Callers must pass ELF machine, byte swapping and symbol LSB clearing
information to ELF loader. A.out loader needs page size information, pass
that too as a parameter.
Extract prototypes to a separate file. Move loader.[ch] and elf_ops.h under hw.
Adjust callers. Also use target_phys_addr_t instead of target_ulong for
addresses: loader addresses aren't virtual.
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
While i386, x86_64 and Sparc64/OpenBSD still worked after
df70204db5, Sparc32 and Sparc64 Linux hosts
broke.
Partially revert the commit: make the restored code conditional to
!CONFIG_USER_PIE.
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
There is a link hack in linux-user which produces an executable that
looks like PIE, but always has text relocations since all object files
isn't position-independent (compiled without -fpic/-fpie). Dynamic loader
has to do more work to load a binary with text relocations.
The best way to keep this functionality is to build a true PIE without
text relocations.
Signed-off-by: Kirill A. Shutemov <kirill@shutemov.name>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Problem: Our file sys-queue.h is a copy of the BSD file, but there are
some additions and it's not entirely compatible. Because of that, there have
been conflicts with system headers on BSD systems. Some hacks have been
introduced in the commits 15cc923584,
f40d753718,
96555a96d7 and
3990d09adf but the fixes were fragile.
Solution: Avoid the conflict entirely by renaming the functions and the
file. Revert the previous hacks.
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
qemu's page table can be incomple if /proc/self/maps is unavailable or
host allocating a memory with mmap(), so we can't use it to find free
memory area.
New version mmap_find_vma() uses mmap() without MAP_FIXED to find free
memory.
Tested-by: Martin Mohring <martin.mohring@opensuse.org> :
quite some time ago this patch had been sent by Kirill to the QEMU ml.
At that time, the patch was rejected. Now we found out why the current
user mode memory allocator sometimes fails:
- Kernel Bug linux/fs/proc/task_mmu.c (fixed after 2.6.27)
http://bugzilla.kernel.org/attachment.cgi?id=17219
- use of proc file system to find memory mappings => bad idea
So I please apply the attached patch from Kirill to qemu to fix this
longstanding bug, because it causes all older linux distros (using
kernel 2.6.26 or older) to fail the QEMU memory allocator in user mode.
Signed-off-by: Kirill A. Shutemov <kirill@shutemov.name>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
The fstat implementation does not initialize the nanosecond fields in the
stat buffer; this caused funny values to turn up there, preventing, for
instance, cp -p from preserving timestamps because utimensat rejected
the out-of-bounds nanosecond values. Resetting the entire structure
to zero fixes that.
Signed-off-by: Ulrich Hecht <uli@suse.de>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
works perfectly fine with the example from getdents(2) and passes the LTP
tests (tested with s390x on x86_64 emulation)
Signed-off-by: Ulrich Hecht <uli@suse.de>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
Fixes swaps on l_pid which were pretty much of random size. Implements
F_SETLEASE, F_GETLEASE. Now passes all LTP fcntl tests.
Signed-off-by: Ulrich Hecht <uli@suse.de>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
mqueue.h is only available if __NR_mq_open is defined. So don't include
it unconditionally. Similarly, the mq_* family of syscalls depend on
__NR_mq_open. Finally, the copy_{from,to}_user_mq_attr functions should
not be defined unconditionally, but only if we're going to use the mq_*
syscalls.
Signed-off-by: Nathan Froyd <froydnj@codesourcery.com
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
As setup_frame() and setup_rt_frame() are now implemented we can now
enable sigaltstack().
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
This patch adds signals management for linux-user.
It implements setup_frame() which allows to call the user signal
handler.
setup_rt_frame() is always unimplemented.
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
Straightforward implementation. This syscall is rare enough that we
don't need to support the odder cases, just disable it if host glibc
is too old.
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
this patch fixes an issue in symfind.
Assume you have the following symbols:
Address Size
0045bca0 00000080 T s0
0045bd20 00000112 T s1
You'll notice that s1 is s0 + size.
So the current symfind will find that address 0045bd20 belongs to s0
instead of s1.
Laurent
Signed-off-by: Laurent Desnogues <laurent.desnogues@gmail.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Message-Id:
tl and tsptr of members sparc64 cpu state must be changed
simultaneously to keep trap state window in sync with current
trap level. Currently translation of store to tl does not change
tsptr, which leads to corrupt trap state on corresponding
trap level.
This patch removes tsptr from sparc64 cpu state and replaces
all uses with call to helper routine.
Changes v0->v1:
- reimplemented helper routine with tcg generator
- on cpu reset trap type and pstate are populated with power-on reset
values, including tl=maxtl
Signed-off-by: igor.v.kovalenko@gmail.com
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
We handle conditional stores as an exception so we can ensure that no
other thread is changing memory out from underneath us.
