r[id]tlb[01], [iw][id]tlb opcodes use TLB way index passed in a register
by the guest. The host uses 3 bits of the index for ITLB indexing and 4
bits for DTLB, but there's only 7 entries in the ITLB array and 10 in
the DTLB array, so a malicious guest may trigger out-of-bound access to
these arrays.
Change split_tlb_entry_spec return type to bool to indicate whether TLB
way passed to it is valid. Change get_tlb_entry to return NULL in case
invalid TLB way is requested. Add assertion to xtensa_tlb_get_entry that
requested TLB way and entry indices are valid. Add checks to the
[rwi]tlb helpers that requested TLB way is valid and return 0 or do
nothing when it's not.
Cc: qemu-stable@nongnu.org
Fixes: b67ea0cd74 ("target-xtensa: implement memory protection options")
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20231215120307.545381-1-jcmvbkbc@gmail.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
(cherry picked from commit 604927e357)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
On a loaded system with --enable-debug, this test can take longer than
5 minutes. Raising the timeout to 6 minutes gives greater headroom for
such situations.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
[thuth: Increase the timeout to 6 minutes for very loaded systems]
Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20231215070357.10888-11-thuth@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
(cherry picked from commit e8a12fe31f)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(Mjt: context fixup in tests/qtest/meson.build)
monitor_qmp_dispatcher_co() runs in the iohandler AioContext that is not
polled during nested event loops. The coroutine currently reschedules
itself in the main loop's qemu_aio_context AioContext, which is polled
during nested event loops. One known problem is that QMP device-add
calls drain_call_rcu(), which temporarily drops the BQL, leading to all
sorts of havoc like other vCPU threads re-entering device emulation code
while another vCPU thread is waiting in device emulation code with
aio_poll().
Paolo Bonzini suggested running non-coroutine QMP handlers in the
iohandler AioContext. This avoids trouble with nested event loops. His
original idea was to move coroutine rescheduling to
monitor_qmp_dispatch(), but I resorted to moving it to qmp_dispatch()
because we don't know if the QMP handler needs to run in coroutine
context in monitor_qmp_dispatch(). monitor_qmp_dispatch() would have
been nicer since it's associated with the monitor implementation and not
as general as qmp_dispatch(), which is also used by qemu-ga.
A number of qemu-iotests need updated .out files because the order of
QMP events vs QMP responses has changed.
Solves Issue #1933.
Cc: qemu-stable@nongnu.org
Fixes: 7bed89958b ("device_core: use drain_call_rcu in in qmp_device_add")
Buglink: https://bugzilla.redhat.com/show_bug.cgi?id=2215192
Buglink: https://bugzilla.redhat.com/show_bug.cgi?id=2214985
Buglink: https://issues.redhat.com/browse/RHEL-17369
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-ID: <20240118144823.1497953-4-stefanha@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Tested-by: Fiona Ebner <f.ebner@proxmox.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit effd60c878)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(Mjt: omit changes to tests missing in 7.2)
The common.qemu bash functions allow tests to interact with the QMP
monitor of a QEMU process. I spent two days trying to update 141 when
the order of the test output changed, but found it would still fail
occassionally because printf() and QMP events race with synchronous QMP
communication.
I gave up and ported 141 to the existing Python API for QMP tests. The
Python API is less affected by the order in which QEMU prints output
because it does not print all QMP traffic by default.
The next commit changes the order in which QMP messages are received.
Make 141 reliable first.
Cc: Hanna Czenczek <hreitz@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-ID: <20240118144823.1497953-3-stefanha@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit 9ee2dd4c22)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Add a filter function for QMP responses that contain QEMU's
automatically generated node ids. The ids change between runs and must
be masked in the reference output.
The next commit will use this new function.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-ID: <20240118144823.1497953-2-stefanha@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit da62b507a2)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(Mjt: context fix in tests/qemu-iotests/iotests.py due to
v7.2.0-939-gbcc6777ad6 "iotests: Filter child node information")
There is a bug in the blklogwrites driver pertaining to logging "write
zeroes" operations, causing log corruption. This can be easily observed
by setting detect-zeroes to something other than "off" for the driver.
The issue is caused by a concurrency bug pertaining to the fact that
"write zeroes" operations have to be logged in two parts: first the log
entry metadata, then the zeroed-out region. While the log entry
metadata is being written by bdrv_co_pwritev(), another operation may
begin in the meanwhile and modify the state of the blklogwrites driver.
This is as intended by the coroutine-driven I/O model in QEMU, of
course.
Unfortunately, this specific scenario is mishandled. A short example:
1. Initially, in the current operation (#1), the current log sector
number in the driver state is only incremented by the number of sectors
taken by the log entry metadata, after which the log entry metadata is
written. The current operation yields.
2. Another operation (#2) may start while the log entry metadata is
being written. It uses the current log position as the start offset for
its log entry. This is in the sector right after the operation #1 log
entry metadata, which is bad!
3. After bdrv_co_pwritev() returns (#1), the current log sector
number is reread from the driver state in order to find out the start
offset for bdrv_co_pwrite_zeroes(). This is an obvious blunder, as the
offset will be the sector right after the (misplaced) operation #2 log
entry, which means that the zeroed-out region begins at the wrong
offset.
4. As a result of the above, the log is corrupt.
Fix this by only reading the driver metadata once, computing the
offsets and sizes in one go (including the optional zeroed-out region)
and setting the log sector number to the appropriate value for the next
operation in line.
Signed-off-by: Ari Sundholm <ari@tuxera.com>
Cc: qemu-stable@nongnu.org
Message-ID: <20240109184646.1128475-1-megari@gmx.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit a9c8ea9547)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
When HASH_REPORT is negotiated, the guest_hdr_len might be larger than
the size of the mergeable rx buffer header. Using
virtio_net_hdr_mrg_rxbuf during the header swap might lead a stack
overflow in this case. Fixing this by using virtio_net_hdr_v1_hash
instead.
Reported-by: Xiao Lei <leixiao.nop@zju.edu.cn>
Cc: Yuri Benditovich <yuri.benditovich@daynix.com>
Cc: qemu-stable@nongnu.org
Cc: Mauro Matteo Cascella <mcascell@redhat.com>
Fixes: CVE-2023-6693
Fixes: e22f0603fb ("virtio-net: reference implementation of hash report")
Reviewed-by: Michael Tokarev <mjt@tls.msk.ru>
Signed-off-by: Jason Wang <jasowang@redhat.com>
(cherry picked from commit 2220e8189f)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Using fleecing backup like in [0] on a qcow2 image (with metadata
preallocation) can lead to the following assertion failure:
> bdrv_co_do_block_status: Assertion `!(ret & BDRV_BLOCK_ZERO)' failed.
In the reproducer [0], it happens because the BDRV_BLOCK_RECURSE flag
will be set by the qcow2 driver, so the caller will recursively check
the file child. Then the BDRV_BLOCK_ZERO set too. Later up the call
chain, in bdrv_co_do_block_status() for the snapshot-access driver,
the assertion failure will happen, because both flags are set.
To fix it, clear the recurse flag after the recursive check was done.
In detail:
> #0 qcow2_co_block_status
Returns 0x45 = BDRV_BLOCK_RECURSE | BDRV_BLOCK_DATA |
BDRV_BLOCK_OFFSET_VALID.
> #1 bdrv_co_do_block_status
Because of the data flag, bdrv_co_do_block_status() will now also set
BDRV_BLOCK_ALLOCATED. Because of the recurse flag,
bdrv_co_do_block_status() for the bdrv_file child will be called,
which returns 0x16 = BDRV_BLOCK_ALLOCATED | BDRV_BLOCK_OFFSET_VALID |
BDRV_BLOCK_ZERO. Now the return value inherits the zero flag.
Returns 0x57 = BDRV_BLOCK_RECURSE | BDRV_BLOCK_DATA |
BDRV_BLOCK_OFFSET_VALID | BDRV_BLOCK_ALLOCATED | BDRV_BLOCK_ZERO.
> #2 bdrv_co_common_block_status_above
> #3 bdrv_co_block_status_above
> #4 bdrv_co_block_status
> #5 cbw_co_snapshot_block_status
> #6 bdrv_co_snapshot_block_status
> #7 snapshot_access_co_block_status
> #8 bdrv_co_do_block_status
Return value is propagated all the way up to here, where the assertion
failure happens, because BDRV_BLOCK_RECURSE and BDRV_BLOCK_ZERO are
both set.
