A quick search for 'qemu_strtosz' in the code base shows that outside
of the testsuite, the ONLY place that passes a non-NULL pointer to
@endptr of any variant of a size parser is in hmp.c (the 'o' parser of
monitor_parse_arguments), and that particular caller warns of
"extraneous characters at the end of line" unless the trailing bytes
are purely whitespace. Thus, it makes no semantic difference at the
high level whether we parse "1.5e1k" as "1" + ".5e1" + "k" (an attempt
to use scientific notation in strtod with a scaling suffix of 'k' with
no trailing junk, but which qemu_strtosz says should fail with
EINVAL), or as "1.5e" + "1k" (a valid size with scaling suffix of 'e'
for exabytes, followed by two junk bytes) - either way, any user
passing such a string will get an error message about a parse failure.
However, an upcoming patch to qemu_strtosz will fix other corner case
bugs in handling the fractional portion of a size, and in doing so, it
is easier to declare that qemu_strtosz() itself stops parsing at the
first 'e' rather than blindly consuming whatever strtod() will
recognize. Once that is fixed, the difference will be visible at the
low level (getting a valid parse with trailing garbage when @endptr is
non-NULL, while continuing to get -EINVAL when @endptr is NULL); this
is easier to demonstrate by moving the affected strings from
test_qemu_strtosz_invalid() (which declares them as always -EINVAL) to
test_qemu_strtosz_trailing() (where @endptr affects behavior, for now
with FIXME comments).
Note that a similar argument could be made for having "0x1.5" or
"0x1M" parse as 0x1 with ".5" or "M" as trailing junk, instead of
blindly treating it as -EINVAL; however, as these cases do not suffer
from the same problems as floating point, they are not worth changing
at this time.
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Hanna Czenczek <hreitz@redhat.com>
Message-Id: <20230522190441.64278-11-eblake@redhat.com>
It's hard to tweak code for consistency if I can't prove what will or
won't break from those tweaks. Time to add unit tests for
qemu_strtod() and qemu_strtod_finite().
Among other things, I wrote a check whether we have C99 semantics for
strtod("0x1") (which MUST parse hex numbers) rather than C89 (which
must stop parsing at 'x'). These days, I suspect that is okay; but if
it fails CI checks, knowing the difference will help us decide what we
want to do about it. Note that C2x, while not final at the time of
this patch, has been considering whether to make strtol("0b1") parse
as 1 with no slop instead of the C17 parse of 0 with slop "b1"; that
decision may also bleed over to strtod(). But for now, I didn't think
it worth adding unit tests on that front (to strtol or strtod) as
things may still change.
Likewise, there are plenty more corner cases of strtod proper that I
don't explicitly test here, but there are enough unit tests added here
that it covers all the branches reached in our wrappers. In
particular, it demonstrates the difference on when *value is left
uninitialized, which an upcoming patch will normalize.
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Hanna Czenczek <hreitz@redhat.com>
Message-Id: <20230522190441.64278-10-eblake@redhat.com>
All the qemu_strto*() functions permit a NULL endptr, just like their
libc counterparts, leaving parse_uint() as the oddball that caused
SEGFAULT on NULL and required the user to call parse_uint_full()
instead. Relax things for consistency, even though the testsuite is
the only impacted caller. Add one more unit test to ensure even
parse_uint_full(NULL, 0, &value) works. This also fixes our code to
uniformly favor EINVAL over ERANGE when both apply.
Also fixes a doc mismatch @v vs. a parameter named value.
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Hanna Czenczek <hreitz@redhat.com>
Message-Id: <20230522190441.64278-9-eblake@redhat.com>
It's already confusing that we have two very similar functions for
wrapping the parse of a 64-bit unsigned value, differing mainly on
whether they permit leading '-'. Adjust the signature of parse_uint()
and parse_uint_full() to be like all of qemu_strto*(): put the result
parameter last, use the same types (uint64_t and unsigned long long
have the same width, but are not always the same type), and mark
endptr const (this latter change only affects the rare caller of
parse_uint). Adjust all callers in the tree.
While at it, note that since cutils.c already includes:
QEMU_BUILD_BUG_ON(sizeof(int64_t) != sizeof(long long));
we are guaranteed that the result of parse_uint* cannot exceed
UINT64_MAX (or the build would have failed), so we can drop
pre-existing dead comparisons in opts-visitor.c that were never false.
