perf_output_read_group may respond to IPI request of other cores and invoke
__perf_install_in_context function. As a result, hwc configuration is modified.
causing inconsistency and unexpected consequences.
Syzkaller has reported a problem with this scenario:
Internal error: Oops - undefined instruction: 0 [#1] SMP
Modules linked in:
CPU: 1 PID: 15523 Comm: syz-executor.3 Not tainted 5.10.0 #6
Hardware name: linux,dummy-virt (DT)
pstate: 80400005 (Nzcv daif +PAN -UAO -TCO BTYPE=--)
pc : read_pmevcntrn+0x1e4/0x1ec arch/arm64/kernel/perf_event.c:423
lr : read_pmevcntrn+0x1e4/0x1ec arch/arm64/kernel/perf_event.c:423
sp : ffffa000174d6ee0
x29: ffffa000174d6ee0 x28: ffffedbca6b17a40
x27: ffffa000174d7310 x26: ffff4b3ec102e398
x25: 00000000ffffffff x24: 00000000ffffffff
x23: ffff4b3ed8afb000 x22: ffff4b3ed8afb160
x21: ffff4b3ed8afb184 x20: ffffedbca46aa3e0
x19: ffffedbca2a524bc x18: 0000000000000000
x17: 0000000000000000 x16: 0000000000000000
x15: 0000000020000100 x14: 0000000000000000
x13: 0000000000000000 x12: ffff8967db15f63d
x11: 1fffe967db15f63c x10: ffff8967db15f63c
x9 : ffffedbca2a5273c x8 : ffff4b3ed8afb1e7
x7 : 0000000000000001 x6 : ffff8967db15f63c
x5 : ffff4b3f08b89400 x4 : 0000000000000000
x3 : ffffedbca2a00000 x2 : ffffedbca4690000
x1 : ffff4b3f08b89400 x0 : 0000000000000000
Call trace:
read_pmevcntrn+0x1e4/0x1ec arch/arm64/kernel/perf_event.c:423
armv8pmu_read_evcntr arch/arm64/kernel/perf_event.c:467 [inline]
armv8pmu_read_hw_counter arch/arm64/kernel/perf_event.c:475 [inline]
armv8pmu_read_counter+0x10c/0x1f0 arch/arm64/kernel/perf_event.c:528
armpmu_event_update+0x9c/0x1bc drivers/perf/arm_pmu.c:247
armpmu_read+0x24/0x30 drivers/perf/arm_pmu.c:264
perf_output_read_group+0x4cc/0x71c kernel/events/core.c:6806
perf_output_read+0x78/0x1c4 kernel/events/core.c:6845
perf_output_sample+0xafc/0x1000 kernel/events/core.c:6892
__perf_event_output kernel/events/core.c:7273 [inline]
perf_event_output_forward+0xd8/0x130 kernel/events/core.c:7287
__perf_event_overflow+0xbc/0x20c kernel/events/core.c:8943
perf_swevent_overflow kernel/events/core.c:9019 [inline]
perf_swevent_event+0x274/0x2c0 kernel/events/core.c:9047
do_perf_sw_event kernel/events/core.c:9160 [inline]
___perf_sw_event+0x150/0x1b4 kernel/events/core.c:9191
__perf_sw_event+0x58/0x7c kernel/events/core.c:9203
perf_sw_event include/linux/perf_event.h:1177 [inline]
mm_account_fault mm/memory.c:4707 [inline]
handle_mm_fault+0x364/0x3f0 mm/memory.c:4758
__do_page_fault arch/arm64/mm/fault.c:438 [inline]
do_page_fault+0x334/0x8f0 arch/arm64/mm/fault.c:537
do_translation_fault+0x188/0x1e0 arch/arm64/mm/fault.c:619
do_mem_abort+0x68/0x120 arch/arm64/mm/fault.c:743
el1_abort+0xc0/0x150 arch/arm64/kernel/entry-common.c:119
el1_sync_handler+0x118/0x150 arch/arm64/kernel/entry-common.c:202
el1_sync+0x74/0x100 arch/arm64/kernel/entry.S:665
__arch_clear_user+0x20/0xa0 arch/arm64/lib/clear_user.S:25
read_iter_zero+0x90/0x16c drivers/char/mem.c:718
call_read_iter include/linux/fs.h:1954 [inline]
do_iter_readv_writev+0x394/0x414 fs/read_write.c:735
do_iter_read+0x1b0/0x280 fs/read_write.c:798
vfs_readv+0xf0/0x150 fs/read_write.c:918
do_readv+0x108/0x270 fs/read_write.c:955
__do_sys_readv fs/read_write.c:1046 [inline]
__se_sys_readv fs/read_write.c:1043 [inline]
__arm64_sys_readv+0x54/0x64 fs/read_write.c:1043
__invoke_syscall arch/arm64/kernel/syscall.c:36 [inline]
invoke_syscall arch/arm64/kernel/syscall.c:48 [inline]
el0_svc_common.constprop.0+0xf4/0x414 arch/arm64/kernel/syscall.c:155
do_el0_svc+0x50/0x11c arch/arm64/kernel/syscall.c:217
el0_svc+0x20/0x30 arch/arm64/kernel/entry-common.c:353
el0_sync_handler+0xe4/0x1e0 arch/arm64/kernel/entry-common.c:369
el0_sync+0x148/0x180 arch/arm64/kernel/entry.S:683
Code: 940c387b d53be813 17ffff9c 940c3878 (d53bebd3)
---[ end trace 6aab9f4b33ebf0aa ]---
----------------
Code disassembly (best guess):
0: 940c387b bl 0x30e1ec
4: d53be813 mrs x19, pmevcntr0_el0
8: 17ffff9c b 0xfffffffffffffe78
c: 940c3878 bl 0x30e1ec
* 10: d53bebd3 mrs x19, pmevcntr30_el0 <-- trapping instruction
Interrupts are not disabled when perf_output_read_group reads PMU counter.