> #9 bdrv_co_common_block_status_above
> #10 bdrv_co_block_status_above
> #11 block_copy_block_status
> #12 block_copy_dirty_clusters
> #13 block_copy_common
> #14 block_copy_async_co_entry
> #15 coroutine_trampoline
[0]:
> #!/bin/bash
> rm /tmp/disk.qcow2
> ./qemu-img create /tmp/disk.qcow2 -o preallocation=metadata -f qcow2 1G
> ./qemu-img create /tmp/fleecing.qcow2 -f qcow2 1G
> ./qemu-img create /tmp/backup.qcow2 -f qcow2 1G
> ./qemu-system-x86_64 --qmp stdio \
> --blockdev qcow2,node-name=node0,file.driver=file,file.filename=/tmp/disk.qcow2 \
> --blockdev qcow2,node-name=node1,file.driver=file,file.filename=/tmp/fleecing.qcow2 \
> --blockdev qcow2,node-name=node2,file.driver=file,file.filename=/tmp/backup.qcow2 \
> <<EOF
> {"execute": "qmp_capabilities"}
> {"execute": "blockdev-add", "arguments": { "driver": "copy-before-write", "file": "node0", "target": "node1", "node-name": "node3" } }
> {"execute": "blockdev-add", "arguments": { "driver": "snapshot-access", "file": "node3", "node-name": "snap0" } }
> {"execute": "blockdev-backup", "arguments": { "device": "snap0", "target": "node1", "sync": "full", "job-id": "backup0" } }
> EOF
Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
Message-id: 20240116154839.401030-1-f.ebner@proxmox.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
(cherry picked from commit 8a9be79924)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
This is now expected by rtd so I've expanded using their example as
22.04 is one of our supported platforms. I tried to work out if there
was an easy way to re-generate a requirements.txt from our
pythondeps.toml but in the end went for the easier solution.
Cc: <qemu-stable@nongnu.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20231221174200.2693694-1-alex.bennee@linaro.org>
(cherry picked from commit b16a45bc5e)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Even though the BLAST command isn't fully implemented in QEMU, the DMA_STAT_BCMBLT
bit should be set after the command has been issued to indicate that the command
has completed.
This fixes an issue with the DC390 DOS driver which issues the BLAST command as
part of its normal error recovery routine at startup, and otherwise sits in a
tight loop waiting for DMA_STAT_BCMBLT to be set before continuing.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Guenter Roeck <linux@roeck-us.net>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Message-ID: <20240112131529.515642-5-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
(cherry picked from commit c2d7de557d)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
The setting of DMA_STAT_DONE at the end of a DMA transfer can be configured to
generate an interrupt, however the Linux driver manually checks for DMA_STAT_DONE
being set and if it is, considers that a DMA transfer has completed.
If DMA_STAT_DONE is set but the ESP device isn't indicating an interrupt then
the Linux driver considers this to be a spurious interrupt. However this can
occur in QEMU as there is a delay between the end of DMA transfer where
DMA_STAT_DONE is set, and the ESP device raising its completion interrupt.
This appears to be an incorrect assumption in the Linux driver as the ESP and
PCI DMA interrupt sources are separate (and may not be raised exactly
together), however we can work around this by synchronising the setting of
DMA_STAT_DONE at the end of a DMA transfer with the ESP completion interrupt.
In conjunction with the previous commit Linux is now able to correctly boot
from an am53c974 PCI SCSI device on the hppa C3700 machine without emitting
"iget: checksum invalid" and "Spurious irq, sreg=10" errors.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Guenter Roeck <linux@roeck-us.net>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Message-ID: <20240112131529.515642-4-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
(cherry picked from commit 1e8e6644e0)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
The am53c974/dc390 PCI interrupt has two separate sources: the first is from the
internal ESP device, and the second is from the PCI DMA transfer logic.
Update the ESP interrupt handler so that it sets DMA_STAT_SCSIINT rather than
driving the PCI IRQ directly, and introduce a new esp_pci_update_irq() function
to generate the correct PCI IRQ level. In particular this fixes spurious interrupts
being generated by setting DMA_STAT_DONE at the end of a transfer if DMA_CMD_INTE_D
isn't set in the DMA_CMD register.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Guenter Roeck <linux@roeck-us.net>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Message-ID: <20240112131529.515642-3-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
(cherry picked from commit 6b41417d93)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(Mjt: fixup in hw/scsi/esp-pci.c due to v8.0.0-1556-g7d5b0d6864
"bulk: Remove pointless QOM casts")
The current code in esp_pci_dma_memory_rw() sets the DMA address to the value
of the DMA_SPA (Starting Physical Address) register which is incorrect: this
means that for each callback from the SCSI layer the DMA address is set back
to the starting address.
In the case where only a single SCSI callback occurs (currently for transfer
lengths < 128kB) this works fine, however for larger transfers the DMA address
wraps back to the initial starting address, corrupting the buffer holding the
data transferred to the guest.
Fix esp_pci_dma_memory_rw() to use the DMA_WAC (Working Address Counter) for
the DMA address which is correctly incremented across multiple SCSI layer
transfers.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Guenter Roeck <linux@roeck-us.net>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Message-ID: <20240112131529.515642-2-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
(cherry picked from commit 84a6835e00)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Add an update buffer where all block updates are staged.
Flush or discard updates properly, so we should never see
half-completed block writes in pflash storage.
Drop a bunch of FIXME comments ;)
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20240108160900.104835-4-kraxel@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
(cherry picked from commit 284a7ee2e2)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(Mjt: drop const in hw/block/pflash_cfi01.c for before
v8.2.0-220-g7d5dc0a367 "hw/block: Constify VMState")
Move the offset calculation, do it once at the start of the function and
let the 'p' variable point directly to the memory location which should
be updated. This makes it simpler to update other buffers than
pfl->storage in an upcoming patch. No functional change.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20240108160900.104835-2-kraxel@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
(cherry picked from commit 3b14a555fd)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
For PC-relative translation blocks, env->eip changes during the
execution of a translation block, Therefore, QEMU must be able to
recover an instruction's PC just from the TranslationBlock struct and
the instruction data with. Because a TB will not span two pages, QEMU
stores all the low bits of EIP in the instruction data and replaces them
in x86_restore_state_to_opc. Bits 12 and higher (which may vary between
executions of a PCREL TB, since these only use the physical address in
the hash key) are kept unmodified from env->eip. The assumption is that
these bits of EIP, unlike bits 0-11, will not change as the translation
block executes.
Unfortunately, this is incorrect when the CS base is not aligned to a page.
Then the linear address of the instructions (i.e. the one with the
CS base addred) indeed will never span two pages, but bits 12+ of EIP
can actually change. For example, if CS base is 0x80262200 and EIP =
0x6FF4, the first instruction in the translation block will be at linear
address 0x802691F4. Even a very small TB will cross to EIP = 0x7xxx,
while the linear addresses will remain comfortably within a single page.
The fix is simply to use the low bits of the linear address for data[0],
since those don't change. Then x86_restore_state_to_opc uses tb->cs_base
to compute a temporary linear address (referring to some unknown
instruction in the TB, but with the correct values of bits 12 and higher);
the low bits are replaced with data[0], and EIP is obtained by subtracting
again the CS base.
Huge thanks to Mark Cave-Ayland for the image and initial debugging,
and to Gitlab user @kjliew for help with bisecting another occurrence
of (hopefully!) the same bug.
It should be relatively easy to write a testcase that performs MMIO on
an EIP with different bits 12+ than the first instruction of the translation
block; any help is welcome.
Fixes: e3a79e0e87 ("target/i386: Enable TARGET_TB_PCREL", 2022-10-11)
Cc: qemu-stable@nongnu.org
Cc: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Cc: Richard Henderson <richard.henderson@linaro.org>
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1759
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1964
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2012
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit 729ba8e933)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(Mjt: fixup in target/i386/tcg/tcg-cpu.c target/i386/tcg/translate.c for
v7.2.0-1839-g2e3afe8e19 "target/i386: Replace `TARGET_TB_PCREL` with `CF_PCREL`")
With PCREL, we have a page-relative view of EIP, and an
approximation of PC = EIP+CSBASE that is good enough to
detect page crossings. If we try to recompute PC after
masking EIP, we will mess up that approximation and write
a corrupt value to EIP.
We already handled masking properly for PCREL, so the
fix in b5e0d5d2 was only needed for the !PCREL path.
Cc: qemu-stable@nongnu.org
Fixes: b5e0d5d22f ("target/i386: Fix 32-bit wrapping of pc/eip computation")
Reported-by: Michael Tokarev <mjt@tls.msk.ru>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-ID: <20240101230617.129349-1-richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit a58506b748)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
In 32-bit mode, pc = eip + cs_base is also 32-bit, and must wrap.
Failure to do so results in incorrect memory exceptions to the guest.
Before 732d548732, this was implicitly done via truncation to
target_ulong but only in qemu-system-i386, not qemu-system-x86_64.
To fix this, we must add conditional zero-extensions.
Since we have to test for 32 vs 64-bit anyway, note that cs_base
is always zero in 64-bit mode.
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2022
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20231212172510.103305-1-richard.henderson@linaro.org>
(cherry picked from commit b5e0d5d22f)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(Mjt: context fix in target/i386/tcg/tcg-cpu.c for v8.1.0-1190-gb77af26e97
"accel/tcg: Replace CPUState.env_ptr with cpu_env()")
(Mjt: fixup in target/i386/tcg/tcg-cpu.c for v7.2.0-1854-g34a39c2443
"target/i386: Replace `tb_pc()` with `tb->pc`")
j is used while loading an ELF file to byteswap segments'
data. If data is larger than 2GB an overflow may happen.
