When executing a 64bit target chroot on 64bit host,
the ioctl() command can mismatch.
It seems the previous commit doesn't solve the problem in
my case:
9c6bf9c7 linux-user: Fix ioctl cmd type mismatch on 64-bit targets
For example, a ppc64 chroot on an x86_64 host:
bash-4.3# ls
Unsupported ioctl: cmd=0x80087467
Unsupported ioctl: cmd=0x802c7415
The origin of the problem is in syscall.c:do_ioctl().
static abi_long do_ioctl(int fd, abi_long cmd, abi_long arg)
In this case (ppc64) abi_long is long (on the x86_64), and
cmd = 0x0000000080087467
then
if (ie->target_cmd == cmd)
target_cmd is int, so target_cmd = 0x80087467
and to compare an int with a long, the sign is extended to 64bit,
so the comparison is:
if (0xffffffff80087467 == 0x0000000080087467)
which doesn't match whereas it should.
This patch uses int in the case of the target command type
instead of abi_long.
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
The target payloads in cmsg conversions may not have the alignment
required by the host. Using the get_user and put_user functions is
the easiest way to handle this and also do the byte-swapping we
require.
(Note that prior to this commit target_to_host_cmsg was incorrectly
using __put_user() rather than __get_user() for the SCM_CREDENTIALS
conversion, which meant it wasn't getting the benefit of the
misalignment handling.)
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
The previous code for handling payload length when converting
cmsg structures from host to target had a number of problems:
* we required the msg->msg_controllen to declare the buffer
to have enough space for final trailing padding (we were
checking against CMSG_SPACE), whereas the kernel does not
require this, and common userspace code assumes this. (In
particular, glibc's "try to talk to nscd" code that it will
run on startup will receive a cmsg with a 4 byte payload and
only allocate 4 bytes for it, which was causing us to do
the wrong thing on architectures that need 8-alignment.)
* we weren't correctly handling the fact that the SO_TIMESTAMP
payload may be larger for the target than the host
* we weren't marking the messages with MSG_CTRUNC when we did
need to truncate a message that wasn't truncated by the host,
but were instead logging a QEMU message; since truncation is
always the result of a guest giving us an insufficiently
sized buffer, we should report it to the guest as the kernel
does and don't log anything
Rewrite the parts of the function that deal with length to
fix these issues, and add a comment in target_to_host_cmsg
to explain why the overflow logging it does is a QEMU bug,
not a guest issue.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
We store all struct types in an array of static size without ever
checking whether we overrun it. Of course some day someone (like me
in another, ancient ALSA enabling patch set) will run into the limit
without realizing it.
So let's make the allocation dynamic. We already know the number of
structs that we want to allocate, so we only need to pass the variable
into the respective piece of code.
Also, to ensure we don't accidently overwrite random memory, add some
asserts to sanity check whether a thunk is actually part of our array.
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
If QEMU forks after the CPU threads have been created, qemu_mutex_lock_iothread
will not be able to do qemu_cpu_kick_thread. There is no solution other than
assuming that forks after the CPU threads have been created will end up in an
exec. Forks before the CPU threads have been created (such as -daemonize)
have to call rcu_after_fork manually.
Notably, the oxygen theme for GTK+ forks and shows a "No such process" error
without this patch.
This patch can be reverted once the iothread loses the "kick the TCG thread"
magic.
User-mode emulation does not use the iothread, so it can also call
rcu_after_fork.
Reported by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Tested by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The second and fourth argument are in/out parameters, store them back
after the syscall. Also, the fourth argument was mishandled, and EFAULT
handling was missing.
Signed-off-by: Andreas Schwab <schwab@suse.de>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
In abi_long do_ioctl_dm(), after lock_user() call, the code does
not call unlock_user() before going to failure return in default case.
Signed-off-by: Chen Gang <gang.chen.5i5j@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
It is only a typo issue, need use tswapal(target_vec[i].iov_len) for the
len.
Signed-off-by: Chen Gang <gang.chen.5i5j@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
When failure occurs during locking of vec[i], we also need to unlock all
already locked vec[i] in failure processing code block before return.
Code in unlock_user() checks vec[i].iov_base for NULL, so there's no
need not check it .
