Indeed, i did not add those for the cover letter (i used the wiki's
commit hook though)
Anyway, I sent a V2 after discovering the rcu usage hada significant
performance impact.
It was not sent to the New email fo Fam Zheng though. i'll forward it to
its new address, thanks.
Regards.
Remy
On 12/7/18 11:09 AM, Philippe Mathieu-Daudé wrote:
Hi Remy,
On 11/16/18 8:02 PM, remy.n...@blade-group.com wrote:
From: Remy Noel <remy.n...@blade-group.com>
It is possible for an io_poll callback to be concurrently executed along
with an aio_set_fd_handlers. This can cause all sorts of problems, like
a NULL callback or a bad opaque pointer.
We fixes that by using an remove/insert RCU scheme.
You forgot to Cc the relevant maintainers, that's probably why your
series went unnoticed.
From the wiki:
https://wiki.qemu.org/Contribute/SubmitAPatch#CC_the_relevant_maintainer
$ ./scripts/get_maintainer.pl -f include/block/aio.h util/aio-posix.c
util/aio-win32.c util/async.c
Stefan Hajnoczi <stefa...@redhat.com> (supporter:Block I/O path)
Fam Zheng <f...@redhat.com> (supporter:Block I/O path)
Kevin Wolf <kw...@redhat.com> (supporter:Block layer core)
Max Reitz <mre...@redhat.com> (supporter:Block layer core)
Stefan Weil <s...@weilnetz.de> (maintainer:W32, W64)
qemu-bl...@nongnu.org (open list:Block I/O path)
qemu-devel@nongnu.org (open list:All patches CC here)
Please Cc them in your v2.
(Note Fam has a new email: Fam Zheng <f...@euphon.net>)
Regards,
Phil.
Please note that i did not test the win32 version.
Remy Noel (3):
aio-posix: Fix concurrent aio_poll/set_fd_handler.
util/aio-posix: Use RCU for handler insertion.
aio: Do not use list_lock as a sync mechanism for aio_handlers
anymore.
include/block/aio.h | 4 +-
util/aio-posix.c | 148 ++++++++++++++++++++------------------------
util/aio-win32.c | 82 +++++++++---------------
util/async.c | 7 ++-
4 files changed, 103 insertions(+), 138 deletions(-)
