In linux-user/signal.c we have two FIXME comments claiming that
parts of the signal-handling code are not threadsafe. These are
very old, as they were first introduced in commit 624f797905
in 2008. Since then we've radically overhauled the signal-handling
logic, while carefully preserving these FIXME comments.
It's unclear exactly what thread-safety issue the original
author was trying to point out -- the relevant data structures
are in the TaskStruct, which makes them per-thread and only
operated on by that thread. The old code at the time of that
commit did have various races involving signal handlers being
invoked at awkward times; possibly this was what was meant.
Delete these FIXME comments:
* they were written at a time when the way we handled
signals was completely different
* the code today appears to us to not have thread-safety issues
* nobody knows what the problem the comments were trying to
point out was
so they are serving no useful purpose for us today.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Warner Losh <imp@bsdimp.com>
Message-Id: <20220114155032.3767771-1-peter.maydell@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Generally all BMCs will use the fmc_model to hold their own flash
and most will have a spi_model to hold the managed system's flash,
but not all systems do. Add a simple NULL check to allow a system
to set the spi_model as NULL to indicate it should not be instantiated.
Signed-off-by: Patrick Williams <patrick@stwcx.xyz>
Message-Id: <20220305000656.1944589-1-patrick@stwcx.xyz>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
If no id is provided, qdev automatically assigns an unique name with
the following pattern "<type>.<index>" which avoids bus name collision
when using multiple buses.
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-Id: <20220307071856.1410731-6-clg@kaod.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Currently, the allocation of the flash devices uses the number of
slave selects configured in the SoC realize routine. It is simpler to
use directly the number of FMC devices defined in the machine class
and 1 for spi devices (which is what the SoC does in the back of the
machine).
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-Id: <20220307071856.1410731-3-clg@kaod.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
The Aspeed SMC model uses the 'num_cs' field to allocate resources
fitting the number of devices of the machine. This is a small
optimization without real need in the controller. Simplify modelling
and use the max_peripherals field instead.
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-Id: <20220307071856.1410731-2-clg@kaod.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
A memory chunk is allocated with g_new0() and assigned to the variable
'storage'. However, if the branch takes true, there will be only an
error report but not a free operation for 'storage' before function
returns. As a result, a memory leak bug is triggered.
Use g_autofree to fix the issue.
Suggested-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Wentao_Liang <Wentao_Liang_g@163.com>
[ clg: reworked the commit log ]
Signed-off-by: Cédric Le Goater <clg@kaod.org>
When writing zeroes can fall back to a slow write, permitting an
overly large request can become an amplification denial of service
attack in triggering a large amount of work from a small request. But
the whole point of the no fallback flag is to quickly determine if
writing an entire device to zero can be done quickly (such as when it
is already known that the device started with zero contents); in those
cases, artificially capping things at 2G in qemu-io itself doesn't
help us.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <20211203231539.3900865-4-eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
The block layer has supported 64-bit block status from drivers since
commit 86a3d5c688 ("block: Add .bdrv_co_block_status() callback",
v2.12) and friends, with individual driver callbacks responsible for
capping things where necessary. Artificially capping things below 2G
in the qemu-io 'map' command, added in commit d6a644bbfe ("block: Make
bdrv_is_allocated() byte-based", v2.10) is thus no longer necessary.
One way to test this is with qemu-nbd as server on a raw file larger
than 4G (the entire file should show as allocated), plus 'qemu-io -f
raw -c map nbd://localhost --trace=nbd_\*' as client. Prior to this
patch, the NBD_CMD_BLOCK_STATUS requests are fragmented at 0x7ffffe00
distances; with this patch, the fragmenting changes to 0x7fffffff
(since the NBD protocol is currently still limited to 32-bit
transactions - see block/nbd.c:nbd_client_co_block_status). Then in
later patches, once I add an NBD extension for a 64-bit block status,
the same map command completes with just one NBD_CMD_BLOCK_STATUS.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <20211203231539.3900865-3-eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
This validates that connections to an NBD server where the certificate
hostname does not match will fail. It further validates that using the
new 'tls-hostname' override option can solve the failure.
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20220304193610.3293146-11-berrange@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
Introduce a filter for the output of qemu-nbd export list so it can be
reused in multiple tests.
The filter is a bit more permissive that what test 241 currently uses,
as its allows printing of the export count, along with any possible
error messages that might be emitted.
