The goal of this module is to profile synchronization primitives (i.e.
mutexes, recursive mutexes and condition variables) so that scalability
issues can be quickly diagnosed.
Sync primitives are profiled by QSP based on the vaddr of the object accessed
as well as the call site (file:line_nr). That means the same object called
from two different call sites will be tracked in separate entries, which
might be reported together or separately (see subsequent commit on
call site coalescing).
Some perf numbers:
Host: Intel(R) Core(TM) i7-6700K CPU @ 4.00GHz
Command: taskset -c 0 tests/atomic_add-bench -d 5 -m
- Before: 54.80 Mops/s
- After: 54.75 Mops/s
That is, a negligible slowdown due to the now indirect call to
qemu_mutex_lock. Note that using a branch instead of an indirect
call introduces a more severe slowdown (53.65 Mops/s, i.e. 2% slowdown).
Enabling the profiler (with -p, added in this series) is more interesting:
- No profiling: 54.75 Mops/s
- W/ profiling: 12.53 Mops/s
That is, a 4.36X slowdown.
We can break down this slowdown by removing the get_clock calls or
the entry lookup:
- No profiling: 54.75 Mops/s
- W/o get_clock: 25.37 Mops/s
- W/o entry lookup: 19.30 Mops/s
- W/ profiling: 12.53 Mops/s
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
We have had some tracing tools for mutex but it's not easy to use them
for e.g. dead locks. Let's provide "--enable-debug-mutex" parameter
when configure to allow QemuMutex to store the last owner that took
specific lock. It will be easy to use this tool to debug deadlocks
since we can directly know who took the lock then as long as we can have
a debugger attached to the process.
Reviewed-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180425025459.5258-4-peterx@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Not all platforms check whether a lock is initialized before used. In
particular Linux seems to be more permissive than OSX.
Check initialization state explicitly in our code to catch such bugs
earlier.
Signed-off-by: Fam Zheng <famz@redhat.com>
Message-Id: <20170704122325.25634-1-famz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The multithreaded TCG implementation exposed deadlocks in the win32
condition variables: as implemented, qemu_cond_broadcast waited on
receivers, whereas the pthreads API it was intended to emulate does
not. This was causing a deadlock because broadcast was called while
holding the IO lock, as well as all possible waiters blocked on the
same lock.
This patch replaces all the custom synchronisation code for mutexes
and condition variables with native Windows primitives (SRWlocks and
condition variables) with the same semantics as their POSIX
equivalents. To enable that, it requires a Windows Vista or newer host
OS.
Signed-off-by: Andrey Shedel <ashedel@microsoft.com>
[AB: edited commit message]
Signed-off-by: Andrew Baumann <Andrew.Baumann@microsoft.com>
Message-Id: <20170324220141.10104-1-Andrew.Baumann@microsoft.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
GRecMutex is new in glib 2.32, so we cannot use it. Introduce
a recursive mutex in qemu-thread instead, which will be used
instead of RFifoLock.
Reviewed-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <1477565348-5458-20-git-send-email-pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>
Tracked down with an ugly, brittle and probably buggy Perl script.
Also move includes converted to <...> up so they get included before
ours where that's obviously okay.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Tested-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
QemuEvents are used heavily by call_rcu. We do not want them to be slow,
but the current implementation does a kernel call on every invocation
of qemu_event_* and won't cut it.
So, wrap a Win32 manual-reset event with a fast userspace path. The
states and transitions are the same as for the futex and mutex/condvar
implementations, but the slow path is different of course. The idea
is to reset the Win32 event lazily, as part of a test-reset-test-wait
sequence. Such a sequence is, indeed, how QemuEvents are used by
RCU and other subsystems!
The patch includes a formal model of the algorithm.
Tested-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefan Weil <sw@weilnetz.de>
This emulates Win32 manual-reset events using futexes or conditional
variables. Typical ways to use them are with multi-producer,
single-consumer data structures, to test for a complex condition whose
elements come from different threads:
for (;;) {
qemu_event_reset(ev);
... test complex condition ...
if (condition is true) {
break;
}
qemu_event_wait(ev);
}
Or more efficiently (but with some duplication):
... evaluate condition ...
while (!condition) {
qemu_event_reset(ev);
... evaluate condition ...
if (!condition) {
qemu_event_wait(ev);
... evaluate condition ...
}
}
QemuEvent provides a very fast userspace path in the common case when
no other thread is waiting, or the event is not changing state.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
