linux-user patches for upstream
cover linux-user patches for upstream
From: Riku Voipio <[email protected]>
Also available from:
git pull git://gitorious.org/qemu-maemo/qemu-maemo.git linux-user-for-upstream
Mostly trivial new syscalls, but more urgently fix the 0 sized write which is a regression from Richard Henderson's multilevel page table patches.
Michael Casadevall (2):
linux-user: add pselect syscall
linux-user: Add the syscall id for pselect6 on ARM
Riku Voipio (2):
Add inotify_init1 syscall support
fix 0 sized write syscall
configure | 18 ++++++ exec.c | 5 +- linux-user/arm/syscall_nr.h | 2 +- linux-user/syscall.c | 133 +++++++++++++++++++++++++++++++++++++++++++ 4 files changed, 156 insertions(+), 2 deletions(-)
PATCH 1/4 linux-user: add pselect syscall
From: Michael Casadevall <[email protected]>
This patch adds support for the pselect syscall in linux-user emulation and also adds several support functions required to translate the timespec structs between the target and the host.
--- linux-user/syscall.c | 119 ++++++++++++++++++++++++++++++++++++++++++++++++++ 1 files changed, 119 insertions(+), 0 deletions(-)
linux-user/syscall.c +119 -0
@@ -850,6 +850,38 @@ static inline abi_long copy_to_user_timeval(abi_ulong target_tv_addr,return 0 ; } +static inline abi_long copy_from_user_timespec (struct timespec *ts, +abi_ulong target_ts_addr) +{ +struct target_timespec *target_ts; + +if (!lock_user_struct (VERIFY_READ , target_ts, target_ts_addr,1 )) +return -TARGET_EFAULT ; + +__get_user (ts->tv_sec, &target_ts->tv_sec); +__get_user (ts->tv_nsec, &target_ts->tv_nsec); + +unlock_user_struct (target_ts, target_ts_addr,0 ); + +return 0 ; +} + + +static inline abi_long copy_to_user_timespec (abi_ulong target_ts_addr, +const struct timespec *ts) +{ +struct target_timespec *target_ts; + +if (!lock_user_struct (VERIFY_WRITE , target_ts, target_ts_addr,0 )) +return -TARGET_EFAULT ; + +__put_user (ts->tv_sec, &target_ts->tv_sec); +__put_user (ts->tv_nsec, &target_ts->tv_nsec); + +unlock_user_struct (target_ts, target_ts_addr,1 ); + +return 0 ; +}#if defined(TARGET_NR_mq_open) && defined(__NR_mq_open)#include <mqueue.h>
@@ -949,6 +981,75 @@ static abi_long do_select(int n,return ret; } +#ifdef TARGET_NR_pselect6 +/* do_pselect() must return target values and target errnos. */ +static abi_long do_pselect (int n, +abi_ulong rfd_addr,abi_ulong wfd_addr, +abi_ulong efd_addr,abi_ulong target_tv_addr, +abi_ulong set_addr) +{ +fd_set rfds, wfds, efds; +fd_set *rfds_ptr, *wfds_ptr, *efds_ptr; +struct timespec tv, *tv_ptr; +sigset_t set, *set_ptr; +abi_long ret; + +if (rfd_addr) { +if (copy_from_user_fdset (&rfds, rfd_addr, n)) +return -TARGET_EFAULT ; + rfds_ptr = &rfds; + }else { + rfds_ptr =NULL ; + } +if (wfd_addr) { +if (copy_from_user_fdset (&wfds, wfd_addr, n)) +return -TARGET_EFAULT ; + wfds_ptr = &wfds; + }else { + wfds_ptr =NULL ; + } +if (efd_addr) { +if (copy_from_user_fdset (&efds, efd_addr, n)) +return -TARGET_EFAULT ; + efds_ptr = &efds; + }else { + efds_ptr =NULL ; + } + +if (target_tv_addr) { +if (copy_from_user_timespec (&tv, target_tv_addr)) +return -TARGET_EFAULT ; + tv_ptr = &tv; + }else { + tv_ptr =NULL ; + } + +/* We don't need to return sigmask to target */ +if (set_addr) { +target_to_host_old_sigset (&set, &set_addr); + set_ptr = &set; + }else { + set_ptr =NULL ; + } + + ret =get_errno (pselect (n, rfds_ptr, wfds_ptr, efds_ptr, tv_ptr, set_ptr)); + +if (!is_error (ret)) { +if (rfd_addr &©_to_user_fdset (rfd_addr, &rfds, n)) +return -TARGET_EFAULT ; +if (wfd_addr &©_to_user_fdset (wfd_addr, &wfds, n)) +return -TARGET_EFAULT ; +if (efd_addr &©_to_user_fdset (efd_addr, &efds, n)) +return -TARGET_EFAULT ; + +if (target_tv_addr &©_to_user_timespec (target_tv_addr, &tv)) +return -TARGET_EFAULT ; + } + +return ret; +} +#endif +static abi_long do_pipe2 (int host_pipe[],int flags) {#ifdef CONFIG_PIPE2
@@ -5186,6 +5287,24 @@ abi_long do_syscall(void *cpu_env, int num, abi_long arg1, }break ;#endif + +#ifdef TARGET_NR_pselect6 +case TARGET_NR_pselect6: + { +abi_ulong inp, outp, exp, tvp, set; +long nsel; + + nsel =tswapl (arg1); + inp =tswapl (arg2); + outp =tswapl (arg3); + exp =tswapl (arg4); + tvp =tswapl (arg5); + set =tswapl (arg6); + + ret =do_pselect (nsel, inp, outp, exp, tvp, set); + } +break ; +#endif case TARGET_NR_symlink: {void *p2;
Signature · 1 line
1.6.5
PATCH 2/4 linux-user: Add the syscall id for pselect6 on ARM
From: Michael Casadevall <[email protected]>
As this is now supported in newer linux kernels.