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20220304193610.3293146-9-berrange@redhat.com>
Tested-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
The second bit of IAPC_BOOT_ARCH in FADT table indicates the presence of
keyboard controller implemented as 8042 or equivalent micro controller. This
change enables this flag for microvms if such a device exists (for example,
when added explicitly from the QEMU commandline). Change
654701e292d98b308b0 ("hw/acpi: add indication for i8042 in IA-PC boot flags of the FADT table")
enabled this flag for i386 q35 based machines. The reason for doing the same
for micro-vms is to make sure we provide the correct tables to the guest OS
uniformly in all cases when an i8042 device is present. When this bit is not
enabled, guest OSes has to find other indirect methods to detect the device
which we would like to avoid.
Signed-off-by: Ani Sinha <ani@anisinha.ca>
Message-Id: <20220304154032.2071585-5-ani@anisinha.ca>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
When developing an I/O test it is typical to add some logic to the
test script, run it to view the output diff, and then apply the
output diff to the reference file. This can be drastically simplified
by letting the test runner update the reference file in place.
By setting 'QEMU_IOTESTS_REGEN=1', the test runner will report the
failure and show the diff, but at the same time update the reference
file. So next time the I/O test is run it will succeed.
Continuing to display the diff when updating the reference gives the
developer a chance to review what was changed.
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20220304193610.3293146-7-berrange@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
The TLS usage for NBD was restricted to IP sockets because validating
x509 certificates requires knowledge of the hostname that the client
is connecting to.
TLS does not have to use x509 certificates though, as PSK (pre-shared
keys) provide an alternative credential option. These have no
requirement for a hostname and can thus be trivially used for UNIX
sockets.
Furthermore, with the ability to overide the default hostname for
TLS validation in the previous patch, it is now also valid to want
to use x509 certificates with FD passing and UNIX sockets.
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20220304193610.3293146-6-berrange@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
When using the --list option, qemu-nbd acts as an NBD client rather
than a server. As such when using TLS, it has a need to validate
the server certificate. This adds a --tls-hostname option which can
be used to override the default hostname used for certificate
validation.
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20220304193610.3293146-5-berrange@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
When connecting to an NBD server with TLS and x509 credentials,
the client must validate the hostname it uses for the connection,
against that published in the server's certificate. If the client
is tunnelling its connection over some other channel, however, the
hostname it uses may not match the info reported in the server's
certificate. In such a case, the user needs to explicitly set an
override for the hostname to use for certificate validation.
This is achieved by adding a 'tls-hostname' property to the NBD
block driver.
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20220304193610.3293146-4-berrange@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
In
commit a71d597b98
Author: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Date: Thu Jun 10 13:08:00 2021 +0300
block/nbd: reuse nbd_co_do_establish_connection() in nbd_open()
the use of the 'hostname' field from the BDRVNBDState struct was
lost, and 'nbd_connect' just hardcoded it to match the IP socket
address. This was a harmless bug at the time since we block use
with anything other than IP sockets.
Shortly though, we want to allow the caller to override the hostname
used in the TLS certificate checks. This is to allow for TLS
when doing port forwarding or tunneling. Thus we need to reinstate
the passing along of the 'hostname'.
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20220304193610.3293146-3-berrange@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
Currently the TLS session object assumes that the caller will always
provide a hostname when using x509 creds on a client endpoint. This
relies on the caller to detect and report an error if the user has
configured QEMU with x509 credentials on a UNIX socket. The migration
code has such a check, but it is too broad, reporting an error when
the user has configured QEMU with PSK credentials on a UNIX socket,
where hostnames are irrelevant.
Putting the check into the TLS session object credentials validation
code ensures we report errors in only the scenario that matters.
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20220304193610.3293146-2-berrange@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
_bind_hack() was a quick fix to allow async QMP to call bind(2) prior to
calling listen(2) and accept(2). This wasn't sufficient to fully address
the race condition present in synchronous clients.
With the race condition in legacy.py fixed (see the previous commit),
there are no longer any users of _bind_hack(). Drop it.
Fixes: b0b662bb2b
Signed-off-by: John Snow <jsnow@redhat.com>
Acked-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-id: 20220225205948.3693480-11-jsnow@redhat.com
[Expanded commit message. --js]
Signed-off-by: John Snow <jsnow@redhat.com>
legacy.py provides a synchronous model. iotests frequently uses this
paradigm:
- create QMP client object
- start QEMU process
- await connection from QEMU process
In the switch from sync to async QMP, the QMP client object stopped
calling bind() and listen() during the QMP object creation step, which
creates a race condition if the QEMU process dials in too quickly.
With refactoring out of the way, restore the former behavior of calling
bind() and listen() during __init__() to fix this race condition.
