- fix build with musl libc
- fix potential deadlock of QEMU main event loop (cannot be hit with linux
client)
- revert 9pfs reply truncation (LP 1877688)
- xen backend waits for client to free space on the reply ring instead of
truncating or disconnecting
# gpg: Signature made Tue 26 May 2020 10:36:23 BST
# gpg: using RSA key B4828BAF943140CEF2A3491071D4D5E5822F73D6
# gpg: Good signature from "Greg Kurz <groug@kaod.org>" [full]
# gpg: aka "Gregory Kurz <gregory.kurz@free.fr>" [full]
# gpg: aka "[jpeg image of size 3330]" [full]
# Primary key fingerprint: B482 8BAF 9431 40CE F2A3 4910 71D4 D5E5 822F 73D6
* remotes/gkurz/tags/9p-next-2020-05-26:
xen/9pfs: increase max ring order to 9
xen/9pfs: yield when there isn't enough room on the ring
Revert "9p: init_in_iov_from_pdu can truncate the size"
9p: Lock directory streams with a CoMutex
9pfs: include linux/limits.h for XATTR_SIZE_MAX
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Commit 571a8c522e caused the HMP wavcapture command to segfault when
processing audio data in audio_pcm_sw_write(), where a NULL
sw->hw->pcm_ops is dereferenced. This fix checks that the pointer is
valid before dereferincing it. A similar fix is also made in the
parallel function audio_pcm_sw_read().
Fixes: 571a8c522e (audio: split ctl_* functions into enable_* and
volume_*)
Signed-off-by: Bruce Rogers <brogers@suse.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-id: 20200521172931.121903-1-brogers@suse.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The code in CONV_NATURAL_FLOAT() and CLIP_NATURAL_FLOAT()
seems to use the constant 2^31-0.5 to convert float to integer
and back. But the float type lacks the required precision and
the constant used for the conversion is 2^31. This is equiva-
lent to a [-1.f, 1.f] <-> [INT32_MIN, INT32_MAX + 1] mapping.
This patch explicitly writes down the used constant. The
compiler generated code doesn't change.
The constant 2^31 has an exact float representation and the
clang 10 compiler stops complaining about an implicit int to
float conversion with a changed value.
A few notes:
- The conversion of 1.f to INT32_MAX + 1 doesn't overflow. The
type of the destination variable is int64_t.
- At a later stage one of the clip_* functions in
audio/mixeng_template.h limits INT32_MAX + 1 to the integer
range.
- The clip_natural_float_* functions in audio/mixeng.c convert
INT32_MAX and INT32_MAX + 1 to 1.f.
Buglink: https://bugs.launchpad.net/bugs/1878627
Signed-off-by: Volker Rümelin <vr_qemu@t-online.de>
Message-id: 20200523201712.23908-1-vr_qemu@t-online.de
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Instead of truncating replies, which is problematic, wait until the
client reads more data and frees bytes on the reply ring.
Do that by calling qemu_coroutine_yield(). The corresponding
qemu_coroutine_enter_if_inactive() is called from xen_9pfs_bh upon
receiving the next notification from the client.
We need to be careful to avoid races in case xen_9pfs_bh and the
coroutine are both active at the same time. In xen_9pfs_bh, wait until
either the critical section is over (ring->co == NULL) or until the
coroutine becomes inactive (qemu_coroutine_yield() was called) before
continuing. Then, simply wake up the coroutine if it is inactive.
Signed-off-by: Stefano Stabellini <stefano.stabellini@xilinx.com>
Reviewed-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Message-Id: <20200521192627.15259-2-sstabellini@kernel.org>
Signed-off-by: Greg Kurz <groug@kaod.org>
A guest user may set channel frame count via es1370_write()
such that, in es1370_transfer_audio(), total frame count
'size' is lesser than the number of frames that are processed
'cnt'.
int cnt = d->frame_cnt >> 16;
int size = d->frame_cnt & 0xffff;
if (size < cnt), it results in incorrect calculations leading
to OOB access issue(s). Add check to avoid it.
Reported-by: Ren Ding <rding@gatech.edu>
Reported-by: Hanqing Zhao <hanqing@gatech.edu>
Signed-off-by: Prasad J Pandit <pjp@fedoraproject.org>
Message-id: 20200514200608.1744203-1-ppandit@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Locking was introduced in QEMU 2.7 to address the deprecation of
readdir_r(3) in glibc 2.24. It turns out that the frontend code is
the worst place to handle a critical section with a pthread mutex:
the code runs in a coroutine on behalf of the QEMU mainloop and then
yields control, waiting for the fsdev backend to process the request
in a worker thread. If the client resends another readdir request for
the same fid before the previous one finally unlocked the mutex, we're
deadlocked.