--- linux-user/arm/syscall_nr.h | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)
linux-user/arm/syscall_nr.h +1 -1
@@ -338,7 +338,7 @@#define TARGET_NR_readlinkat (332)#define TARGET_NR_fchmodat (333)#define TARGET_NR_faccessat (334) -/* 335 for pselect6 */ +#define TARGET_NR_pselect6 (335)/* 336 for ppoll */ #define TARGET_NR_unshare (337)#define TARGET_NR_set_robust_list (338)
Signature · 1 line
1.6.5
PATCH 3/4 Add inotify_init1 syscall support
From: Riku Voipio <[email protected]>
New syscall which gets actively used when you have a fresh kernel.
--- configure | 18 ++++++++++++++++++ linux-user/syscall.c | 14 ++++++++++++++ 2 files changed, 32 insertions(+), 0 deletions(-)
configure +18 -0
@@ -1629,6 +1629,21 @@ if compile_prog "" "" ; then inotify=yes fi +inotify1=no +cat > $TMPC << EOF +#include <sys/inotify.h> + +int +main(void) +{ + /* try to start inotify */ + return inotify_init1(0); +} +EOF +if compile_prog "" "" ; then + inotify1=yes +fi + # check if utimensat and futimens are supported utimens=no cat > $TMPC << EOF
@@ -2136,6 +2151,9 @@ fi if test "$inotify" = "yes" ; then echo "CONFIG_INOTIFY=y" >> $config_host_mak fi +if test "$inotify1" = "yes" ; then + echo "CONFIG_INOTIFY1=y" >> $config_host_mak +fi if test "$byteswap_h" = "yes" ; then echo "CONFIG_BYTESWAP_H=y" >> $config_host_mak fi
linux-user/syscall.c +14 -0
@@ -506,9 +506,18 @@ static int sys_inotify_rm_watch(int fd, int32_t wd)return (inotify_rm_watch (fd, wd)); }#endif +#ifdef CONFIG_INOTIFY1 +#if defined(TARGET_NR_inotify_init1) && defined(__NR_inotify_init1) +static int sys_inotify_init1 (int flags) +{ +return (inotify_init1 (flags)); +} +#endif +#endif #else /* Userspace can usually survive runtime without inotify */ #undef TARGET_NR_inotify_init +#undef TARGET_NR_inotify_init1#undef TARGET_NR_inotify_add_watch#undef TARGET_NR_inotify_rm_watch#endif /* CONFIG_INOTIFY */
@@ -7174,6 +7183,11 @@ abi_long do_syscall(void *cpu_env, int num, abi_long arg1, ret =get_errno (sys_inotify_init ());break ;#endif +#if defined(TARGET_NR_inotify_init1) && defined(__NR_inotify_init1) +case TARGET_NR_inotify_init1: + ret =get_errno (sys_inotify_init1 (arg1)); +break ; +#endif #if defined(TARGET_NR_inotify_add_watch) && defined(__NR_inotify_add_watch)case TARGET_NR_inotify_add_watch: p =lock_user_string (arg2);
Signature · 1 line
1.6.5
PATCH 4/4 fix 0 sized write syscall
From: Riku Voipio <[email protected]>
fixes running ldconfig under qemu linux-user
--- exec.c | 5 ++++- 1 files changed, 4 insertions(+), 1 deletions(-)
exec.c +4 -1
@@ -2409,8 +2409,11 @@ int page_check_range(target_ulong start, target_ulong len, int flags)#if TARGET_ABI_BITS >L1_MAP_ADDR_SPACE_BITS assert (start < ((abi_ulong )1 <<L1_MAP_ADDR_SPACE_BITS ));#endif +if (len ==0 ) { +return 0 ; + } -if (start + len -1 < start) { +if (start + len < start) {/* We've wrapped around. */ return -1 ; }
Signature · 1 line
1.6.5
Re: [Qemu-devel] [PATCH 4/4] fix 0 sized write syscall
exec.c int page_check_range(target_ulong start, target_ulong len, int flags) · line 2414
Quoted 3 lines · Riku Voipio
exec.c +3
-0+if (len ==0 ) { +return 0 ; + }
This part is ok.
