This patch fixes the lockdep warning of "unannotated irqs-off"[1].
After entering __irq_usr, arm core will disable interrupt automatically,
but __irq_usr does not annotate the irq disable, so lockdep may complain
the warning if it has chance to check this in irq handler.
This patch adds trace_hardirqs_off in __irq_usr before entering irq_handler
to handle the irq, also calls ret_to_user_from_irq to avoid calling
disable_irq again.
This is also a fix for irq off tracer.
[1], lockdep warning log of "unannotated irqs-off"
[ 13.804687] ------------[ cut here ]------------
[ 13.809570] WARNING: at kernel/lockdep.c:3335 check_flags+0x78/0x1d0()
[ 13.816467] Modules linked in:
[ 13.819732] Backtrace:
[ 13.822357] [<c01cb42c>] (dump_backtrace+0x0/0x100) from [<c06abb14>] (dump_stack+0x20/0x24)
[ 13.831268] r6:c07d8c2c r5:00000d07 r4:00000000 r3:00000000
[ 13.837280] [<c06abaf4>] (dump_stack+0x0/0x24) from [<c01ffc04>] (warn_slowpath_common+0x5c/0x74)
[ 13.846649] [<c01ffba8>] (warn_slowpath_common+0x0/0x74) from [<c01ffc48>] (warn_slowpath_null+0x2c/0x34)
[ 13.856781] r8:00000000 r7:00000000 r6:c18b8194 r5:60000093 r4:ef182000
[ 13.863708] r3:00000009
[ 13.866485] [<c01ffc1c>] (warn_slowpath_null+0x0/0x34) from [<c0237d84>] (check_flags+0x78/0x1d0)
[ 13.875823] [<c0237d0c>] (check_flags+0x0/0x1d0) from [<c023afc8>] (lock_acquire+0x4c/0x150)
[ 13.884704] [<c023af7c>] (lock_acquire+0x0/0x150) from [<c06af638>] (_raw_spin_lock+0x4c/0x84)
[ 13.893798] [<c06af5ec>] (_raw_spin_lock+0x0/0x84) from [<c01f9a44>] (sched_ttwu_pending+0x58/0x8c)
[ 13.903320] r6:ef92d040 r5:00000003 r4:c18b8180
[ 13.908233] [<c01f99ec>] (sched_ttwu_pending+0x0/0x8c) from [<c01f9a90>] (scheduler_ipi+0x18/0x1c)
[ 13.917663] r6:ef183fb0 r5:00000003 r4:00000000 r3:00000001
[ 13.923645] [<c01f9a78>] (scheduler_ipi+0x0/0x1c) from [<c01bc458>] (do_IPI+0x9c/0xfc)
[ 13.932006] [<c01bc3bc>] (do_IPI+0x0/0xfc) from [<c06b0888>] (__irq_usr+0x48/0xe0)
[ 13.939971] Exception stack(0xef183fb0 to 0xef183ff8)
[ 13.945281] 3fa0: ffffffc3 0001500c 00000001 0001500c
[ 13.953948] 3fc0: 00000050 400b45f0 400d9000 00000000 00000001 400d9600 6474e552 bea05b3c
[ 13.962585] 3fe0: 400d96c0 bea059c0 400b6574 400b65d8 20000010 ffffffff
[ 13.969573] r6:00000403 r5:fa240100 r4:ffffffff r3:20000010
[ 13.975585] ---[ end trace efc4896ab0fb62cb ]---
[ 13.980468] possible reason: unannotated irqs-off.
[ 13.985534] irq event stamp: 1610
[ 13.989044] hardirqs last enabled at (1610): [<c01c703c>] no_work_pending+0x8/0x2c
[ 13.997131] hardirqs last disabled at (1609): [<c01c7024>] ret_slow_syscall+0xc/0x1c
[ 14.005371] softirqs last enabled at (0): [<c01fe5e4>] copy_process+0x2cc/0xa24
[ 14.013183] softirqs last disabled at (0): [< (null)>] (null)
Signed-off-by: Ming Lei <ming.lei@canonical.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
If the irqsoff tracer is in use, stop tracing the interrupt disable
interval when returning to userspace. Tracing userspace execution time
as interrupts disabled time is not helpful for kernel performance
analysis purposes. Only do so if the irqsoff tracer is enabled, to
avoid overhead for lockdep, which doesn't care.
