It is incorrect to have the VIRTIO_NET_HDR_F_NEEDS_CSUM set when
checksum offloading is disabled so clear the bit.
TCP/UDP checksum is usually offloaded when the peer requires virtio
headers because they can instruct the peer to compute checksum. However,
igb disables TX checksum offloading when a VF is enabled whether the
peer requires virtio headers because a transmitted packet can be routed
to it and it expects the packet has a proper checksum. Therefore, it
is necessary to have a correct virtio header even when checksum
offloading is disabled.
A real TCP/UDP checksum will be computed and saved in the buffer when
checksum offloading is disabled. The virtio specification requires to
set the packet checksum stored in the buffer to the TCP/UDP pseudo
header when the VIRTIO_NET_HDR_F_NEEDS_CSUM bit is set so the bit must
be cleared in that case.
Fixes: ffbd2dbd8e ("e1000e: Perform software segmentation for loopback")
Buglink: https://issues.redhat.com/browse/RHEL-23067
Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
(cherry picked from commit 89a8de364b)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
virtio_net_guest_notifier_pending() and virtio_net_guest_notifier_mask()
checked VIRTIO_NET_F_MQ to know there are multiple queues, but
VIRTIO_NET_F_RSS also enables multiple queues. Refer to n->multiqueue,
which is set to true either of VIRTIO_NET_F_MQ or VIRTIO_NET_F_RSS is
enabled.
Fixes: 68b0a6395f ("virtio-net: align ctrl_vq index for non-mq guest for vhost_vdpa")
Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
(cherry picked from commit 1c188fc8cb)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Currently, QEMU only sets the iforce register to 0 and returns early
when claiming the iforce register. However, this may leave mip.meip
remains at 1 if a spurious external interrupt triggered by iforce
register is the only pending interrupt to be claimed, and the interrupt
cannot be lowered as expected.
This commit fixes this issue by calling riscv_aplic_idc_update() to
update the IDC status after the iforce register is claimed.
Signed-off-by: Frank Chang <frank.chang@sifive.com>
Reviewed-by: Jim Shu <jim.shu@sifive.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240321104951.12104-1-frank.chang@sifive.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
(cherry picked from commit 078189b327)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
We're going to make changes that will required each helper to be
responsible for the 'vstart' management, i.e. we will relieve the
'vstart < vl' assumption that helpers have today.
Helpers are usually able to deal with vstart >= vl, i.e. doing nothing
aside from setting vstart = 0 at the end, but the tail update functions
will update the tail regardless of vstart being valid or not. Unifying
the tail update process in a single function that would handle the
vstart >= vl case isn't trivial (see [1] for more info).
This patch takes a blunt approach: do an early exit in every single
vector helper if vstart >= vl, unless the helper is guarded with
vstart_eq_zero in the translation. For those cases the helper is ready
to deal with cases where vl might be zero, i.e. throwing exceptions
based on it like vcpop_m() and first_m().
Helpers that weren't changed:
- vcpop_m(), vfirst_m(), vmsetm(), GEN_VEXT_VIOTA_M(): these are guarded
directly with vstart_eq_zero;
- GEN_VEXT_VCOMPRESS_VM(): guarded with vcompress_vm_check() that checks
vstart_eq_zero;
- GEN_VEXT_RED(): guarded with either reduction_check() or
reduction_widen_check(), both check vstart_eq_zero;
- GEN_VEXT_FRED(): guarded with either freduction_check() or
freduction_widen_check(), both check vstart_eq_zero.
Another exception is vext_ldst_whole(), who operates on effective vector
length regardless of the current settings in vtype and vl.
[1] https://lore.kernel.org/qemu-riscv/1590234b-0291-432a-a0fa-c5a6876097bc@linux.alibaba.com/
Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240314175704.478276-7-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
(cherry picked from commit df4252b2ec)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
These insns have 2 paths: we'll either have vstart already cleared if
vstart_eq_zero or we'll do a brcond to check if vstart >= maxsz to call
the 'vmvr_v' helper. The helper will clear vstart if it executes until
the end, or if vstart >= vl.
