The patch titled
lock validator: reenable NMIs
has been removed from the -mm tree. Its filename is
lock-validator-v3-lock-validator-reenable-nmis.patch
This patch was dropped because lockdep is being redone
------------------------------------------------------
Subject: lock validator: reenable NMIs
From: Ingo Molnar <mingo@xxxxxxx>
- not touch the irqtrace state in NMI handlers
- re-enable NMIs and oprofile on i386 and x86_64
I have tested this with high-rate (10 KHz per CPU) NMI load on the kernel,
and while previously we'd quickly get the irqtrace state confused and a
lock validator bug message, with this patch applied it's now working fine.
(Note that on x86_64 we only build one variant of the paranoid-exit
codepath if IRQTRACE is disabled - so there should be no difference in code
generated, as long as the validator is disabled)
Signed-off-by: Ingo Molnar <mingo@xxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxx>
---
arch/i386/kernel/nmi.c | 10 --
arch/i386/oprofile/Kconfig | 2
arch/x86_64/kernel/entry.S | 133 +++++++++++++++++++--------------
arch/x86_64/kernel/nmi.c | 11 --
arch/x86_64/oprofile/Kconfig | 2
include/linux/hardirq.h | 4
kernel/lockdep.c | 16 +--
7 files changed, 89 insertions(+), 89 deletions(-)
diff -puN arch/i386/kernel/nmi.c~lock-validator-v3-lock-validator-reenable-nmis arch/i386/kernel/nmi.c
--- a/arch/i386/kernel/nmi.c~lock-validator-v3-lock-validator-reenable-nmis
+++ a/arch/i386/kernel/nmi.c
@@ -747,16 +747,6 @@ void setup_apic_nmi_watchdog (void *unus
{
struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk);
-#ifdef CONFIG_LOCKDEP
- /*
- * The NMI watchdog uses spinlocks (notifier chains, etc.),
- * so it's not lockdep-safe:
- */
- nmi_watchdog = NMI_NONE;
- printk("lockdep: disabled NMI watchdog.\n");
-
- return;
-#endif
/* only support LOCAL and IO APICs for now */
if ((nmi_watchdog != NMI_LOCAL_APIC) &&
(nmi_watchdog != NMI_IO_APIC))
diff -puN arch/i386/oprofile/Kconfig~lock-validator-v3-lock-validator-reenable-nmis arch/i386/oprofile/Kconfig
--- a/arch/i386/oprofile/Kconfig~lock-validator-v3-lock-validator-reenable-nmis
+++ a/arch/i386/oprofile/Kconfig
@@ -7,7 +7,7 @@ config PROFILING
config OPROFILE
tristate "OProfile system profiling (EXPERIMENTAL)"
- depends on PROFILING && !LOCKDEP
+ depends on PROFILING
help
OProfile is a profiling system capable of profiling the
whole system, include the kernel, kernel modules, libraries,
diff -puN arch/x86_64/kernel/entry.S~lock-validator-v3-lock-validator-reenable-nmis arch/x86_64/kernel/entry.S
--- a/arch/x86_64/kernel/entry.S~lock-validator-v3-lock-validator-reenable-nmis
+++ a/arch/x86_64/kernel/entry.S
@@ -718,7 +718,7 @@ END(spurious_interrupt)
/* error code is on the stack already */
/* handle NMI like exceptions that can happen everywhere */
- .macro paranoidentry sym, ist=0
+ .macro paranoidentry sym, ist=0, irqtrace=1
SAVE_ALL
cld
movl $1,%ebx
@@ -743,9 +743,73 @@ END(spurious_interrupt)
addq $EXCEPTION_STKSZ, per_cpu__init_tss + TSS_ist + (\ist - 1) * 8(%rbp)
.endif
cli
+ .if \irqtrace
TRACE_IRQS_OFF
+ .endif
.endm
-
+
+ /*
+ * "Paranoid" exit path from exception stack.
+ * Paranoid because this is used by NMIs and cannot take
+ * any kernel state for granted.
+ * We don't do kernel preemption checks here, because only
+ * NMI should be common and it does not enable IRQs and
+ * cannot get reschedule ticks.
