This reverts commit a7077b8e35,
and add comments to explain why child sources cannot be used.
When a GSource is added as a child of another GSource, if its
'prepare' function indicates readiness, then the parent's
'prepare' function will never be run. The io_watch_poll_prepare
absolutely *must* be run on every iteration of the main loop,
to ensure that the chardev backend doesn't feed data to the
frontend that it is unable to consume.
At the time a7077b8e35 was made,
all the child GSource impls were relying on poll'ing an FD,
so their 'prepare' functions would never indicate readiness
ahead of poll() being invoked. So the buggy behaviour was
not noticed and lay dormant.
Relatively recently the QIOChannelTLS impl introduced a
level 2 child GSource, which checks with GNUTLS whether it
has cached any data that was decoded but not yet consumed:
commit ffda5db65a
Author: Antoine Damhet <antoine.damhet@shadow.tech>
Date: Tue Nov 15 15:23:29 2022 +0100
io/channel-tls: fix handling of bigger read buffers
Since the TLS backend can read more data from the underlying QIOChannel
we introduce a minimal child GSource to notify if we still have more
data available to be read.
Signed-off-by: Antoine Damhet <antoine.damhet@shadow.tech>
Signed-off-by: Charles Frey <charles.frey@shadow.tech>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
With this, it is now quite common for the 'prepare' function
on a QIOChannelTLS GSource to indicate immediate readiness,
bypassing the parent GSource 'prepare' function. IOW, the
critical 'io_watch_poll_prepare' is being skipped on some
iterations of the main loop. As a result chardev frontend
asserts are now being triggered as they are fed data they
are not ready to consume.
A reproducer is as follows:
* In terminal 1 run a GNUTLS *echo* server
$ gnutls-serv --echo \
--x509cafile ca-cert.pem \
--x509keyfile server-key.pem \
--x509certfile server-cert.pem \
-p 9000
* In terminal 2 run a QEMU guest
$ qemu-system-s390x \
-nodefaults \
-display none \
-object tls-creds-x509,id=tls0,dir=$PWD,endpoint=client \
-chardev socket,id=con0,host=localhost,port=9000,tls-creds=tls0 \
-device sclpconsole,chardev=con0 \
-hda Fedora-Cloud-Base-39-1.5.s390x.qcow2
After the previous patch revert, but before this patch revert,
this scenario will crash:
qemu-system-s390x: ../hw/char/sclpconsole.c:73: chr_read: Assertion
`size <= SIZE_BUFFER_VT220 - scon->iov_data_len' failed.
This assert indicates that 'tcp_chr_read' was called without
'tcp_chr_read_poll' having first been checked for ability to
receive more data
QEMU's use of a 'prepare' function to create/delete another
GSource is rather a hack and not normally the kind of thing that
is expected to be done by a GSource. There is no mechanism to
force GLib to always run the 'prepare' function of a parent
GSource. The best option is to simply not use the child source
concept, and go back to the functional approach previously
relied on.
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Tested-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
This commit results in unexpected termination of the TLS connection.
When 'fd_can_read' returns 0, the code goes on to pass a zero length
buffer to qio_channel_read. The TLS impl calls into gnutls_recv()
with this zero length buffer, at which point GNUTLS returns an error
GNUTLS_E_INVALID_REQUEST. This is treated as fatal by QEMU's TLS code
resulting in the connection being torn down by the chardev.
Simply skipping the qio_channel_read when the buffer length is zero
is also not satisfactory, as it results in a high CPU burn busy loop
massively slowing QEMU's functionality.
The proper solution is to avoid tcp_chr_read being called at all
unless the frontend is able to accept more data. This will be done
in a followup commit.
This reverts commit 462945cd22
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
The socket chardev often has 2 GSource object registered against the
same FD. One is registered all the time and is just intended to handle
POLLHUP events, while the other gets registered & unregistered on the
fly as the frontend is ready to receive more data or not.
It is very common for poll() to signal a POLLHUP event at the same time
as there is pending incoming data from the disconnected client. It is
therefore essential to process incoming data prior to processing HUP.