So j should be elf_word.
This commit fixes a minor bug: it's unlikely anybody is trying to
load ELF files with 2GB+ segments for wrong-endianness targets,
but if they did, it wouldn't work correctly.
Found by Linux Verification Center (linuxtesting.org) with SVACE.
Cc: qemu-stable@nongnu.org
Fixes: 7ef295ea5b ("loader: Add data swap option to load-elf")
Signed-off-by: Anastasia Belova <abelova@astralinux.ru>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
(cherry picked from commit 410c2a4d75)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Sometimes the CI "pages" job fails with a message like this from
htags:
$ htags -anT --tree-view=filetree -m qemu_init -t "Welcome to the QEMU sourcecode"
htags: Negative exec line limit = -371
This is due to a bug in hflags where if the environment is too large it
falls over:
https://lists.gnu.org/archive/html/bug-global/2024-01/msg00000.html
This happens to us because GitLab CI puts the commit message of the
commit under test into the CI_COMMIT_MESSAGE and/or CI_COMMIT_TAG_MESSAGE
environment variables, so the job will fail if the commit happens to
have a verbose commit message.
Work around the htags bug by unsetting these variables while running
htags.
Cc: qemu-stable@nongnu.org
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2080
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20240111125543.1573473-1-peter.maydell@linaro.org>
Signed-off-by: Thomas Huth <thuth@redhat.com>
(cherry picked from commit 52a21689cd)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
LAE should set the access register corresponding to the first operand,
instead, it always modifies access register 1.
Co-developed-by: Ido Plat <Ido.Plat@ibm.com>
Cc: qemu-stable@nongnu.org
Fixes: a1c7610a68 ("target-s390x: implement LAY and LAEY instructions")
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Message-ID: <20240111092328.929421-2-iii@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
(cherry picked from commit e358a25a97)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(Mjt: target/s390x/tcg/translate.c: fixup for
v8.1.0-1189-gad75a51e84 "tcg: Rename cpu_env to tcg_env" and
v7.2.0-2636-g3ac6f91bca "target/s390x: Drop tcg_temp_free from translate.c")
The hypervisor can deliver (virtual) LPIs to a guest by setting up a
list register to have an intid which is an LPI. The GIC has to treat
these a little differently to standard interrupt IDs, because LPIs
have no Active state, and so the guest will only EOI them, it will
not also deactivate them. So icv_eoir_write() must do two things:
* if the LPI ID is not in any list register, we drop the
priority but do not increment the EOI count
* if the LPI ID is in a list register, we immediately deactivate
it, regardless of the split-drop-and-deactivate control
This can be seen in the VirtualWriteEOIR0() and VirtualWriteEOIR1()
pseudocode in the GICv3 architecture specification.
Without this fix, potentially a hypervisor guest might stall because
LPIs get stuck in a bogus Active+Pending state.
Cc: qemu-stable@nongnu.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Miguel Luis <miguel.luis@oracle.com>
(cherry picked from commit 82a65e3188)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
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>
(cherry picked from commit 4ad87cd4b2)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
The mcycle/minstret counter's stop flag is mistakenly updated on a copy
on stack. Thus the counter increments even when the CY/IR bit in the
mcountinhibit register is set. This commit corrects its behavior.
Fixes: 3780e33732 (target/riscv: Support mcycle/minstret write operation)
Signed-off-by: Xu Lu <luxu.kernel@bytedance.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(cherry picked from commit 5cb0e7abe1)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
We have a few test cases that include tests for corner case aspects of
internal snapshots, but nothing that tests that they actually function
as snapshots or that involves deleting a snapshot. Add a test for this
kind of basic internal snapshot functionality.
The error cases include a regression test for the crash we just fixed
with snapshot operations on inactive images.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-ID: <20231201142520.32255-4-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit bb6e2511eb)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Currently, the conflict between -incoming and -loadvm is only detected
when loading the snapshot fails because the image is still inactive for
the incoming migration. This results in a suboptimal error message:
$ ./qemu-system-x86_64 -hda /tmp/test.qcow2 -loadvm foo -incoming defer
qemu-system-x86_64: Device 'ide0-hd0' is writable but does not support snapshots
Catch the situation already in qemu_validate_options() to improve the
message:
$ ./qemu-system-x86_64 -hda /tmp/test.qcow2 -loadvm foo -incoming defer
qemu-system-x86_64: 'incoming' and 'loadvm' options are mutually exclusive
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-ID: <20231201142520.32255-3-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit 5a7f21efaf)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
bdrv_is_read_only() only checks if the node is configured to be
read-only eventually, but even if it returns false, writing to the node
may not be permitted at the moment (because it's inactive).
bdrv_is_writable() checks that the node can be written to right now, and
this is what the snapshot operations really need.
Change bdrv_can_snapshot() to use bdrv_is_writable() to fix crashes like
the following:
$ ./qemu-system-x86_64 -hda /tmp/test.qcow2 -loadvm foo -incoming defer
qemu-system-x86_64: ../block/io.c:1990: int bdrv_co_write_req_prepare(BdrvChild *, int64_t, int64_t, BdrvTrackedRequest *, int): Assertion `!(bs->open_flags & BDRV_O_INACTIVE)' failed.
The resulting error message after this patch isn't perfect yet, but at
least it doesn't crash any more:
$ ./qemu-system-x86_64 -hda /tmp/test.qcow2 -loadvm foo -incoming defer
qemu-system-x86_64: Device 'ide0-hd0' is writable but does not support snapshots
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-ID: <20231201142520.32255-2-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit d3007d348a)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
There is no architectural requirement that SME implies SVE, but
our implementation currently assumes it. (FEAT_SME_FA64 does
imply SVE.) So if you try to run a CPU with eg "-cpu max,sve=off"
you quickly run into an assert when the guest tries to write to
SMCR_EL1:
#6 0x00007ffff4b38e96 in __GI___assert_fail
(assertion=0x5555566e69cb "sm", file=0x5555566e5b24 "../../target/arm/helper.c", line=6865, function=0x5555566e82f0 <__PRETTY_FUNCTION__.31> "sve_vqm1_for_el_sm") at ./assert/assert.c:101
#7 0x0000555555ee33aa in sve_vqm1_for_el_sm (env=0x555557d291f0, el=2, sm=false) at ../../target/arm/helper.c:6865
#8 0x0000555555ee3407 in sve_vqm1_for_el (env=0x555557d291f0, el=2) at ../../target/arm/helper.c:6871
#9 0x0000555555ee3724 in smcr_write (env=0x555557d291f0, ri=0x555557da23b0, value=2147483663) at ../../target/arm/helper.c:6995
#10 0x0000555555fd1dba in helper_set_cp_reg64 (env=0x555557d291f0, rip=0x555557da23b0, value=2147483663) at ../../target/arm/tcg/op_helper.c:839
#11 0x00007fff60056781 in code_gen_buffer ()
Avoid this unsupported and slightly odd combination by
disabling SME when SVE is not present.
Cc: qemu-stable@nongnu.org
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2005
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20231127173318.674758-1-peter.maydell@linaro.org
(cherry picked from commit f7767ca301)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Commit d921fea338 ("ui/vnc-clipboard: fix infinite loop in
inflate_buffer (CVE-2023-3255)") removed this hunk, but it is still
required, because it can happen that stream.avail_in becomes zero
before coming across a return value of Z_STREAM_END in the loop.
This fixes the host->guest direction of the clipboard with noVNC and
TigerVNC as clients.
Fixes: d921fea338 ("ui/vnc-clipboard: fix infinite loop in inflate_buffer (CVE-2023-3255)")
Reported-by: Friedrich Weber <f.weber@proxmox.com>
Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
Acked-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20231122125826.228189-1-f.ebner@proxmox.com>
(cherry picked from commit ebfbf39467)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Commit 6f189a08c1 ("ui/gtk-egl: Check EGLSurface before doing
scanout") introduced a regression when QEMU is running with a
virtio-gpu-gl-device on a host under X11. After the guest has
initialized the virtio-gpu-gl-device, the guest screen only
shows "Display output is not active.".
Commit 6f189a08c1 moved all function calls in
gd_egl_scanout_texture() to a code path which is only called
once after gd_egl_init() succeeds in gd_egl_scanout_texture().
Move all function calls in gd_egl_scanout_texture() back to
the regular code path so they get always called if one of the
gd_egl_init() calls was successful.
Fixes: 6f189a08c1 ("ui/gtk-egl: Check EGLSurface before doing scanout")
Signed-off-by: Volker Rümelin <vr_qemu@t-online.de>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20231111104020.26183-1-vr_qemu@t-online.de>
(cherry picked from commit 53a939f1bf)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
The first time gd_egl_scanout_texture() is called, there's a possibility
that the GTK drawing area might not be realized yet, in which case its
associated GdkWindow is NULL. This means gd_egl_init() was also skipped
and the EGLContext and EGLSurface stored in the VirtualGfxConsole are
not valid yet.