Signed-off-by: Nathan Froyd <froydnj@codesourcery.com>
Signed-off-by: malc <av1474@comtv.ru>
Use was not consistent, in Makefile was TARGET_GPROF and in *h HAVE_GPROF
Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
I used the following command to enable debugging:
perl -p -i -e 's/^\/\/#define DEBUG/#define DEBUG/g' * */* */*/*
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
All elf64_note structure members are Elf64_Word (which is 32bit value)
hence using bswaptls to byteswap it on 64bit platforms is incorrect.
Signed-off-by: malc <av1474@comtv.ru>
Re-implement GUEST_BASE support.
Offset guest ddress space by default if the guest binary contains
regions below the host mmap_min_addr.
Implement support for i386, x86-64 and arm hosts.
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
Signed-off-by: Paul Brook <paul@codesourcery.com>
Fixes getrlimit implementation that overwrote the result of the syscall
instead of converting it
Signed-off-by: Ulrich Hecht <uli@suse.de>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
makes socketcall 64-bit clean so it works on 64-bit big-endian systems
Signed-off-by: Ulrich Hecht <uli@suse.de>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
pipedes is an address, it should not be signed (breaks for addresses
> 0x80000000)
Signed-off-by: Ulrich Hecht <uli@suse.de>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
Removes the following warning
CC i386-linux-user/syscall.o
cc1: warnings being treated as errors
/media/nfs/qemu/linux-user/syscall.c: In function ‘do_syscall’:
/media/nfs/qemu/linux-user/syscall.c:2219: warning: ‘array’ may be used uninitialized in this function
Signed-off-by: Vibi Sreenivasan <vibi_sreenivasan@cms.com>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
This patch is fixing following issues :
- commit 8fea36025b was applied to
do_getsockname instead of do_accept.
- Some syscalls were not checking properly the memory addresses passed
as argument
- Add check before syscalls made for cases like do_getpeername() where
we're using the address parameter after doing the syscall
- Fix do_accept to return EINVAL instead of EFAULT when parameters
invalid to match with linux behaviour
Signed-off-by: Arnaud Patard <arnaud.patard@rtp-net.org>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
There's a error When doing something like that :
find / -type f -print0 | xargs -0 echo
[ done in a arm chroot with qemu-arm and linux binfmt stuff or with
find / -type f -print0 | qemu-arm -L <path> <path>/usr/bin/xargs -0
echo ]
Doing this outsite qemu is fine. The problem was the huge number of
parameters. Increasing MAX_ARG_PAGES is fixing that.
While I was at it, I've modified linux-user/main.c to report error code
of loader_exec. It helps to debug/know what's wrong.
Signed-off-by: Arnaud Patard <arnaud.patard@rtp-net.org>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
Fix missing strnlen (a GNU extension) problems by using qemu_strnlen
used for user emulators also for system emulators.
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Microblaze recently changed their ABI. The new is not backwards compatible
and there doesn't seem to be a way to distinguish old/new binaries.
Let's support the latest ABI for now and hope someone figures out a way to
hande both ABI's later.
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>
On Tue, Jun 16, 2009 at 08:19:23PM -0500, Anthony Liguori wrote:
> malc wrote:
>>
>> On my system the above line causes gcc to emit:
>>
>> In file included from /home/malc/x/rcs/git/qemu/linux-user/strace.c:12:
>> /usr/include/linux/futex.h:48: error: field `__user' has incomplete type
>> /usr/include/linux/futex.h:48: error: syntax error before '*' token
>> /usr/include/linux/futex.h:63: error: field `list' has incomplete type
>> /usr/include/linux/futex.h:83: error: field `__user' has incomplete type
>> /usr/include/linux/futex.h:83: error: syntax error before '*' token
>> make[1]: *** [strace.o] Error 1
> We had the same problem with usb-linux.c. It's broken system headers,
> the __user stuff is supposed to get removed as part of the headers
> installation.
> It builds fine on my system (Fedora 10).
Howabout something like this:
commit eb8387cb0eda32a18880664eb5f0ca5c8bf05b45
Author: Riku Voipio <riku.voipio@iki.fi>
Date: Thu Jun 18 22:44:31 2009 +0300
Subject: linux-user: include futex defines directly
Since some common distributions have broken linux/futex.h, stop
including it. Instead add the defines directly.
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
This issue has been detected with tests/linux-tests.c:
linux-test.c:330: getsockopt
327 len = sizeof(val);
328 chk_error(getsockopt(server_fd, SOL_SOCKET, SO_TYPE, &val, &len));
329 if (val != SOCK_STREAM)
330 error("getsockopt");
In linux-user/syscall.c:do_getsockopt(), we have:
...
val = tswap32(val);
...
if (put_user_u32(val, optval_addr))
...
whereas "put_user_u32" calls in the end "__put_user" which uses "tswap32".
