Syscalls are not allowed inside restartable sequences, so add a call to
rseq_syscall() at the very beginning of the system call exit path when
CONFIG_DEBUG_RSEQ=y. This will help us to detect whether there is a
syscall issued erroneously inside a restartable sequence.
Signed-off-by: Paul Burton <paul.burton@mips.com>
Reviewed-by: James Hogan <jhogan@kernel.org>
Patchwork: https://patchwork.linux-mips.org/patch/19522/
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Boqun Feng <boqun.feng@gmail.com>
Cc: linux-mips@linux-mips.org
Cc: linux-kernel@vger.kernel.org
When the scheduler sets TIF_NEED_RESCHED & we call into the scheduler
from arch/mips/kernel/entry.S we disable interrupts. This is true
regardless of whether we reach work_resched from syscall_exit_work,
resume_userspace or by looping after calling schedule(). Although we
disable interrupts in these paths we don't call trace_hardirqs_off()
before calling into C code which may acquire locks, and we therefore
leave lockdep with an inconsistent view of whether interrupts are
disabled or not when CONFIG_PROVE_LOCKING & CONFIG_DEBUG_LOCKDEP are
both enabled.
Without tracing this interrupt state lockdep will print warnings such
as the following once a task returns from a syscall via
syscall_exit_partial with TIF_NEED_RESCHED set:
[ 49.927678] ------------[ cut here ]------------
[ 49.934445] WARNING: CPU: 0 PID: 1 at kernel/locking/lockdep.c:3687 check_flags.part.41+0x1dc/0x1e8
[ 49.946031] DEBUG_LOCKS_WARN_ON(current->hardirqs_enabled)
[ 49.946355] CPU: 0 PID: 1 Comm: init Not tainted 4.10.0-00439-gc9fd5d362289-dirty #197
[ 49.963505] Stack : 0000000000000000 ffffffff81bb5d6a 0000000000000006 ffffffff801ce9c4
[ 49.974431] 0000000000000000 0000000000000000 0000000000000000 000000000000004a
[ 49.985300] ffffffff80b7e487 ffffffff80a24498 a8000000ff160000 ffffffff80ede8b8
[ 49.996194] 0000000000000001 0000000000000000 0000000000000000 0000000077c8030c
[ 50.007063] 000000007fd8a510 ffffffff801cd45c 0000000000000000 a8000000ff127c88
[ 50.017945] 0000000000000000 ffffffff801cf928 0000000000000001 ffffffff80a24498
[ 50.028827] 0000000000000000 0000000000000001 0000000000000000 0000000000000000
[ 50.039688] 0000000000000000 a8000000ff127bd0 0000000000000000 ffffffff805509bc
[ 50.050575] 00000000140084e0 0000000000000000 0000000000000000 0000000000040a00
[ 50.061448] 0000000000000000 ffffffff8010e1b0 0000000000000000 ffffffff805509bc
[ 50.072327] ...
[ 50.076087] Call Trace:
[ 50.079869] [<ffffffff8010e1b0>] show_stack+0x80/0xa8
[ 50.086577] [<ffffffff805509bc>] dump_stack+0x10c/0x190
[ 50.093498] [<ffffffff8015dde0>] __warn+0xf0/0x108
[ 50.099889] [<ffffffff8015de34>] warn_slowpath_fmt+0x3c/0x48
[ 50.107241] [<ffffffff801c15b4>] check_flags.part.41+0x1dc/0x1e8
[ 50.114961] [<ffffffff801c239c>] lock_is_held_type+0x8c/0xb0
[ 50.122291] [<ffffffff809461b8>] __schedule+0x8c0/0x10f8
[ 50.129221] [<ffffffff80946a60>] schedule+0x30/0x98
[ 50.135659] [<ffffffff80106278>] work_resched+0x8/0x34
[ 50.142397] ---[ end trace 0cb4f6ef5b99fe21 ]---
[ 50.148405] possible reason: unannotated irqs-off.