Continuing with the scanout in this conditions would result in hitting
an assert in libepoxy: "Couldn't find current GLX or EGL context".
A possible workaround is to just ignore the scanout request, giving the
the GTK drawing area some time to finish its realization. At that point,
the gd_egl_init() will succeed and the EGLContext and EGLSurface stored
in the VirtualGfxConsole will be valid.
Signed-off-by: Antonio Caggiano <quic_acaggian@quicinc.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20231016123215.2699269-1-quic_acaggian@quicinc.com>
(cherry picked from commit 6f189a08c1)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
erst_realizefn() passes @errp to functions without checking for
failure. If it runs into another failure, it trips error_setv()'s
assertion.
Use the ERRP_GUARD() macro and check *errp, as suggested in commit
ae7c80a7bd ("error: New macro ERRP_GUARD()").
Cc: qemu-stable@nongnu.org
Fixes: f7e26ffa59 ("ACPI ERST: support for ACPI ERST feature")
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20231120130017.81286-1-philmd@linaro.org>
Reviewed-by: Ani Sinha <anisinha@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
(cherry picked from commit 20bc50137f)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
After a relatively short time, there is an multiplication overflow
when multiplying (now - buft_start) with hda_bytes_per_second().
While the uptime now - buft_start only overflows after 2**63 ns
= 292.27 years, this happens hda_bytes_per_second() times faster
with the multiplication. At 44100 samples/s * 2 channels
* 2 bytes/channel = 176400 bytes/s that is 14.52 hours. After the
multiplication overflow the affected audio stream stalls.
Replace the multiplication and following division with muldiv64()
to prevent a multiplication overflow.
Fixes: 280c1e1cdb ("audio/hda: create millisecond timers that handle IO")
Reported-by: M_O_Bz <m_o_bz@163.com>
Signed-off-by: Volker Rümelin <vr_qemu@t-online.de>
Message-Id: <20231105172552.8405-1-vr_qemu@t-online.de>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
(cherry picked from commit 74e8593e7e)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Booting a Linux kernel with the malta machine is currently broken
on big endian hosts. The cpu_to_gt32 macro wants to byteswap a value
for little endian targets only, but uses the wrong way to do this:
cpu_to_[lb]e32 works the other way round on big endian hosts! Fix
it by using the same ways on both, big and little endian hosts.
Fixes: 0c8427baf0 ("hw/mips/malta: Use bootloader helper to set BAR registers")
Cc: qemu-stable@nongnu.org
Message-Id: <20230330152613.232082-1-thuth@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Thomas Huth <thuth@redhat.com>
(cherry picked from commit dc96009afd)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(Mjt: adjust context for before v7.2.0-677-g0e45355c5c)
If the text description file is larger than DESC_SIZE, we force the last
byte in the buffer to be 0 and write it out.
This results in a corruption.
Try to allocate a big buffer in this case.
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1923
Signed-off-by: Fam Zheng <fam@euphon.net>
Message-ID: <20231124115654.3239137-1-fam@euphon.net>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit 9fb7b350ba)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(Mjt: fixups in block/vmdk.c due to missing-in-7.2 v8.0.0-2084-g28944f99c4
"vmdk: mark more functions as coroutine_fns and GRAPH_RDLOCK")
Commit 0be6bfac62 ("qdev: Implement variable length array properties")
added the DEFINE_PROP_ARRAY() macro with the following comment:
* It is the responsibility of the device deinit code to free the
* @_arrayfield memory.
Commit 9e4aa1fafe added:
DEFINE_PROP_ARRAY("pg0-lock",
XlnxVersalEFuseCtrl, extra_pg0_lock_n16,
extra_pg0_lock_spec, qdev_prop_uint16, uint16_t),
but forgot to free the 'extra_pg0_lock_spec' array. Do it in the
instance_finalize() handler.
Cc: qemu-stable@nongnu.org
Fixes: 9e4aa1fafe ("hw/nvram: Xilinx Versal eFuse device") # v6.2.0+
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-id: 20231121174051.63038-6-philmd@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
(cherry picked from commit 4f10c66077)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Commit 0be6bfac62 ("qdev: Implement variable length array properties")
added the DEFINE_PROP_ARRAY() macro with the following comment:
* It is the responsibility of the device deinit code to free the
* @_arrayfield memory.
Commit 68fbcc344e added:
DEFINE_PROP_ARRAY("read-only", XlnxEFuse, ro_bits_cnt, ro_bits,
qdev_prop_uint32, uint32_t),
but forgot to free the 'ro_bits' array. Do it in the instance_finalize
handler.
Cc: qemu-stable@nongnu.org
Fixes: 68fbcc344e ("hw/nvram: Introduce Xilinx eFuse QOM") # v6.2.0+
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-id: 20231121174051.63038-5-philmd@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
(cherry picked from commit 49b3e28b7b)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Commit 0be6bfac62 ("qdev: Implement variable length array properties")
added the DEFINE_PROP_ARRAY() macro with the following comment:
* It is the responsibility of the device deinit code to free the
* @_arrayfield memory.
Commit 4fb013afcc added:
DEFINE_PROP_ARRAY("oscclk", MPS2SCC, num_oscclk, oscclk_reset,
qdev_prop_uint32, uint32_t),
but forgot to free the 'oscclk_reset' array. Do it in the
instance_finalize() handler.
Cc: qemu-stable@nongnu.org
Fixes: 4fb013afcc ("hw/misc/mps2-scc: Support configurable number of OSCCLK values") # v6.0.0+
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-id: 20231121174051.63038-4-philmd@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
(cherry picked from commit 896dd6ff7b)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Commit 0be6bfac62 ("qdev: Implement variable length array properties")
added the DEFINE_PROP_ARRAY() macro with the following comment:
* It is the responsibility of the device deinit code to free the
* @_arrayfield memory.
Commit 8077b8e549 added:
DEFINE_PROP_ARRAY("reserved-regions", VirtIOIOMMUPCI,
vdev.nb_reserved_regions, vdev.reserved_regions,
qdev_prop_reserved_region, ReservedRegion),
but forgot to free the 'vdev.reserved_regions' array. Do it in the
instance_finalize() handler.
Cc: qemu-stable@nongnu.org
Fixes: 8077b8e549 ("virtio-iommu-pci: Add array of Interval properties") # v5.1.0+
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Message-id: 20231121174051.63038-3-philmd@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
(cherry picked from commit c9a4aa06df)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(Mjt: fixup hw/virtio/virtio-iommu-pci.c for before v8.1.0-2552-g41cc70cdf5,
"virtio-iommu: Rename reserved_regions into prop_resv_regions" -- so now
patch subject matches actual change again)
The syndrome register value always has an IL field at bit 25, which
is 0 for a trap on a 16 bit instruction, and 1 for a trap on a 32
bit instruction (or for exceptions which aren't traps on a known
instruction, like PC alignment faults). This means that our
syn_*() functions should always either take an is_16bit argument to
determine whether to set the IL bit, or else unconditionally set it.
We missed setting the IL bit for the syndrome for three kinds of trap:
* an SVE access exception
* a pointer authentication check failure
* a BTI (branch target identification) check failure
All of these traps are AArch64 only, and so the instruction causing
the trap is always 64 bit. This means we can unconditionally set
the IL bit in the syn_*() function.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20231120150121.3458408-1-peter.maydell@linaro.org
Cc: qemu-stable@nongnu.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
(cherry picked from commit 11a3c4a286)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Recently MemReentrancyGuard was added to DeviceState to record that the
device is engaging in I/O. The network device backend needs to update it
when delivering a packet to a device.
This implementation follows what bottom half does, but it does not add
a tracepoint for the case that the network device backend started
delivering a packet to a device which is already engaging in I/O. This
is because such reentrancy frequently happens for
qemu_flush_queued_packets() and is insignificant.
Fixes: CVE-2023-3019
Reported-by: Alexander Bulekov <alxndr@bu.edu>
Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Acked-by: Alexander Bulekov <alxndr@bu.edu>
Signed-off-by: Jason Wang <jasowang@redhat.com>
(cherry picked from commit 9050f976e4)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Recently MemReentrancyGuard was added to DeviceState to record that the
device is engaging in I/O. The network device backend needs to update it
when delivering a packet to a device.
In preparation for such a change, add MemReentrancyGuard * as a
parameter of qemu_new_nic().
Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Reviewed-by: Alexander Bulekov <alxndr@bu.edu>
Signed-off-by: Jason Wang <jasowang@redhat.com>
(cherry picked from commit 7d0fefdf81)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(Mjt: fixup in hw/net/xen_nic.c due to lack of v8.1.0-2771-g25967ff69f
"hw/xen: update Xen PV NIC to XenDevice model"
and removed hw/net/igb.c bits)
Legacy software contains a standard mechanism for generating a reset to a
Serial ATA device - setting the SRST (software reset) bit in the Device
Control register.
Serial ATA has a more robust mechanism called COMRESET, also referred to
as port reset. A port reset is the preferred mechanism for error
recovery and should be used in place of software reset.