The problem with having 2 GSource on the same FD is that there is no
guaranteed ordering of execution between them, so the chardev code may
process HUP first and thus discard data.
This failure scenario is non-deterministic but can be seen fairly
reliably by reverting a7077b8e35, and
then running 'tests/unit/test-char', which will sometimes fail with
missing data.
Ideally QEMU would only have 1 GSource, but that's a complex code
refactoring job. The next best solution is to try to ensure ordering
between the 2 GSource objects. This can be achieved by lowering the
priority of the HUP GSource, so that it is never dispatched if the
main GSource is also ready to dispatch. Counter-intuitively, lowering
the priority of a GSource is done by raising its priority number.
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Both
$ qemu-system-x86_64 -chardev null,id=chr0,mux=on -mon chardev=chr0 -mon chardev=chr0 -mon chardev=chr0 -mon chardev=chr0 -mon chardev=chr0
and
$ qemu-system-x86_64 -chardev null,id=chr0 -mon chardev=chr0 -mon chardev=chr0
fail with
qemu-system-x86_64: -mon chardev=chr0: Device 'chr0' is in use
Improve to
qemu-system-x86_64: -mon chardev=chr0: too many uses of multiplexed chardev 'chr0' (maximum is 4)
and
qemu-system-x86_64: -mon chardev=chr0: chardev 'chr0' is already in use
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Michael Tokarev <mjt@tls.msk.ru>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Commit ffda5db65a ("io/channel-tls: fix handling of bigger read buffers")
changed the behavior of the TLS io channels to schedule a second reading
attempt if there is still incoming data pending. This caused a regression
with backends like the sclpconsole that check in their read function that
the sender does not try to write more bytes to it than the device can
currently handle.
The problem can be reproduced like this:
1) In one terminal, do this:
mkdir qemu-pki
cd qemu-pki
openssl genrsa 2048 > ca-key.pem
openssl req -new -x509 -nodes -days 365000 -key ca-key.pem -out ca-cert.pem
# enter some dummy value for the cert
openssl genrsa 2048 > server-key.pem
openssl req -new -x509 -nodes -days 365000 -key server-key.pem \
-out server-cert.pem
# enter some other dummy values for the cert
gnutls-serv --echo --x509cafile ca-cert.pem --x509keyfile server-key.pem \
--x509certfile server-cert.pem -p 8338
2) In another terminal, do this:
wget https://download.fedoraproject.org/pub/fedora-secondary/releases/39/Cloud/s390x/images/Fedora-Cloud-Base-39-1.5.s390x.qcow2
qemu-system-s390x -nographic -nodefaults \
-hda Fedora-Cloud-Base-39-1.5.s390x.qcow2 \
-object tls-creds-x509,id=tls0,endpoint=client,verify-peer=false,dir=$PWD/qemu-pki \
-chardev socket,id=tls_chardev,host=localhost,port=8338,tls-creds=tls0 \
-device sclpconsole,chardev=tls_chardev,id=tls_serial
QEMU then aborts after a second or two with:
qemu-system-s390x: ../hw/char/sclpconsole.c:73: chr_read: Assertion
`size <= SIZE_BUFFER_VT220 - scon->iov_data_len' failed.
Aborted (core dumped)
It looks like the second read does not trigger the chr_can_read() function
to be called before the second read, which should normally always be done
before sending bytes to a character device to see how much it can handle,
so the s->max_size in tcp_chr_read() still contains the old value from the
previous read. Let's make sure that we use the up-to-date value by calling
tcp_chr_read_poll() again here.