Reviewed-by: Hanna Czenczek <hreitz@redhat.com>
Message-Id: <20230522190441.64278-8-eblake@redhat.com>
[eblake: Drop dead code spotted by Markus]
Signed-off-by: Eric Blake <eblake@redhat.com>
These two functions are subtly different, and not just because of
swapped parameter order. It took me adding better unit tests to
figure out why. Document the differences to make it more obvious to
developers trying to pick which one to use, as well as to aid in
upcoming semantic changes.
While touching the documentation, adjust a mis-statement: parse_uint
does not return -EINVAL on invalid base, but assert()s, like all the
other qemu_strto* functions that take a base argument.
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Hanna Czenczek <hreitz@redhat.com>
Message-Id: <20230522190441.64278-7-eblake@redhat.com>
While we were matching 32-bit strtol in qemu_strtoi, our use of a
64-bit parse was leaking through for some inaccurate answers in
qemu_strtoui in comparison to a 32-bit strtoul (see the unit test for
examples). The comment for that function even described what we have
to do for a correct parse, but didn't implement it correctly: since
strtoull checks for overflow against the wrong values and then
negates, we have to temporarily undo negation before checking for
overflow against our desired value.
Our int wrappers would be a lot easier to write if libc had a
guaranteed 32-bit parser even on platforms with 64-bit long.
Whether we parse C2x binary strings like "0b1000" is currently up to
what libc does; our unit tests intentionally don't cover that at the
moment, though.
Fixes: 473a2a331e ("cutils: add qemu_strtoi & qemu_strtoui parsers for int/unsigned int types", v2.12.0)
Signed-off-by: Eric Blake <eblake@redhat.com>
CC: qemu-stable@nongnu.org
Message-Id: <20230522190441.64278-6-eblake@redhat.com>
Reviewed-by: Hanna Czenczek <hreitz@redhat.com>
We have quite a few undertested and underdocumented integer parsing
corner cases. To ensure that any changes we make in the code are
intentional rather than accidental semantic changes, it is time to add
more unit tests of existing behavior.
In particular, this demonstrates that parse_uint() and qemu_strtou64()
behave differently. For "-0", it's hard to argue why parse_uint needs
to reject it (it's not a negative integer), but the documentation sort
of mentions it; but it is intentional that all other negative values
are treated as ERANGE with value 0 (compared to qemu_strtou64()
treating "-2" as success and UINT64_MAX-1, for example).
Also, when mixing overflow/underflow with a check for no trailing
junk, parse_uint_full favors ERANGE over EINVAL, while qemu_strto[iu]*
favor EINVAL. This behavior is outside the C standard, so we can pick
whatever we want, but it would be nice to be consistent.
Note that C requires that "9223372036854775808" fail strtoll() with
ERANGE/INT64_MAX, but "-9223372036854775808" pass with INT64_MIN; we
weren't testing this. For strtol(), the behavior depends on whether
long is 32- or 64-bits (the cutoff point either being the same as
strtoll() or at "-2147483648"). Meanwhile, C is clear that
"-18446744073709551615" pass stroull() (but not strtoll) with value 1,
even though we want it to fail parse_uint(). And although
qemu_strtoui() has no C counterpart, it makes more sense if we design
it like 32-bit strtoul() (that is, where "-4294967296" be an alternate
acceptable spelling for "1", but "-0xffffffff00000001" should be
treated as overflow and return 0xffffffff rather than 1). We aren't
there yet, so some of the tests added in this patch have FIXME
comments.
However, note that C2x will (likely) be adding a SILENT semantic
change, where C17 strtol("0b1", &ep, 2) returns 0 with ep="b1", but
C2x will have it return 1 with ep="". I did not feel like adding
testing for those corner cases, in part because the next version of C
is not standard and libc support for binary parsing is not yet
wide-spread (as of this patch, glibc.git still misparses bare "0b":
https://sourceware.org/bugzilla/show_bug.cgi?id=30371).