For starters, the check itself is wrong: we're checking vstart >= maxsz,
when in fact we should use vstart in bytes, or 'startb' like 'vmvr_v' is
calling, to do the comparison. But even after fixing the comparison we'll
still need to clear vstart in the end, which isn't happening too.
We want to make the helpers responsible to manage vstart, including
these corner cases, precisely to avoid these situations:
- remove the wrong vstart >= maxsz cond from the translation;
- add a 'startb >= maxsz' cond in 'vmvr_v', and clear vstart if that
happens.
This way we're now sure that vstart is being cleared in the end of the
execution, regardless of the path taken.
Fixes: f714361ed7 ("target/riscv: rvv-1.0: implement vstart CSR")
Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: LIU Zhiwei <zhiwei_liu@linux.alibaba.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-ID: <20240314175704.478276-5-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
(cherry picked from commit 7e53e3ddf6)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
trans_vmv_x_s, trans_vmv_s_x, trans_vfmv_f_s and trans_vfmv_s_f aren't
setting vstart = 0 after execution. This is usually done by a helper in
vector_helper.c but these functions don't use helpers.
We'll set vstart after any potential 'over' brconds, and that will also
mandate a mark_vs_dirty() too.
Fixes: dedc53cbc9 ("target/riscv: rvv-1.0: integer scalar move instructions")
Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240314175704.478276-3-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
(cherry picked from commit 0848f7c18e)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
In qemu monitor mode, when we use gpa2hva command to print the host
virtual address corresponding to a guest physical address, if the gpa is
not in RAM, the error message is below:
(qemu) gpa2hva 0x750000000
Memory at address 0x750000000is not RAM
A space is missed between '0x750000000' and 'is'.
Signed-off-by: Yao Xingtao <yaoxt.fnst@fujitsu.com>
Fixes: e9628441df ("hmp: gpa2hva and gpa2hpa hostaddr command")
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Dr. David Alan Gilbert <dave@treblig.org>
Message-ID: <20240319021610.2423844-1-ruansy.fnst@fujitsu.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
(cherry picked from commit a158c63b3b)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
The io_timeout property, introduced in c9b6609 (part of 6.0) is
silently overwritten by the hardcoded default value of 30 seconds
(DEFAULT_IO_TIMEOUT) in scsi_generic_realize because that function is
being called after the properties have already been applied.
The property definition already has a default value which is applied
correctly when no value is explicitly set, so we can just remove the
code which overrides the io_timeout completely.
This has been tested by stracing SG_IO operations with the io_timeout
property set and unset and now sets the timeout field in the ioctl
request to the proper value.
Fixes: c9b6609b69 ("scsi: make io_timeout configurable")
Signed-off-by: Lorenz Brun <lorenz@brun.one>
Message-ID: <20240315145831.2531695-1-lorenz@brun.one>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
(cherry picked from commit 7c7a9f578e)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
CXL emulation of interleave requires read and write hooks due to
requirement for subpage granularity. The Linux kernel stack now enables
using this memory as conventional memory in a separate NUMA node. If a
process is deliberately forced to run from that node
$ numactl --membind=1 ls
the page table walk on i386 fails.
