Now all linux-user targets support building with NPTL, we can make it
mandatory. This is a good idea because:
 * NPTL is no longer new and experimental; it is completely standard
 * in practice, linux-user without NPTL is nearly useless for
   binaries built against non-ancient glibc
 * it allows us to delete the rather untested code for handling
   the non-NPTL configuration

Note that this patch leaves the CONFIG_USE_NPTL ifdefs in the
bsd-user codebase alone. This makes no change for bsd-user, since
our configure test for NPTL had a "#include <linux/futex.h>"
which means bsd-user would never have been compiled with
CONFIG_USE_NPTL defined, and it still is not.
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>
Reviewed-by: NRichard Henderson <rth@twiddle.net>
Signed-off-by: NRiku Voipio <riku.voipio@linaro.org>

Add x86-64 implementation of cpu_set_tls() (like the kernel, we
just have to call do_arch_prctl() to set FS); this allows us to
enable NPTL.
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>
Signed-off-by: NRiku Voipio <riku.voipio@linaro.org>

We can easily set the TLS on i386. Add code to do so.
Signed-off-by: NAlexander Graf <agraf@suse.de>
[PMM: also remove "target_nptl=no" line from configure, for
 consistency with other patches in this series]
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>
Signed-off-by: NRiku Voipio <riku.voipio@linaro.org>

Linux manages to have three separate orderings of the arguments to
the clone() syscall on different architectures. In the kernel these
are selected via CONFIG_CLONE_BACKWARDS and CONFIG_CLONE_BACKWARDS2.
Clean up our implementation of this to use similar #define names
rather than a TARGET_* ifdef ladder.

This includes behaviour changes fixing bugs on cris, x86-64, m68k,
openrisc and unicore32.  cris had explicit but wrong handling; the
others were just incorrectly using QEMU's default, which happened
to be the equivalent of CONFIG_CLONE_BACKWARDS. (unicore32 appears
to be broken in the mainline kernel in that it tries to use arg3 for
both parent_tidptr and newtls simultaneously -- we don't attempt
to emulate this bug...)
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>
Signed-off-by: NRiku Voipio <riku.voipio@linaro.org>

The i386 code for the get_thread_area syscall was missing a
'break' which meant it would have fallen through into the
implementation of the following syscall; add it.
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>
Reviewed-by: NRichard Henderson <rth@twiddle.net>
Signed-off-by: NRiku Voipio <riku.voipio@linaro.org>

For m68k, per-thread data is a purely kernel construct with no
CPU level support. Implement it via a field in the TaskState structure,
used by cpu_set_tls() and the set_thread_area/get_thread_area
syscalls. This allows us to enable compilation with NPTL.
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>
Reviewed-by: NLaurent Vivier <laurent@vivier.eu>
Signed-off-by: NRiku Voipio <riku.voipio@linaro.org>

Signed-off-by: NAndreas Färber <afaerber@suse.de>

Move next_cpu from CPU_COMMON to CPUState.
Move first_cpu variable to qom/cpu.h.

gdbstub needs to use CPUState::env_ptr for now.
cpu_copy() no longer needs to save and restore cpu_next.
Acked-by: NPaolo Bonzini <pbonzini@redhat.com>
[AF: Rebased, simplified cpu_copy()]
Signed-off-by: NAndreas Färber <afaerber@suse.de>

In particular fix 6-/10-char indentation.
Signed-off-by: NAndreas Färber <afaerber@suse.de>

Previous implementation has failed to take into account different value of
SOCK_NONBLOCK on target and host, and existence of SOCK_CLOEXEC.
The same conversion has to be applied both for do_socket and do_socketpair,
so the code has been isolated in a static inline function.

enum sock_type in linux-user/socket.h has been extended to include
TARGET_SOCK_CLOEXEC and TARGET_SOCK_NONBLOCK, similar to definition in libc.
The patch also includes necessary code style changes (tab to spaces) in the
header file since most of the file has been touched by this change.
Signed-off-by: NPetar Jovanovic <petar.jovanovic@imgtec.com>
Message-id: 1372639454-7560-1-git-send-email-petar.jovanovic@rt-rk.com
Reviewed-by: NPeter Maydell <peter.maydell@linaro.org>

