From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:54499) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1cvkvq-0004VS-Pw for qemu-devel@nongnu.org; Wed, 05 Apr 2017 09:25:07 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1cvkvn-0000CV-LI for qemu-devel@nongnu.org; Wed, 05 Apr 2017 09:25:06 -0400 Received: from mail-wr0-x229.google.com ([2a00:1450:400c:c0c::229]:34132) by eggs.gnu.org with esmtps (TLS1.0:RSA_AES_128_CBC_SHA1:16) (Exim 4.71) (envelope-from ) id 1cvkvn-0000CH-DZ for qemu-devel@nongnu.org; Wed, 05 Apr 2017 09:25:03 -0400 Received: by mail-wr0-x229.google.com with SMTP id t20so14810844wra.1 for ; Wed, 05 Apr 2017 06:25:03 -0700 (PDT) From: =?UTF-8?q?Alex=20Benn=C3=A9e?= Date: Wed, 5 Apr 2017 14:24:51 +0100 Message-Id: <20170405132503.32125-1-alex.bennee@linaro.org> MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Subject: [Qemu-devel] [PATCH v2 00/12] icount and misc MTTCG fixes for 2.9-rc4 List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: dovgaluk@ispras.ru, rth@twiddle.net, pbonzini@redhat.com Cc: peter.maydell@linaro.org, qemu-devel@nongnu.org, mttcg@listserver.greensocs.com, fred.konrad@greensocs.com, a.rigo@virtualopensystems.com, cota@braap.org, bobby.prani@gmail.com, nikunj@linux.vnet.ibm.com, =?UTF-8?q?Alex=20Benn=C3=A9e?= Hi, Here is my current queue for icount and miscellaneous MTTCG related fixes. The main difference from the RFC series is I have dropped the graceful handling of backward time events in the replay log in favour of updating the global qemu_icount more frequently. The graceful handling is now replaced with an assertion about the arrow of time. With these changes I was able to dispenses with additional BQL locking which I had been applying during cpu_handle_exception. It doesn't completely solve the problem of keeping the main-loop and vCPU events stream in sync but it works pretty reliably in my test case. Given the current release time-line I think this is the best solution for now. Options for post-2.9 include either replacing the BQL mediated sequencing with a new lock (pushing replay_lock up the tree?) or marshalling events through a single thread which can then keep things in order. Aside from that other patches have had various r-b and a-b tags applied from the original posting. I hope to roll a pull request with these fixes by the end of the week so we there isn't a rush for next Tuesday's rc4. Expedient comments and reviews are therefor gratefully received ;-) Regards, Alex Bennée (12): scripts/qemugdb/mtree.py: fix up mtree dump scripts/qemu-gdb/timers.py: new helper to dump timer state scripts/replay-dump.py: replay log dumper target/i386/misc_helper: wrap BQL around another IRQ generator cpus: remove icount handling from qemu_tcg_cpu_thread_fn cpus: check cpu->running in cpu_get_icount_raw() cpus: move icount preparation out of tcg_exec_cpu cpus: don't credit executed instructions before they have run cpus: introduce cpu_update_icount helper cpu-exec: update icount after each TB_EXIT cpus: call cpu_update_icount on read replay: assert time only goes forward cpu-exec.c | 14 +-- cpus.c | 98 ++++++++++++----- include/qemu/timer.h | 1 + include/qom/cpu.h | 1 + replay/replay-internal.c | 4 + replay/replay.c | 4 + scripts/qemu-gdb.py | 3 +- scripts/qemugdb/mtree.py | 12 +- scripts/qemugdb/timers.py | 54 +++++++++ scripts/replay-dump.py | 272 ++++++++++++++++++++++++++++++++++++++++++++++ target/i386/misc_helper.c | 3 + 11 files changed, 426 insertions(+), 40 deletions(-) create mode 100644 scripts/qemugdb/timers.py create mode 100755 scripts/replay-dump.py -- 2.11.0