Older versions of Clang (before 3.5) and GCC (before 4.1) do not
support the "__attribute__((flatten))" yet. We don't care about
such old versions of GCC anymore, but since Clang 3.4 is still
used in EPEL for RHEL7 / CentOS 7, we should not use this attribute
directly but with a wrapper macro instead.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Thomas Huth <thuth@redhat.com>
If QEMU is compiled with clang-7 it results in the warning:
hw/display/qxl.c:1884:19: error: misaligned or large atomic operation
may incur significant performance penalty [-Werror,-Watomic-alignment]
old_pending = atomic_fetch_or(&d->ram->int_pending, le_events);
^
This is because the Spice headers forgot to define the QXLRam struct
with the '__aligned__(4)' attribute. clang 7 and newer will thus
warn that the access here to int_pending might not be 4-aligned
(because the QXLRam object d->ram points at might start at a
misaligned address). In fact we set up d->ram in init_qxl_ram() so
it always starts at a 4K boundary, so we know the atomic access here
is OK.
Newer Spice versions (with Spice commit
beda5ec7a6848be20c0cac2a9a8ef2a41e8069c1) will fix the bug;
for older Spice versions, work around it by telling the compiler
explicitly that the alignment is OK using __builtin_assume_aligned().
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180927155538.699-1-peter.maydell@linaro.org
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
This patch moves definitions of Windows dump structures to
include/qemu/win_dump_defs.h to keep create_win_dump() prototype separate.
Signed-off-by: Viktor Prutyanov <viktor.prutyanov@virtuozzo.com>
Message-Id: <1535546488-30208-2-git-send-email-viktor.prutyanov@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Slirp and VNC modules use virtual clock for processing some events that
are related to the guest execution speed.
But virtual clock-related events are consideres to be deterministic and
are recorded/replayed by icount mechanism. But slirp and VNC lie outside
the recorded guest core (which includes CPU and peripherals).
Therefore slirp and VNC are external for the guest, but should work at
guest speed.
This patch introduces new virtual clock which can be used for external
subsystems for running timers that are synchronized with the guest.
Signed-off-by: Pavel Dovgalyuk <Pavel.Dovgaluk@ispras.ru>
Message-Id: <20180912082002.3228.82417.stgit@pasha-VirtualBox>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Run some memfd-related checks before registering hostmem-memfd &
various properties. This will help libvirt to figure out what the host
is supposed to be capable of.
qemu_memfd_check() is changed to a less optimized version, since it is
used with various flags, it no longer caches the result.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20180906161415.8543-1-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This introduces read/set accessors for int64_t and uint64_t.
Signed-off-by: Emilio G. Cota <cota@braap.org>
Message-Id: <20180910232752.31565-3-cota@braap.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Message-Id: <20180903171831.15446-4-cota@braap.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Coverity does not see anymore that qemu_mutex_lock is taking a lock.
Hide all the QSP magic so that static analysis works again.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Adding a lookup table for the powers of two, with the appropriate size
prefixes. This is needed when a size has to be stringified, in which
case something like '(1 * KiB)' would become a literal '(1 * (1L << 10))'
string. Powers of two are used very often for sizes, so such a table
will also make it easier and more intuitive to write them.
This table is generatred using the following AWK script:
BEGIN {
suffix="KMGTPE";
for(i=10; i<64; i++) {
val=2**i;
s=substr(suffix, int(i/10), 1);
n=2**(i%10);
pad=21-int(log(n)/log(10));
printf("#define S_%d%siB %*d\n", n, s, pad, val);
}
}
Signed-off-by: Leonid Bloch <lbloch@janustech.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
seqlock_read_begin takes a const param since c04649eeea
("seqlock: constify seqlock_read_begin", 2018-08-23), so
we can constify the entire lookup.
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Accessing the HT from an iterator results almost always
in a deadlock. Given that only one qht-internal function
uses this argument, drop it from the interface.
Suggested-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
This currently has no users, but the use case is so common that I
think we must support it.
Note that without the appended we cannot safely remove a set of
elements; a 2-step approach (i.e. qht_iter first, keep track of
the to-be-deleted elements, and then a bunch of qht_remove calls)
would be racy, since between the iteration and the removals other
threads might insert additional elements.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Block jobs claim in .drained_poll() that they are in a quiescent state
as soon as job->deferred_to_main_loop is true. This is obviously wrong,
they still have a completion BH to run. We only get away with this
because commit 91af091f92 added an unconditional aio_poll(false) to the
drain functions, but this is bypassing the regular drain mechanisms.
However, just removing this and telling that the job is still active
doesn't work either: The completion callbacks themselves call drain
functions (directly, or indirectly with bdrv_reopen), so they would
deadlock then.