Commit c0d472b1 accidentally dropped the definition of
__NR_SYS_utimensat even though its use is guarded by
CONFIG_UTIMENSAT, not CONFIG_ATFILE. Some older glibc don't
have utimensat() (even if they have the other *at() functions).
Fix this by correctly cleaning up the sys_utimensat()
implementation and #defines, so that we always provide the
syscall if needed whether we're doing it via glibc or not.
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>
Tested-by: NLaurent Desnogues <laurent.desnogues@gmail.com>
Message-id: 1371743841-26110-1-git-send-email-peter.maydell@linaro.org

This allows to pass the device name.

You can test this with the "route" command.

WITHOUT this patch:

$ sudo route add -net default gw 10.0.3.1 eth0
SIOCADDRT: Bad address
$ netstat -nr
Kernel IP routing table
Destination     Gateway         Genmask         Flags   MSS Window  irtt Ifa
10.0.3.0        0.0.0.0         255.255.255.0   U         0 0          0 eth

WITH this patch:

$ sudo route add -net default gw 10.0.3.1 eth0
$ netstat -nr
Kernel IP routing table
Destination     Gateway         Genmask         Flags   MSS Window  irtt Ifa
0.0.0.0         10.0.3.1        0.0.0.0         UG        0 0          0 eth
10.0.3.0        0.0.0.0         255.255.255.0   U         0 0          0 eth
Signed-off-by: NLaurent Vivier <laurent@vivier.eu>
Reviewed-by: NPeter Maydell <peter.maydell@linaro.org>

Some applications use /proc/$$/... (where $$ is the own pid) instead of
/proc/self/... to refer to their own proc files.  Extend the interception
for open and readlink to handle this case.  Also, do the same interception
in readlinkat.
Signed-off-by: NAndreas Schwab <schwab@suse.de>
Reviewed-by: NPeter Maydell <peter.maydell@linaro.org>

The linux-user syscall emulation layer currently supports the
openat family of syscalls via two mechanisms: simply calling
the corresponding libc functions, and making direct syscalls.
Since glibc has supported these functions since at least glibc
2.5, there's no real need to retain the (essentially untested)
direct syscall fallback code, so simply delete it. This allows
us to remove some ifdeffery that was attempting to disable
provision of some of the syscalls if the host didn't seem to
support them, which in some cases was actually wrong (eg where
there are several flavours of the syscall and we only need
one of them, not necessarily the exact one the guest has,
as with the fstatat* calls).
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>
Reviewed-by: NRichard Henderson <rth@twiddle.net>
Tested-by: NClaudio Fontana <claudio.fontana@huawei.com>
Message-id: 1370126121-22975-2-git-send-email-peter.maydell@linaro.org

Newer architectures may only implement the getdents64 syscall, not
getdents. Provide an implementation of getdents in terms of getdents64
so that we can run getdents-using targets on a getdents64-only host.
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>
Reviewed-by: NRichard Henderson <rth@twiddle.net>
Tested-by: NClaudio Fontana <claudio.fontana@huawei.com>
Message-id: 1370344377-27445-1-git-send-email-peter.maydell@linaro.org
Message-id: 1370193044-24535-1-git-send-email-peter.maydell@linaro.org

Signed-off-by: NLei Li <lilei@linux.vnet.ibm.com>
Reviewed-by: NStefan Weil <sw@weilnetz.de>
Signed-off-by: NMichael Tokarev <mjt@tls.msk.ru>

Add a space at end of line when there is no filename to print, to
conform to linux kernel format (see show_map_vma() in
fs/proc/task_mmu.c).
Signed-off-by: NChristophe Lyon <christophe.lyon@linaro.org>
Reviewed-by: NPeter Maydell <peter.maydell@linaro.org>
Signed-off-by: NMichael Tokarev <mjt@tls.msk.ru>

