[Qemu-devel] [PATCH v2 00/12] icount and misc MTTCG fixes for 2.9-rc4

Alex Bennée posted 12 patches 7 years ago
Only 11 patches received!
There is a newer version of this series
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
[Qemu-devel] [PATCH v2 00/12] icount and misc MTTCG fixes for 2.9-rc4
Posted by Alex Bennée 7 years ago
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