So the "val = tswap32(val);" is useless and wrong.
This patch removes it.
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
Needed to make sure the xxxat() functions are available.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
We initialize mmap_mutex in any child threads/processes, but we need to
correctly statically initialize it for the original process.
Signed-off-by: Nathan Froyd <froydnj@codesourcery.com>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
Hi,
This is a new try to fix the fcntl support in linux-user. I tried to
adress all comments but as the previous version is several weeks old,
it's possible that I've missed some.
This patch doesn't handle linux specific fcntl flags. My plan is to get
this version of the patch reviewed/fixed and then, add them if wanted.
Thanks,
Arnaud
Signed-off-by: Arnaud Patard (Rtp) <arnaud.patard@rtp-net.org>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
Add support for tee, splice and vmsplice.
Originally from: vibi sreenivasan <vibi_sreenivasan@cms.com>
Riku: squashed patches together, added a test to configure
and removed compliler warning by picking up correct type for
splice param
Signed-off-by: vibisreenivasan <vibi_sreenivasan@cms.com>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
implement pipe2 syscall.
[v2] fix do_pipe on mips and sh4
[v3] use pipe2 to ensure atomicity, but only when it is available.
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
defines FIGETBSZ FIBMAP, allowing the respective ioctl's to
be implemented.
From: Martin Mohring <martin.mohring@opensuse.org>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
In current code, we're sending ENOSYS to target when a syscall for the
xattrs is done. This makes applications like ls complain loudly about
that and breaks scripts parsing the output. Moreover, iirc, implemented
features of filesystems are are sending EOPNOTSUPP (I've not checked so
I may be a little bit wrong on that...).
So, I'm proposing to return -EOPNOTSUPP and make ls happy
Signed-off-by: Arnaud Patard <arnaud.patard@rtp-net.org>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
When converting the termios structure between host and target in
target_to_host_termios and host_to_target_termios, the c_cc[] array is
never initialised.
Calling memset() before using it allows to run successfully "stty echo /
stty -echo" on arm-linux-user target (host being x86 and mips).
Signed-off-by: Arnaud Patard <arnaud.patard@rtp-net.org>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
The glibc function for utimensat glibc returns -EINVAL when the path is null
which is a different behaviour with the syscall.
path can be null because internally the glibc is using utimensat with
path null when implmenting futimens. If path is null, call futimes
instead.
don't try to copy timespec from user if is NULL.
Add configure check for older systems
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
- to not to break strace with GUEST_BASE is set:
- Strace now can load and print guest strings correctly.
- Added printing support for commonly used flags in some syscalls
(e.g open, creat, mmap etc.)
v2:
- fix strace.c build on etch
- add futex print to strace
Signed-off-by: Mika Westerberg <mika.westerberg@iki.fi>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
When target process is killed with signal (such signal that
should dump core) a coredump file is created. This file is
similar than coredump generated by Linux (there are few exceptions
though).
Riku Voipio: added support for rlimit
Signed-off-by: Mika Westerberg <mika.westerberg@iki.fi>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
No regressions were observed on either 64bit or 32bit
IA hosts.
Patch based on original patches by:
Kirill A. Shutemov <kirill@shutemov.name>
- Implement shm* syscalls
- Fix and cleanup IPCOP_shm* ipc calls handling
Depends on "export mmap_find_vma for shmat" patch.
Various whitespace uglifications applied to minimize
patch size.
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
When debugging multi-threaded programs, QEMU's gdb stub would report the
correct number of threads (the qfThreadInfo and qsThreadInfo packets).
However, the stub was unable to actually switch between threads (the T
packet), since it would report every thread except the first as being
dead. Furthermore, the stub relied upon cpu_index as a reliable means
of assigning IDs to the threads. This was a bad idea; if you have this
sequence of events:
initial thread created
new thread #1
new thread #2
thread #1 exits
new thread #3
thread #3 will have the same cpu_index as thread #1, which would confuse
GDB. (This problem is partly due to the remote protocol not having a
good way to send thread creation/destruction events.)
We fix this by using the host thread ID for the identifier passed to GDB
when debugging a multi-threaded userspace program. The thread ID might
wrap, but the same sort of problems with wrapping thread IDs would come
up with debugging programs natively, so this doesn't represent a
problem.
Signed-off-by: Nathan Froyd <froydnj@codesourcery.com>
Implement setup_{,rt_}frame and do_{,rt_}sigreturn for PPC 32-bit. Use
the same TARGET_QEMU_ESIGRETURN hack as for MIPS to avoid clobbering
register state on a sigreturn.
Signed-off-by: Nathan Froyd <froydnj@codesourcery.com>
Signed-off-by: malc <av1474@comtv.ru>
Also fixes a register corruption bug in do_sigreturn. When "returning"
from sigreturn we are actually restoring the virtual cpu state from the
signal frame. This is actually surprisingly hard to observe in practice.