Signed-off-by: Todd Poynor <toddpoynor@google.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: Tim Bird <tim.bird@am.sony.com>
[rabin@rab.in: rebase on top of latest code,
keep code in ftrace.c instead of separate file,
check for ftrace_graph_entry also]
Signed-off-by: Rabin Vincent <rabin@rab.in>
Use assembler macros to avoid copy/pasting code between the
implementations of the two variants of the mcount call.
Signed-off-by: Rabin Vincent <rabin@rab.in>
If a signal hits us outside of a syscall and another gets delivered
when we are in sigreturn (e.g. because it had been in sa_mask for
the first one and got sent to us while we'd been in the first handler),
we have a chance of returning from the second handler to location one
insn prior to where we ought to return. If r0 happens to contain -513
(-ERESTARTNOINTR), sigreturn will get confused into doing restart
syscall song and dance.
Incredible joy to debug, since it manifests as random, infrequent and
very hard to reproduce double execution of instructions in userland
code...
The fix is simple - mark it "don't bother with restarts" in wrapper,
i.e. set r8 to 0 in sys_sigreturn and sys_rt_sigreturn wrappers,
suppressing the syscall restart handling on return from these guys.
They can't legitimately return a restart-worthy error anyway.
Testcase:
#include <unistd.h>
#include <signal.h>
#include <stdlib.h>
#include <sys/time.h>
#include <errno.h>
void f(int n)
{
__asm__ __volatile__(
"ldr r0, [%0]\n"
"b 1f\n"
"b 2f\n"
"1:b .\n"
"2:\n" : : "r"(&n));
}
void handler1(int sig) { }
void handler2(int sig) { raise(1); }
void handler3(int sig) { exit(0); }
main()
{
struct sigaction s = {.sa_handler = handler2};
struct itimerval t1 = { .it_value = {1} };
struct itimerval t2 = { .it_value = {2} };
signal(1, handler1);
sigemptyset(&s.sa_mask);
sigaddset(&s.sa_mask, 1);
sigaction(SIGALRM, &s, NULL);
signal(SIGVTALRM, handler3);
setitimer(ITIMER_REAL, &t1, NULL);
setitimer(ITIMER_VIRTUAL, &t2, NULL);
f(-513); /* -ERESTARTNOINTR */
write(1, "buggered\n", 9);
return 1;
}
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Al Viro reports that calling "sys_sigsuspend(-ERESTARTNOHAND, 0, 0)"
with two signals coming and being handled in kernel space results
in the syscall restart being done twice.
Avoid this by clearing the 'why' flag when we call the signal handling
code to prevent further syscall restarts after the first.
Acked-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
This adds mcount recording and updates dynamic ftrace for ARM to work
with the new ftrace dyamic tracing implementation. It also adds support
for the mcount format used by newer ARM compilers.
With dynamic tracing, mcount() is implemented as a nop. Callsites are
patched on startup with nops, and dynamically patched to call to the
ftrace_caller() routine as needed.
Acked-by: Steven Rostedt <rostedt@goodmis.org> [recordmcount.pl change]
Signed-off-by: Rabin Vincent <rabin@rab.in>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Fix the mcount routines to build and run on a kernel built with the
Thumb-2 instruction set by correcting the following errors using the
fixes suggested by Catalin Marinas:
- Problem: The following assembler errors appear at the "adr r0,
ftrace_stub" instruction:
entry-common.S: Assembler messages:
entry-common.S:179: Error: invalid immediate for address calculation (value = 0x00000004)
Fix: The errors don't occur with a non-global symbol, so use one.
- Problem: The "mov lr, pc" does not set the lsb when storing the pc in
lr. The called function returns with "bx lr", and the mode changes
to ARM.
Fix: Add a label on the return address and use "adr lr, BSYM(label)".
We don't modify the old mcount because it won't be built when using
Thumb-2.
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Rabin Vincent <rabin@rab.in>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
When building as Thumb-2, the ".type foo, %function" annotation in
ENDPROC seems to be required in order for the assembly routines to be
recognized as Thumb-2 code. If the ENDPROC annotations are not present,
calls to these routines are generated as BLX instead of BL.