In this case, IPI request may be received from other cores.
As a result, PMU configuration is modified and an error occurs when
reading PMU counter:
CPU0 CPU1
__se_sys_perf_event_open
perf_install_in_context
perf_output_read_group smp_call_function_single
for_each_sibling_event(sub, leader) { generic_exec_single
if ((sub != event) && remote_function
(sub->state == PERF_EVENT_STATE_ACTIVE)) |
<enter IPI handler: __perf_install_in_context> <----RAISE IPI-----+
__perf_install_in_context
ctx_resched
event_sched_out
armpmu_del
...
hwc->idx = -1; // event->hwc.idx is set to -1
...
<exit IPI>
sub->pmu->read(sub);
armpmu_read
armv8pmu_read_counter
armv8pmu_read_hw_counter
int idx = event->hw.idx; // idx = -1
u64 val = armv8pmu_read_evcntr(idx);
u32 counter = ARMV8_IDX_TO_COUNTER(idx); // invalid counter = 30
read_pmevcntrn(counter) // undefined instruction
Signed-off-by: Yang Jihong <[email protected]>
---
Changes since v3:
- Modify commit message according to Peter's suggestion.
Changes since v2:
- Update commit message.
Changes since v1:
- Adapt patch to mainline branch to solve compilation problem.
kernel/events/core.c | 9 +++++++++
1 file changed, 9 insertions(+)
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 2621fd24ad26..752a9079a807 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -6895,6 +6895,13 @@ static void perf_output_read_group(struct perf_output_handle *handle,
u64 read_format = event->attr.read_format;
u64 values[6];
int n = 0;
+ unsigned long flags;
+
+ /*
+ * Disabling interrupts avoids all counter scheduling
+ * (context switches, timer based rotation and IPIs).
+ */
+ local_irq_save(flags);
values[n++] = 1 + leader->nr_siblings;
@@ -6931,6 +6938,8 @@ static void perf_output_read_group(struct perf_output_handle *handle,
__output_copy(handle, values, n * sizeof(u64));
}
+
+ local_irq_restore(flags);
}
#define PERF_FORMAT_TOTAL_TIMES (PERF_FORMAT_TOTAL_TIME_ENABLED|\
--
2.30.GIT
On Fri, Sep 02, 2022 at 04:29:18PM +0800, Yang Jihong wrote:
> perf_output_read_group may respond to IPI request of other cores and invoke
> __perf_install_in_context function. As a result, hwc configuration is modified.
> causing inconsistency and unexpected consequences.
>
> Interrupts are not disabled when perf_output_read_group reads PMU counter.
> In this case, IPI request may be received from other cores.
> As a result, PMU configuration is modified and an error occurs when
> reading PMU counter:
>
> CPU0 CPU1
> __se_sys_perf_event_open
> perf_install_in_context
> perf_output_read_group smp_call_function_single
> for_each_sibling_event(sub, leader) { generic_exec_single
> if ((sub != event) && remote_function
> (sub->state == PERF_EVENT_STATE_ACTIVE)) |
> <enter IPI handler: __perf_install_in_context> <----RAISE IPI-----+
> __perf_install_in_context
> ctx_resched
> event_sched_out
> armpmu_del
> ...
> hwc->idx = -1; // event->hwc.idx is set to -1
> ...
> <exit IPI>
> sub->pmu->read(sub);
> armpmu_read
> armv8pmu_read_counter
> armv8pmu_read_hw_counter
> int idx = event->hw.idx; // idx = -1
> u64 val = armv8pmu_read_evcntr(idx);
> u32 counter = ARMV8_IDX_TO_COUNTER(idx); // invalid counter = 30
> read_pmevcntrn(counter) // undefined instruction
>
> Signed-off-by: Yang Jihong <[email protected]>
> ---
Right; thanks for reminding me. I had hoped you'd pick up and test the
extra assertion some, but I'll just do that on top.