This never bit us because the linux client serializes readdir requests
for the same fid, but it is quite easy to demonstrate with a custom
client.
A good solution could be to narrow the critical section in the worker
thread code and to return a copy of the dirent to the frontend, but
this causes quite some changes in both 9p.c and codir.c. So, instead
of that, in order for people to easily backport the fix to older QEMU
versions, let's simply use a CoMutex since all the users for this
sit in coroutines.
Fixes: 7cde47d4a8 ("9p: add locking to V9fsDir")
Reviewed-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Message-Id: <158981894794.109297.3530035833368944254.stgit@bahia.lan>
Signed-off-by: Greg Kurz <groug@kaod.org>
When updating the PFLASH file contents, we should check for a
possible failure of blk_pwrite(). Similar to commit 3a688294e.
Reported-by: Coverity (CID 1357678 CHECKED_RETURN)
Signed-off-by: Mansour Ahmadi <mansourweb@gmail.com>
Message-Id: <20200408003552.58095-1-mansourweb@gmail.com>
[PMD: Add missing "qemu/error-report.h" include and TODO comment]
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Rename the 'reset_flash' as 'mode_read_array' to make explicit we
do not reset the device, we simply set its internal state machine
in the READ_ARRAY mode. We do not reset the status register error
bits, as a device reset would do.
Reviewed-by: John Snow <jsnow@redhat.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-Id: <20190716221555.11145-5-philmd@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
The command 0x00 is used by this model since its origin (commit
05ee37ebf6). In this commit the command is described with a
amusing '/* ??? */' comment, probably meaning 'FIXME'.
switch (cmd) {
case 0x00: /* ??? */
...
This comment survived 12 years because the 0x00 value is indeed
not specified by the CFI open standard (as of this commit).
The 'cmd' field is transfered during migration. To keep the
migration feature working with older QEMU version, we have to
take a lot of care with migrated field. We figured out it is
too late to remove a non-specified value from this model
(this would make migration review very complex). It is however
not too late to improve the documentation.
Add few comments to remember this is a special value related
to QEMU, and we won't find information about it on the CFI
spec.
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-Id: <20190716221555.11145-3-philmd@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
The Arm signal-handling code has some parts ifdeffed with a
TARGET_CONFIG_CPU_32, which is always defined. This is a leftover
from when this code's structure was based on the Linux kernel
signal handling code, where it was intended to support 26-bit
Arm CPUs. The kernel dropped its CONFIG_CPU_32 in kernel commit
4da8b8208eded0ba21e3 in 2009.
QEMU has never had 26-bit CPU support and is unlikely to ever
add it; we certainly aren't going to support 26-bit Linux
binaries via linux-user mode. The ifdef is just unhelpful
noise, so remove it entirely.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200518143014.20689-1-peter.maydell@linaro.org
Using the MSR instruction to write to CPSR.E is deprecated, but it is
required to work from any mode including unprivileged code. We were
incorrectly forbidding usermode code from writing it because
CPSR_USER did not include the CPSR_E bit.
We use CPSR_USER in only three places:
* as the mask of what to allow userspace MSR to write to CPSR
* when deciding what bits a linux-user signal-return should be
able to write from the sigcontext structure
* in target_user_copy_regs() when we set up the initial
registers for the linux-user process
In the first two cases not being able to update CPSR.E is a bug, and
in the third case it doesn't matter because CPSR.E is always 0 there.
So we can fix both bugs by adding CPSR_E to CPSR_USER.
Because the cpsr_write() in restore_sigcontext() is now changing
a CPSR bit which is cached in hflags, we need to add an
arm_rebuild_hflags() call there; the callsite in
target_user_copy_regs() was already rebuilding hflags for other
reasons.
(The recommended way to change CPSR.E is to use the 'SETEND'
instruction, which we do correctly allow from usermode code.)
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200518142801.20503-1-peter.maydell@linaro.org
hw_error() calls exit(). This a bit overkill when we can log
the accesses as unimplemented or guest error.
When fuzzing the devices, we don't want the whole process to
exit. Replace some hw_error() calls by qemu_log_mask().
Per the datasheet "Exynos 4412 RISC Microprocessor Rev 1.00"
Chapter 25 "Multi Core Timer (MCT)" figure 1 and table 4,
the default value on the APB bus is 0.