Useful part of backtrace:
(cpu=cpu@entry=0x555556fd9000, fmt=fmt@entry=0x555555fe3378 "cpu_io_recompile: could not find TB for pc=%p")
at ../../cpu-target.c:359
(retaddr=0, addr=19595792376, attrs=..., xlat=<optimized out>, cpu=0x555556fd9000, out_offset=<synthetic pointer>)
at ../../accel/tcg/cputlb.c:1339
(cpu=0x555556fd9000, full=0x7fffee0d96e0, ret_be=ret_be@entry=0, addr=19595792376, size=size@entry=8, mmu_idx=4, type=MMU_DATA_LOAD, ra=0) at ../../accel/tcg/cputlb.c:2030
(cpu=cpu@entry=0x555556fd9000, p=p@entry=0x7ffff56fddc0, mmu_idx=<optimized out>, type=type@entry=MMU_DATA_LOAD, memop=<optimized out>, ra=ra@entry=0) at ../../accel/tcg/cputlb.c:2356
(cpu=cpu@entry=0x555556fd9000, addr=addr@entry=19595792376, oi=oi@entry=52, ra=ra@entry=0, access_type=access_type@entry=MMU_DATA_LOAD) at ../../accel/tcg/cputlb.c:2439
at ../../accel/tcg/ldst_common.c.inc:301
at ../../target/i386/tcg/sysemu/excp_helper.c:173
(err=0x7ffff56fdf80, out=0x7ffff56fdf70, mmu_idx=0, access_type=MMU_INST_FETCH, addr=18446744072116178925, env=0x555556fdb7c0)
at ../../target/i386/tcg/sysemu/excp_helper.c:578
(cs=0x555556fd9000, addr=18446744072116178925, size=<optimized out>, access_type=MMU_INST_FETCH, mmu_idx=0, probe=<optimized out>, retaddr=0) at ../../target/i386/tcg/sysemu/excp_helper.c:604
Avoid this by plumbing the address all the way down from
x86_cpu_tlb_fill() where is available as retaddr to the actual accessors
which provide it to probe_access_full() which already handles MMIO accesses.
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2180
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2220
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Suggested-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Gregory Price <gregory.price@memverge.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Message-ID: <20240307155304.31241-2-Jonathan.Cameron@huawei.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
(cherry picked from commit 9dab7bbb01)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Previously, bdrv_pad_request() could not deal with a NULL qiov when
a read needed to be aligned. During prefetch, a stream job will pass a
NULL qiov. Add a test case to cover this scenario.
By accident, also covers a previous race during shutdown, where block
graph changes during iteration in bdrv_flush_all() could lead to
unreferencing the wrong block driver state and an assertion failure
later.
Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
Message-ID: <20240322095009.346989-5-f.ebner@proxmox.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit 12d7b3bbd3)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Same rationale as for commit "block-backend: fix edge case in
bdrv_next() where BDS associated to BB changes". The block graph might
change between the bdrv_next() call and the bdrv_next_cleanup() call,
so it could be that the associated BDS is not the same that was
referenced previously anymore. Instead, rely on bdrv_next() to set
it->bs to the BDS it referenced and unreference that one in any case.
Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
Message-ID: <20240322095009.346989-4-f.ebner@proxmox.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit bac09b093e)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Some operations, e.g. block-stream, perform reads while discarding the
results (only copy-on-read matters). In this case, they will pass NULL
as the target QEMUIOVector, which will however trip bdrv_pad_request,
since it wants to extend its passed vector. In particular, this is the
case for the blk_co_preadv() call in stream_populate().
If there is no qiov, no operation can be done with it, but the bytes
and offset still need to be updated, so the subsequent aligned read
will actually be aligned and not run into an assertion failure.
In particular, this can happen when the request alignment of the top
node is larger than the allocated part of the bottom node, in which
case padding becomes necessary. For example:
> ./qemu-img create /tmp/backing.qcow2 -f qcow2 64M -o cluster_size=32768
> ./qemu-io -c "write -P42 0x0 0x1" /tmp/backing.qcow2
> ./qemu-img create /tmp/top.qcow2 -f qcow2 64M -b /tmp/backing.qcow2 -F qcow2
> ./qemu-system-x86_64 --qmp stdio \
> --blockdev qcow2,node-name=node0,file.driver=file,file.filename=/tmp/top.qcow2 \
> <<EOF
> {"execute": "qmp_capabilities"}
> {"execute": "blockdev-add", "arguments": { "driver": "compress", "file": "node0", "node-name": "node1" } }
> {"execute": "block-stream", "arguments": { "job-id": "stream0", "device": "node1" } }
> EOF
Originally-by: Stefan Reiter <s.reiter@proxmox.com>
Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
[FE: do update bytes and offset in any case
add reproducer to commit message]
Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
Message-ID: <20240322095009.346989-2-f.ebner@proxmox.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit 3f934817c8)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
VDUSE requires that virtqueues are first enabled before the DRIVER_OK
status flag is set; with the current API of the kernel module, it is
impossible to enable the opposite order in our block export code because
userspace is not notified when a virtqueue is enabled.