+ *
+ * "trace" is 0 for the NMI handler only, because irq-tracing
+ * is fundamentally NMI-unsafe. (we cannot change the soft and
+ * hard flags at once, atomically)
+ */
+ .macro paranoidexit trace=1
+ /* ebx: no swapgs flag */
+paranoid_exit\trace:
+ testl %ebx,%ebx /* swapgs needed? */
+ jnz paranoid_restore\trace
+ testl $3,CS(%rsp)
+ jnz paranoid_userspace\trace
+paranoid_swapgs\trace:
+ TRACE_IRQS_IRETQ 0
+ swapgs
+paranoid_restore\trace:
+ RESTORE_ALL 8
+ iretq
+paranoid_userspace\trace:
+ GET_THREAD_INFO(%rcx)
+ movl threadinfo_flags(%rcx),%ebx
+ andl $_TIF_WORK_MASK,%ebx
+ jz paranoid_swapgs\trace
+ movq %rsp,%rdi /* &pt_regs */
+ call sync_regs
+ movq %rax,%rsp /* switch stack for scheduling */
+ testl $_TIF_NEED_RESCHED,%ebx
+ jnz paranoid_schedule\trace
+ movl %ebx,%edx /* arg3: thread flags */
+ .if \trace
+ TRACE_IRQS_ON
+ .endif
+ sti
+ xorl %esi,%esi /* arg2: oldset */
+ movq %rsp,%rdi /* arg1: &pt_regs */
+ call do_notify_resume
+ cli
+ .if \trace
+ TRACE_IRQS_OFF
+ .endif
+ jmp paranoid_userspace\trace
+paranoid_schedule\trace:
+ .if \trace
+ TRACE_IRQS_ON
+ .endif
+ sti
+ call schedule
+ cli
+ .if \trace
+ TRACE_IRQS_OFF
+ .endif
+ jmp paranoid_userspace\trace
+ CFI_ENDPROC
+ .endm
+
/*
* Exception entry point. This expects an error code/orig_rax on the stack
* and the exception handler in %rax.
@@ -974,8 +1038,7 @@ KPROBE_ENTRY(debug)
pushq $0
CFI_ADJUST_CFA_OFFSET 8
paranoidentry do_debug, DEBUG_STACK
- jmp paranoid_exit
- CFI_ENDPROC
+ paranoidexit
END(debug)
.previous .text
@@ -984,54 +1047,12 @@ KPROBE_ENTRY(nmi)
INTR_FRAME
pushq $-1
CFI_ADJUST_CFA_OFFSET 8
- paranoidentry do_nmi
- /*
- * "Paranoid" exit path from exception stack.
- * Paranoid because this is used by NMIs and cannot take
- * any kernel state for granted.
- * We don't do kernel preemption checks here, because only
- * NMI should be common and it does not enable IRQs and
- * cannot get reschedule ticks.
- */
- /* ebx: no swapgs flag */
-paranoid_exit:
- testl %ebx,%ebx /* swapgs needed? */
- jnz paranoid_restore
- testl $3,CS(%rsp)
- jnz paranoid_userspace
-paranoid_swapgs:
- TRACE_IRQS_IRETQ 0
- swapgs
-paranoid_restore:
- RESTORE_ALL 8
- iretq
-paranoid_userspace:
- GET_THREAD_INFO(%rcx)
- movl threadinfo_flags(%rcx),%ebx
- andl $_TIF_WORK_MASK,%ebx
- jz paranoid_swapgs
- movq %rsp,%rdi /* &pt_regs */
- call sync_regs
- movq %rax,%rsp /* switch stack for scheduling */
- testl $_TIF_NEED_RESCHED,%ebx
- jnz paranoid_schedule
- movl %ebx,%edx /* arg3: thread flags */
- TRACE_IRQS_ON
- sti
- xorl %esi,%esi /* arg2: oldset */
- movq %rsp,%rdi /* arg1: &pt_regs */
- call do_notify_resume
- cli
- TRACE_IRQS_OFF
- jmp paranoid_userspace
-paranoid_schedule:
- TRACE_IRQS_ON
- sti
- call schedule
- cli
- TRACE_IRQS_OFF
- jmp paranoid_userspace
- CFI_ENDPROC
+ paranoidentry do_nmi, 0, 0
+#ifdef CONFIG_TRACE_IRQFLAGS
+ paranoidexit 0
+#else
+ paranoidexit 1
+#endif
END(nmi)
.previous .text
@@ -1040,7 +1061,7 @@ KPROBE_ENTRY(int3)
pushq $0
CFI_ADJUST_CFA_OFFSET 8
paranoidentry do_int3, DEBUG_STACK
- jmp paranoid_exit
+ jmp paranoid_exit1
CFI_ENDPROC
END(int3)
.previous .text
@@ -1069,7 +1090,7 @@ END(reserved)
ENTRY(double_fault)
XCPT_FRAME
paranoidentry do_double_fault
- jmp paranoid_exit
+ jmp paranoid_exit1
