unix_listen/connect_saddr now support abstract address types
two aditional BOOL switches are introduced:
tight: whether to set @addrlen to the minimal string length,
or the maximum sun_path length. default is TRUE
abstract: whether we use abstract address. default is FALSE
cli example:
-monitor unix:/tmp/unix.socket,abstract,tight=off
OR
-chardev socket,path=/tmp/unix.socket,id=unix1,abstract,tight=on
Signed-off-by: xiaoqiang zhao <zxq_yx_007@163.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
The glib event loop does not call fdmon_io_uring_wait() so fd handlers
waiting to be submitted build up in the list. There is no benefit is
using io_uring when the glib GSource is being used, so disable it
instead of implementing a more complex fix.
This fixes a memory leak where AioHandlers would build up and increasing
amounts of CPU time were spent iterating them in aio_pending(). The
symptom is that guests become slow when QEMU is built with io_uring
support.
Buglink: https://bugs.launchpad.net/qemu/+bug/1877716
Fixes: 73fd282e7b ("aio-posix: add io_uring fd monitoring implementation")
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Tested-by: Oleksandr Natalenko <oleksandr@redhat.com>
Message-id: 20200511183630.279750-3-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The io_uring file descriptor monitoring implementation has an internal
list of fd handlers that are pending submission to io_uring.
fdmon_io_uring_destroy() deletes all fd handlers on the list.
Don't delete fd handlers directly in fdmon_io_uring_destroy() for two
reasons:
1. This duplicates the aio-posix.c AioHandler deletion code and could
become outdated if the struct changes.
2. Only handlers with the FDMON_IO_URING_REMOVE flag set are safe to
remove. If the flag is not set then something still has a pointer to
the fd handler. Let aio-posix.c and its user worry about that. In
practice this isn't an issue because fdmon_io_uring_destroy() is only
called when shutting down so all users have removed their fd
handlers, but the next patch will need this!
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Tested-by: Oleksandr Natalenko <oleksandr@redhat.com>
Message-id: 20200511183630.279750-2-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Blindly setting FD_CLOEXEC without a read-modify-write will
inadvertently clear any other intentionally-set bits, such as a
proposed new bit for designating a fd that must behave in 32-bit mode.
However, we cannot use our wrapper qemu_set_cloexec(), because that
wrapper intentionally abort()s on failure, whereas the probe here
intentionally tolerates failure to deal with incorrect socket
activation gracefully. Instead, fix the code to do the proper
read-modify-write.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <20200420175309.75894-3-eblake@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
- ran regexp "qemu_mutex_lock\(.*\).*\n.*if" to find targets
- replaced result with QEMU_LOCK_GUARD if all unlocks at function end
- replaced result with WITH_QEMU_LOCK_GUARD if unlock not at end
Signed-off-by: Daniel Brodsky <dnbrdsky@gmail.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-id: 20200404042108.389635-3-dnbrdsky@gmail.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Compress two lines into a single line if immediate return statement is found.
It also remove variables progress, val, data, ret and sock
as they are no longer needed.
Remove space between function "mixer_load" and '(' to fix the
checkpatch.pl error:-
ERROR: space prohibited between function name and open parenthesis '('
Done using following coccinelle script:
@@
local idexpression ret;
expression e;
@@
-ret =
+return
e;
-return ret;
Signed-off-by: Simran Singhal <singhalsimran0@gmail.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20200401165314.GA3213@simran-Inspiron-5558>
[lv: in handle_aiocb_write_zeroes_unmap() move "int ret" inside the #ifdef]
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20200415083048.14339-6-armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
qdict_iter() has just three uses and no test coverage. Replace by
qdict_first(), qdict_next() for more concise code and less type
punning.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20200415083048.14339-5-armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
is_valid_option_list()'s purpose is ensuring qemu-img.c's can safely
join multiple parameter strings separated by ',' like this:
g_strdup_printf("%s,%s", params1, params2);
How it does that is anything but obvious. A close reading of the code
reveals that it fails exactly when its argument starts with ',' or
ends with an odd number of ','. Makes sense, actually, because when
the argument starts with ',', a separating ',' preceding it would get
escaped, and when it ends with an odd number of ',', a separating ','
following it would get escaped.