[ 50.154600] irq event stamp: 400463
[ 50.159566] hardirqs last enabled at (400463): [<ffffffff8094edc8>] _raw_spin_unlock_irqrestore+0x40/0xa8
[ 50.171981] hardirqs last disabled at (400462): [<ffffffff8094eb98>] _raw_spin_lock_irqsave+0x30/0xb0
[ 50.183897] softirqs last enabled at (400450): [<ffffffff8016580c>] __do_softirq+0x4ac/0x6a8
[ 50.195015] softirqs last disabled at (400425): [<ffffffff80165e78>] irq_exit+0x110/0x128
Fix this by using the TRACE_IRQS_OFF macro to call trace_hardirqs_off()
when CONFIG_TRACE_IRQFLAGS is enabled. This is done before invoking
schedule() following the work_resched label because:
1) Interrupts are disabled regardless of the path we take to reach
work_resched() & schedule().
2) Performing the tracing here avoids the need to do it in paths which
disable interrupts but don't call out to C code before hitting a
path which uses the RESTORE_SOME macro that will call
trace_hardirqs_on() or trace_hardirqs_off() as appropriate.
We call trace_hardirqs_on() using the TRACE_IRQS_ON macro before calling
syscall_trace_leave() for similar reasons, ensuring that lockdep has a
consistent view of state after we re-enable interrupts.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Fixes: 1da177e4c3 ("Linux-2.6.12-rc2")
Cc: linux-mips@linux-mips.org
Cc: stable <stable@vger.kernel.org>
Patchwork: https://patchwork.linux-mips.org/patch/15385/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
The r2_emul_return field in struct thread_info was used in order to take
an alternate codepath when returning to userland, which (besides not
implementing certain features) effectively used the eretnc instruction
in place of eret. The difference is that eretnc doesn't clear LLBit, and
therefore doesn't cause a linked load & store sequence to fail due to
emulation like eret would.
The reason eret would usually be used to clear LLBit is so that after
context switching we ensure that a load performed by one task doesn't
influence another task. However commit 7c151d3d5d ("MIPS: Make use of
the ERETNC instruction on MIPS R6") which introduced the r2_emul_return
field and conditional use of eretnc also for some reason began
explicitly clearing LLBit during context switches - despite retaining
the use of eret for everything but returns from the pre-r6 instruction
emulation code.
As LLBit is cleared upon context switches anyway, simplify this by using
eretnc unconditionally for MIPSr6 kernels. This allows us to remove the
4 byte r2_emul_return boolean from struct thread_info, simplify the
return to user code in entry.S and avoid the overhead of tracking &
checking state which we don't need.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/14408/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Commit 6ebb496ffc7e("MIPS: kernel: entry.S: Add MIPS R6 related
definitions") added the MIPSR6 definition but it did not update the
ISA level of the actual assembly code so a pre-MIPSR6 jr.hb instruction
was generated instead. Fix this by using the MISP_ISA_LEVEL_RAW macro.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Fixes: 6ebb496ffc7e("MIPS: kernel: entry.S: Add MIPS R6 related definitions")
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9386/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
The ERETNC instruction, introduced in MIPS R5, is similar to the ERET
one, except it does not clear the LLB bit in the LLADDR register.
This feature is necessary to safely emulate R2 LL/SC instructions.
However, on context switches, we need to clear the LLAddr/LLB bit
in order to make sure that an SC instruction from the new thread
will never succeed if it happens to interrupt an LL operation on the
same address from the previous thread.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Nobody is maintaining SMTC anymore and there also seems to be no userbase.
Which is a pity - the SMTC technology primarily developed by Kevin D.
Kissell <kevink@paralogos.com> is an ingenious demonstration for the MT
ASE's power and elegance.
Based on Markos Chandras <Markos.Chandras@imgtec.com> patch
https://patchwork.linux-mips.org/patch/6719/ which while very similar did
no longer apply cleanly when I tried to merge it plus some additional
post-SMTC cleanup - SMTC was a feature as tricky to remove as it was to
merge once upon a time.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Pull big execve/kernel_thread/fork unification series from Al Viro:
"All architectures are converted to new model. Quite a bit of that
stuff is actually shared with architecture trees; in such cases it's
literally shared branch pulled by both, not a cherry-pick.
A lot of ugliness and black magic is gone (-3KLoC total in this one):
- kernel_thread()/kernel_execve()/sys_execve() redesign.
We don't do syscalls from kernel anymore for either kernel_thread()
or kernel_execve():
kernel_thread() is essentially clone(2) with callback run before we
return to userland, the callbacks either never return or do
successful do_execve() before returning.
kernel_execve() is a wrapper for do_execve() - it doesn't need to
do transition to user mode anymore.