CFI_ENDPROC
END(double_fault)
@@ -1085,7 +1106,7 @@ END(segment_not_present)
ENTRY(stack_segment)
XCPT_FRAME
paranoidentry do_stack_segment
- jmp paranoid_exit
+ jmp paranoid_exit1
CFI_ENDPROC
END(stack_segment)
@@ -1113,7 +1134,7 @@ ENTRY(machine_check)
pushq $0
CFI_ADJUST_CFA_OFFSET 8
paranoidentry do_machine_check
- jmp paranoid_exit
+ jmp paranoid_exit1
CFI_ENDPROC
END(machine_check)
#endif
diff -puN arch/x86_64/kernel/nmi.c~lock-validator-v3-lock-validator-reenable-nmis arch/x86_64/kernel/nmi.c
--- a/arch/x86_64/kernel/nmi.c~lock-validator-v3-lock-validator-reenable-nmis
+++ a/arch/x86_64/kernel/nmi.c
@@ -674,17 +674,6 @@ void setup_apic_nmi_watchdog(void *unuse
{
struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk);
-#ifdef CONFIG_LOCKDEP
- /*
- * The NMI watchdog uses spinlocks (notifier chains, etc.),
- * so it's not lockdep-safe:
- */
- nmi_watchdog = NMI_NONE;
- printk("lockdep: disabled NMI watchdog.\n");
-
- return;
-#endif
-
/* only support LOCAL and IO APICs for now */
if ((nmi_watchdog != NMI_LOCAL_APIC) &&
(nmi_watchdog != NMI_IO_APIC))
diff -puN arch/x86_64/oprofile/Kconfig~lock-validator-v3-lock-validator-reenable-nmis arch/x86_64/oprofile/Kconfig
--- a/arch/x86_64/oprofile/Kconfig~lock-validator-v3-lock-validator-reenable-nmis
+++ a/arch/x86_64/oprofile/Kconfig
@@ -7,7 +7,7 @@ config PROFILING
config OPROFILE
tristate "OProfile system profiling (EXPERIMENTAL)"
- depends on PROFILING && !LOCKDEP
+ depends on PROFILING
help
OProfile is a profiling system capable of profiling the
whole system, include the kernel, kernel modules, libraries,
diff -puN include/linux/hardirq.h~lock-validator-v3-lock-validator-reenable-nmis include/linux/hardirq.h
--- a/include/linux/hardirq.h~lock-validator-v3-lock-validator-reenable-nmis
+++ a/include/linux/hardirq.h
@@ -127,7 +127,7 @@ static inline void account_system_vtime(
*/
extern void irq_exit(void);
-#define nmi_enter() irq_enter()
-#define nmi_exit() __irq_exit()
+#define nmi_enter() do { lockdep_off(); irq_enter(); } while (0)
+#define nmi_exit() do { __irq_exit(); lockdep_on(); } while (0)
#endif /* LINUX_HARDIRQ_H */
diff -puN kernel/lockdep.c~lock-validator-v3-lock-validator-reenable-nmis kernel/lockdep.c
--- a/kernel/lockdep.c~lock-validator-v3-lock-validator-reenable-nmis
+++ a/kernel/lockdep.c
@@ -1869,7 +1869,7 @@ void trace_hardirqs_on(void)
struct task_struct *curr = current;
unsigned long ip;
- if (unlikely(!debug_locks))
+ if (unlikely(!debug_locks || current->lockdep_recursion))
return;
if (DEBUG_WARN_ON(unlikely(!early_boot_irqs_enabled)))
@@ -1916,7 +1916,7 @@ void trace_hardirqs_off(void)
{
struct task_struct *curr = current;
- if (unlikely(!debug_locks))
+ if (unlikely(!debug_locks || current->lockdep_recursion))
return;
if (DEBUG_WARN_ON(!irqs_disabled()))
@@ -2488,16 +2488,16 @@ void lock_acquire(struct lockdep_map *lo
{
unsigned long flags;
+ if (unlikely(current->lockdep_recursion))
+ return;
+
raw_local_irq_save(flags);
check_flags(flags);
- if (unlikely(current->lockdep_recursion))
- goto out;
current->lockdep_recursion = 1;
__lock_acquire(lock, subtype, trylock, read, check,
irqs_disabled_flags(flags), ip);
current->lockdep_recursion = 0;
-out:
raw_local_irq_restore(flags);
}
@@ -2507,14 +2507,14 @@ void lock_release(struct lockdep_map *lo