Remove a stray colon from the end of a #ifdef line. Some versions
of gcc complain about this:
 linux-user/syscall.c: In function ‘do_syscall’:
 linux-user/syscall.c:7606:28: error: extra tokens at end of #ifdef directive [-Werror]
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>
Acked-By: NRiku Voipio <riku.voipio@linaro.org>
Signed-off-by: NStefan Hajnoczi <stefanha@redhat.com>

do_semop() is called from two places, and one of these fails to convert
return error to target errno when semop fails. This patch changes the
function to always return target errno in case of an unsuccessful call.
Signed-off-by: NPetar Jovanovic <petar.jovanovic@imgtec.com>
Reviewed-by: NPeter Maydell <peter.maydell@linaro.org>
Signed-off-by: NStefan Hajnoczi <stefanha@redhat.com>

Don't assume target_id is a short.
Signed-off-by: NAndreas Schwab <schwab@suse.de>
Signed-off-by: NRiku Voipio <riku.voipio@linaro.org>

If TARGET_ABI_BITS is bigger than 32 we shift by more than the size of int.
Signed-off-by: NAndreas Schwab <schwab@suse.de>
Reviewed-by: NPeter Maydell <peter.maydell@linaro.org>
Signed-off-by: NRiku Voipio <riku.voipio@linaro.org>

do_semop() is called from two places, and one of these fails to convert
return error to target errno when semop fails. This patch changes the
function to always return target errno in case of an unsuccessful call.
Signed-off-by: NPetar Jovanovic <petar.jovanovic@imgtec.com>
Reviewed-by: NPeter Maydell <peter.maydell@linaro.org>
Signed-off-by: NRiku Voipio <riku.voipio@linaro.org>

This change makes conversion of TARGET_O_NONBLOCK and TARGET_O_CLOEXEC flags
to host flags before calling eventfd for TARGET_NR_eventfd2.
Signed-off-by: NPetar Jovanovic <petar.jovanovic@imgtec.com>
Reviewed-by: NPeter Maydell <peter.maydell@linaro.org>
Signed-off-by: NStefan Hajnoczi <stefanha@redhat.com>

The nature of the kernel ABI for the get_robust_list and set_robust_list
syscalls means we cannot implement them in QEMU. Make get_robust_list
silently return ENOSYS rather than using the default "print message and
then fail ENOSYS" code path, in the same way we already do for
set_robust_list, and add a comment documenting why we do this.

This silences warnings which were being produced for emulating
even trivial programs like 'ls' in x86-64-on-x86-64.
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>
Reviewed-by: NRichard Henderson <rth@twiddle.net>
Signed-off-by: NRiku Voipio <riku.voipio@linaro.org>

Implement the accept4 syscall (which is identical to accept
but has an additional flags argument).
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>
Reviewed-by: NRichard Henderson <rth@twiddle.net>
Signed-off-by: NRiku Voipio <riku.voipio@linaro.org>

Implement the sendfile and sendfile64 syscalls. This implementation
passes all the LTP test cases for these syscalls.
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>
Reviewed-by: NRichard Henderson <rth@twiddle.net>
Signed-off-by: NRiku Voipio <riku.voipio@linaro.org>

If the guest passes us a bogus negative length for an iovec, fail
EINVAL rather than proceeding blindly forward. This fixes some of
the error cases tests for readv and writev in the LTP.
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>
Reviewed-by: NRichard Henderson <rth@twiddle.net>
Signed-off-by: NRiku Voipio <riku.voipio@linaro.org>

Signed-off-by: NDillon Amburgey <dillona@dillona.com>
Reviewed-by: NPeter Maydell <peter.maydell@linaro.org>
Signed-off-by: NRiku Voipio <riku.voipio@linaro.org>

Upstream libc has recently changed to start using
FUTEX_WAIT_BITSET instead of FUTEX_WAIT and this
is causing do_futex to return -TARGET_ENOSYS.