Move it to qemu-img.c and rewrite it the obvious way.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20200415074927.19897-9-armbru@redhat.com>
When opts_parse() sets @invalidp to true, qemu_opts_parse_noisily()
uses has_help_option() to decide whether to print help. This parses
the input string a second time.
Easy to avoid: replace @invalidp by @help_wanted.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20200415074927.19897-7-armbru@redhat.com>
has_help_option() uses its own parser. It's inconsistent with
qemu_opts_parse(), as demonstrated by test-qemu-opts case
/qemu-opts/has_help_option. Fix by reusing the common parser.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20200415074927.19897-5-armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20200415074927.19897-4-armbru@redhat.com>
The next commits will put it to use.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20200415074927.19897-3-armbru@redhat.com>
With the module upgrades code change, the statically sized dirs array
can now overflow. Increase it's size by one, according to the new
maximum possible usage.
Fixes: bd83c861c0 ("modules: load modules from versioned /var/run dir")
Signed-off-by: Bruce Rogers <brogers@suse.com>
Message-Id: <20200411010746.472295-1-brogers@suse.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
In touch_all_pages, if the mutex is not taken around qemu_cond_broadcast,
qemu_cond_broadcast may be called before all touch page threads enter
qemu_cond_wait. In this case, the touch page threads wait forever for the
main thread to wake them up, causing a deadlock.
Signed-off-by: Bauerchen <bauerchen@tencent.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
When using C11 atomics, non-seqcst reads and writes do not participate
in the total order of seqcst operations. In util/async.c and util/aio-posix.c,
in particular, the pattern that we use
write ctx->notify_me write bh->scheduled
read bh->scheduled read ctx->notify_me
if !bh->scheduled, sleep if ctx->notify_me, notify
needs to use seqcst operations for both the write and the read. In
general this is something that we do not want, because there can be
many sources that are polled in addition to bottom halves. The
alternative is to place a seqcst memory barrier between the write
and the read. This also comes with a disadvantage, in that the
memory barrier is implicit on strongly-ordered architectures and
it wastes a few dozen clock cycles.
Fortunately, ctx->notify_me is never written concurrently by two
threads, so we can assert that and relax the writes to ctx->notify_me.
The resulting solution works and performs well on both aarch64 and x86.
Note that the atomic_set/atomic_read combination is not an atomic
read-modify-write, and therefore it is even weaker than C11 ATOMIC_RELAXED;
on x86, ATOMIC_RELAXED compiles to a locked operation.
Analyzed-by: Ying Fang <fangying1@huawei.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Tested-by: Ying Fang <fangying1@huawei.com>
Message-Id: <20200407140746.8041-6-pbonzini@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The io_uring_enter(2) syscall returns with errno=EINTR when interrupted
by a signal. Retry the syscall in this case.
It's essential to do this in the io_uring_submit_and_wait() case. My
interpretation of the Linux v5.5 io_uring_enter(2) code is that it
shouldn't affect the io_uring_submit() case, but there is no guarantee
this will always be the case. Let's check for -EINTR around both APIs.
Note that the liburing APIs have -errno return values.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-id: 20200408091139.273851-1-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Unfortunately reading /proc/self/maps is still considered the gold
standard for a process finding out about it's own memory layout. As we
will want this data in other contexts soon factor out the code to read
and parse the data. Rather than just blindly copying the existing
sscanf based code we use a more modern glib version of the parsing
code to make a more general purpose map structure.
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20200403191150.863-9-alex.bennee@linaro.org>
When a file descriptor becomes ready we must re-arm POLL_ADD. This is
done by adding an sqe to the io_uring sq ring. The ->need_wait()
function wasn't taking pending sqes into account and therefore
io_uring_submit_and_wait() was not being called. Polling for cqes
failed to detect fd readiness since we hadn't submitted the sqe to
io_uring.
This patch fixes the following tests/test-aio -p /aio/event/wait
failure:
ok 11 /aio/event/wait
**
ERROR:tests/test-aio.c:374:test_flush_event_notifier: assertion failed: (aio_poll(ctx, false))
Reported-by: Cole Robinson <crobinso@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Tested-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 20200402145434.99349-1-stefanha@redhat.com
Fixes: 73fd282e7b
("aio-posix: add io_uring fd monitoring implementation")
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
By increasing avx2 length_to_accel to 128, we can simplify its logic and reduce a
branch.
The authorship of this patch actually belongs to Richard Henderson
<richard.henderson@linaro.org>, I just fixed a boundary case on his
original patch.
Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Robert Hoo <robert.hu@linux.intel.com>
Message-Id: <1585119021-46593-2-git-send-email-robert.hu@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Because in unit test, init_accel() will be called several times, each with
different accelerator type.
Signed-off-by: Robert Hoo <robert.hu@linux.intel.com>
Message-Id: <1585119021-46593-1-git-send-email-robert.hu@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
When external event sources are disabled fdmon-io_uring falls back to
fdmon-poll. The ->need_wait() callback needs to watch for this so it
can return true when external event sources are disabled.
It is also necessary to call ->wait() when AioHandlers have changed
because io_uring is asynchronous and we must submit new sqes.
Both of these changes to ->need_wait() together fix tests/test-aio -p
/aio/external-client, which failed with:
test-aio: tests/test-aio.c:404: test_aio_external_client: Assertion `aio_poll(ctx, false)' failed.
Reported-by: Julia Suvorova <jusual@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Message-id: 20200319163559.117903-1-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Firstly, _next_dirty_area is for scenarios when we may contiguously
search for next dirty area inside some limited region, so it is more
comfortable to specify "end" which should not be recalculated on each
iteration.
Secondly, let's add a possibility to limit resulting area size, not
limiting searching area. This will be used in NBD code in further
commit. (Note that now bdrv_dirty_bitmap_next_dirty_area is unused)
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Message-id: 20200205112041.6003-8-vsementsov@virtuozzo.com
Signed-off-by: John Snow <jsnow@redhat.com>
We have bdrv_dirty_bitmap_next_zero, let's add corresponding
bdrv_dirty_bitmap_next_dirty, which is more comfortable to use than
bitmap iterators in some cases.
For test modify test_hbitmap_next_zero_check_range to check both
next_zero and next_dirty and add some new checks.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Message-id: 20200205112041.6003-7-vsementsov@virtuozzo.com
Signed-off-by: John Snow <jsnow@redhat.com>
We are going to introduce bdrv_dirty_bitmap_next_dirty so that same
variable may be used to store its return value and to be its parameter,
so it would int64_t.
Similarly, we are going to refactor hbitmap_next_dirty_area to use
hbitmap_next_dirty together with hbitmap_next_zero, therefore we want
hbitmap_next_zero parameter type to be int64_t too.
So, for convenience update all parameters of *_next_zero and
*_next_dirty_area to be int64_t.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Message-id: 20200205112041.6003-6-vsementsov@virtuozzo.com
Signed-off-by: John Snow <jsnow@redhat.com>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Message-id: 20200205112041.6003-5-vsementsov@virtuozzo.com
Signed-off-by: John Snow <jsnow@redhat.com>
Function is internal and even commented as internal. Drop its
definition from .h file.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Message-id: 20200205112041.6003-4-vsementsov@virtuozzo.com
Signed-off-by: John Snow <jsnow@redhat.com>
The function is definitely internal (it's not used by third party and
it has complicated interface). Move it to .c file.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Message-id: 20200205112041.6003-3-vsementsov@virtuozzo.com
Signed-off-by: John Snow <jsnow@redhat.com>
We have APIs which returns signed int64_t, to be able to return error.
Therefore we can't handle bitmaps with absolute size larger than
(INT64_MAX+1). Still, keep maximum to be INT64_MAX which is a bit
safer.
Note, that bitmaps are used to represent disk images, which can't
exceed INT64_MAX anyway.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Message-id: 20200205112041.6003-2-vsementsov@virtuozzo.com
Signed-off-by: John Snow <jsnow@redhat.com>
This patch introduces two lock guard macros that automatically unlock a
lock object (QemuMutex and others):
void f(void) {
QEMU_LOCK_GUARD(&mutex);
if (!may_fail()) {
return; /* automatically unlocks mutex */
}
...
}
and:
WITH_QEMU_LOCK_GUARD(&mutex) {
if (!may_fail()) {
return; /* automatically unlocks mutex */
}
}
/* automatically unlocks mutex here */
...
Convert qemu-timer.c functions that benefit from these macros as an
example. Manual qemu_mutex_lock/unlock() callers are left unmodified in
cases where clarity would not improve by switching to the macros.
Many other QemuMutex users remain in the codebase that might benefit
from lock guards. Over time they can be converted, if that is
desirable.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
[Use QEMU_MAKE_LOCKABLE_NONNULL. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
On upgrades the old .so files usually are replaced. But on the other
hand since a qemu process represents a guest instance it is usually kept
around.