Commit e2a5d9b3d9 ("hw/ide/ahci: simplify and document PxCI handling")
(mjt: 1e5ad6b06b in stable-7.2 series, v7.2.6)
improved the handling of PxCI, such that PxCI gets cleared after handling
a non-NCQ, or NCQ command (instead of incorrectly clearing PxCI after
receiving anything - even a FIS that failed to parse, which should NOT
clear PxCI, so that you can see which command slot that caused an error).
However, simply clearing PxCI after a non-NCQ, or NCQ command, is not
enough, we also need to clear PxCI when receiving a SRST in the Device
Control register.
A legacy software reset is performed by the host sending two H2D FISes,
the first H2D FIS asserts SRST, and the second H2D FIS deasserts SRST.
The first H2D FIS will not get a D2H reply, and requires the FIS to have
the C bit set to one, such that the HBA itself will clear the bit in PxCI.
The second H2D FIS will get a D2H reply once the diagnostic is completed.
The clearing of the bit in PxCI for this command should ideally be done
in ahci_init_d2h() (if it was a legacy software reset that caused the
reset (a COMRESET does not use a command slot)). However, since the reset
value for PxCI is 0, modify ahci_reset_port() to actually clear PxCI to 0,
that way we can avoid complex logic in ahci_init_d2h().
This fixes an issue for FreeBSD where the device would fail to reset.
The problem was not noticed in Linux, because Linux uses a COMRESET
instead of a legacy software reset by default.
Fixes: e2a5d9b3d9 ("hw/ide/ahci: simplify and document PxCI handling")
Reported-by: Marcin Juszkiewicz <marcin.juszkiewicz@linaro.org>
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Message-ID: <20231108222657.117984-1-nks@flawful.org>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Tested-by: Marcin Juszkiewicz <marcin.juszkiewicz@linaro.org>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit eabb921250)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(Mjt: mention 1e5ad6b06b for stable-7.2)
LAALG uses op_laa() and wout_addu64(). The latter expects cc_src to be
set, but the former does not do it. This can lead to assertion failures
if something sets cc_src to neither 0 nor 1 before.
Fix by introducing op_laa_addu64(), which sets cc_src, and using it for
LAALG.
Fixes: 4dba4d6fef ("target/s390x: Use atomic operations for LOAD AND OP")
Cc: qemu-stable@nongnu.org
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-ID: <20231106093605.1349201-4-iii@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
(cherry picked from commit bea402482a)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Before commit "hw/ide: reset: cancel async DMA operation before
resetting state", this test would fail, because a reset with a
pending write operation would lead to an unsolicited write to the
first sector of the disk.
The test writes a pattern to the beginning of the disk and verifies
that it is still intact after a reset with a pending operation. It
also checks that the pending operation actually completes correctly.
Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
Message-ID: <20230906130922.142845-2-f.ebner@proxmox.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
(cherry picked from commit cc610857bb)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
If there is a pending DMA operation during ide_bus_reset(), the fact
that the IDEState is already reset before the operation is canceled
can be problematic. In particular, ide_dma_cb() might be called and
then use the reset IDEState which contains the signature after the
reset. When used to construct the IO operation this leads to
ide_get_sector() returning 0 and nsector being 1. This is particularly
bad, because a write command will thus destroy the first sector which
often contains a partition table or similar.
Traces showing the unsolicited write happening with IDEState
0x5595af6949d0 being used after reset:
> ahci_port_write ahci(0x5595af6923f0)[0]: port write [reg:PxSCTL] @ 0x2c: 0x00000300
> ahci_reset_port ahci(0x5595af6923f0)[0]: reset port
> ide_reset IDEstate 0x5595af6949d0
> ide_reset IDEstate 0x5595af694da8
> ide_bus_reset_aio aio_cancel
> dma_aio_cancel dbs=0x7f64600089a0
> dma_blk_cb dbs=0x7f64600089a0 ret=0
> dma_complete dbs=0x7f64600089a0 ret=0 cb=0x5595acd40b30
> ahci_populate_sglist ahci(0x5595af6923f0)[0]
> ahci_dma_prepare_buf ahci(0x5595af6923f0)[0]: prepare buf limit=512 prepared=512
> ide_dma_cb IDEState 0x5595af6949d0; sector_num=0 n=1 cmd=DMA WRITE
> dma_blk_io dbs=0x7f6420802010 bs=0x5595ae2c6c30 offset=0 to_dev=1
> dma_blk_cb dbs=0x7f6420802010 ret=0
> (gdb) p *qiov
> $11 = {iov = 0x7f647c76d840, niov = 1, {{nalloc = 1, local_iov = {iov_base = 0x0,
> iov_len = 512}}, {__pad = "\001\000\000\000\000\000\000\000\000\000\000",
> size = 512}}}
> (gdb) bt
> #0 blk_aio_pwritev (blk=0x5595ae2c6c30, offset=0, qiov=0x7f6420802070, flags=0,
> cb=0x5595ace6f0b0 <dma_blk_cb>, opaque=0x7f6420802010)
> at ../block/block-backend.c:1682
> #1 0x00005595ace6f185 in dma_blk_cb (opaque=0x7f6420802010, ret=<optimized out>)
> at ../softmmu/dma-helpers.c:179
> #2 0x00005595ace6f778 in dma_blk_io (ctx=0x5595ae0609f0,
> sg=sg@entry=0x5595af694d00, offset=offset@entry=0, align=align@entry=512,
> io_func=io_func@entry=0x5595ace6ee30 <dma_blk_write_io_func>,
> io_func_opaque=io_func_opaque@entry=0x5595ae2c6c30,
> cb=0x5595acd40b30 <ide_dma_cb>, opaque=0x5595af6949d0,
> dir=DMA_DIRECTION_TO_DEVICE) at ../softmmu/dma-helpers.c:244
> #3 0x00005595ace6f90a in dma_blk_write (blk=0x5595ae2c6c30,
> sg=sg@entry=0x5595af694d00, offset=offset@entry=0, align=align@entry=512,
> cb=cb@entry=0x5595acd40b30 <ide_dma_cb>, opaque=opaque@entry=0x5595af6949d0)
> at ../softmmu/dma-helpers.c:280
> #4 0x00005595acd40e18 in ide_dma_cb (opaque=0x5595af6949d0, ret=<optimized out>)
> at ../hw/ide/core.c:953
> #5 0x00005595ace6f319 in dma_complete (ret=0, dbs=0x7f64600089a0)
> at ../softmmu/dma-helpers.c:107
> #6 dma_blk_cb (opaque=0x7f64600089a0, ret=0) at ../softmmu/dma-helpers.c:127
> #7 0x00005595ad12227d in blk_aio_complete (acb=0x7f6460005b10)
> at ../block/block-backend.c:1527
> #8 blk_aio_complete (acb=0x7f6460005b10) at ../block/block-backend.c:1524
> #9 blk_aio_write_entry (opaque=0x7f6460005b10) at ../block/block-backend.c:1594
> #10 0x00005595ad258cfb in coroutine_trampoline (i0=<optimized out>,
> i1=<optimized out>) at ../util/coroutine-ucontext.c:177
Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Tested-by: simon.rowe@nutanix.com
Message-ID: <20230906130922.142845-1-f.ebner@proxmox.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
(cherry picked from commit 7d7512019f)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
In a two-stage translation, the result of the BTI guarded bit should
be the guarded bit from the first stage of translation, as there is
no BTI guard information in stage two. Our code tried to do this,
but got it wrong, because we currently have two fields where the GP
bit information might live (ARMCacheAttrs::guarded and
CPUTLBEntryFull::extra::arm::guarded), and we were storing the GP bit
in the latter during the stage 1 walk but trying to copy the former
in combine_cacheattrs().
Remove the duplicated storage, and always use the field in
CPUTLBEntryFull; correctly propagate the stage 1 value to the output
in get_phys_addr_twostage().
Note for stable backports: in v8.0 and earlier the field is named
result->f.guarded, not result->f.extra.arm.guarded.
Cc: qemu-stable@nongnu.org
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1950
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20231031173723.26582-1-peter.maydell@linaro.org
(cherry picked from commit 4c09abeae8)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(Mjt: replace f.extra.arm.guarded -> f.guarded due to v8.1.0-1179-ga81fef4b64)
We currently don't correctly handle the VSTCR_EL2.SW and VTCR_EL2.NSW
configuration bits. These allow configuration of whether the stage 2
page table walks for Secure IPA and NonSecure IPA should do their
descriptor reads from Secure or NonSecure physical addresses. (This
is separate from how the translation table base address and other
parameters are set: an NS IPA always uses VTTBR_EL2 and VTCR_EL2
for its base address and walk parameters, regardless of the NSW bit,
and similarly for Secure.)
Provide a new function ptw_idx_for_stage_2() which returns the
MMU index to use for descriptor reads, and use it to set up
the .in_ptw_idx wherever we call get_phys_addr_lpae().