Pass bitset in val3 to sys_futex which will be
ignored by kernel for the FUTEX_WAIT case.
Signed-off-by: NJohn Rigby <john.rigby@linaro.org>
Signed-off-by: NRiku Voipio <riku.voipio@linaro.org>

Change gdbstub's cpu_index() argument to CPUState now that CPUArchState
is no longer used.
Signed-off-by: NAndreas Färber <afaerber@suse.de>

CPUs are never added to the composition tree, so delete is achieved
simply by removing the last references to them.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NAnthony Liguori <aliguori@us.ibm.com>

According to man reboot(2), the 4th argument is only used with
LINUX_REBOOT_CMD_RESTART2. In other cases, trying to convert
the value can generate EFAULT.
Signed-off-by: NLaurent Vivier <laurent@vivier.eu>
Reviewed-by: NPeter Maydell <peter.maydell@linaro.org>

SO_SNDTIMEO and SO_RCVTIMEO take a struct timeval, not an int

To test this, you can use :

QEMU_STRACE= ping localhost 2>&1 |grep TIMEO
568 setsockopt(3,SOL_SOCKET,SO_SNDTIMEO,{1,0},8) = 0
568 setsockopt(3,SOL_SOCKET,SO_RCVTIMEO,{1,0},8) = 0
Signed-off-by: NLaurent Vivier <laurent@vivier.eu>
Reviewed-by: NPeter Maydell <peter.maydell@linaro.org>

All parameters must be swapped before the call of do_msgrcv().

Allow faked (debian fakeroot daemon) to work properly.

WITHOUT this patch:

$ faked-sysv --foreground --debug
using 1723744788 as msg key
msg_key=1723744788
1723744788:431
FAKEROOT: msg=131072, key=1723744788
FAKEROOT: r=-1, received message type=-150996052, message=-160219330
FAKEROOT, get_msg: Bad address
r=14, EINTR=4
fakeroot: clearing up message queues and semaphores, signal=-1
fakeroot: database save FAILED

WITH this patch:

$ faked-sysv --foreground --debug
using 1569385744 as msg key
msg_key=1569385744
1569385744:424
FAKEROOT: msg=0, key=1569385744
^C
fakeroot: clearing up message queues and semaphores, signal=2
fakeroot: database save FAILED
Signed-off-by: NLaurent Vivier <laurent@vivier.eu>
Reviewed-by: NPeter Maydell <peter.maydell@linaro.org>

Alpha, like s390x, passes all select arguments in registers.
Signed-off-by: NLaurent Vivier <laurent@vivier.eu>
Signed-off-by: NRichard Henderson <rth@twiddle.net>

The values of F_RDLCK, F_WRLCK, F_UNLCK, F_EXLCK, F_SHLCK
differ between alpha and other linux architectures.

This patch allows to run "dpkg" (database lock).
Signed-off-by: NLaurent Vivier <laurent@vivier.eu>
Signed-off-by: NRichard Henderson <rth@twiddle.net>

instead use the correct headers that define these functions.
Requested-by: NStefan Weil <sw@weilnetz.de>
Signed-off-by: NJohn Spencer <maillist-qemu@barfooze.de>
Reviewed-by: Amos Kong <kongjianjun@gmail.com>
Reviewed-by: NStefan Weil <sw@weilnetz.de>
Signed-off-by: NStefan Hajnoczi <stefanha@redhat.com>

Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

this declaration is wrong:
the correct prototype on linux is:
int setgroups(size_t size, const gid_t *list);

since by default musl libc exposes this symbol in unistd.h
additionally to grp.h, the wrong declaration causes a build error.

the proper fix is to simply include the correct header.
Signed-off-by: NJohn Spencer <maillist-qemu@barfooze.de>
Reviewed-by: NStefan Weil <sw@weilnetz.de>
Signed-off-by: NStefan Hajnoczi <stefanha@redhat.com>

The Linux syscalls underlying pread() and pwrite() take a 64 bit
offset on all architectures, even if some of them name the syscall
"pread/pwrite" rather than "pread64/pwrite64" for historical reasons.
So move the four QEMU target architectures (arm, i386, sparc,
unicore32) which were defining TARGET_NR_pread/pwrite to define
TARGET_NR_pread64/pwrite64 instead, and drop the TARGET_NR_pread/pwrite
implementation code completely.

(Based on examination of the kernel sources for the four architectures
this patch affects.)
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>
Signed-off-by: NBlue Swirl <blauwirbel@gmail.com>