Fixes: ffda5db65a ("io/channel-tls: fix handling of bigger read buffers")
Buglink: https://issues.redhat.com/browse/RHEL-24614
Reviewed-by: "Daniel P. Berrangé" <berrange@redhat.com>
Message-ID: <20240229104339.42574-1-thuth@redhat.com>
Reviewed-by: Antoine Damhet <antoine.damhet@blade-group.com>
Tested-by: Antoine Damhet <antoine.damhet@blade-group.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
The __linux__ version of qemu_chr_open_pp_fd() tries to claim the
parport device with a PPCLAIM ioctl(). On success, it stores the file
descriptor in the chardev object, and returns success. On failure, it
closes the file descriptor, and returns failure.
chardev_new() then passes the Chardev to object_unref(). This duly
calls char_parallel_finalize(), which closes the file descriptor
stored in the chardev object. Since qemu_chr_open_pp_fd() didn't
store it, it's still zero, so this closes standard input. Ooopsie.
To demonstate, add a unit test. With the bug above unfixed, running
this test closes standard input. char_hotswap_test() happens to run
next. It opens a socket, duly gets file descriptor 0, and since it
tests for success with > 0 instead of >= 0, it fails.
The new unit test needs to be conditional exactly like the chardev it
tests. Since the condition is rather complicated, steal the solution
from the serial chardev: define HAVE_CHARDEV_PARALLEL in qemu/osdep.h.
This also permits simplifying chardev/meson.build a bit.
The bug fix is easy enough: store the file descriptor, and leave
closing it to char_parallel_finalize().
The next commit will fix char_hotswap_test()'s test for success.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-ID: <20240203080228.2766159-2-armbru@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[Test fixed up for BSDs, indentation fixed up, commit message improved]
SocketAddress branch @fd is documented in enum SocketAddressType,
unlike the other branches. That's because the branch's type is String
from common.json.
Use a local copy of String, so we can put the documentation in the
usual place.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-ID: <20240205074709.3613229-14-armbru@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
The chardev socket backend will unref the QIOChannel object while
it is still potentially open. When using TLS there could be a
pending TLS handshake taking place. If the channel is left open
then when the TLS handshake callback runs, it can end up accessing
free'd memory in the tcp_chr_tls_handshake method.
Closing the QIOChannel will unregister any pending handshake
source.
Reported-by: jiangyegen <jiangyegen@huawei.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
The function qemu_chr_fe_init already treats be->fe_open as a bool and
if it acts like a bool it should be one. While we are at it make the
variable name more descriptive and add kdoc decorations.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20231211145959.93759-1-alex.bennee@linaro.org>
Record/replay fixes for replay_kernel tests
- add a 32 bit x86 replay test case
- fix some typos
- use modern snapshot setting for tests
- update replay_dump for current ABI
- remove stale replay variables
- improve kdoc for ReplayState
- introduce common error path for replay
- always fully drain chardevs when in replay
- catch unexpected waitio on playback
- remove flaky tags from replay_kernel tests
# -----BEGIN PGP SIGNATURE-----
#
# iQEzBAABCgAdFiEEZoWumedRZ7yvyN81+9DbCVqeKkQFAmWcAJgACgkQ+9DbCVqe
# KkS/TQf+PuIPtuX71ENajfRBjz6450IbGqLUJ1HEaPGYGRj+fR6rg5g5u8qaBrT7
# TUv9ef9L22NtyL+Gbs1OGpGDWKoqV6RQc+A/MHa8IKFpcS24nUo3k4psIC6NSGRH
# 6w3++fPC1Q5cDk9Lei3Qt8fXzcnUZz+NTiIK05aC0xh7D6uGfdADvKqHeLav7qi+
# X2ztNdBsy/WJWCuWcMVzb/dGwDBtuyyxvqTD4EF+zn+gSYq9od2G8XdF+0o6ZVLM
# mXEHwNwB6UjOkLt2cYaay59SXcJFvwxKbEGTDnA7T+kgd3rknuBaWdVBIazoSPQh
# +522nPz5qq/3wO1l7+iQXuvd38fWyw==
# =nKRx
# -----END PGP SIGNATURE-----
# gpg: Signature made Mon 08 Jan 2024 14:03:04 GMT
# gpg: using RSA key 6685AE99E75167BCAFC8DF35FBD0DB095A9E2A44
# gpg: Good signature from "Alex Bennée (Master Work Key) <alex.bennee@linaro.org>" [full]
# Primary key fingerprint: 6685 AE99 E751 67BC AFC8 DF35 FBD0 DB09 5A9E 2A44
* tag 'pull-replay-fixes-080124-1' of https://gitlab.com/stsquad/qemu:
tests/avocado: remove skips from replay_kernel
chardev: force write all when recording replay logs
replay: stop us hanging in rr_wait_io_event
replay/replay-char: use report_sync_error
replay: introduce a central report point for sync errors
replay: make has_unread_data a bool
replay: add proper kdoc for ReplayState
replay: remove host_clock_last
scripts/replay_dump: track total number of instructions
scripts/replay-dump: update to latest format
tests/avocado: modernise the drive args for replay_linux
tests/avocado: fix typo in replay_linux
tests/avocado: add a simple i386 replay kernel test
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Current error message:
qemu-system-x86_64: -chardev spice,id=foo: Parameter 'driver' expects an abstract device type
while in fact the meaning is in reverse, -chardev expects
a non-abstract device type.