If error is EFAULT when "i == 0", vec[i].iov_base is NULL, we can just
skip it, so can still use "while (--i >= 0)" loop condition.
Signed-off-by: Chen Gang <gang.chen.5i5j@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
When passing ancillary data through a unix socket, handle
credentials properly instead of doing a simple copy and
issuing a warning.
Signed-off-by: Alex Suykov <alex.suykov@gmail.com>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
linux-user passes the cmd argument of the ioctl syscall as a signed long,
but compares it to an unsigned int when iterating through the ioctl_entries
list. When the cmd is a large value like 0x80047476 (TARGET_TIOCSWINSZ on
mips64) it gets sign-extended to 0xffffffff80047476, causing the comparison
to fail and resulting in lots of spurious "Unsupported ioctl" errors.
Changing the target_cmd field in the ioctl_entries list to a signed int
causes those values to be sign-extended as well during the comparison.
Signed-off-by: Ed Swierk <eswierk@skyportsystems.com>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
The resource argument is translated from host to target for
[gs]etprlimit but not for prlimit64. Fix this.
Signed-off-by: Felix Janda <felix.janda@posteo.de>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
When creating a timer handle, we give the timer id a special magic offset
of 0xcafe0000. However, we never mask that offset out of the timer id before
we start using it to dereference our timer array. So we always end up aborting
timer operations because the timer id is out of bounds.
This was not an issue before my patch e52a99f756 ("linux-user: Simplify
timerid checks on g_posix_timers range") because before we would blindly mask
anything above the first 16 bits.
This patch simplifies the code around timer id creation by introducing a proper
target_timer_id typedef that is s32, just like Linux has it. It also changes the
magic offset to a value that makes all timer ids be positive.
Reported-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Tom Musta <tommusta@gmail.com>
Tested-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
Without this, builds on older systems fail with:
qemu/linux-user/syscall.c:61:25: warning: sys/timerfd.h: No such file or directory
v2: fix the usual case where CONFIG_TIMERFD is enabled..
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
We check whether the passed in timer id is negative on all calls
that involve g_posix_timers.
However, these checks are bogus. First off we limit the timer_id to
16 bits which is not what Linux does. Then we check whether it's negative
which it can't be because we masked it.
We can safely remove the masking. For the negativity check we can just
treat the timerid as unsigned and only check for upper boundaries.
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
The blkpg ioctl can take different payloads depending on the opcode in
its payload structure. Create a new special ioctl handler that can only
deal with partition style ones for now.
This patch fixes running parted for me.
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
Handle variable "fd_orig" going out of scope leaks the handle.
Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
Reviewed-by: Gonglei <arei.gonglei@huawei.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Although not technically not required by POSIX, the writev system call will
typically write out its buffers individually. That is, if the first buffer
is written successfully, but the second buffer pointer is invalid, then
the first chuck will be written and its size is returned.
Signed-off-by: Tom Musta <tommusta@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
The argument to the mlockall system call is not necessarily the same on
all platforms and thus may require translation prior to passing to the
host.
For example, PowerPC 64 bit platforms define values for MCL_CURRENT
(0x2000) and MCL_FUTURE (0x4000) which are different from Intel platforms
(0x1 and 0x2, respectively)
Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
The clock_nanosleep syscall is unusual in that it returns positive
numbers in error handling situations, versus returning -1 and setting
errno, or returning a negative errno value. On POWER, the kernel will
set the SO bit of CR0 to indicate failure in a syscall. QEMU has
generic handling to do this for syscalls with standard return values.
Add special case code for clock_nanosleep to handle CR0 properly.
Signed-off-by: Tom Musta <tommusta@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
Properly detect a fault when attempting to store into an invalid
struct timespec pointer.
Signed-off-by: Tom Musta <tommusta@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
The sched_getparam, sched_setparam and sched_setscheduler system
calls take a pointer argument to a sched_param structure. When
this pointer is null, errno should be set to EINVAL.
Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
The msgsnd system call takes an argument that describes the message
size (msgsz) and is of type size_t. The system call should set
errno to EINVAL in the event that a negative message size is passed.