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Rabin Vincent <rabin@rab.in>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
With a new enough GCC, ARM function tracing can be supported without the
need for frame pointers. This is essential for Thumb-2 support, since
frame pointers aren't available then.
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Rabin Vincent <rabin@rab.in>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Add a comment describing the mcount variants and how the callsites look
like.
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Rabin Vincent <rabin@rab.in>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
The mcount implementation currently uses a different indentation style
from the rest of the file (and the rest of the ARM assembly in the
kernel). Clean it up.
Acked-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Rabin Vincent <rabin@rab.in>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
New helper - sys_mmap_pgoff(); switch syscalls to using it.
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Instruction fault status register, IFSR, was introduced on ARMv6 to
provide status information about the last insturction fault. It
needed for proper prefetch abort handling.
Now we have three prefetch abort model:
* legacy - for CPUs before ARMv6. They doesn't provide neither
IFSR nor IFAR. We simulate IFSR with section translation fault
status for them to generalize code;
* ARMv6 - provides IFSR, but not IFAR;
* ARMv7 - provides both IFSR and IFAR.
Signed-off-by: Kirill A. Shutemov <kirill@shutemov.name>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Commit 181f817eaa introduced some new code to entry-common.S
Sadly, this new code uses 'bx' instruction which is available only on
ARMv5 and higher CPUs. This causes following compilation errors when
building kernel for StrongARM (ARMv4):
arch/arm/kernel/entry-common.S: Assembler messages:
arch/arm/kernel/entry-common.S:129: Error: selected processor does not
support `bx ip'
arch/arm/kernel/entry-common.S:138: Error: selected processor does not
support `bx ip'
Fix these errors by using 'mov pc' instead of 'bx'.
Signed-off-by: Dmitry Artamonow <mad_soft@inbox.ru>
Acked-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Implement TIF_NOTIFY_RESUME for most of those architectures in which isn't yet
available, and, whilst we're at it, have it call the appropriate tracehook.
After this patch, blackfin, m68k* and xtensa still lack support and need
alteration of assembly code to make it work.
Resume notification can then be used (by a later patch) to install a new
session keyring on the parent of a process.
Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: Russell King <rmk+kernel@arm.linux.org.uk>
cc: linux-arch@vger.kernel.org
Signed-off-by: James Morris <jmorris@namei.org>
This patch adds support for TIF_RESTORE_SIGMASK to ARM's
signal handling, which allows to hook up the pselect6, ppoll,
and epoll_pwait syscalls on ARM.
Tested here with eabi userspace and a test program with a
deliberate race between a child's exit and the parent's
sigprocmask/select sequence. Using sys_pselect6() instead
of sigprocmask/select reliably prevents the race.
The other arch's support for TIF_RESTORE_SIGMASK has evolved
over time:
In 2.6.16:
- add TIF_RESTORE_SIGMASK which parallels TIF_SIGPENDING
- test both when checking for pending signal [changed later]
- reimplement sys_sigsuspend() to use current->saved_sigmask,
TIF_RESTORE_SIGMASK [changed later], and -ERESTARTNOHAND;
ditto for sys_rt_sigsuspend(), but drop private code and
use common code via __ARCH_WANT_SYS_RT_SIGSUSPEND;
- there are now no "extra" calls to do_signal() so its oldset
parameter is always ¤t->blocked so need not be passed,
also its return value is changed to void
- change handle_signal() to return 0/-errno
- change do_signal() to honor TIF_RESTORE_SIGMASK:
+ get oldset from current->saved_sigmask if TIF_RESTORE_SIGMASK
is set
+ if handle_signal() was successful then clear TIF_RESTORE_SIGMASK
+ if no signal was delivered and TIF_RESTORE_SIGMASK is set then
clear it and restore the sigmask
- hook up sys_pselect6() and sys_ppoll()
In 2.6.19:
- hook up sys_epoll_pwait()
In 2.6.26:
- allow archs to override how TIF_RESTORE_SIGMASK is implemented;
default set_restore_sigmask() sets both TIF_RESTORE_SIGMASK and
TIF_SIGPENDING; archs need now just test TIF_SIGPENDING again
when checking for pending signal work; some archs now implement
TIF_RESTORE_SIGMASK as a secondary/non-atomic thread flag bit
- call set_restore_sigmask() in sys_sigsuspend() instead of setting
TIF_RESTORE_SIGMASK
In 2.6.29-rc:
- kill sys_pselect7() which no arch wanted
So for 2.6.31-rc6/ARM this patch does the following:
- Add TIF_RESTORE_SIGMASK. Use the generic set_restore_sigmask()
which sets both TIF_SIGPENDING and TIF_RESTORE_SIGMASK, so
TIF_RESTORE_SIGMASK need not claim one of the scarce low thread
flags, and existing TIF_SIGPENDING and _TIF_WORK_MASK tests need
not be extended for TIF_RESTORE_SIGMASK.