As a better lie, tell that the job is active as long as the BH is
pending, but falsely call it quiescent from the point in the BH when the
completion callback is called. At this point, nested drain calls won't
deadlock because they ignore the job, and outer drains will wait for the
job to really reach a quiescent state because the callback is already
running.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
bdrv_do_drained_begin/end() assume that they are called with the
AioContext lock of bs held. If we call drain functions from a coroutine
with the AioContext lock held, we yield and schedule a BH to move out of
coroutine context. This means that the lock for the home context of the
coroutine is released and must be re-acquired in the bottom half.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
All callers in QEMU proper hold the AioContext lock when calling
job_finish_sync(). test-blockjob should do the same when it calls the
function indirectly through job_cancel_sync().
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
In the context of draining a BDS, the .drained_poll callback of block
jobs is called. If this returns true (i.e. there is still some activity
pending), the drain operation may call aio_poll() with blocking=true to
wait for completion.
As soon as the pending activity is completed and the job finally arrives
in a quiescent state (i.e. its coroutine either yields with busy=false
or terminates), the block job must notify the aio_poll() loop to wake
up, otherwise we get a deadlock if both are running in different
threads.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Now that all of the jobs use the component finalization callbacks,
there's no use for the heavy-hammer .exit callback anymore.
job_exit becomes a glorified type shim so that we can call
job_completed from aio_bh_schedule_oneshot.
Move these three functions down into job.c to eliminate a
forward reference.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 20180906130225.5118-12-jsnow@redhat.com
Reviewed-by: Jeff Cody <jcody@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
{error,warn}_report_once() are a special case of the new functions
and can simply switch to them.
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
Message-Id: <20180830145902.27376-3-cohuck@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
[Dispense with unlikely() to keep the macros as simple as possible]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Add two functions to print an error/warning report once depending
on a passed-in condition variable and flip it if printed. This is
useful if you want to print a message not once-globally, but e.g.
once-per-device.
Inspired by warn_once() in hw/vfio/ccw.c, which has been replaced
with warn_report_once_cond().
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
Message-Id: <20180830145902.27376-2-cohuck@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
[Function comments reworded]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Now that the job infrastructure is handling the job_completed call for
all implemented jobs, we can remove the interface that allowed jobs to
schedule their own completion.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 20180830015734.19765-10-jsnow@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
Jobs are now expected to return their retcode on the stack, from the
.run callback, so we can remove that argument.
job_cancel does not need to set -ECANCELED because job_completed will
update the return code itself if the job was canceled.
While we're here, make job_completed static to job.c and remove it from
job.h; move the documentation of return code to the .run() callback and
to the job->ret property, accordingly.
Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 20180830015734.19765-9-jsnow@redhat.com
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
All jobs do the same thing when they leave their running loop:
- Store the return code in a structure
- wait to receive this structure in the main thread
- signal job completion via job_completed
Few jobs do anything beyond exactly this. Consolidate this exit
logic for a net reduction in SLOC.
More seriously, when we utilize job_defer_to_main_loop_bh to call
a function that calls job_completed, job_finalize_single will run
in a context where it has recursively taken the aio_context lock,
which can cause hangs if it puts down a reference that causes a flush.
You can observe this in practice by looking at mirror_exit's careful
placement of job_completed and bdrv_unref calls.
If we centralize job exiting, we can signal job completion from outside
of the aio_context, which should allow for job cleanup code to run with
only one lock, which makes cleanup callbacks less tricky to write.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 20180830015734.19765-4-jsnow@redhat.com
Reviewed-by: Jeff Cody <jcody@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Jobs presently use both an Error object in the case of the create job,
and char strings in the case of generic errors elsewhere.
Unify the two paths as just j->err, and remove the extra argument from
job_completed. The integer error code for job_completed is kept for now,
to be removed shortly in a separate patch.
Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 20180830015734.19765-3-jsnow@redhat.com
[mreitz: Dropped a superfluous g_strdup()]
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Presently we codify the entry point for a job as the "start" callback,
but a more apt name would be "run" to clarify the idea that when this
function returns we consider the job to have "finished," except for
any cleanup which occurs in separate callbacks later.
As part of this clarification, change the signature to include an error
object and a return code. The error ptr is not yet used, and the return
code while captured, will be overwritten by actions in the job_completed
function.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 20180830015734.19765-2-jsnow@redhat.com
Reviewed-by: Jeff Cody <jcody@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
vhost-user-gpu will share the same code to open a DRM node.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20180713130916.4153-20-marcandre.lureau@redhat.com>
[ kraxel: buildfix: util/drm.o must be CONFIG_OPENGL not CONFIG_LINUX ]
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
There are many error_report()s that can be used in frequently called
functions, especially on IO paths. That can be unideal in that
malicious guest can try to trigger the error tons of time which might
use up the log space on the host (e.g., libvirt can capture the stderr
of QEMU and put it persistently onto disk). In VT-d emulation code, we
have trace_vtd_error() tracer. AFAIU all those places can be replaced
by something like error_report() but trace points are mostly used to
avoid the DDOS attack that mentioned above. However using trace points
mean that errors are not dumped if trace not enabled.