Signed-off-by: Tom Musta <tommusta@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
The mq_open system call takes an optional struct mq_attr pointer
argument in the fourth position. This pointer is used when O_CREAT
is specified in the flags (second) argument. It may be NULL, in
which case the queue is created with implementation defined attributes.
Change the code to properly handle the case when NULL is passed in the
arg4 position.
Signed-off-by: Tom Musta <tommusta@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
For those target ABIs that use the ipc system call (e.g. POWER),
the third argument is used in the shmat path as a pointer. It
therefore must be declared as an abi_long (versus int) so that
the address bits are not lost in truncation. In fact, all arguments
to do_ipc should be declared as abit_long.
In fact, it makes more sense for all of the arguments to be declaried
as abi_long (except call).
Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
The semun union used in the semctl system call contains both an int (val) and
pointers. In cross-endian situations on 64 bit targets, the value passed to
semctl is an 8 byte (abi_long) value and thus does not have the 4-byte val
field in the correct location. In order to rectify this, the other half
of the union must be accessed. This is achieved in code by performing
a byte swap on the entire 8 byte union, followed by a 4-byte swap of the
first half.
Also, eliminate an extraneous (dead) line of code that sets target_su.val in
the IPC_SET/IPC_GET case.
Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
When the ipc system call is used to wrap a semctl system call,
the ptr argument to ipc needs to be dereferenced prior to passing
it to the semctl handler. This is because the fourth argument to
semctl is a union and not a pointer to a union.
Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
The 64 bit PowerPC platforms eliminate the _unused1 and _unused2
elements of the semid_ds structure from <sys/sem.h>. So eliminate
these from the target_semid_ds structure.
Signed-off-by: Tom Musta <tommusta@gmail.com>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
Add support for the setns and unshare syscalls, trivially passed through to
the host. Based on patches by Paul Burton, added configure check.
Signed-off-by: Paul Burton <paul@archlinuxmips.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
Add support for the ioprio_get & ioprio_set syscalls, allowing their
use by target programs.
Signed-off-by: Paul Burton <paul@archlinuxmips.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
Adds support for the timerfd_create, timerfd_gettime & timerfd_settime
syscalls, allowing use of timerfds by target programs.
v2: By Riku - added configure check for timerfd and ifdefs
for benefit of old distributions like RHEL5.
Signed-off-by: Paul Burton <paul@archlinuxmips.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
The current code always returns the length of the path when it should
be returning the number of bytes it wrote to the output string.
Further, readlink is not supposed to append a NUL byte, but the current
snprintf logic will always do just that.
Even further, if you pass in a length of 0, you're suppoesd to get back
an error (EINVAL), but the current logic just returns 0.
Further still, if there was an error reading the symlink, we should not
go ahead and try to read the target buffer as it is garbage.
Simple test for the first two issues:
$ cat test.c
int main() {
char buf[50];
size_t len;
for (len = 0; len < 10; ++len) {
memset(buf, '!', sizeof(buf));
ssize_t ret = readlink("/proc/self/exe", buf, len);
buf[20] = '\0';
printf("readlink(/proc/self/exe, {%s}, %zu) = %zi\n", buf, len, ret);
}
return 0;
}
Now compare the output of the native:
$ gcc test.c -o /tmp/x
$ /tmp/x
$ strace /tmp/x
With what qemu does:
$ armv7a-cros-linux-gnueabi-gcc test.c -o /tmp/x -static
$ qemu-arm /tmp/x
$ qemu-arm -strace /tmp/x
Signed-off-by: Mike Frysinger <vapier@chromium.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
There were a number of bugs in the conversion of the sigevent
argument to timer_create from target to host format:
* signal number not converted from target to host
* thread ID not copied across
* sigev_value not copied across
* we never unlocked the struct when we were done
Between them, these problems meant that SIGEV_THREAD_ID
timers (and the glibc-implemented SIGEV_THREAD timers which
depend on them) didn't work.
Fix these problems and clean up the code a little by pulling
the struct conversion out into its own function, in line with
how we convert various other structs. This allows the test
program in bug LP:1042388 to run.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
While Mikhail fixed /proc/self/maps, it was noticed openat calls are
not redirected currently. Some archs don't have open at all, so
openat needs to be redirected.