- sys_sigsuspend() is reimplemented to use current->saved_sigmask
and set_restore_sigmask(), making it identical to most other archs
- The private code for sys_rt_sigsuspend() is removed, instead
generic code supplies it via __ARCH_WANT_SYS_RT_SIGSUSPEND.
- sys_sigsuspend() and sys_rt_sigsuspend() no longer need a pt_regs
parameter, so their assembly code wrappers are removed.
- handle_signal() is changed to return 0 on success or -errno.
- The oldset parameter to do_signal() is now redundant and removed,
and the return value is now also redundant and changed to void.
- do_signal() is changed to honor TIF_RESTORE_SIGMASK:
+ get oldset from current->saved_sigmask if TIF_RESTORE_SIGMASK
is set
+ if handle_signal() was successful then clear TIF_RESTORE_SIGMASK
+ if no signal was delivered and TIF_RESTORE_SIGMASK is set then
clear it and restore the sigmask
- Hook up sys_pselect6, sys_ppoll, and sys_epoll_pwait.
Signed-off-by: Mikael Pettersson <mikpe@it.uu.se>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Since gcc 4.4 the name and calling convention for function profiling
on ARM changed. With this patch both types are supported.
See http://sourceware.org/ml/libc-ports/2008-04/msg00009.html for some
details.
Lightly-Tested-by: Anand Gadiyar <gadiyar@ti.com>
Tested-by: Kevin Hilman <khilman@deeprootsystems.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
After ftrace_trace_function is called r1 is probably clobbered so don't
try to use its value for restoring.
This was introduced in v2.6.29~38^2~7
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Starting with ARMv6, the CPUs support the BE-8 variant of big-endian
(byte-invariant). This patch adds the core support:
- setting of the BE-8 mode via the CPSR.E register for both kernel and
user threads
- big-endian page table walking
- REV used to rotate instructions read from memory during fault
processing as they are still little-endian format
- Kconfig and Makefile support for BE-8. The --be8 option must be passed
to the final linking stage to convert the instructions to
little-endian
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
gcc seems to expect that lr isn't clobbered by mcount, because for a
function starting with:
static int func(void)
{
void *ra = __builtin_return_address(0);
printk(KERN_EMERG "__builtin_return_address(0) = %pS\n", ra)
...
the following assembler is generated by gcc 4.3.2:
0: e1a0c00d mov ip, sp
4: e92dd810 push {r4, fp, ip, lr, pc}
8: e24cb004 sub fp, ip, #4 ; 0x4
c: ebfffffe bl 0 <mcount>
10: e59f0034 ldr r0, [pc, #52]
14: e1a0100e mov r1, lr
18: ebfffffe bl 0 <printk>
Without this patch obviously __builtin_return_address(0) yields
func+0x10 instead of the return address of the caller.
Note this patch fixes a similar issue for the routines used with dynamic
ftrace even though this isn't currently selectable for ARM.
Cc: Abhishek Sagar <sagar.abhishek@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
This is needed to allow or stop the unwinding at certain points in the
kernel like exception entries.
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Due to confusion between the ftrace infrastructure and the gcc profiling
tracer "ftrace", this patch renames the config options from FTRACE to
FUNCTION_TRACER. The other two names that are offspring from FTRACE
DYNAMIC_FTRACE and FTRACE_MCOUNT_RECORD will stay the same.
This patch was generated mostly by script, and partially by hand.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This declaration specifies the "function" type and size for various
assembly functions, mainly needed for generating the correct branch
instructions in Thumb-2.