Message-Id: <20230522190441.64278-5-eblake@redhat.com>
[eblake: fix a few typos spotted by Hanna]
Reviewed-by: Hanna Czenczek <hreitz@redhat.com>
[eblake: fix typo on platforms with 32-bit long]
Signed-off-by: Eric Blake <eblake@redhat.com>
We are inconsistent on the contents of *value after a strto* parse
failure. I found the following behaviors:
- parse_uint() and parse_uint_full(), which document that *value is
slammed to 0 on all EINVAL failures and 0 or UINT_MAX on ERANGE
failures, and has unit tests for that (note that parse_uint requires
non-NULL endptr, and does not fail with EINVAL for trailing junk)
- qemu_strtosz(), which leaves *value untouched on all failures (both
EINVAL and ERANGE), and has unit tests but not documentation for
that
- qemu_strtoi() and other integral friends, which document *value on
ERANGE failures but is unspecified on EINVAL (other than implicitly
by comparison to libc strto*); there, *value is untouched for NULL
string, slammed to 0 on no conversion, and left at the prefix value
on NULL endptr; unit tests do not consistently check the value
- qemu_strtod(), which documents *value on ERANGE failures but is
unspecified on EINVAL; there, *value is untouched for NULL string,
slammed to 0.0 for no conversion, and left at the prefix value on
NULL endptr; there are no unit tests (other than indirectly through
qemu_strtosz)
- qemu_strtod_finite(), which documents *value on ERANGE failures but
is unspecified on EINVAL; there, *value is left at the prefix for
'inf' or 'nan' and untouched in all other cases; there are no unit
tests (other than indirectly through qemu_strtosz)
Upcoming patches will change behaviors for consistency, but it's best
to first have more unit test coverage to see the impact of those
changes.
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Hanna Czenczek <hreitz@redhat.com>
Message-Id: <20230522190441.64278-4-eblake@redhat.com>
When debugging test failures, seeing unsigned values as large positive
values rather than negative values matters (assuming glib 2.78+; given
that I just fixed a bug in glib 2.76 [1] where g_assert_cmpuint
displays signed instead of unsigned values). No impact when the test
is passing, but using a consistent style will matter more in upcoming
test additions. Also, some tests are better with cmphex.
While at it, fix some spacing and minor typing issues spotted nearby.
[1] https://gitlab.gnome.org/GNOME/glib/-/issues/2997
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Hanna Czenczek <hreitz@redhat.com>
Message-Id: <20230522190441.64278-3-eblake@redhat.com>
glib documentation[1] is clear: g_assert() should be avoided in unit
tests because it is ineffective if G_DISABLE_ASSERT is defined; unit
tests should stick to constructs based on g_assert_true() instead.
Note that since commit 262a69f428, we intentionally state that you
cannot define G_DISABLE_ASSERT while building qemu; but our code can
be copied to other projects without that restriction, so we should be
consistent.
For most of the replacements in this patch, using g_assert_cmpstr()
would be a regression in quality - although it would helpfully display
the string contents of both pointers on test failure, here, we really
do care about pointer equality, not just string content equality. But
when a NULL pointer is expected, g_assert_null works fine.
[1] https://libsoup.org/glib/glib-Testing.html#g-assert
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Hanna Czenczek <hreitz@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20230522190441.64278-2-eblake@redhat.com>
Although we already covered the need for padding bytes with our
changes in commit 3ae3fcfa, commit 66fcbca5 (both v5.0.0) added one
byte and relied on the rest of the text for implicitly covering 7
padding bytes. For consistency with other parts of the header (such
as the header extension format listing padding from n - m, or the
snapshot table entry listing variable padding), we might as well call
out the remaining 7 bytes as padding until such time (as any) as they
gain another meaning.
Signed-off-by: Eric Blake <eblake@redhat.com>
CC: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
Message-Id: <20230522184631.47211-1-eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
In the past, commit a231cb27 ("iotests: Fix 104 for NBD", v2.3.0)
added an additional filter to _filter_img_info to rewrite NBD URIs
into the expected output form. This recently broke when we tweaked
tests to run in a per-format directory, which did not match the regex,
because _img_info itself is now already changing
SOCK_DIR=/tmp/tmpphjfbphd/raw-nbd-104 into
/tmp/tmpphjfbphd/IMGFMT-nbd-104 prior to _img_info_filter getting a
chance to further filter things.
While diagnosing the problem, I also noticed some filter lines
rendered completely useless by a typo when we switched from TCP to
Unix sockets for NBD (in shell, '\\+' is different from "\\+" (one
gives two backslash to the regex, matching the literal 2-byte sequence
<\+> after a single digit; the other gives one backslash to the regex,
as the metacharacter \+ to match one or more of <[0-9]>); since the
literal string <nbd://127.0.0.1:0\+> is not a valid URI, that regex
hasn't been matching anything for years so it is fine to just drop it
rather than fix the typo.