This requirement also mathces the normal initialisation order as done by
the generic vhost code in QEMU. However, commit 6c482547 accidentally
changed the order for vdpa-dev and broke access to VDUSE devices with
this.
This changes vdpa-dev to use the normal order again and use the standard
vhost callback .vhost_set_vring_enable for this. VDUSE devices can be
used with vdpa-dev again after this fix.
vhost_net intentionally avoided enabling the vrings for vdpa and does
this manually later while it does enable them for other vhost backends.
Reflect this in the vhost_net code and return early for vdpa, so that
the behaviour doesn't change for this device.
Cc: qemu-stable@nongnu.org
Fixes: 6c4825476a ('vdpa: move vhost_vdpa_set_vring_ready to the caller')
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-ID: <20240315155949.86066-1-kwolf@redhat.com>
Reviewed-by: Eugenio Pérez <eperezma@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit 2c66de61f8)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
The macOS jobs in our CI recently started failing, complaining that
the distutils module is not available anymore. And indeed, according to
https://peps.python.org/pep-0632/ it's been deprecated since a while
and now likely got removed in recent Python versions.
Fortunately, we only use it for a version check via LooseVersion here
which we don't really need anymore - according to Repology.org, these
are the versions of sphinx-rtd-theme that are currently used by the
various distros:
centos_stream_8: 0.3.1
centos_stream_9: 0.5.1
fedora_38: 1.1.1
fedora_39: 1.2.2
freebsd: 1.0.0
haikuports_master: 1.2.1
openbsd: 1.2.2
opensuse_leap_15_5: 0.5.1
pkgsrc_current: 2.0.0
debian_11: 0.5.1
debian_12: 1.2.0
ubuntu_20_04: 0.4.3
ubuntu_22_04: 1.0.0
ubuntu_24_04: 2.0.0
So except for CentOS 8, all distros are using a newer version of
sphinx-rtd-theme, and for CentOS 8 we don't support compiling with
the Sphinx of the distro anymore anyway, since it's based on the
Python 3.6 interpreter there. For compiling on CentOS 8, you have
to use the alternative Python 3.8 interpreter which comes without
Sphinx, so that needs the Sphinx installed via pip in the venv
instead, and that is using a newer version, too, according to our
pythondeps.toml file.
Thus we can simply drop the version check now to get rid of the
distutils dependency here.
Signed-off-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Michael Tokarev <mjt@tls.msk.ru>
Message-id: 20240304130403.129543-1-thuth@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
(cherry picked from commit eb844330bd)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
monitor_puts() doesn't check the monitor pointer, but do_inject_x86_mce()
may have a parameter with NULL monitor pointer. Revert monitor_puts() in
do_inject_x86_mce() to fix, then the fact that we send the same message to
monitor and log is again more obvious.
Fixes: bf0c50d4aa (monitor: expose monitor_puts to rest of code)
Reviwed-by: Xiaoyao Li <xiaoyao.li@intel.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Tao Su <tao1.su@linux.intel.com>
Message-ID: <20240320083640.523287-1-tao1.su@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit 7fd226b047)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Building dbus-display1.c explicitly as a static library drops -fPIC by
default, which may not be correct if it ends up linked to a shared
library.
Let the target decide how to build the unit, with or without -fPIC. This
makes commit 186acfbaf7 ("tests/qtest: Depend on dbus_display1_dep") no
longer relevant, as dbus-display1.c will be recompiled.