exec.c int page_check_range(target_ulong start, target_ulong len, int flags) · line 2416
Quoted 3 lines
exec.c +1
-1-if (start + len -1 < start) { +if (start + len < start) {
This part re-introduces the bug I fixed.
start = 0xffffff00
len = 256
should succeed, and it won't reverting the (len-1) change.
r~
Aurelien Jarno
AppliedRe: [Qemu-devel] [PATCH 3/4] Add inotify_init1 syscall support
In reply to Riku VoipioQuoted 6 lines · Riku Voipio
From: Riku Voipio <[email protected]>
New syscall which gets actively used when you have a fresh kernel.
Signed-off-by Riku Voipio <[email protected]>
Thanks, applied.
Quoted 82 lines
--- configure | 18 ++++++++++++++++++ linux-user/syscall.c | 14 ++++++++++++++ 2 files changed, 32 insertions(+), 0 deletions(-)configure +18
-0@@ -1629,6 +1629,21 @@ if compile_prog "" "" ; then inotify=yes fi +inotify1=no +cat > $TMPC << EOF +#include <sys/inotify.h> + +int +main(void) +{ + /* try to start inotify */ + return inotify_init1(0); +} +EOF +if compile_prog "" "" ; then + inotify1=yes +fi + # check if utimensat and futimens are supported utimens=no cat > $TMPC << EOF@@ -2136,6 +2151,9 @@ fi if test "$inotify" = "yes" ; then echo "CONFIG_INOTIFY=y" >> $config_host_mak fi +if test "$inotify1" = "yes" ; then + echo "CONFIG_INOTIFY1=y" >> $config_host_mak +fi if test "$byteswap_h" = "yes" ; then echo "CONFIG_BYTESWAP_H=y" >> $config_host_mak filinux-user/syscall.c +14
-0@@ -506,9 +506,18 @@ static int sys_inotify_rm_watch(int fd, int32_t wd)return (inotify_rm_watch (fd, wd)); }#endif +#ifdef CONFIG_INOTIFY1 +#if defined(TARGET_NR_inotify_init1) && defined(__NR_inotify_init1) +static int sys_inotify_init1 (int flags) +{ +return (inotify_init1 (flags)); +} +#endif +#endif #else /* Userspace can usually survive runtime without inotify */ #undef TARGET_NR_inotify_init +#undef TARGET_NR_inotify_init1#undef TARGET_NR_inotify_add_watch#undef TARGET_NR_inotify_rm_watch#endif /* CONFIG_INOTIFY */ @@ -7174,6 +7183,11 @@ abi_long do_syscall(void *cpu_env, int num, abi_long arg1, ret =get_errno (sys_inotify_init ());break ;#endif +#if defined(TARGET_NR_inotify_init1) && defined(__NR_inotify_init1) +case TARGET_NR_inotify_init1: + ret =get_errno (sys_inotify_init1 (arg1)); +break ; +#endif #if defined(TARGET_NR_inotify_add_watch) && defined(__NR_inotify_add_watch)case TARGET_NR_inotify_add_watch: p =lock_user_string (arg2);Signature · 1 line
1.6.5
Signature · 2 lines
Aurelien Jarno GPG: 1024D/F1BCDB73 [email protected] http://www.aurel32.net
Aurelien Jarno
AppliedRe: [Qemu-devel] [PATCH 2/4] linux-user: Add the syscall id for pselect6 on ARM
In reply to Riku VoipioQuoted 9 lines · Riku Voipio
From: Michael Casadevall <[email protected]>
As this is now supported in newer linux kernels.
Signed-off-by Michael Casadevall <[email protected]>Signed-off-by Riku Voipio <[email protected]>--- linux-user/arm/syscall_nr.h | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)
Thanks, applied.