That makes late addition of dynamic features e.g. 'hot-attach of a ceph
disk' fail by trying to load a new version of e.f. block-rbd.so into an
old still running qemu binary.
This adds a fallback to also load modules from a versioned directory in the
temporary /var/run path. That way qemu is providing a way for packaging
to store modules of an upgraded qemu package as needed until the next reboot.
An example how that can then be used in packaging can be seen in:
https://git.launchpad.net/~paelzer/ubuntu/+source/qemu/log/?h=bug-1847361-miss-old-so-on-upgrade-UBUNTU
Fixes: https://bugs.launchpad.net/ubuntu/+source/qemu/+bug/1847361
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20200310145806.18335-2-christian.ehrhardt@canonical.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The mutex and condition variable were never initialized, causing
-mem-prealloc to abort with an assertion failure.
Fixes: 037fb5eb39
Reported-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Cc: bauerchen <bauerchen@tencent.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
And intialize buffer_is_zero() with it, when Intel AVX512F is
available on host.
This function utilizes Intel AVX512 fundamental instructions which
is faster than its implementation with AVX2 (in my unit test, with
4K buffer, on CascadeLake SP, ~36% faster, buffer_zero_avx512() V.S.
buffer_zero_avx2()).
Signed-off-by: Robert Hoo <robert.hu@linux.intel.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
When there are many poll handlers it's likely that some of them are idle
most of the time. Remove handlers that haven't had activity recently so
that the polling loop scales better for guests with a large number of
devices.
This feature only takes effect for the Linux io_uring fd monitoring
implementation because it is capable of combining fd monitoring with
userspace polling. The other implementations can't do that and risk
starving fds in favor of poll handlers, so don't try this optimization
when they are in use.
IOPS improves from 10k to 105k when the guest has 100
virtio-blk-pci,num-queues=32 devices and 1 virtio-blk-pci,num-queues=1
device for rw=randread,iodepth=1,bs=4k,ioengine=libaio on NVMe.
[Clarified aio_poll_handlers locking discipline explanation in comment
after discussion with Paolo Bonzini <pbonzini@redhat.com>.
--Stefan]
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Link: https://lore.kernel.org/r/20200305170806.1313245-8-stefanha@redhat.com
Message-Id: <20200305170806.1313245-8-stefanha@redhat.com>
Unlike ppoll(2) and epoll(7), Linux io_uring completions can be polled
from userspace. Previously userspace polling was only allowed when all
AioHandler's had an ->io_poll() callback. This prevented starvation of
fds by userspace pollable handlers.
Add the FDMonOps->need_wait() callback that enables userspace polling
even when some AioHandlers lack ->io_poll().
For example, it's now possible to do userspace polling when a TCP/IP
socket is monitored thanks to Linux io_uring.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Link: https://lore.kernel.org/r/20200305170806.1313245-7-stefanha@redhat.com
Message-Id: <20200305170806.1313245-7-stefanha@redhat.com>
The recent Linux io_uring API has several advantages over ppoll(2) and
epoll(2). Details are given in the source code.
Add an io_uring implementation and make it the default on Linux.
Performance is the same as with epoll(7) but later patches add
optimizations that take advantage of io_uring.
It is necessary to change how aio_set_fd_handler() deals with deleting
AioHandlers since removing monitored file descriptors is asynchronous in
io_uring. fdmon_io_uring_remove() marks the AioHandler deleted and
aio_set_fd_handler() will let it handle deletion in that case.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Link: https://lore.kernel.org/r/20200305170806.1313245-6-stefanha@redhat.com
Message-Id: <20200305170806.1313245-6-stefanha@redhat.com>
The AioHandler *node, bool is_new arguments are more complicated to
think about than simply being given AioHandler *old_node, AioHandler
*new_node.
Furthermore, the new Linux io_uring file descriptor monitoring mechanism
added by the new patch requires access to both the old and the new
nodes. Make this change now in preparation.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Link: https://lore.kernel.org/r/20200305170806.1313245-5-stefanha@redhat.com
Message-Id: <20200305170806.1313245-5-stefanha@redhat.com>
The ppoll(2) and epoll(7) file descriptor monitoring implementations are
mixed with the core util/aio-posix.c code. Before adding another
implementation for Linux io_uring, extract out the existing
ones so there is a clear interface and the core code is simpler.