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Record the address of the mcount call-site. Currently all archs except sparc64
record the address of the instruction following the mcount call-site. Some
general cleanups are entailed. Storing mcount addresses in rec->ip enables
looking them up in the kprobe hash table later on to check if they're kprobe'd.
Signed-off-by: Abhishek Sagar <sagar.abhishek@gmail.com>
Cc: davem@davemloft.net
Cc: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Core ftrace support for the ARM architecture, which includes support
for dynamic function tracing.
Signed-off-by: Abhishek Sagar <sagar.abhishek@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This patch adds a prefetch abort handler similar to the data abort one
and renames the latter for consistency. Initial implementation by Paul
Brook with some renaming by Catalin Marinas.
Signed-off-by: Paul Brook <paul@codesourcery.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Remove unused TIF_NOTIFY_RESUME flag for all processor architectures. The
flag was not used excecpt on IA-64 where the patch replaces it with
TIF_PERFMON_WORK.
Signed-off-by: stephane eranian <eranian@hpl.hp.com>
Cc: <linux-arch@vger.kernel.org>
Cc: "Luck, Tony" <tony.luck@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
get_irqnr_preamble allows machines to take some action before entering the
get_irqnr_and_base loop. On iop we enable cp6 access.
arch_ret_to_user is added to the userspace return path to allow individual
architectures to take actions, like disabling coprocessor access, before
the final return to userspace.
Per Nicolas Pitre's note, there is no need to cp_wait on the return to user
as the latency to return is sufficient.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
RETINSTR is a left-over from the days when we had 26-bit and
32-bit CPU support integrated into the same tree. Since this
is no longer the case, we can now remove RETINSTR.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Paul Brook
The old-abi sys_syscall syscall is broken when called from Thumb mode. It
assumes the syscall number is an Arm syscall number (ie. starts from
__NR_OABI_SYSCALL_BASE). In thumb mode syscall numbers start from zero.
The patch below fixes this by clearing the nigh bits of the syscall number
instead of inverting them. Technically this means we accept some invalid
syscall numbers, but I can't see how that could be a problem. The two sets of
numbers far apart that unimplemented syscalls should still be rejected.
Signed-off-by: Paul Brook <paul@codesourcery.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
ARM entry-common.S needs to know syscall table size; in itself that would
not be a problem, but there's an additional constraint - some of the
instructions using it want a constant that would be a multiple of 4.
So we have to pad syscall table with sys_ni_syscall and that's where
the trouble begins. .rept pseudo-op wants a constant expression for
number of repetitions and subtraction of two labels (before and after
syscall table) doesn't always get simplified to constant early enough
for .rept. If labels end up in different frags, we lose. And while
the frag size is large enough (slightly below 4Kb), the syscall table
is about 1/3 of that. We used to get away with that, but the recent
changes had been enough to trigger the breakage.
Proper fix is simple: have a macro (CALL(x)) to populate the table
instead of using explicit .long x and the first time we include calls.S
have it defined to .equ NR_syscalls,NR_syscalls+1. Then we can find
the proper amount of padding on the first inclusion simply by looking
at NR_syscalls at that time. And that will be constant, no matter what.
Moreover, the same trick kills the need of having an estimate of padded
NR_syscalls - it will be calculated for free at the same time.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Nicolas Pitre
This is needed by strace to properly handle the tracing of some system
calls. It could be useful for other applications as well.
Based on an earlier patch from Daniel Jacobowitz.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Daniel Jacobowitz <dan@debian.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Nicolas Pitre
This patch adds the required code to support both user space ABIs at
the same time. A second syscall table is created to include legacy ABI
syscalls that need an ABI compat wrapper.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Nicolas Pitre
The difference between EABI and the legacy ABI may affect either
structure member alignment and/or argument register selection.
The patch has the details.
Included are wrappers for the following syscalls:
sys_stat64
sys_lstat64
sys_fstat64
sys_fcntl64
sys_epoll_ctl
sys_epoll_wait
sys_ipc
sys_semop
sys_semtimedop
sys_pread64
sys_pwrite64
sys_truncate64
sys_ftruncate64
sys_readahead
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Nicolas Pitre
struct statfs64 has extra padding with EABI growing its size from 84 to
88. This struct is now __attribute__((packed,aligned(4))) with a small
assembly wrapper to force the sz argument to 84 if it is 88 to avoid
copying the extra padding over user space memory unexpecting it.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Nicolas Pitre
For a while we wanted to change the way syscalls were called on ARM.