Quoted 19 lines
linux-user/arm/syscall_nr.h +1
-1@@ -338,7 +338,7 @@#define TARGET_NR_readlinkat (332)#define TARGET_NR_fchmodat (333)#define TARGET_NR_faccessat (334) -/* 335 for pselect6 */ +#define TARGET_NR_pselect6 (335)/* 336 for ppoll */ #define TARGET_NR_unshare (337)#define TARGET_NR_set_robust_list (338)Signature · 1 line
1.6.5
Signature · 2 lines
Aurelien Jarno GPG: 1024D/F1BCDB73 [email protected] http://www.aurel32.net
Re: [Qemu-devel] [PATCH 1/4] linux-user: add pselect syscall
In reply to Riku VoipioQuoted 11 lines · Riku Voipio
From: Michael Casadevall <[email protected]>
This patch adds support for the pselect syscall in linux-user emulation and also adds several support functions required to translate the timespec structs between the target and the host.
Signed-off-by Riku Voipio <[email protected]>Signed-off-by Michael Casadevall <[email protected]>--- linux-user/syscall.c | 119 ++++++++++++++++++++++++++++++++++++++++++++++++++ 1 files changed, 119 insertions(+), 0 deletions(-)
This patch basically needs ok, but need to conform to CODING_STYLE (see below).
Quoted 5 lines
linux-user/syscall.c +0
-0
linux-user/syscall.c static inline abi_long copy_to_user_timeval(abi_ulong target_tv_addr, · line 859
Quoted 10 lines
linux-user/syscall.c +7
-0@@ -850,6 +850,38 @@ static inline abi_long copy_to_user_timeval(abi_ulong target_tv_addr,return 0 ; } +static inline abi_long copy_from_user_timespec (struct timespec *ts, +abi_ulong target_ts_addr) +{ +struct target_timespec *target_ts; + +if (!lock_user_struct (VERIFY_READ , target_ts, target_ts_addr,1 )) +return -TARGET_EFAULT ;
You should use curly braces around the return;
Quoted 38 lines
linux-user/syscall.c +31
-0+ +__get_user (ts->tv_sec, &target_ts->tv_sec); +__get_user (ts->tv_nsec, &target_ts->tv_nsec); + +unlock_user_struct (target_ts, target_ts_addr,0 ); + +return 0 ; +} + + +static inline abi_long copy_to_user_timespec (abi_ulong target_ts_addr, +const struct timespec *ts) +{ +struct target_timespec *target_ts; + +if (!lock_user_struct (VERIFY_WRITE , target_ts, target_ts_addr,0 )) +return -TARGET_EFAULT ; + +__put_user (ts->tv_sec, &target_ts->tv_sec); +__put_user (ts->tv_nsec, &target_ts->tv_nsec); + +unlock_user_struct (target_ts, target_ts_addr,1 ); + +return 0 ; +}#if defined(TARGET_NR_mq_open) && defined(__NR_mq_open)#include <mqueue.h> @@ -949,6 +981,75 @@ static abi_long do_select(int n,return ret; } +#ifdef TARGET_NR_pselect6 +/* do_pselect() must return target values and target errnos. */ +static abi_long do_pselect (int n, +abi_ulong rfd_addr,abi_ulong wfd_addr, +abi_ulong efd_addr,abi_ulong target_tv_addr, +abi_ulong set_addr)
linux-user/syscall.c static abi_long do_select(int n, · line 999
Quoted 10 lines
linux-user/syscall.c +10
-0+{ +fd_set rfds, wfds, efds; +fd_set *rfds_ptr, *wfds_ptr, *efds_ptr; +struct timespec tv, *tv_ptr; +sigset_t set, *set_ptr; +abi_long ret; + +if (rfd_addr) { +if (copy_from_user_fdset (&rfds, rfd_addr, n)) +return -TARGET_EFAULT ;
Same here, and later on.