Fix this by consolidating open/openat code to do_openat - open
is implemented using openat(AT_FDCWD, ... ), which according
to open(2) man page is identical.
Since all targets now have openat, remove the ifdef around sys_openat
and openat: case in do_syscall.
Cc: Mikhail Ilin <m.ilin@samsung.com>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
Build /proc/self/maps doing a match against guest memory translation table.
Output only that map records which are valid for guest memory layout.
Signed-off-by: Mikhail Ilyin <m.ilin@samsung.com>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
target_to_host_sockaddr() may increase the lenth with 1 byte
for AF_UNIX sockets so allocate 1 extra byte.
Signed-off-by: Joakim Tjernlund <Joakim.Tjernlund@transmode.se>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
Add a definition of the KDSIGACCEPT ioctl & allow its use by target
programs.
Signed-off-by: Paul Burton <paul@archlinuxmips.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
The tv argument to the settimeofday syscall is allowed to be NULL, if
the program only wishes to provide the timezone. QEMU previously
returned -EFAULT when tv was NULL. Instead, execute the syscall &
provide NULL to the kernel as the target program expected.
Signed-off-by: Paul Burton <paul@archlinuxmips.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
The settimeofday syscall accepts a tz argument indicating the desired
timezone to the kernel. QEMU previously ignored any argument provided
by the target program & always passed NULL to the kernel. Instead,
translate the argument & pass along the data userland provided.
Although this argument is described by the settimeofday man page as
obsolete, it is used by systemd as of version 213.
Signed-off-by: Paul Burton <paul@archlinuxmips.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
Calls to the mount syscall can legitimately provide NULL as the value
for the source of filesystemtype arguments, which QEMU would previously
reject & return -EFAULT to the target program. An example of this is
remounting an already mounted filesystem with different properties.
Instead of rejecting such syscalls with -EFAULT, pass NULL along to the
kernel as the target program expects.
Additionally this patch fixes a potential memory leak when DEBUG_REMAP
is enabled and lock_user_string fails on the target or filesystemtype
arguments but a prior argument was non-NULL and already locked.
Since the patch already touched most lines of the TARGET_NR_mount case,
it fixes the indentation & coding style for good measure.
Signed-off-by: Paul Burton <paul@archlinuxmips.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
Translate the SO_PASSSEC option to setsockopt to the host value &
perform the syscall as expected, allowing use of the option by target
programs.
Signed-off-by: Paul Burton <paul@archlinuxmips.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
Translate the SO_SNDBUFFORCE & SO_RCVBUFFORCE options to setsockopt to
the host values & perform the syscall as expected, allowing use of those
options by target programs.
Signed-off-by: Paul Burton <paul@archlinuxmips.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
Translate the SO_ACCEPTCONN option to the host value & execute the
syscall as expected.
Signed-off-by: Paul Burton <paul@archlinuxmips.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
QEMU previously passed the result of the host syscall directly to the
target program. This is a problem if the host & target have different
representations of socket types, as is the case when running a MIPS
target program on an x86 host. Introduce a host_to_target_sock_type
helper function mirroring the existing target_to_host_sock_type, and
call it to translate the value provided by getsockopt when called for
the SO_TYPE option.
Signed-off-by: Paul Burton <paul@archlinuxmips.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
mmap_flags_tbl contains a list of mmap flags, and how to map them to
the target. This patch adds MAP_NORESERVE, which was missing to the
list.
Signed-off-by: Christophe Lyon <christophe.lyon@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
From MIPS documentation (Volume III):
UserLocal Register (CP0 Register 4, Select 2)
Compliance Level: Recommended.
The UserLocal register is a read-write register that is not interpreted by
the hardware and conditionally readable via the RDHWR instruction.
This register only exists if the Config3-ULRI register field is set.
Privileged software may write this register with arbitrary information and
make it accessible to unprivileged software via register 29 (ULR) of the
RDHWR instruction. To do so, bit 29 of the HWREna register must be set to a
1 to enable unprivileged access to the register.
Signed-off-by: Petar Jovanovic <petar.jovanovic@imgtec.com>
Reviewed-by: Andreas Färber <afaerber@suse.de>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Laurent Vivier