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20200518140309.5220-5-f4bug@amsat.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
hw_error() calls exit(). This a bit overkill when we can log
the accesses as unimplemented or guest error.
When fuzzing the devices, we don't want the whole process to
exit. Replace some hw_error() calls by qemu_log_mask().
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20200518140309.5220-4-f4bug@amsat.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
hw_error() calls exit(). This a bit overkill when we can log
the accesses as unimplemented or guest error.
When fuzzing the devices, we don't want the whole process to
exit. Replace some hw_error() calls by qemu_log_mask().
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20200518140309.5220-3-f4bug@amsat.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
hw_error() calls exit(). This a bit overkill when we can log
the accesses as unimplemented or guest error.
When fuzzing the devices, we don't want the whole process to
exit. Replace some hw_error() calls by qemu_log_mask().
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20200518140309.5220-2-f4bug@amsat.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Our code to identify syscall numbers has some issues:
* for Thumb mode, we never need the immediate value from the insn,
but we always read it anyway
* bad immediate values in the svc insn should cause a SIGILL, but we
were abort()ing instead (via "goto error")
We can fix both these things by refactoring the code that identifies
the syscall number to more closely follow the kernel COMPAT_OABI code:
* for Thumb it is always r7
* for Arm, if the immediate value is 0, then this is an EABI call
with the syscall number in r7
* otherwise, we XOR the immediate value with 0x900000
(ARM_SYSCALL_BASE for QEMU; __NR_OABI_SYSCALL_BASE in the kernel),
which converts valid syscall immediates into the desired value,
and puts all invalid immediates in the range 0x100000 or above
* then we can just let the existing "value too large, deliver
SIGILL" case handle invalid numbers, and drop the 'goto error'
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 20200420212206.12776-5-peter.maydell@linaro.org
The kernel has different handling for syscalls with invalid
numbers that are in the "arm-specific" range 0x9f0000 and up:
* 0x9f0000..0x9f07ff return -ENOSYS if not implemented
* other out of range syscalls cause a SIGILL
(see the kernel's arch/arm/kernel/traps.c:arm_syscall())
Implement this distinction. (Note that our code doesn't look
quite like the kernel's, because we have removed the
0x900000 prefix by this point, whereas the kernel retains
it in arm_syscall().)
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20200420212206.12776-4-peter.maydell@linaro.org
We incorrectly treat SVC 0xf0002 as a cacheflush request (which is a
NOP for QEMU). This is the wrong syscall number, because in the
svc-immediate OABI syscall numbers are all offset by the
ARM_SYSCALL_BASE value and so the correct insn is SVC 0x9f0002.
(This is handled further down in the code with the other Arm-specific
syscalls like NR_breakpoint.)
When this code was initially added in commit 6f1f31c069 in
2004, ARM_NR_cacheflush was defined as (ARM_SYSCALL_BASE + 0xf0000 + 2)
so the value in the comparison took account of the extra 0x900000
offset. In commit fbb4a2e371 in 2008, the ARM_SYSCALL_BASE
was removed from the definition of ARM_NR_cacheflush and handling
for this group of syscalls was added below the point where we subtract
ARM_SYSCALL_BASE from the SVC immediate value. However that commit
forgot to remove the now-obsolete earlier handling code.
Remove the spurious ARM_NR_cacheflush condition.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 20200420212206.12776-3-peter.maydell@linaro.org
In linux-user/arm/cpu-loop.c we incorrectly treat EXCP_BKPT similarly
to EXCP_SWI, which means that if the guest executes a BKPT insn then
QEMU will perform a syscall for it (which syscall depends on what
value happens to be in r7...). The correct behaviour is that the
guest process should take a SIGTRAP.
This code has been like this (more or less) since commit
06c949e62a in 2006 which added BKPT in the first place. This is
probably because at the time the same code path was used to handle
both Linux syscalls and semihosting calls, and (on M profile) BKPT
with a suitable magic number is used for semihosting calls. But
these days we've moved handling of semihosting out to an entirely
different codepath, so we can fix this bug by simply removing this
handling of EXCP_BKPT and instead making it deliver a SIGTRAP like
EXCP_DEBUG (as we do already on aarch64).
Reported-by: <omerg681@gmail.com>
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20200420212206.12776-2-peter.maydell@linaro.org
Fixes: https://bugs.launchpad.net/qemu/+bug/1873898
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