For a stage 2 walk, wherever we call get_phys_addr_lpae():
* .in_ptw_idx should be ptw_idx_for_stage_2() of the .in_mmu_idx
* .in_secure should be true if .in_mmu_idx is Stage2_S
This allows us to correct S1_ptw_translate() so that it consistently
always sets its (out_secure, out_phys) to the result it gets from the
S2 walk (either by calling get_phys_addr_lpae() or by TLB lookup).
This makes better conceptual sense because the S2 walk should return
us an (address space, address) tuple, not an address that we then
randomly assign to S or NS.
Our previous handling of SW and NSW was broken, so guest code
trying to use these bits to put the s2 page tables in the "other"
address space wouldn't work correctly.
Cc: qemu-stable@nongnu.org
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1600
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20230504135425.2748672-3-peter.maydell@linaro.org
(cherry picked from commit fcc0b0418f)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Bit 63 in a Table descriptor is only the NSTable bit for stage 1
translations; in stage 2 it is RES0. We were incorrectly looking at
it all the time.
This causes problems if:
* the stage 2 table descriptor was incorrectly setting the RES0 bit
* we are doing a stage 2 translation in Secure address space for
a NonSecure stage 1 regime -- in this case we would incorrectly
do an immediate downgrade to NonSecure
A bug elsewhere in the code currently prevents us from getting
to the second situation, but when we fix that it will be possible.
Cc: qemu-stable@nongnu.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-id: 20230504135425.2748672-2-peter.maydell@linaro.org
(cherry picked from commit 21a4ab8318)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
When TCG is disabled this part of the code should not be reachable, so
wrap it with an ifdef for now.
Signed-off-by: Fabiano Rosas <farosas@suse.de>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
(cherry picked from commit 0d3de77a07)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(Mjt: trivial change which makes two subsequent cherry-picks to apply cleanly)
In case when we're rebasing within one backing chain, and when target image
is larger than old backing file, bdrv_is_allocated_above() ends up setting
*pnum = 0. As a result, target offset isn't getting incremented, and we
get stuck in an infinite for loop. Let's detect this case and proceed
further down the loop body, as the offsets beyond the old backing size need
to be explicitly zeroed.
Signed-off-by: Andrey Drobyshev <andrey.drobyshev@virtuozzo.com>
Reviewed-by: Denis V. Lunev <den@openvz.org>
Reviewed-by: Hanna Czenczek <hreitz@redhat.com>
Message-ID: <20230919165804.439110-2-andrey.drobyshev@virtuozzo.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit 8b097fd6b0)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
A build of GCC 13.2 will have stack protector enabled by default if it
was configured with --enable-default-ssp option. For such a compiler,
it is necessary to explicitly disable stack protector when linking
without standard libraries.
Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20230731091042.139159-2-akihiko.odaki@daynix.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
(cherry picked from commit 7a06a8fec9)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Block Size Register bits [14:12] is SDMA Buffer Boundary, it is missed
in register write, but it is needed in SDMA transfer. e.g. it will be
used in sdhci_sdma_transfer_multi_blocks to calculate boundary_ variables.
Missing this field will cause wrong operation for different SDMA Buffer
Boundary settings.
Fixes: d7dfca0807 ("hw/sdhci: introduce standard SD host controller")
Fixes: dfba99f17f ("hw/sdhci: Fix DMA Transfer Block Size field")
Signed-off-by: Lu Gao <lu.gao@verisilicon.com>
Signed-off-by: Jianxian Wen <jianxian.wen@verisilicon.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-ID: <20220321055618.4026-1-lu.gao@verisilicon.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
(cherry picked from commit ae5f70baf5)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Those PS/2 ports are created with the LASI controller when
a 32-bit PA-RISC machine is created.
Mark them not user-createable to avoid showing them in
the qemu device list.
Signed-off-by: Helge Deller <deller@gmx.de>
Cc: qemu-stable@nongnu.org
(cherry picked from commit a1e6a5c462)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
sh4 uses gUSA (general UserSpace Atomicity) to provide atomicity on CPUs
that don't have atomic instructions. A gUSA region that adds 1 to an
atomic variable stored in @R2 looks like this:
4004b6: 03 c7 mova 4004c4 <gusa+0x10>,r0
4004b8: f3 61 mov r15,r1
4004ba: 09 00 nop
4004bc: fa ef mov #-6,r15
4004be: 22 63 mov.l @r2,r3
4004c0: 01 73 add #1,r3
4004c2: 32 22 mov.l r3,@r2
4004c4: 13 6f mov r1,r15
R0 contains a pointer to the end of the gUSA region
R1 contains the saved stack pointer
R15 contains negative length of the gUSA region
When this region is interrupted by a signal, the kernel detects if
R15 >= -128U. If yes, the kernel rolls back PC to the beginning of the
region and restores SP by copying R1 to R15.
The problem happens if we are interrupted by a signal at address 4004c4.
R15 still holds the value -6, but the atomic value was already written by
an instruction at address 4004c2. In this situation we can't undo the
gUSA. The function unwind_gusa does nothing, the signal handler attempts
to push a signal frame to the address -6 and crashes.
This patch fixes it, so that if we are interrupted at the last instruction
in a gUSA region, we copy R1 to R15 to restore the correct stack pointer
and avoid crashing.
There's another bug: if we are interrupted in a delay slot, we save the
address of the instruction in the delay slot. We must save the address of
the previous instruction.
Cc: qemu-stable@nongnu.org
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Reviewed-by: Yoshinori Sato <ysato@users.sourcefoege.jp>
Message-Id: <b16389f7-6c62-70b7-59b3-87533c0bcc@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
(cherry picked from commit 3b894b699c)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
The migration code uses unsigned values for 16, 32 and 64-bit
operations. Fix the script to do the same.
This was causing an issue when parsing the migration stream generated
on the ppc64 target because one of instance_ids was larger than the
32bit signed maximum:
Traceback (most recent call last):
File "/home/fabiano/kvm/qemu/build/scripts/analyze-migration.py", line 658, in <module>
dump.read(dump_memory = args.memory)
File "/home/fabiano/kvm/qemu/build/scripts/analyze-migration.py", line 592, in read
classdesc = self.section_classes[section_key]
KeyError: ('spapr_iommu', -2147483648)
Signed-off-by: Fabiano Rosas <farosas@suse.de>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Message-ID: <20231009184326.15777-6-farosas@suse.de>
(cherry picked from commit caea03279e)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Guest driver allocates and initialize page tables to be used as a ring
of descriptors for CQ and async events.
The page table that represents the ring, along with the number of pages
in the page table is passed to the device.
Currently our device supports only one page table for a ring.
Let's make sure that the number of page table entries the driver
reports, do not exceeds the one page table size.
Reported-by: Soul Chen <soulchen8650@gmail.com>
Signed-off-by: Yuval Shaia <yuval.shaia.ml@gmail.com>
Fixes: CVE-2023-1544
Message-ID: <20230301142926.18686-1-yuval.shaia.ml@gmail.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
(cherry picked from commit 85fc35afa9)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Reset the current sample counter when writing the Channel Sample
Count Register. The Linux ens1370 driver and the AROS sb128
driver expect the current sample counter counts down from sample
count to 0 after a write to the Channel Sample Count Register.
Currently the current sample counter starts from 0 after a reset
or the last count when the counter was stopped.
The current sample counter is used to raise an interrupt whenever
a complete buffer was transferred. When the counter starts with a
value lower than the reload value, the interrupt triggeres before
the buffer was completly transferred. This may lead to corrupted
audio streams.
Tested-by: Rene Engel <ReneEngel80@emailn.de>
Signed-off-by: Volker Rümelin <vr_qemu@t-online.de>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Tested-by: BALATON Zoltan <balaton@eik.bme.hu>
Message-Id: <20230917065813.6692-1-vr_qemu@t-online.de>
(cherry picked from commit 00e3b29d06)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
QEMU will crash if anyone tries to set tls-authz (which is a type
StrOrNull) with 'null' value. Fix it in the easy way by converting it to
qstring just like the other two tls parameters.
Cc: qemu-stable@nongnu.org # v4.0+
Fixes: d2f1d29b95 ("migration: add support for a "tls-authz" migration parameter")
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Message-ID: <20230905162335.235619-2-peterx@redhat.com>
(cherry picked from commit 86dec715a7)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(Mjt: the function is in m/migration.c, not m/option.c; minor context tweak)
An MSI from I/O APIC may not exactly equal to APIC_DEFAULT_ADDRESS. In
fact, Windows 17763.3650 configures I/O APIC to set the dest_mode bit.
Cover the range assigned to APIC.
Fixes: 577c470f43 ("x86_iommu/amd: Prepare for interrupt remap support")
Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Message-Id: <20230921114612.40671-1-akihiko.odaki@daynix.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
(cherry picked from commit 0114c45130)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Use abi_ullong not uint64_t so that the alignment of the field
and therefore the layout of the struct is correct.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
(cherry picked from commit 33bc4fa78b)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
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>
(cherry picked from commit 4f7689f081)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(Mjt: use TFR() instead of RETRY_ON_EINTR() before v7.2.0-538-g8b6aa69365)
The fw_cfg DMA write callback in ramfb prepares a new display surface in
QEMU; this new surface is put to use ("swapped in") upon the next display
update. At that time, the old surface (if any) is released.