Typically an thread be blocked in a FUTEX_WAIT call when the signal arrives,
so the effect is a spurious syscall success and the introduction of a
subtle race condition.
On x86/arm a syscall modifies a single word sized register, so
do_sigreturn can just return that value. On MIPS a syscall clobbers
multiple registers, so we need additional smarts. My solution is to
invent a magic errno value that means "don't touch CPU state".
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@7194 c046a42c-6fe2-441c-8c8c-71466251a162
In Linux kernel, fchmodat() and faccessat() take tree args.
4th value <int flags> is only processed by libc.
Signed-off-by: Takashi YOSHII <takasi-y@ops.dti.ne.jp>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@7187 c046a42c-6fe2-441c-8c8c-71466251a162
Fix and cleanup IPCOP_sem* ipc calls handling and
implement sem* syscalls.
Riku:
1) Uglify whitespace so that diff gets smaller and easier
to review
2) use __get_user in target_to_host_sembuf
Signed-off-by: Kirill A. Shutemov <kirill@shutemov.name>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@7184 c046a42c-6fe2-441c-8c8c-71466251a162
Configure test was broken, so the breakage of the #ifdef'd
code was not noticed.
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@7134 c046a42c-6fe2-441c-8c8c-71466251a162
The patch called "prefer glibc over direct syscalls" (commit 7118) has
replaced the getcwd syscall with a call to the glibc. With this change,
the syscall is returning -1 in error case and 0 otherwise.
This is problematic as the sys_getcwd syscall should return the number
of bytes written to the buffer including the '\0'.
Signed-off-by: Arnaud Patard <arnaud.patard@rtp-net.org>
Acked-By: Riku Voipio <riku.voipio@iki.fi>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@7130 c046a42c-6fe2-441c-8c8c-71466251a162
The proper exit code for dieing from an uncaught signal is -<signal>.
The kernel doesn't allow exit() or _exit() to pass a negative value.
To get the proper exit code we need to actually die from an uncaught signal.
A default signal handler is installed, we send ourself a signal
and we wait for it to arrive.
Patch originates from Scratchbox
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@7119 c046a42c-6fe2-441c-8c8c-71466251a162
The openat/*at syscalls are incredibly common with modern coreutils,
calling them directly via syscalls breaks for example fakeroot. Use
glibc stubs whenever directly available and provide old syscall
calling for people still using older libc.
Patch originally from Mika Westerberg, Adapted to
apply to current trunk and cleaned up by Riku Voipio.
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@7118 c046a42c-6fe2-441c-8c8c-71466251a162
- This check is not needed because kernel will check whether given
buffer is too small and there is no upper limit for size of the buffer.
From: Mika Westerberg <mika.westerberg@iki.fi>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@7117 c046a42c-6fe2-441c-8c8c-71466251a162
dbus sends too short (according to man 7 unix) addrlen for it's
unix socket. I've been told that happens with other applications
as well. Linux kernel doesn't appear to mind, so I guess
we whould be tolerant as well. Expand sockaddr with +1 to fit
the \0 of the pathname passed.
(scratchbox1 qemu had a very different workaround for the same issue).
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@7116 c046a42c-6fe2-441c-8c8c-71466251a162
Added switch -0 (zero) which can be used to pass argv[0] to
target process. The main use is for a binfmt_misc wrapper when
the "P - preserve-argv[0]" setting is used.
From: Mika Westerberg
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@7115 c046a42c-6fe2-441c-8c8c-71466251a162
There are two different syscall names for the same goal.
On systems with sizeof(long) == 64 it calls newfstatat.
On systems with sizeof(long) == 32 it calls fstatat64.
Signed-off-by: Kirill A. Shutemov <kirill@shutemov.name>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@7050 c046a42c-6fe2-441c-8c8c-71466251a162
The sys_inotify* calls are defined if the target supports them and the
host supports the necessary syscalls. But the syscalls are handled if
the target supports them. This situation leads to compilation failures
when the host doesn't support the necessary syscalls, as the linker will
complain about undefined functions.
Fix this state of affairs by making the handling conditions the same as
the call definition conditions.
Signed-off-by: Nathan Froyd <froydnj@codesourcery.com>
Acked-By: Riku Voipio <riku.voipio@iki.fi>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@7038 c046a42c-6fe2-441c-8c8c-71466251a162
Allows distributors to identify their builds without needing to hack the
sources.
Signed-off-by: Paul Brook <paul@codesourcery.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@7036 c046a42c-6fe2-441c-8c8c-71466251a162
This replaces a compile time option for some targets and adds
this feature to targets which did not have a compile time option.
Add monitor command to enable or disable single step mode.
Modify monitor command "info status" to display single step mode.
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@7004 c046a42c-6fe2-441c-8c8c-71466251a162