Fixes: f3923a72 ("iotests: Switch nbd tests to use Unix rather than TCP", v4.2.0)
Fixes: 5ba7db09 ("iotests: always use a unique sub-directory per test", v8.0.0)
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <20230519150216.2599189-1-eblake@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Document every single live=true use cases on why it should be done in the
live manner. Also document on the parameter so new precopy cases should
always use live=off unless with explicit reasonings.
Cc: Thomas Huth <thuth@redhat.com>
Cc: Juan Quintela <quintela@redhat.com>
Cc: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20230601172935.175726-1-peterx@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
There are 27 pre-copy live migration scenarios being tested. In all of
these we force non-convergence and run for one iteration, then let it
converge and wait for completion during the second (or following)
iterations. At 3 mbps bandwidth limit the first iteration takes a very
long time (~30 seconds).
While it is important to test the migration passes and convergence
logic, it is overkill to do this for all 27 pre-copy scenarios. The
TLS migration scenarios in particular are merely exercising different
code paths during connection establishment.
To optimize time taken, switch most of the test scenarios to run
non-live (ie guest CPUs paused) with no bandwidth limits. This gives
a massive speed up for most of the test scenarios.
For test coverage the following scenarios are unchanged
* Precopy with UNIX sockets
* Precopy with UNIX sockets and dirty ring tracking
* Precopy with XBZRLE
* Precopy with UNIX compress
* Precopy with UNIX compress (nowait)
* Precopy with multifd
On a test machine this reduces execution time from 13 minutes to
8 minutes.
Tested-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20230601161347.1803440-10-berrange@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
The 'got_stop' and 'got_resume' global variables apply to the src and
dst migration VM respectively. Change their names to make this explicit
to developers.
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20230601161347.1803440-9-berrange@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
When running migration tests we monitor for a STOP event so we can skip
redundant waits. This will be needed for the RESUME event too shortly.
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20230601161347.1803440-8-berrange@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Most usage of wait_command() is followed by qobject_unref(), which
is just a verbose re-implementation of qtest_qmp_assert_success().
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20230601161347.1803440-7-berrange@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Change the migration test to use the new qtest event callback to watch
for the stop event. This ensures that we only watch for the STOP event
on the source QEMU. The previous code would set the single 'got_stop'
flag when either source or dest QEMU got the STOP event.
Reviewed-by: Juan Quintela <quintela@redhat.com>
Acked-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20230601161347.1803440-6-berrange@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Some of the usage is just a verbose way of re-inventing the
qtest_qmp_assert_success(_ref) methods.
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20230601161347.1803440-5-berrange@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
This function duplicates logic of qtest_qmp_assert_success_ref.
The qtest_qmp_assert_success_ref method has better diagnostics
on failure because it prints the entire QMP response, instead
of just asserting on existance of the 'error' key.
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20230601161347.1803440-4-berrange@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Currently code must call one of the qtest_qmp_event* functions to
fetch events. These are only usable if the immediate caller knows
the particular event they want to capture, and are only interested
in one specific event type. Adding ability to register an event
callback lets the caller capture a range of events over any period
of time.
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20230601161347.1803440-3-berrange@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Add several counterparts of qtest_qmp_assert_success() that can
* Use va_list instead of ...
* Accept a list of FDs to send
* Return the response data
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20230601161347.1803440-2-berrange@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Currently, it is only done when the iteration finishes successfully.
Not cleaning up the userfaultfd write protection can lead to
symptoms/issues such as the process hanging in memmove or GDB not
being able to attach.
Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
Message-Id: <20230526115908.196171-1-f.ebner@proxmox.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
1. Otherwise failed migration just drops guest-panicked state, which is
not good for management software.
2. We do keep different paused states like guest-panicked during
migration with help of global_state state.
3. We do restore running state on source when migration is cancelled or
failed.
4. "postmigrate" state is documented as "guest is paused following a
successful 'migrate'", so originally it's only for successful path
and we never documented current behavior.
Let's restore paused states like guest-panicked in case of cancel or
fail too. Allow same transitions like for inmigrate state.