As a result kernel_thread() and kernel_execve() are
arch-independent now - they live in kernel/fork.c and fs/exec.c
resp. sys_execve() is also in fs/exec.c and it's completely
architecture-independent.
- daemonize() is gone, along with its parts in fs/*.c
- struct pt_regs * is no longer passed to do_fork/copy_process/
copy_thread/do_execve/search_binary_handler/->load_binary/do_coredump.
- sys_fork()/sys_vfork()/sys_clone() unified; some architectures
still need wrappers (ones with callee-saved registers not saved in
pt_regs on syscall entry), but the main part of those suckers is in
kernel/fork.c now."
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/signal: (113 commits)
do_coredump(): get rid of pt_regs argument
print_fatal_signal(): get rid of pt_regs argument
ptrace_signal(): get rid of unused arguments
get rid of ptrace_signal_deliver() arguments
new helper: signal_pt_regs()
unify default ptrace_signal_deliver
flagday: kill pt_regs argument of do_fork()
death to idle_regs()
don't pass regs to copy_process()
flagday: don't pass regs to copy_thread()
bfin: switch to generic vfork, get rid of pointless wrappers
xtensa: switch to generic clone()
openrisc: switch to use of generic fork and clone
unicore32: switch to generic clone(2)
score: switch to generic fork/vfork/clone
c6x: sanitize copy_thread(), get rid of clone(2) wrapper, switch to generic clone()
take sys_fork/sys_vfork/sys_clone prototypes to linux/syscalls.h
mn10300: switch to generic fork/vfork/clone
h8300: switch to generic fork/vfork/clone
tile: switch to generic clone()
...
Conflicts:
arch/microblaze/include/asm/Kbuild
The problem occurs [1] when a kernel-mode task returns from a system
call with a pending signal.
A real-life scenario is a child of 'khelper' returning from a failed
kernel_execve() in ____call_usermodehelper() [ kernel/kmod.c ].
kernel_execve() fails due to a pending SIGKILL, which is the result of
"kill -9 -1" (at least, busybox's init does it upon reboot).
The loop is as follows:
* syscall_exit_work:
- work_pending: // start_of_the_loop
- work_notifysig:
- do_notify_resume()
- do_signal()
- if (!user_mode(regs)) return;
- resume_userspace // TIF_SIGPENDING is still set
- work_pending // so we call work_pending => goto
// start_of_the_loop
More information can be found in another LKML thread:
http://www.serverphorums.com/read.php?12,457826
[1] The problem was also reproduced on !CONFIG_VM86 x86, and the
following fix was accepted.
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=29a2e2836ff9ea65a603c89df217f4198973a74f
Signed-off-by: Dmitry Adamushko <dmitry.adamushko@gmail.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/3571/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
This version contains a few updates by David Daney, in particular it's
now using __builtin_frame_address() instead of asm() which depending
on personal taste, is slightly more appealing.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
No need to keep 4 copies of that stuff; merged and taken to
entry.S, unused public symbols there killed off.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
To support multiple TC microthreads acting as "CPUs" within a VPE,
VPE-wide interrupt mask bits must be specially manipulated during
interrupt handling. To support legacy drivers and interrupt controller
management code, SMTC has a "backstop" to track and if necessary restore
the interrupt mask. This has some performance impact on interrupt service
overhead. Disable it only if you know what you are doing.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Check the IEP bit for R3000 style processors when checking to see if
interrupts will be reenabled in restore_all.
Signed-off-by: Chris Dearman <chris@mips.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
When CONFIG_PREEMPT is not set, it also moves one branch instruction from
ret_from_irq() to ret_from_exception(). Therefore we favour the return
from irq case which should be more common than the other one.
Signed-off-by: Franck Bui-Huu <fbuihuu@gmail.com>
Acked-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Make sure that RA on top of interrupt stack is an address of ret_from_irq,
so that dump_stack etc. can trace info interrupted context.
Also this patch fixes except_vec_vi_handler and __smtc_ipi_vector which
seems broken.
Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
cleanup: remove task_t and convert all the uses to struct task_struct. I
introduced it for the scheduler anno and it was a mistake.
Conversion was mostly scripted, the result was reviewed and all
secondary whitespace and style impact (if any) was fixed up by hand.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.
Let it rip!