Instead of encoding the syscall number in the swi instruction which
requires reading back the instruction from memory to extract that number
and polluting the data cache, it was decided that simply storing the
syscall number into r7 would be more efficient. Since this represents
an ABI change then making that change at the same time as EABI support
is the right thing to do.
It is now expected that EABI user space binaries put the syscall number
into r7 and use "swi 0" to call the kernel. Syscall register argument
are also expected to have "EABI arrangement" i.e. 64-bit arguments
should be put in a pair of registers from an even register number.
Example with long ftruncate64(unsigned int fd, loff_t length):
legacy ABI:
- put fd into r0
- put length into r1-r2
- use "swi #(0x900000 + 194)" to call the kernel
new ARM EABI:
- put fd into r0
- put length into r2-r3 (skipping over r1)
- put 194 into r7
- use "swi 0" to call the kernel
Note that it is important to use 0 for the swi argument as backward
compatibility with legacy ABI user space relies on this.
The syscall macros in asm-arm/unistd.h were also updated to support
both ABIs and implement the right call method automatically.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Rather than providing more wrappers for 6-arg syscalls, arrange for
them to be supported as standard. This just means that we always
store the 6th argument on the stack, rather than in the wrappers.
This means we eliminate the wrappers for:
* sys_futex
* sys_arm_fadvise64_64
* sys_mbind
* sys_ipc
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Daniel Jacobowitz
After delivering a signal (creating its stack frame) we must check for
additional pending unblocked signals before returning to userspace.
Otherwise signals may be delayed past the next syscall or reschedule.
Once that was fixed it became obvious that the ARM signal mask manipulation
was broken. It was a little bit broken before the recent SA_NODEFER
changes, and then very broken after them. We must block the requested
signals before starting the handler or the same signal can be delivered
again before the handler even gets a chance to run.
Signed-off-by: Daniel Jacobowitz <dan@codesourcery.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Nicolas Pitre
Either no one is using an ARM710 with recent kernels, or all ARM710s
still in use are not afflicted by this swi bug. Nevertheless, the code
to work around the ARM710 swi bug is itself currently buggy since it
uses r8 as a pointer to S_PC while in fact it holds the spsr content
these days. Fix that, and simplify the code as well.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from George G. Davis
As pointed out be Matthew Klahn <MKLAHN@motorola.com>, some sys_ipc()
call options require six args, e.g. SEMTIMEDOP. This patch adds an ARM sys_ipc_wrapper to save the sys_ipc() 'fifth' arg on the stack.
Signed-off-by: George G. Davis <gdavis@mvista.com>
arch/arm/kernel/calls.S | 2 +-
arch/arm/kernel/entry-common.S | 5 +++++
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Nicolas Pitre
The prototype for sys_fadvise64_64() is:
long sys_fadvise64_64(int fd, loff_t offset, loff_t len, int advice)
The argument list is therefore as follows on legacy ABI:
fd: type int (r0)
offset: type long long (r1-r2)
len: type long long (r3-sp[0])
advice: type int (sp[4])
With EABI this becomes:
fd: type int (r0)
offset: type long long (r2-r3)
len: type long long (sp[0]-sp[4])
advice: type int (sp[8])
Not only do we have ABI differences here, but the EABI version requires
one additional word on the syscall stack.
To avoid the ABI mismatch and the extra stack space required with EABI
this syscall is now defined with a different argument ordering
on ARM as follows:
long sys_arm_fadvise64_64(int fd, int advice, loff_t offset, loff_t len)
This gives us the following ABI independent argument distribution:
fd: type int (r0)
advice: type int (r1)
offset: type long long (r2-r3)
len: type long long (sp[0]-sp[4])
Now, since the syscall entry code takes care of 5 registers only by
default including the store of r4 to the stack, we need a wrapper to
store r5 to the stack as well. Because that wrapper was missing and was
always required this means that sys_fadvise64_64 never worked on ARM and
therefore we can safely reuse its syscall number for our new
sys_arm_fadvise64_64 interface.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Ming Lei