Signed-off-by: John Snow <jsnow@redhat.com>
Acked-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-id: 20220225205948.3693480-10-jsnow@redhat.com
[Expanded commit message. --js]
Signed-off-by: John Snow <jsnow@redhat.com>
Add start_server() and accept() methods that can be used instead of
start_server_and_accept() to allow more fine-grained control over the
incoming connection process.
(Eagle-eyed reviewers will surely notice that it's a bit weird that
"CONNECTING" is a state that's shared between both the start_server()
and connect() states. That's absolutely true, and it's very true that
checking on the presence of _accepted as an indicator of state is a
hack. That's also very certainly true. But ... this keeps client code an
awful lot simpler, as it doesn't have to care exactly *how* the
connection is being made, just that it *is*. Is it worth disrupting that
simplicity in order to provide a better state guard on `accept()`? Hm.)
Signed-off-by: John Snow <jsnow@redhat.com>
Acked-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-id: 20220225205948.3693480-9-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
Refactor _do_accept() into _do_start_server() and _do_accept(). As of
this commit, the former calls the latter, but in subsequent commits
they'll be split apart.
(So please forgive the misnomer for _do_start_server(); it will live up
to its name shortly, and the docstring will be updated then too. I'm
just cutting down on some churn.)
Signed-off-by: John Snow <jsnow@redhat.com>
Acked-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-id: 20220225205948.3693480-7-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
As part of disentangling the monolithic nature of _do_accept(), split
out the incoming callback to prepare for factoring out the "wait for a
peer" step. Namely, this means using an event signal we can wait on from
outside of this method.
Signed-off-by: John Snow <jsnow@redhat.com>
Acked-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-id: 20220225205948.3693480-5-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
These two methods attempted to entirely envelop the logic of
establishing a connection to a peer start to finish. However, we need to
break apart the incoming connection step into more granular steps. We
will no longer be able to reasonably constrain the logic inside of these
helper functions.
So, remove them - with _session_guard(), they no longer serve a real
purpose.
Although the public API doesn't change, the internal API does. Now that
there are no intermediary methods between e.g. connect() and
_do_connect(), there's no hook where the runstate is set. As a result,
the test suite changes a little to cope with the new semantics of
_do_accept() and _do_connect().
Lastly, take some pieces of the now-deleted docstrings and move
them up to the public interface level. They were a little more detailed,
and it won't hurt to keep them.
Signed-off-by: John Snow <jsnow@redhat.com>
Acked-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-id: 20220225205948.3693480-4-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
Previously, I had a method named "accept()" that under-the-hood calls
bind(2), listen(2) *and* accept(2). I meant this as a simplification and
counterpart to the one-shot "connect()" method.
This is confusing to readers who expect accept() to mean *just*
accept(2). Since I need to split apart the "accept()" method into
multiple methods anyway (one of which strongly resembling accept(2)), it
feels pertinent to rename this method *now*.
Rename this all-in-one method "start_server_and_accept()" instead.
Signed-off-by: John Snow <jsnow@redhat.com>
Acked-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-id: 20220225205948.3693480-3-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
In _new_session, there's a fairly complex except clause that's used to
give semantic errors to callers of accept() and connect(). We need to
create a new two-step replacement for accept(), so factoring out this
piece of logic will be useful.
Bolster the comments and docstring here to try and demystify what's
going on in this fairly delicate piece of Python magic.
(If we were using Python 3.7+, this would be an @asynccontextmanager. We
don't have that very nice piece of magic, however, so this must take an
Awaitable to manage the Exception contexts properly. We pay the price
for platform compatibility.)
Signed-off-by: John Snow <jsnow@redhat.com>
Acked-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-id: 20220225205948.3693480-2-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
Since the previous commit 9ea89876f9d ("target/mips: Fix cycle
counter timing calculations"), MIPSCPU::cp0_count_rate is not
used anymore. We don't need it since it is already expressed
as mips_def_t::CCRes. Remove the duplicate and clean.
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <>20211213102340.1847248-1-f4bug@amsat.org>
The cp0_count_ns value is calculated from the CP0_COUNT_RATE_DEFAULT
constant in target/mips/cpu.c. The cycle counter resolution is defined
per-CPU in target/mips/cpu-defs.c.inc; use this value for calculating
cp0_count_ns. Fixings timing problems on guest OSs for the 20Kc CPU
which has a CCRes of 1.
Signed-off-by: Simon Burge <simonb@NetBSD.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20211213135125.18378-1-simonb@NetBSD.org>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
- Re-org accel/ and softmmu/ to have more target-agnostic objects.