{
unsigned long flags;
+ if (unlikely(current->lockdep_recursion))
+ return;
+
raw_local_irq_save(flags);
check_flags(flags);
- if (unlikely(current->lockdep_recursion))
- goto out;
current->lockdep_recursion = 1;
__lock_release(lock, nested, ip);
current->lockdep_recursion = 0;
-out:
raw_local_irq_restore(flags);
}
_
Patches currently in -mm which might be from mingo@xxxxxxx are
origin.patch
disable-debugging-version-of-write_lock.patch
fix-drivers-mfd-ucb1x00-corec-irq-probing-bug.patch
cifs-remove-f_ownerlock-use.patch
lock-validator-fix-ns83820c-irq-flags-bug.patch
revert-gregkh-pci-pci-test-that-drivers-properly-call-pci_set_master.patch
x86-re-enable-generic-numa.patch
vdso-randomize-the-i386-vdso-by-moving-it-into-a-vma.patch
vdso-randomize-the-i386-vdso-by-moving-it-into-a-vma-tidy.patch
vdso-randomize-the-i386-vdso-by-moving-it-into-a-vma-arch_vma_name-fix.patch
vdso-randomize-the-i386-vdso-by-moving-it-into-a-vma-vs-x86_64-mm-reliable-stack-trace-support-i386.patch
vdso-randomize-the-i386-vdso-by-moving-it-into-a-vma-vs-x86_64-mm-reliable-stack-trace-support-i386-2.patch
vdso-randomize-the-i386-vdso-by-moving-it-into-a-vma-vs-x86_64-mm-reliable-stack-trace-support-i386-2-revert-maxmem-change.patch
work-around-ppc64-bootup-bug-by-making-mutex-debugging-save-restore-irqs.patch
kernel-kernel-cpuc-to-mutexes.patch
define-__raw_get_cpu_var-and-use-it.patch
ide-cd-end-of-media-error-fix.patch
spin-rwlock-init-cleanups.patch
lock-validator-introduce-warn_on_oncecond.patch
lock-validator-introduce-warn_on_oncecond-speedup.patch
emu10k1-mark-midi_spinlock-as-used.patch
epoll-use-unlocked-wqueue-operations.patch
reintegrate-irqreturnh-into-hardirqh.patch
time-clocksource-infrastructure.patch
sched-fix-smt-nice-lock-contention-and-optimization.patch
sched-fix-smt-nice-lock-contention-and-optimization-tidy.patch
sched-comment-bitmap-size-accounting.patch
sched-fix-interactive-ceiling-code.patch
sched-cpu-hotplug-race-vs-set_cpus_allowed.patch
sched-implement-smpnice.patch
sched-protect-calculation-of-max_pull-from-integer-wrap.patch
sched-store-weighted-load-on-up.patch
sched-add-discrete-weighted-cpu-load-function.patch
sched-prevent-high-load-weight-tasks-suppressing-balancing.patch
sched-improve-stability-of-smpnice-load-balancing.patch
sched-improve-smpnice-load-balancing-when-load-per-task.patch
smpnice-dont-consider-sched-groups-which-are-lightly-loaded-for-balancing.patch
smpnice-dont-consider-sched-groups-which-are-lightly-loaded-for-balancing-fix.patch
sched-modify-move_tasks-to-improve-load-balancing-outcomes.patch
sched-avoid-unnecessarily-moving-highest-priority-task-move_tasks.patch
sched-avoid-unnecessarily-moving-highest-priority-task-move_tasks-fix-2.patch
sched_domain-handle-kmalloc-failure.patch
sched_domain-handle-kmalloc-failure-fix.patch
sched_domain-dont-use-gfp_atomic.patch
sched_domain-use-kmalloc_node.patch
sched_domain-allocate-sched_group-structures-dynamically.patch
sched-mc-smt-power-savings-sched-policy.patch
sched_exit-fix-parent-time_slice-calculation.patch
sched_exit-move-the-callsite-to-do_exit.patch
sched-uninline-task_rq_lock.patch
bug-if-setscheduler-is-called-from-interrupt-context.patch
sched-add-above-background-load-function.patch