The new interface is AioContext->fdmon_ops, a pointer to a FDMonOps
struct. See the patch for details.
Semantic changes:
1. ppoll(2) now reflects events from pollfds[] back into AioHandlers
while we're still on the clock for adaptive polling. This was
already happening for epoll(7), so if it's really an issue then we'll
need to fix both in the future.
2. epoll(7)'s fallback to ppoll(2) while external events are disabled
was broken when the number of fds exceeded the epoll(7) upgrade
threshold. I guess this code path simply wasn't tested and no one
noticed the bug. I didn't go out of my way to fix it but the correct
code is simpler than preserving the bug.
I also took some liberties in removing the unnecessary
AioContext->epoll_available (just check AioContext->epollfd != -1
instead) and AioContext->epoll_enabled (it's implicit if our
AioContext->fdmon_ops callbacks are being invoked) fields.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Link: https://lore.kernel.org/r/20200305170806.1313245-4-stefanha@redhat.com
Message-Id: <20200305170806.1313245-4-stefanha@redhat.com>
Now that run_poll_handlers_once() is only called by run_poll_handlers()
we can improve the CPU time profile by moving the expensive
RCU_READ_LOCK() out of the polling loop.
This reduces the run_poll_handlers() from 40% CPU to 10% CPU in perf's
sampling profiler output.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Link: https://lore.kernel.org/r/20200305170806.1313245-3-stefanha@redhat.com
Message-Id: <20200305170806.1313245-3-stefanha@redhat.com>
One iteration of polling is always performed even when polling is
disabled. This is done because:
1. Userspace polling is cheaper than making a syscall. We might get
lucky.
2. We must poll once more after polling has stopped in case an event
occurred while stopping polling.
However, there are downsides:
1. Polling becomes a bottleneck when the number of event sources is very
high. It's more efficient to monitor fds in that case.
2. A high-frequency polling event source can starve non-polling event
sources because ppoll(2)/epoll(7) is never invoked.
This patch removes the forced polling iteration so that poll_ns=0 really
means no polling.
IOPS increases from 10k to 60k when the guest has 100
virtio-blk-pci,num-queues=32 devices and 1 virtio-blk-pci,num-queues=1
device because the large number of event sources being polled slows down
the event loop.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Link: https://lore.kernel.org/r/20200305170806.1313245-2-stefanha@redhat.com
Message-Id: <20200305170806.1313245-2-stefanha@redhat.com>
QLIST_SAFE_REMOVE() is confusing here because the node must be on the
list. We actually just wanted to clear the linked list pointers when
removing it from the list. QLIST_REMOVE() now does this, so switch to
it.
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Link: https://lore.kernel.org/r/20200224103406.1894923-3-stefanha@redhat.com
Message-Id: <20200224103406.1894923-3-stefanha@redhat.com>
Use error_setg_win32() which adds a hint similar to strerror(errno)).
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20200228100726.8414-3-philmd@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
[desc]:
Large memory VM starts slowly when using -mem-prealloc, and
there are some areas to optimize in current method;
1、mmap will be used to alloc threads stack during create page
clearing threads, and it will attempt mm->mmap_sem for write
lock, but clearing threads have hold read lock, this competition
will cause threads createion very slow;
2、methods of calcuating pages for per threads is not well;if we use
64 threads to split 160 hugepage,63 threads clear 2page,1 thread
clear 34 page,so the entire speed is very slow;
to solve the first problem,we add a mutex in thread function,and
start all threads when all threads finished createion;
and the second problem, we spread remainder to other threads,in
situation that 160 hugepage and 64 threads, there are 32 threads
clear 3 pages,and 32 threads clear 2 pages.
[test]:
320G 84c VM start time can be reduced to 10s
680G 84c VM start time can be reduced to 18s
Signed-off-by: bauerchen <bauerchen@tencent.com>
Reviewed-by: Pan Rui <ruippan@tencent.com>
Reviewed-by: Ivan Ren <ivanren@tencent.com>
[Simplify computation of the number of pages per thread. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The virtual-device fuzzer must initialize QOM, prior to running
vl:qemu_init, so that it can use the qos_graph to identify the arguments
required to initialize a guest for libqos-assisted fuzzing. This change
prevents errors when vl:qemu_init tries to (re)initialize the previously
initialized QOM module.