Fixes: c172136ea3 ("meson: ensure dbus-display generated code is built
before other units")
Reported-by: Olaf Hering <olaf@aepfle.de>
Reviewed-by: Michael Tokarev <mjt@tls.msk.ru>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
(cherry picked from commit d4069a84a3)
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>
(cherry picked from commit e8ee827ffd)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
The low bit of MMU indices for x86 TCG indicates whether the processor is
in 32-bit mode and therefore linear addresses have to be masked to 32 bits.
However, the index was computed incorrectly, leading to possible conflicts
in the TLB for any address above 4G.
Analyzed-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Fixes: b1661801c1 ("target/i386: Fix physical address truncation", 2024-02-28)
Fixes: a28b6b4e74 ("target/i386: Fix physical address truncation" in stable-8.2)
Cc: qemu-stable@nongnu.org
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2206
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit 2cc68629a6)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(Mjt: move changes for x86_cpu_mmu_index() to cpu_mmu_index() due to missing
v8.2.0-1030-gace0c5fe59 "target/i386: Populate CPUClass.mmu_index")
Accesses from a 32-bit environment (32-bit code segment for instruction
accesses, EFER.LMA==0 for processor accesses) have to mask away the
upper 32 bits of the address. While a bit wasteful, the easiest way
to do so is to use separate MMU indexes. These days, QEMU anyway is
compiled with a fixed value for NB_MMU_MODES. Split MMU_USER_IDX,
MMU_KSMAP_IDX and MMU_KNOSMAP_IDX in two.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit 90f641531c)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(Mjt: move changes for x86_cpu_mmu_index() to cpu_mmu_index() due to missing
v8.2.0-1030-gace0c5fe59 "target/i386: Populate CPUClass.mmu_index")
Remove knowledge of specific MMU indexes (other than MMU_NESTED_IDX and
MMU_PHYS_IDX) from mmu_translate(). This will make it possible to split
32-bit and 64-bit MMU indexes.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit 5f97afe254)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(Mjt: context fixup in target/i386/cpu.h due to other changes in that area)
We're seeing timeouts for this test on CI runs (specifically for
ubuntu-20.04-s390x-all). It doesn't fail consistently, but even the
successful runs take about 27 or 28 seconds, which is not very far from
the 30 seconds timeout.
Bump the timeout a bit to make failure less likely even on this CI host.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-ID: <20240125165803.48373-1-kwolf@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit 63b18312d1)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
When running the tests in slow mode on a very loaded system and with
--enable-debug, the test-crypto-block can take longer than 4 minutes.
Bump the timeout to 5 minutes to make sure that it also passes in
such situations.
Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20231215070357.10888-15-thuth@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
(cherry picked from commit e1b363e328)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
When running the tests in slow mode on a very loaded system and with
--enable-debug, the test-aio-multithread can take longer than 1 minute.
Bump the timeout to two minutes to make sure that it also passes in
such situations.
Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20231215070357.10888-14-thuth@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
(cherry picked from commit c45f8f1aef)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
This replicates the scenario in which the bug was reported.
Unfortunately this relies on actually executing a guest (so that the
firmware initialises the virtio-blk device and moves it to its
configured iothread), so this can't make use of the qtest accelerator
like most other test cases. I tried to find a different easy way to
trigger the bug, but couldn't find one.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-ID: <20240314165825.40261-3-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit e8fce34ecc)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
When draining an NBD export, nbd_drained_begin() first sets
client->quiescing so that nbd_client_receive_next_request() won't start
any new request coroutines. Then nbd_drained_poll() tries to makes sure
that we wait for any existing request coroutines by checking that
client->nb_requests has become 0.
However, there is a small window between creating a new request
coroutine and increasing client->nb_requests. If a coroutine is in this
state, it won't be waited for and drain returns too early.
In the context of switching to a different AioContext, this means that
blk_aio_attached() will see client->recv_coroutine != NULL and fail its
assertion.
Fix this by increasing client->nb_requests immediately when starting the
coroutine. Doing this after the checks if we should create a new
coroutine is okay because client->lock is held.