It's not a big deal in most modern server managements since we have
things like logrotate to maintain the logs and make sure the quota is
expected. However it'll still be nice that we just provide another way
to restrict message generations. In most cases, this kind of
error_report()s will only provide valid information on the first message
sent, and all the rest of similar messages will be mostly talking about
the same thing. This patch introduces *_report_once() helpers to allow
a message to be dumped only once during one QEMU process's life cycle.
It will make sure: (1) it's on by deffault, so we can even get something
without turning the trace on and reproducing, and (2) it won't be
affected by DDOS attack.
To implement it, I stole the printk_once() macro from Linux.
CC: Eric Blake <eblake@redhat.com>
CC: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180815095328.32414-2-peterx@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
[Whitespace adjusted, comments improved]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
We reject bytes that can't occur in valid UTF-8 (\xC0..\xC1,
\xF5..\xFF in the lexer. That's insufficient; there's plenty of
invalid UTF-8 not containing these bytes, as demonstrated by
check-qjson:
* Malformed sequences
- Unexpected continuation bytes
- Missing continuation bytes after start bytes other than
\xC0..\xC1, \xF5..\xFD.
* Overlong sequences with start bytes other than \xC0..\xC1,
\xF5..\xFD.
* Invalid code points
Fixing this in the lexer would be bothersome. Fixing it in the parser
is straightforward, so do that.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20180823164025.12553-23-armbru@redhat.com>
-----BEGIN PGP SIGNATURE-----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=GExd
-----END PGP SIGNATURE-----
Merge remote-tracking branch 'remotes/juanquintela/tags/migration/20180822-1' into staging
migration/next for 20180822
# gpg: Signature made Wed 22 Aug 2018 12:07:59 BST
# gpg: using RSA key F487EF185872D723
# gpg: Good signature from "Juan Quintela <quintela@redhat.com>"
# gpg: aka "Juan Quintela <quintela@trasno.org>"
# Primary key fingerprint: 1899 FF8E DEBF 58CC EE03 4B82 F487 EF18 5872 D723
* remotes/juanquintela/tags/migration/20180822-1:
migration: hold the lock only if it is really needed
migration: move handle of zero page to the thread
migration: drop the return value of do_compress_ram_page
migration: introduce save_zero_page_to_file
migration: fix counting normal page for compression
migration: do not wait for free thread
migration: poll the cm event for destination qemu
tests/migration-test: Silence the kvm_hv message by default
migration: implement the shutdown for RDMA QIOChannel
migration: poll the cm event while wait RDMA work request completion
migration: invoke qio_channel_yield only when qemu_in_coroutine()
migration: implement io_set_aio_fd_handler function for RDMA QIOChannel
migration: Stop rdma yielding during incoming postcopy
migration: implement bi-directional RDMA QIOChannel
migration: create a dedicated connection for rdma return path
migration: disable RDMA WRITE after postcopy started
migrate/cpu-throttle: Add max-cpu-throttle migration parameter
docs/migration: Clarify pre_load in subsections
migration: Correctly handle subsections with no 'needed' function
qapi/migration.json: fix the description for "query-migrate" output
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
It's unnecessary because the pointer isn't dereferenced.
Signed-off-by: Emilio G. Cota <cota@braap.org>
Message-Id: <20180819091335.22863-3-cota@braap.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
To avoid undefined behaviour.
Signed-off-by: Emilio G. Cota <cota@braap.org>
Message-Id: <20180819091335.22863-2-cota@braap.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The BQL is acquired via qemu_mutex_lock_iothread(), which makes
the profiler assign the associated wait time (i.e. most of
BQL wait time) entirely to that function. This loses the original
call site information, which does not help diagnose BQL contention.
Fix it by tracking the callers explicitly.
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
I first implemented this by deleting all entries in the global
hash table. But doing that safely slows down profiling, since
we'd need to introduce rcu_read_lock/unlock in the fast path.
What's implemented here avoids messing with the thread-local
data in the global hash table. It achieves this by taking a snapshot
of the current state, so that subsequent reports present the delta
wrt to the snapshot.
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
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>
Try to hold src_page_req_mutex only if the queue is not
empty
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Xiao Guangrong <xiaoguangrong@tencent.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Because we need to make sure the pmem kind memory data is synced
after migration, we choose to call pmem_persist() when the migration
finish. This will make sure the data of pmem is safe and will not
lose if power is off.
Signed-off-by: Junyan He <junyan.he@intel.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Guest writes to vNVDIMM labels are intercepted and performed on the
backend by QEMU. When the backend is a real persistent memort, QEMU
needs to take proper operations to ensure its write persistence on the
persistent memory. Otherwise, a host power failure may result in the
loss of guest label configurations.
Signed-off-by: Haozhong Zhang <haozhong.zhang@intel.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Thomas Huth