If the guest triggers the fw_cfg DMA write callback at least twice between
two adjacent display updates, then the second callback (and further such
callbacks) will leak the previously prepared (but not yet swapped in)
display surface.
The issue can be shown by:
(1) starting QEMU with "-trace displaysurface_free", and
(2) running the following program in the guest UEFI shell:
> #include <Library/ShellCEntryLib.h> // ShellAppMain()
> #include <Library/UefiBootServicesTableLib.h> // gBS
> #include <Protocol/GraphicsOutput.h> // EFI_GRAPHICS_OUTPUT_PROTOCOL
>
> INTN
> EFIAPI
> ShellAppMain (
> IN UINTN Argc,
> IN CHAR16 **Argv
> )
> {
> EFI_STATUS Status;
> VOID *Interface;
> EFI_GRAPHICS_OUTPUT_PROTOCOL *Gop;
> UINT32 Mode;
>
> Status = gBS->LocateProtocol (
> &gEfiGraphicsOutputProtocolGuid,
> NULL,
> &Interface
> );
> if (EFI_ERROR (Status)) {
> return 1;
> }
>
> Gop = Interface;
>
> Mode = 1;
> for ( ; ;) {
> Status = Gop->SetMode (Gop, Mode);
> if (EFI_ERROR (Status)) {
> break;
> }
>
> Mode = 1 - Mode;
> }
>
> return 1;
> }
The symptom is then that:
- only one trace message appears periodically,
- the time between adjacent messages keeps increasing -- implying that
some list structure (containing the leaked resources) keeps growing,
- the "surface" pointer is ever different.
> 18566@1695127471.449586:displaysurface_free surface=0x7f2fcc09a7c0
> 18566@1695127471.529559:displaysurface_free surface=0x7f2fcc9dac10
> 18566@1695127471.659812:displaysurface_free surface=0x7f2fcc441dd0
> 18566@1695127471.839669:displaysurface_free surface=0x7f2fcc0363d0
> 18566@1695127472.069674:displaysurface_free surface=0x7f2fcc413a80
> 18566@1695127472.349580:displaysurface_free surface=0x7f2fcc09cd00
> 18566@1695127472.679783:displaysurface_free surface=0x7f2fcc1395f0
> 18566@1695127473.059848:displaysurface_free surface=0x7f2fcc1cae50
> 18566@1695127473.489724:displaysurface_free surface=0x7f2fcc42fc50
> 18566@1695127473.969791:displaysurface_free surface=0x7f2fcc45dcc0
> 18566@1695127474.499708:displaysurface_free surface=0x7f2fcc70b9d0
> 18566@1695127475.079769:displaysurface_free surface=0x7f2fcc82acc0
> 18566@1695127475.709941:displaysurface_free surface=0x7f2fcc369c00
> 18566@1695127476.389619:displaysurface_free surface=0x7f2fcc32b910
> 18566@1695127477.119772:displaysurface_free surface=0x7f2fcc0d5a20
> 18566@1695127477.899517:displaysurface_free surface=0x7f2fcc086c40
> 18566@1695127478.729962:displaysurface_free surface=0x7f2fccc72020
> 18566@1695127479.609839:displaysurface_free surface=0x7f2fcc185160
> 18566@1695127480.539688:displaysurface_free surface=0x7f2fcc23a7e0
> 18566@1695127481.519759:displaysurface_free surface=0x7f2fcc3ec870
> 18566@1695127482.549930:displaysurface_free surface=0x7f2fcc634960
> 18566@1695127483.629661:displaysurface_free surface=0x7f2fcc26b140
> 18566@1695127484.759987:displaysurface_free surface=0x7f2fcc321700
> 18566@1695127485.940289:displaysurface_free surface=0x7f2fccaad100
We figured this wasn't a CVE-worthy problem, as only small amounts of
memory were leaked (the framebuffer itself is mapped from guest RAM, QEMU
only allocates administrative structures), plus libvirt restricts QEMU
memory footprint anyway, thus the guest can only DoS itself.
Plug the leak, by releasing the last prepared (not yet swapped in) display
surface, if any, in the fw_cfg DMA write callback.
Regarding the "reproducer", with the fix in place, the log is flooded with
trace messages (one per fw_cfg write), *and* the trace message alternates
between just two "surface" pointer values (i.e., nothing is leaked, the
allocator flip-flops between two objects in effect).
This issue appears to date back to the introducion of ramfb (995b30179b,
"hw/display: add ramfb, a simple boot framebuffer living in guest ram",
2018-06-18).
Cc: Gerd Hoffmann <kraxel@redhat.com> (maintainer:ramfb)
Cc: qemu-stable@nongnu.org
Fixes: 995b30179b
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Acked-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-ID: <20230919131955.27223-1-lersek@redhat.com>
(cherry picked from commit e0288a7784)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
CVTPS2PD only loads a half-register for memory, unlike the other
operations under 0x0F 0x5A. "Unpack" the group into separate
emission functions instead of using gen_unary_fp_sse.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit abd41884c5)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
CVTPS2PD only loads a half-register for memory, like CVTPH2PS. It can
reuse the "ph" packed half-precision size to load a half-register,
but rename it to "xh" because it is now a variation of "x" (it is not
used only for half-precision values).
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit a48b26978a)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Before this change, MOVNTPS and MOVNTPD were labeled as Exception Class
4 (only requiring alignment for legacy SSE instructions). This changes
them to Exception Class 1 (always requiring memory alignment), as
documented in the Intel manual.
Message-Id: <20230501111428.95998-3-ricky@rzhou.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit 8bf171c2d1)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Fix the exception classes for some SSE/AVX instructions to match what is
documented in the Intel manual.
These changes are expected to have no functional effect on the behavior
that qemu implements (primarily >= 16-byte memory alignment checks). For
instance, since qemu does not implement the AC flag, there is no
difference in behavior between Exception Classes 4 and 5 for
instructions where the SSE version only takes <16 byte memory operands.
Message-Id: <20230501111428.95998-2-ricky@rzhou.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit cab529b0dc)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Adds some comments describing what instructions correspond to decoding
table entries and fixes some existing comments which named the wrong
instruction.
Message-Id: <20230501111428.95998-1-ricky@rzhou.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit afa94dabc5)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
The instructions also use bits 3 and 7 of their 8-byte immediate.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit 9e65829699)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
VRCPSS, VRSQRTSS and VCVTSx2Sx have a 32-bit or 64-bit memory operand,
which is represented in the decoding tables by X86_VEX_REPScalar. Add it
to the tables, and make validate_vex() handle the case of an instruction
that is in exception type 4 without the REP prefix and exception type 5
with it; this is the cas of VRCP and VRSQRT.
Reported-by: yongwoo <https://gitlab.com/yongwoo36>
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1377
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit 3d304620ec)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Otherwise when a FORMAT UNIT command is issued, the SCSI layer can become
confused because it can find itself in the situation where it thinks there
is still data to be transferred which can cause the next emulated SCSI
command to fail.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Fixes: 6ab71761 ("scsi-disk: add FORMAT UNIT command")
Tested-by: Thomas Huth <thuth@redhat.com>
Message-ID: <20230913204410.65650-4-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit be2b619a17)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
In the case where a SCSI layer transfer is incorrectly terminated, it is
possible for a TI command to cause a SCSI buffer overflow due to the
expected transfer data length being less than the available data in the
FIFO. When this occurs the unsigned async_len variable underflows and
becomes a large offset which writes past the end of the allocated SCSI
buffer.
Restrict the non-DMA transfer length to be the smallest of the expected
transfer length and the available FIFO data to ensure that it is no longer
possible for the SCSI buffer overflow to occur.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1810
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-ID: <20230913204410.65650-3-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit 77668e4b9b)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
The call to esp_dma_enable() was being made with the SYSBUS_ESP type instead of
the ESP type. This meant that when GPIO 1 was being used to trigger a DMA
request from an external DMA controller, the setting of ESPState's dma_enabled
field would clobber unknown memory whilst the dma_cb callback pointer would
typically return NULL so the DMA request would never start.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-ID: <20230913204410.65650-2-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit b86dc5cb0b)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Our linker script for optionroms specifies only the placement of the
.text section, leaving the linker free to place the remaining sections
at arbitrary places in the file.
Since at least binutils 2.39, the .note.gnu.build-id section is now
being placed at the start of the file, which causes label addresses to
be shifted. For linuxboot_dma.bin that means that the PnP header
(among others) will not be found when determining the type of ROM at
optionrom_setup():
(0x1c is the label _pnph, where the magic "PnP" is)
$ xxd /usr/share/qemu/linuxboot_dma.bin | grep "PnP"
00000010: 0000 0000 0000 0000 0000 1c00 2450 6e50 ............$PnP
$ xxd pc-bios/optionrom/linuxboot_dma.bin | grep "PnP"
00000010: 0000 0000 0000 0000 0000 4c00 2450 6e50 ............$PnP
^bad
Using a freshly built linuxboot_dma.bin ROM results in a broken boot:
SeaBIOS (version rel-1.16.2-0-gea1b7a073390-prebuilt.qemu.org)
Booting from Hard Disk...