- Use CPUArchState as an abstract type, defined by each target
(CPUState is our interface with generic code, CPUArchState is
our interface with target-specific code).
# gpg: Signature made Sun 06 Mar 2022 23:23:19 GMT
# gpg: using RSA key FAABE75E12917221DCFD6BB2E3E32C2CDEADC0DE
# gpg: Good signature from "Philippe Mathieu-Daudé (F4BUG) <f4bug@amsat.org>" [unknown]
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg: There is no indication that the signature belongs to the owner.
# Primary key fingerprint: FAAB E75E 1291 7221 DCFD 6BB2 E3E3 2C2C DEAD C0DE
* remotes/philmd/tags/abstract-arch-cpu-20220307: (33 commits)
accel/tcg: Remove pointless CPUArchState casts
target/i386: Remove pointless CPUArchState casts
target: Use ArchCPU as interface to target CPU
target: Introduce and use OBJECT_DECLARE_CPU_TYPE() macro
target: Use CPUArchState as interface to target-specific CPU state
target: Use forward declared type instead of structure type
target/hexagon: Add missing 'hw/core/cpu.h' include
target: Include missing 'cpu.h'
Hexagon (target/hexagon) convert to OBJECT_DECLARE_TYPE
target/i386/tcg/sysemu: Include missing 'exec/exec-all.h' header
cpu: Add missing 'exec/exec-all.h' and 'qemu/accel.h' headers
exec/cpu_ldst: Include 'cpu.h' to get target_ulong definition
meson: Display libfdt as disabled when system emulation is disabled
softmmu: Build target-agnostic objects once
softmmu: Add qemu_init_arch_modules()
exec/cpu: Make address_space_init/reloading_memory_map target agnostic
exec/gdbstub: Make gdb_exit() / gdb_set_stop_cpu() target agnostic
misc: Add missing "sysemu/cpu-timers.h" include
misc: Remove unnecessary "sysemu/cpu-timers.h" include
softmmu/cpu-timers: Remove unused 'exec/exec-all.h' header
...
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
The BootLinux tests are currently failing with an ugly python
stack trace on my RHEL8 system since they cannot get a free port
(likely due to the firewall settings on my system). Let's properly
check the return value of find_free_port() instead and cancel the
test gracefully if it cannot get a free port.
Message-Id: <20220228114325.818294-1-thuth@redhat.com>
Reviewed-by: Beraldo Leal <bleal@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Block patches for 7.0-rc0:
- New fleecing backup scheme
- iotest fixes
- Fixes for the curl block driver
- Fix for the preallocate block driver
- IDE fix for zero-length TRIM requests
# gpg: Signature made Mon 07 Mar 2022 10:33:31 GMT
# gpg: using RSA key CB62D7A0EE3829E45F004D34A1FA40D098019CDF
# gpg: issuer "hreitz@redhat.com"
# gpg: Good signature from "Hanna Reitz <hreitz@redhat.com>" [marginal]
# gpg: WARNING: This key is not certified with sufficiently trusted signatures!
# gpg: It is not certain that the signature belongs to the owner.
# Primary key fingerprint: CB62 D7A0 EE38 29E4 5F00 4D34 A1FA 40D0 9801 9CDF
* remotes/hreitz-gitlab/tags/pull-block-2022-03-07: (23 commits)
iotests/image-fleecing: test push backup with fleecing
iotests/image-fleecing: add test case with bitmap
iotests.py: add qemu_io_pipe_and_status()
iotests/image-fleecing: add test-case for fleecing format node
block: copy-before-write: realize snapshot-access API
block: introduce snapshot-access block driver
block/io: introduce block driver snapshot-access API
block/reqlist: add reqlist_wait_all()
block/dirty-bitmap: introduce bdrv_dirty_bitmap_status()
block/reqlist: reqlist_find_conflict(): use ranges_overlap()
block: intoduce reqlist
block/block-copy: add block_copy_reset()
block/copy-before-write: add bitmap open parameter
block/block-copy: block_copy_state_new(): add bitmap parameter
block/dirty-bitmap: bdrv_merge_dirty_bitmap(): add return value
block/block-copy: move copy_bitmap initialization to block_copy_state_new()
iotests: Write test output to TEST_DIR
tests/qemu-iotests/testrunner: Quote "case not run" lines in TAP mode
tests/qemu-iotests/040: Skip TestCommitWithFilters without 'throttle'
block: fix preallocate filter: don't do unaligned preallocate requests
...
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Matheus Ferst