mm-implement-swap-prefetching.patch
pi-futex-futex-code-cleanups.patch
pi-futex-robust-futex-docs-fix.patch
pi-futex-introduce-debug_check_no_locks_freed.patch
pi-futex-introduce-warn_on_smp.patch
pi-futex-add-plist-implementation.patch
pi-futex-scheduler-support-for-pi.patch
pi-futex-rt-mutex-core.patch
pi-futex-rt-mutex-docs.patch
pi-futex-rt-mutex-docs-update.patch
pi-futex-rt-mutex-debug.patch
pi-futex-rt-mutex-tester.patch
pi-futex-rt-mutex-tester-fix.patch
pi-futex-rt-mutex-futex-api.patch
pi-futex-futex_lock_pi-futex_unlock_pi-support.patch
pi-futex-futex_lock_pi-futex_unlock_pi-support-fix.patch
rtmutex-modify-rtmutex-tester-to-test-the-setscheduler.patch
futex_requeue-optimization.patch
fix-ide-deadlock-in-error-reporting-code.patch
genirq-rename-desc-handler-to-desc-chip.patch
genirq-rename-desc-handler-to-desc-chip-power-fix.patch
genirq-rename-desc-handler-to-desc-chip-ia64-fix.patch
genirq-rename-desc-handler-to-desc-chip-ia64-fix-2.patch
genirq-rename-desc-handler-to-desc-chip-terminate_irqs-fix.patch
genirq-sem2mutex-probe_sem-probing_active.patch
genirq-cleanup-merge-irq_affinity-into-irq_desc.patch
genirq-cleanup-remove-irq_descp.patch
genirq-cleanup-remove-irq_descp-fix.patch
genirq-cleanup-remove-fastcall.patch
genirq-cleanup-misc-code-cleanups.patch
genirq-cleanup-reduce-irq_desc_t-use-mark-it-obsolete.patch
genirq-cleanup-include-linux-irqh.patch
genirq-cleanup-merge-irq_dir-smp_affinity_entry-into-irq_desc.patch
genirq-cleanup-merge-pending_irq_cpumask-into-irq_desc.patch
genirq-cleanup-turn-arch_has_irq_per_cpu-into-config_irq_per_cpu.patch
genirq-debug-better-debug-printout-in-enable_irq.patch
genirq-add-retrigger-irq-op-to-consolidate-hw_irq_resend.patch
genirq-doc-comment-include-linux-irqh-structures.patch
genirq-doc-handle_irq_event-and-__do_irq-comments.patch
genirq-cleanup-no_irq_type-cleanups.patch
genirq-doc-add-design-documentation.patch
genirq-add-genirq-sw-irq-retrigger.patch
genirq-add-irq_noprobe-support.patch
genirq-add-irq_norequest-support.patch
genirq-add-irq_noautoen-support.patch
genirq-update-copyrights.patch
genirq-core.patch
genirq-msi-fixes-2.patch
genirq-add-irq-chip-support.patch
genirq-add-irq-chip-support-fix.patch
genirq-add-irq-chip-support-misroute-irq-dont-call-desc-chip-end.patch
genirq-add-handle_bad_irq.patch
genirq-add-irq-wake-power-management-support.patch
genirq-add-sa_trigger-support.patch
genirq-cleanup-no_irq_type-no_irq_chip-rename.patch
genirq-more-verbose-debugging-on-unexpected-irq-vectors.patch
genirq-ia64-build-fix.patch
genirq-add-irq_type_sense_mask.patch
genirq-add-irq-chip-support-fasteoi-handler-handle-interrupt-disabling.patch
genirq-irq-document-what-an-irq-is.patch
genirq-add-chip-eoi-fastack-fasteoi-core.patch
genirq-add-chip-eoi-fastack-fasteoi-fix.patch
acpi-reduce-code-size-clean-up-fix-validator-message.patch
lock-validator-special-locking-kgdb.patch
detect-atomic-counter-underflows.patch
debug-shared-irqs.patch
make-frame_pointer-default=y.patch
mutex-subsystem-synchro-test-module.patch
vdso-print-fatal-signals.patch
vdso-improve-print_fatal_signals-support-by-adding-memory-maps.patch
-
To unsubscribe from this list: send the line "unsubscribe mm-commits" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