Boot failed: could not read the boot disk
Booting from Floppy...
Boot failed: could not read the boot disk
No bootable device.
We're not using the build-id section, so pass the --build-id=none
option to the linker to remove it entirely.
Note: In theory, this same issue could happen with any other
section. The ideal solution would be to have all unused sections
discarded in the linker script. However that would be a larger change,
specially for the pvh rom which uses the .bss and COMMON sections so
I'm addressing only the immediate issue here.
Reported-by: Vasiliy Ulyanov <vulyanov@suse.de>
Signed-off-by: Fabiano Rosas <farosas@suse.de>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-ID: <20230926192502.15986-1-farosas@suse.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit 35ed01ba54)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(mjt: remove unrelated stable@vger)
VNC_FEATURE_XVP was not shifted left before adding it to vs->features,
so it was never enabled; but it was also checked the wrong way with
a logical AND instead of vnc_has_feature. Fix both places.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit 477b301000)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
The debug message was cut and pasted from the invalid audio format
case, but the audio message is at bytes 2-3.
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit 0cb9c5880e)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
We are doing things like
nb_sectors /= (s->qdev.blocksize / BDRV_SECTOR_SIZE);
in the code here (e.g. in scsi_disk_emulate_mode_sense()), so if
the blocksize is smaller than BDRV_SECTOR_SIZE (=512), this crashes
with a division by 0 exception. Thus disallow block sizes of 256
bytes to avoid this situation.
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1813
CVE: 2023-42467
Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-ID: <20230925091854.49198-1-thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit 7cfcc79b0a)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
mttcg asserts that an execution ending with EXCP_HALTED must have
cpu->halted. However between the event or instruction that sets
cpu->halted and requests exit and the assertion here, an
asynchronous event could clear cpu->halted.
This leads to crashes running AIX on ppc/pseries because it uses
H_CEDE/H_PROD hcalls, where H_CEDE sets self->halted = 1 and
H_PROD sets other cpu->halted = 0 and kicks it.
H_PROD could be turned into an interrupt to wake, but several other
places in ppc, sparc, and semihosting follow what looks like a similar
pattern setting halted = 0 directly. So remove this assertion.
Reported-by: Ivan Warren <ivan@vmfacility.fr>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Message-Id: <20230829010658.8252-1-npiggin@gmail.com>
[rth: Keep the case label and adjust the comment.]
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
(cherry picked from commit 0e5903436d)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
The LDRT/STRT "unprivileged load/store" instructions behave like
normal ones if executed at EL0. We handle this correctly for
the load/store semantics, but get the MTE checking wrong.
We always look at s->mte_active[is_unpriv] to see whether we should
be doing MTE checks, but in hflags.c when we set the TB flags that
will be used to fill the mte_active[] array we only set the
MTE0_ACTIVE bit if UNPRIV is true (i.e. we are not at EL0).
This means that a LDRT at EL0 will see s->mte_active[1] as 0,
and will not do MTE checks even when MTE is enabled.
To avoid the translate-time code having to do an explicit check on
s->unpriv to see if it is OK to index into the mte_active[] array,
duplicate MTE_ACTIVE into MTE0_ACTIVE when UNPRIV is false.
(This isn't a very serious bug because generally nobody executes
LDRT/STRT at EL0, because they have no use there.)
Cc: qemu-stable@nongnu.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20230912140434.1333369-2-peter.maydell@linaro.org
(cherry picked from commit 903dbefc2b)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(Mjt: before v7.2.0-1636-g671efad16a this code was in target/arm/helper.c)
qemu-hppa may crash when delivering a signal. It can be demonstrated with
this program. Compile the program with "hppa-linux-gnu-gcc -O2 signal.c"
and run it with "qemu-hppa -one-insn-per-tb a.out". It reports that the
address of the flag is 0xb4 and it crashes when attempting to touch it.
#include <stdio.h>
#include <stdlib.h>
#include <sys/time.h>
#include <signal.h>
sig_atomic_t flag;
void sig(int n)
{
printf("&flag: %p\n", &flag);
flag = 1;
}
int main(void)
{
struct sigaction sa;
struct itimerval it;
sa.sa_handler = sig;
sigemptyset(&sa.sa_mask);
sa.sa_flags = SA_RESTART;
if (sigaction(SIGALRM, &sa, NULL)) perror("sigaction"), exit(1);
it.it_interval.tv_sec = 0;
it.it_interval.tv_usec = 100;
it.it_value.tv_sec = it.it_interval.tv_sec;
it.it_value.tv_usec = it.it_interval.tv_usec;
if (setitimer(ITIMER_REAL, &it, NULL)) perror("setitimer"), exit(1);
while (1) {
}
}
The reason for the crash is that the signal handling routine doesn't clear
the 'N' flag in the PSW. If the signal interrupts a thread when the 'N'
flag is set, the flag remains set at the beginning of the signal handler
and the first instruction of the signal handler is skipped.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Acked-by: Helge Deller <deller@gmx.de>
Cc: qemu-stable@nongnu.org
Signed-off-by: Helge Deller <deller@gmx.de>
(cherry picked from commit 2529497cb6)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
When writing a value to the decrementer that raises an exception, the
irq is raised, but the value is not stored so the store doesn't appear
to have changed the register when it is read again.
Always store the write value to the register.
Fixes: e81a982aa5 ("PPC: Clean up DECR implementation")
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
(cherry picked from commit febb71d543)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
The decrementer store function has logic that short-cuts the timer if a
very small value is stored (0, 1, or 2) and raises an interrupt
directly. There are two problem with this on BookE.
First is that BookE says a decrementer interrupt should not be raised
on a store of 0, only of a decrement from 1. Second is that raising
the irq directly will bypass the auto-reload logic in the booke decr
timer function, breaking autoreload when 1 or 2 is stored.
Fix this by removing that small-value special case. It makes this
tricky logic even more difficult to reason about, and it hardly matters
for performance.
Cc: sdicaro@DDCI.com
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20230530131214.373524-2-npiggin@gmail.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
(cherry picked from commit 17dd1354c1)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
When storing a large decrementer value with the most significant
implemented bit set, it is to be treated as a negative and sign
extended.
This isn't hit for book3s DEC because of another bug, fixing it
in the next patch exposes this one and can cause additional
problems, so fix this first. It can be hit with HDECR and other
edge triggered types.
Fixes: a8dafa5251 ("target/ppc: Implement large decrementer support for TCG")
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
[ clg: removed extra cpu and pcc variables shadowing local variables ]
Signed-off-by: Cédric Le Goater <clg@kaod.org>
(cherry picked from commit c8fbc6b9f2)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
The decrementer register contains a relative time in timebase units.
When writing to DECR this is converted and stored as an absolute value
in nanosecond units, reading DECR converts back to relative timebase.
The tb<->ns conversion of the relative part can cause rounding such that
a value writen to the decrementer can read back a different, with time
held constant. This is a particular problem for a deterministic icount
and record-replay trace.
Fix this by storing the absolute value in timebase units rather than
nanoseconds. The math before:
store: decr_next = now_ns + decr * ns_per_sec / tb_per_sec
load: decr = (decr_next - now_ns) * tb_per_sec / ns_per_sec
load(store): decr = decr * ns_per_sec / tb_per_sec * tb_per_sec /
ns_per_sec
After:
store: decr_next = now_ns * tb_per_sec / ns_per_sec + decr
load: decr = decr_next - now_ns * tb_per_sec / ns_per_sec
load(store): decr = decr
Fixes: 9fddaa0c0c ("PowerPC merge: real time TB and decrementer - faster and simpler exception handling (Jocelyn Mayer)")
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
(cherry picked from commit 8e0a5ac878)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
The rule of timers is typically that they should never expire before the
timeout, but some time afterward. Rounding timer intervals up when doing
conversion is the right thing to do.
Under most circumstances it is impossible observe the decrementer
interrupt before the dec register has triggered. However with icount
timing, problems can arise. For example setting DEC to 0 can schedule
the timer for now, causing it to fire before any more instructions
have been executed and DEC is still 0.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
(cherry picked from commit eab0888418)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
This will be used for converting time intervals in different base units
to host units, for the purpose of scheduling timers to emulate target
timers. Timers typically must not fire before their requested expiry
time but may fire some time afterward, so rounding up is the right way
to implement these.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
[ clg: renamed __muldiv64() to muldiv64_rounding() ]
Signed-off-by: Cédric Le Goater <clg@kaod.org>
(cherry picked from commit 47de6c4c28)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
These calculations are repeated several times, and they will become
a little more complicated with subsequent changes.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
(cherry picked from commit 7798f5c576)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
2023-09-25 23:43:49 +03:00
189 changed files with 1959 additions and 1093 deletions
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.