Cc: qemu-stable@nongnu.org
Fixes: fd6afc501a ("nbd/server: Use drained block ops to quiesce the server")
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-ID: <20240314165825.40261-2-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit 9c707525cb)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
NBDClient has a number of fields that are accessed by both the export
AioContext and the main loop thread. When the AioContext lock is removed
these fields will need another form of protection.
Add NBDClient->lock and protect fields that are accessed by both
threads. Also add assertions where possible and otherwise add doc
comments stating assumptions about which thread and lock holding.
Note this patch moves the client->recv_coroutine assertion from
nbd_co_receive_request() to nbd_trip() where client->lock is held.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-ID: <20231221192452.1785567-7-stefanha@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit 7075d23511)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
The NBD clients list is currently accessed from both the export
AioContext and the main loop thread. When the AioContext lock is removed
there will be nothing protecting the clients list.
Adding a lock around the clients list is tricky because NBDClient
structs are refcounted and may be freed from the export AioContext or
the main loop thread. nbd_export_request_shutdown() -> client_close() ->
nbd_client_put() is also tricky because the list lock would be held
while indirectly dropping references to NDBClients.
A simpler approach is to only allow nbd_client_put() and client_close()
calls from the main loop thread. Then the NBD clients list is only
accessed from the main loop thread and no fancy locking is needed.
nbd_trip() just needs to reschedule itself in the main loop AioContext
before calling nbd_client_put() and client_close(). This costs more CPU
cycles per NBD request so add nbd_client_put_nonzero() to optimize the
common case where more references to NBDClient remain.
Note that nbd_client_get() can still be called from either thread, so
make NBDClient->refcount atomic.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-ID: <20231221192452.1785567-6-stefanha@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit f816310d0c)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Calling job_pause_point() while holding the graph reader lock
potentially results in a deadlock: bdrv_graph_wrlock() first drains
everything, including the mirror job, which pauses it. The job is only
unpaused at the end of the drain section, which is when the graph writer
lock has been successfully taken. However, if the job happens to be
paused at a pause point where it still holds the reader lock, the writer
lock can't be taken as long as the job is still paused.
Mark job_pause_point() as GRAPH_UNLOCKED and fix mirror accordingly.
Cc: qemu-stable@nongnu.org
Buglink: https://issues.redhat.com/browse/RHEL-28125
Fixes: 004915a96a ("block: Protect bs->backing with graph_lock")
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-ID: <20240313153000.33121-1-kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit ae5a40e858)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
The block .save_setup() handler calls a helper routine
init_blk_migration() which builds a list of block devices to take into
account for migration. When one device is found to be empty (sectors
== 0), the loop exits and all the remaining devices are ignored. This
is a regression introduced when bdrv_iterate() was removed.
Change that by skipping only empty devices.
Cc: Markus Armbruster <armbru@redhat.com>
Cc: qemu-stable <qemu-stable@nongnu.org>
Suggested-by: Kevin Wolf <kwolf@redhat.com>
Fixes: fea68bb6e9 ("block: Eliminate bdrv_iterate(), use bdrv_next()")
Signed-off-by: Cédric Le Goater <clg@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Link: https://lore.kernel.org/r/20240312120431.550054-1-clg@redhat.com
[peterx: fix "Suggested-by:"]
Signed-off-by: Peter Xu <peterx@redhat.com>
(cherry picked from commit 2e128776dc)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
The payload size returned by command VIRTIO_SND_R_PCM_INFO is
wrong. The code in process_cmd() assumes that all commands
return only a virtio_snd_hdr payload, but some commands like
VIRTIO_SND_R_PCM_INFO may return an additional payload.
Add a zero initialized payload_size variable to struct
virtio_snd_ctrl_command to allow for additional payloads.
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Volker Rümelin <vr_qemu@t-online.de>
Message-Id: <20240218083351.8524-1-vr_qemu@t-online.de>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
(cherry picked from commit 633487df8d)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