Quoted 87 lines
linux-user/syscall.c +71
-0+ rfds_ptr = &rfds; + }else { + rfds_ptr =NULL ; + } +if (wfd_addr) { +if (copy_from_user_fdset (&wfds, wfd_addr, n)) +return -TARGET_EFAULT ; + wfds_ptr = &wfds; + }else { + wfds_ptr =NULL ; + } +if (efd_addr) { +if (copy_from_user_fdset (&efds, efd_addr, n)) +return -TARGET_EFAULT ; + efds_ptr = &efds; + }else { + efds_ptr =NULL ; + } + +if (target_tv_addr) { +if (copy_from_user_timespec (&tv, target_tv_addr)) +return -TARGET_EFAULT ; + tv_ptr = &tv; + }else { + tv_ptr =NULL ; + } + +/* We don't need to return sigmask to target */ +if (set_addr) { +target_to_host_old_sigset (&set, &set_addr); + set_ptr = &set; + }else { + set_ptr =NULL ; + } + + ret =get_errno (pselect (n, rfds_ptr, wfds_ptr, efds_ptr, tv_ptr, set_ptr)); + +if (!is_error (ret)) { +if (rfd_addr &©_to_user_fdset (rfd_addr, &rfds, n)) +return -TARGET_EFAULT ; +if (wfd_addr &©_to_user_fdset (wfd_addr, &wfds, n)) +return -TARGET_EFAULT ; +if (efd_addr &©_to_user_fdset (efd_addr, &efds, n)) +return -TARGET_EFAULT ; + +if (target_tv_addr &©_to_user_timespec (target_tv_addr, &tv)) +return -TARGET_EFAULT ; + } + +return ret; +} +#endif +static abi_long do_pipe2 (int host_pipe[],int flags) {#ifdef CONFIG_PIPE2 @@ -5186,6 +5287,24 @@ abi_long do_syscall(void *cpu_env, int num, abi_long arg1, }break ;#endif + +#ifdef TARGET_NR_pselect6 +case TARGET_NR_pselect6: + { +abi_ulong inp, outp, exp, tvp, set; +long nsel; + + nsel =tswapl (arg1); + inp =tswapl (arg2); + outp =tswapl (arg3); + exp =tswapl (arg4); + tvp =tswapl (arg5); + set =tswapl (arg6); + + ret =do_pselect (nsel, inp, outp, exp, tvp, set); + } +break ; +#endif case TARGET_NR_symlink: {void *p2;Signature · 1 line
1.6.5
Signature · 2 lines
Aurelien Jarno GPG: 1024D/F1BCDB73 [email protected] http://www.aurel32.net
Re: [Qemu-devel] [PATCH 1/4] linux-user: add pselect syscall
In reply to Riku VoipioQuoted 3 lines
This patch adds support for the pselect syscall in linux-user emulation and also adds several support functions required to translate the timespec structs between the target and the host.
IIUC the whole point of the pselect is that it should be atomic. By emulating this in a non-atomic fasion I think you're re-introducing the race condition that it is designed to avoid.
Wouldn't it be better to just return ENOSYS and let the guest deal with the problem?
Paul
Re: [Qemu-devel] [PATCH 1/4] linux-user: add pselect syscall
Quoted 10 lines · Paul Brook
Quoted 3 lines
This patch adds support for the pselect syscall in linux-user emulation and also adds several support functions required to translate the timespec structs between the target and the host.
IIUC the whole point of the pselect is that it should be atomic. By emulating this in a non-atomic fasion I think you're re-introducing the race condition that it is designed to avoid.
Wouldn't it be better to just return ENOSYS and let the guest deal with the problem?
Imho, it's very important to return ENOSYS if the atomic behaviour is not provided.
The patch actually calls the host's pselect() - it looks almost ok...
But actually, host's pselect() may be Glibc or some other equally buggy libc, which "emulates" pselect wrongly. FreeBSD's pselect has the same problem going back years - with a very recent patch to support it in the kernel. It's not just Glibc.
Careful apps detect and avoid libc's bug by calling the pselect syscall directly. If it returns ENOSYS, they just use a different (but slightly slower) race-free technique, or abort - at least they don't have a quiet corruption bug.
But here we're providing the kernel syscall to guest apps. There's no way for them to detect if it's really atomic or not. And the ones written carefully to avoid buggy libc will wrongly use it, because they think it's a kernel call.
I think it's really important we don't pass on libc's buggy behaviour to the emulated kernel interface, to even break careful apps.
Solution, Riku: Instead of calling pselect(...), qemu should call syscall(SYS_pselect, ...), with a comment like:
/*
* Glibc and some other libcs unhelpfully "emulate" pselect()
* using select() when the real system call isn't available, and
* break the atomic guarantee which is the entire point of
* pselect(). To avoid these libc bugs, go straight to the host
* kernel using syscall(). If the host kernel doesn't support
* pselect() it will return ENOSYS, which is what we want to
* return to the guest.
*/
There might need to be some header file included to get syscall() and SYS_pselect, which sometimes has other names like NR_pselect. And unfortunately the call arguments aren't quite the same, at least on Linux.
Same goes for ppoll() if that is ever added.
-- Jamie