This commit changes the behavior that was introduced by commit
42da5550d6 "migration: set state to post-migrate on failure" and
provides a bit different fix on related
https://bugzilla.redhat.com/show_bug.cgi?id=1355683
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20230517123752.21615-6-vsementsov@yandex-team.ru>
Signed-off-by: Juan Quintela <quintela@redhat.com>
No logic change here, only refactoring. That's a preparation for next
commit where we finally restore the stopped vm state on migration
failure or cancellation.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20230517123752.21615-5-vsementsov@yandex-team.ru>
Signed-off-by: Juan Quintela <quintela@redhat.com>
The function is unused since previous commit. Drop it.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20230517123752.21615-4-vsementsov@yandex-team.ru>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Actually global_state_store() can never fail. Let's get rid of extra
error paths.
To make things clear, use new runstate_get() and use same approach for
global_state_store() and global_state_store_running().
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20230517123752.21615-3-vsementsov@yandex-team.ru>
Signed-off-by: Juan Quintela <quintela@redhat.com>
It's necessary to restore the state after failed/cancelled migration in
further commit.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20230517123752.21615-2-vsementsov@yandex-team.ru>
Signed-off-by: Juan Quintela <quintela@redhat.com>
The virtio-blk-vhost-vdpa driver in libblkio 1.3.0 supports the fd
passing through the new 'fd' property.
Since now we are using qemu_open() on '@path' if the virtio-blk driver
supports the fd passing, let's announce it.
In this way, the management layer can pass the file descriptor of an
already opened vhost-vdpa character device. This is useful especially
when the device can only be accessed with certain privileges.
Add the '@fdset' feature only when the virtio-blk-vhost-vdpa driver
in libblkio supports it.
Suggested-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Message-id: 20230530071941.8954-3-sgarzare@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Some virtio-blk drivers (e.g. virtio-blk-vhost-vdpa) supports the fd
passing. Let's expose this to the user, so the management layer
can pass the file descriptor of an already opened path.
If the libblkio virtio-blk driver supports fd passing, let's always
use qemu_open() to open the `path`, so we can handle fd passing
from the management layer through the "/dev/fdset/N" special path.
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Message-id: 20230530071941.8954-2-sgarzare@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
No need to pass zeros as we have helpers that do that for us.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20230526165401.574474-11-alex.bennee@linaro.org
Message-Id: <20230524133952.3971948-10-alex.bennee@linaro.org>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Now we no longer have dynamic state affecting things we can remove the
additional fields in cpu.h and simplify the TB hash calculation.
For the benchmark:
hyperfine -w 2 -m 20 \
"./arm-softmmu/qemu-system-arm -cpu cortex-a15 \
-machine type=virt,highmem=off \
-display none -m 2048 \
-serial mon:stdio \
-netdev user,id=unet,hostfwd=tcp::2222-:22 \
-device virtio-net-pci,netdev=unet \
-device virtio-scsi-pci \
-blockdev driver=raw,node-name=hd,discard=unmap,file.driver=host_device,file.filename=/dev/zen-disk/debian-bullseye-armhf \
-device scsi-hd,drive=hd -smp 4 \
-kernel /home/alex/lsrc/linux.git/builds/arm/arch/arm/boot/zImage \
-append 'console=ttyAMA0 root=/dev/sda2 systemd.unit=benchmark.service' \
-snapshot"
It has a marginal effect on runtime, before:
Time (mean ± σ): 26.279 s ± 2.438 s [User: 41.113 s, System: 1.843 s]
Range (min … max): 24.420 s … 32.565 s 20 runs
after:
Time (mean ± σ): 24.440 s ± 2.885 s [User: 34.474 s, System: 2.028 s]
Range (min … max): 21.663 s … 29.937 s 20 runs
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1358
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20230526165401.574474-10-alex.bennee@linaro.org
Message-Id: <20230524133952.3971948-9-alex.bennee@linaro.org>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Now we no longer have vcpu controlled trace events we can excise the
code that allows us to query its status.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20230526165401.574474-9-alex.bennee@linaro.org
Message-Id: <20230524133952.3971948-8-alex.bennee@linaro.org>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Now we no longer have any events that are for vcpus we can start
excising the code from the trace control. As the vcpu parameter is
encoded as part of QMP we just stub out the has_vcpu/vcpu parameters
rather than alter the API.
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20230526165401.574474-8-alex.bennee@linaro.org
Message-Id: <20230524133952.3971948-7-alex.bennee@linaro.org>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
I don't think I can remove the parameters directly but certainly mark
them as deprecated.
