to Markus face to face during the forum after all :).
> I would like it to be in 3.1 as well but I'll see my luck.

I acked the series already (except "monitor: resume the monitor
earlier if needed" which we agreed could be dropped for now), so it's
in Markus hand.

Thanks

Found only minor issues.  v3 should do the trick.  Thanks!