Fixes: 777357d758 ("chardev: qom-ify" 2016-12-07)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
This variable is about the host OS, not the target. It is used a lot
more since the Meson conversion, but the original sin dates back to 2003.
Time to fix it.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
config_targetos is now empty and can be removed; its use in sourcesets
that do not involve target-specific files can be replaced with an empty
dictionary.
In fact, at this point *all* sourcesets that do not involve
target-specific files are just glorified mutable arrays. Enforce that
they never test for symbols in "when:" by computing the set of files
without "strict: false".
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
When starting a guest via libvirt with "virsh start --console ...",
the first second of the console output is missing. This is especially
annoying on s390x that only has a text console by default and no graphical
output - if the bios fails to boot here, the information about what went
wrong is completely lost.
One part of the problem (there is also some things to be done on the
libvirt side) is that QEMU only checks with a 1 second timer whether
the other side of the pty is already connected, so the first second of
the console output is always lost.
This likely used to work better in the past, since the code once checked
for a re-connection during write, but this has been removed in commit
f8278c7d74 ("char-pty: remove the check for connection on write") to avoid
some locking.
To ease the situation here at least a little bit, let's check with g_poll()
whether we could send out the data anyway, even if the connection has not
been marked as "connected" yet. The file descriptor is marked as non-blocking
anyway since commit fac6688a18 ("Do not hang on full PTY"), so this should
not cause any trouble if the other side is not ready for receiving yet.
With this patch applied, I can now successfully see the bios output of
a s390x guest when running it with "virsh start --console" (with a patched
version of virsh that fixes the remaining issues there, too).
Reported-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20230816210743.1319018-1-thuth@redhat.com>
Stop applying config-host.mak to the sourcesets, since it does not
have any more CONFIG_* symbols coming from the command line.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
If the monitor or the serial port use STDIO as backend on Windows 11 host,
e.g. -nographic options is used, the monitor or the guest Linux do not
response to arrow keys.
When Windows creates a console, ENABLE_VIRTUAL_PROCESS_INPUT is disabled
by default. Arrow keys cannot be retrieved by ReadFile or ReadConsoleInput
functions.
Add ENABLE_VIRTUAL_PROCESS_INPUT to the flag which is passed to SetConsoleMode,
when opening stdio console.
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1674
Signed-off-by: Zhang Huasen <huasenzhang@foxmail.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <tencent_8DA57B405D427A560FD40F8FB0C0B1ADDE09@qq.com>
We use the user_ss[] array to hold the user emulation sources,
and the softmmu_ss[] array to hold the system emulation ones.
Hold the latter in the 'system_ss[]' array for parity with user
emulation.
Mechanical change doing:
$ sed -i -e s/softmmu_ss/system_ss/g $(git grep -l softmmu_ss)
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230613133347.82210-10-philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Daniel P. Berrangé
Markus ArmbrusterandMichael Tokarev
Thomas Huth
Peter Maydell
Alex Bennée
Paolo Bonzini
Philippe Mathieu-Daudé
Zhang HuasenandMarc-André Lureau