Signed-off-by: Alexander Bulekov <alxndr@bu.edu>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Darren Kenny <darren.kenny@oracle.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-id: 20200220041118.23264-4-alxndr@bu.edu
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
File descriptor monitoring is O(1) with epoll(7), but
aio_dispatch_handlers() still scans all AioHandlers instead of
dispatching just those that are ready. This makes aio_poll() O(n) with
respect to the total number of registered handlers.
Add a local ready_list to aio_poll() so that each nested aio_poll()
builds a list of handlers ready to be dispatched. Since file descriptor
polling is level-triggered, nested aio_poll() calls also see fds that
were ready in the parent but not yet dispatched. This guarantees that
nested aio_poll() invocations will dispatch all fds, even those that
became ready before the nested invocation.
Since only handlers ready to be dispatched are placed onto the
ready_list, the new aio_dispatch_ready_handlers() function provides O(1)
dispatch.
Note that AioContext polling is still O(n) and currently cannot be fully
disabled. This still needs to be fixed before aio_poll() is fully O(1).
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Message-id: 20200214171712.541358-6-stefanha@redhat.com
[Fix compilation error on macOS where there is no epoll(87). The
aio_epoll() prototype was out of date and aio_add_ready_list() needed to
be moved outside the ifdef.
--Stefan]
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
It is not necessary to scan all AioHandlers for deletion. Keep a list
of deleted handlers instead of scanning the full list of all handlers.
The AioHandler->deleted field can be dropped. Let's check if the
handler has been inserted into the deleted list instead. Add a new
QLIST_IS_INSERTED() API for this check.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Message-id: 20200214171712.541358-5-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Don't pass the nanosecond timeout into epoll_wait(), which expects
milliseconds.
The epoll_wait() timeout value does not matter if qemu_poll_ns()
determined that the poll fd is ready, but passing a value in the wrong
units is still ugly. Pass a 0 timeout to epoll_wait() instead.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Message-id: 20200214171712.541358-3-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
epoll_handler is a stack variable and must not be accessed after it goes
out of scope:
if (aio_epoll_check_poll(ctx, pollfds, npfd, timeout)) {
AioHandler epoll_handler;
...
add_pollfd(&epoll_handler);
ret = aio_epoll(ctx, pollfds, npfd, timeout);
} ...
...
/* if we have any readable fds, dispatch event */
if (ret > 0) {
for (i = 0; i < npfd; i++) {
nodes[i]->pfd.revents = pollfds[i].revents;
}
}
nodes[0] is &epoll_handler, which has already gone out of scope.
There is no need to use pollfds[] for epoll. We don't need an
AioHandler for the epoll fd.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Message-id: 20200214171712.541358-2-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The ctx->first_bh list contains all created BHs, including those that
are not scheduled. The list is iterated by the event loop and therefore
has O(n) time complexity with respected to the number of created BHs.
Rewrite BHs so that only scheduled or deleted BHs are enqueued.
Only BHs that actually require action will be iterated.
One semantic change is required: qemu_bh_delete() enqueues the BH and
therefore invokes aio_notify(). The
tests/test-aio.c:test_source_bh_delete_from_cb() test case assumed that
g_main_context_iteration(NULL, false) returns false after
qemu_bh_delete() but it now returns true for one iteration. Fix up the
test case.
This patch makes aio_compute_timeout() and aio_bh_poll() drop from a CPU
profile reported by perf-top(1). Previously they combined to 9% CPU
utilization when AioContext polling is commented out and the guest has 2
virtio-blk,num-queues=1 and 99 virtio-blk,num-queues=32 devices.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 20200221093951.1414693-1-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The first rcu_read_lock/unlock() is expensive. Nested calls are cheap.
This optimization increases IOPS from 73k to 162k with a Linux guest
that has 2 virtio-blk,num-queues=1 and 99 virtio-blk,num-queues=32
devices.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 20200218182708.914552-1-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
xiaoqiang zhao
Stefan Hajnoczi
Eric Blake
Daniel Brodsky
Simran Singhal
Markus Armbruster
Bruce Rogers
Bauerchen
Paolo Bonzini
Alex Bennée
Robert Hoo
Vladimir Sementsov-Ogievskiy
Christian Ehrhardt
Philippe Mathieu-Daudé
Alexander Bulekov