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20230526165401.574474-7-alex.bennee@linaro.org
Message-Id: <20230524133952.3971948-6-alex.bennee@linaro.org>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Also rename the section to make the fact this is part of the
management protocol even clearer.
Suggested-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20230526165401.574474-6-alex.bennee@linaro.org
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
This makes it a little easier for developers to find where things
where being generated.
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20230526165401.574474-5-alex.bennee@linaro.org
Message-Id: <20230524133952.3971948-5-alex.bennee@linaro.org>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
This does involve temporarily stubbing out some helper functions
before we excise the rest of the code.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20230526165401.574474-4-alex.bennee@linaro.org
Message-Id: <20230524133952.3971948-4-alex.bennee@linaro.org>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
While these are all in helper functions being designated vcpu events
complicates the removal of the dynamic vcpu state code. TCG plugins
allow you to instrument vcpu_[init|exit|idle].
We rename cpu_reset and make it a normal trace point.
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20230526165401.574474-3-alex.bennee@linaro.org
Message-Id: <20230524133952.3971948-3-alex.bennee@linaro.org>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
This is pure duplication now. Both bsd-user and linux-user have
builtin strace support and we can also track syscalls via the plugins
system.
Reviewed-by: Warner Losh <imp@bsdimp.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20230526165401.574474-2-alex.bennee@linaro.org
Message-Id: <20230524133952.3971948-2-alex.bennee@linaro.org>
[Remove unused variable in do_freebsd_syscall() reported by Richard
Henderson.
--Stefan]
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
No block driver implements .bdrv_co_io_plug() anymore. Get rid of the
function pointers.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Acked-by: Kevin Wolf <kwolf@redhat.com>
Message-id: 20230530180959.1108766-7-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Stop using the .bdrv_co_io_plug() API because it is not multi-queue
block layer friendly. Use the new blk_io_plug_call() API to batch I/O
submission instead.
Note that a dev_max_batch check is dropped in laio_io_unplug() because
the semantics of unplug_fn() are different from .bdrv_co_unplug():
1. unplug_fn() is only called when the last blk_io_unplug() call occurs,
not every time blk_io_unplug() is called.
2. unplug_fn() is per-thread, not per-BlockDriverState, so there is no
way to get per-BlockDriverState fields like dev_max_batch.
Therefore this condition cannot be moved to laio_unplug_fn(). It is not
obvious that this condition affects performance in practice, so I am
removing it instead of trying to come up with a more complex mechanism
to preserve the condition.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Acked-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Message-id: 20230530180959.1108766-6-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Stop using the .bdrv_co_io_plug() API because it is not multi-queue
block layer friendly. Use the new blk_io_plug_call() API to batch I/O
submission instead.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Acked-by: Kevin Wolf <kwolf@redhat.com>
Message-id: 20230530180959.1108766-5-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Stop using the .bdrv_co_io_plug() API because it is not multi-queue
block layer friendly. Use the new blk_io_plug_call() API to batch I/O
submission instead.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Acked-by: Kevin Wolf <kwolf@redhat.com>
Message-id: 20230530180959.1108766-4-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Stop using the .bdrv_co_io_plug() API because it is not multi-queue
block layer friendly. Use the new blk_io_plug_call() API to batch I/O
submission instead.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Acked-by: Kevin Wolf <kwolf@redhat.com>
Message-id: 20230530180959.1108766-3-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Introduce a new API for thread-local blk_io_plug() that does not
traverse the block graph. The goal is to make blk_io_plug() multi-queue
friendly.
Instead of having block drivers track whether or not we're in a plugged
section, provide an API that allows them to defer a function call until
we're unplugged: blk_io_plug_call(fn, opaque). If blk_io_plug_call() is
called multiple times with the same fn/opaque pair, then fn() is only
called once at the end of the function - resulting in batching.
This patch introduces the API and changes blk_io_plug()/blk_io_unplug().
blk_io_plug()/blk_io_unplug() no longer require a BlockBackend argument
because the plug state is now thread-local.
Later patches convert block drivers to blk_io_plug_call() and then we
can finally remove .bdrv_co_io_plug() once all block drivers have been
converted.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Acked-by: Kevin Wolf <kwolf@redhat.com>
Message-id: 20230530180959.1108766-2-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