Hello,
On 2022/9/5 9:25, Yang Jihong wrote:
> On Fri, Sep 02, 2022 at 04:29:18PM +0800, Yang Jihong wrote:
> > perf_output_read_group may respond to IPI request of other cores and
> > invoke __perf_install_in_context function. As a result, hwc
> configuration is modified.
> > causing inconsistency and unexpected consequences.
> >
> > Interrupts are not disabled when perf_output_read_group reads PMU
> counter.
> > In this case, IPI request may be received from other cores.
> > As a result, PMU configuration is modified and an error occurs when
> > reading PMU counter:
> >
> > CPU0 CPU1
> > __se_sys_perf_event_open
> > perf_install_in_context
> > perf_output_read_group smp_call_function_single
> > for_each_sibling_event(sub, leader) { generic_exec_single
> > if ((sub != event) && remote_function
> > (sub->state == PERF_EVENT_STATE_ACTIVE)) |
> > <enter IPI handler: __perf_install_in_context> <----RAISE IPI-----+
> > __perf_install_in_context
> > ctx_resched
> > event_sched_out
> > armpmu_del
> > ...
> > hwc->idx = -1; // event->hwc.idx is set to -1 ...
> > <exit IPI>
> > sub->pmu->read(sub);
> > armpmu_read
> > armv8pmu_read_counter
> > armv8pmu_read_hw_counter
> > int idx = event->hw.idx; // idx = -1
> > u64 val = armv8pmu_read_evcntr(idx);
> > u32 counter = ARMV8_IDX_TO_COUNTER(idx); //
> invalid counter = 30
> > read_pmevcntrn(counter) // undefined instruction
> >
> > Signed-off-by: Yang Jihong <[email protected]>
> > ---
>
> Right; thanks for reminding me. I had hoped you'd pick up and test the
> extra assertion some, but I'll just do that on top.
> .
OK, if you want me to do assertion testing on any scenarios,
please send me the patch.
Thanks,
Yang
The following commit has been merged into the perf/urgent branch of tip:
Commit-ID: 6b959ba22d34ca793ffdb15b5715457c78e38b1a
Gitweb: https://git.kernel.org/tip/6b959ba22d34ca793ffdb15b5715457c78e38b1a
Author: Yang Jihong <[email protected]>
AuthorDate: Fri, 02 Sep 2022 16:29:18 +08:00
Committer: Peter Zijlstra <[email protected]>
CommitterDate: Tue, 06 Sep 2022 11:33:00 +02:00
perf/core: Fix reentry problem in perf_output_read_group()
perf_output_read_group may respond to IPI request of other cores and invoke
__perf_install_in_context function. As a result, hwc configuration is modified.
causing inconsistency and unexpected consequences.
Interrupts are not disabled when perf_output_read_group reads PMU counter.
In this case, IPI request may be received from other cores.
As a result, PMU configuration is modified and an error occurs when
reading PMU counter:
CPU0 CPU1
__se_sys_perf_event_open
perf_install_in_context
perf_output_read_group smp_call_function_single
for_each_sibling_event(sub, leader) { generic_exec_single
if ((sub != event) && remote_function
(sub->state == PERF_EVENT_STATE_ACTIVE)) |
<enter IPI handler: __perf_install_in_context> <----RAISE IPI-----+
__perf_install_in_context
ctx_resched
event_sched_out
armpmu_del
...
hwc->idx = -1; // event->hwc.idx is set to -1
...
<exit IPI>
sub->pmu->read(sub);
armpmu_read
armv8pmu_read_counter
armv8pmu_read_hw_counter
int idx = event->hw.idx; // idx = -1
u64 val = armv8pmu_read_evcntr(idx);
u32 counter = ARMV8_IDX_TO_COUNTER(idx); // invalid counter = 30
read_pmevcntrn(counter) // undefined instruction
Signed-off-by: Yang Jihong <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Link: https://lkml.kernel.org/r/[email protected]
---
kernel/events/core.c | 9 +++++++++
1 file changed, 9 insertions(+)
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 2621fd2..ff4bffc 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -6893,9 +6893,16 @@ static void perf_output_read_group(struct perf_output_handle *handle,
{
struct perf_event *leader = event->group_leader, *sub;
u64 read_format = event->attr.read_format;
+ unsigned long flags;
u64 values[6];
int n = 0;
+ /*
+ * Disabling interrupts avoids all counter scheduling
+ * (context switches, timer based rotation and IPIs).
+ */
+ local_irq_save(flags);
+
values[n++] = 1 + leader->nr_siblings;
if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED)
@@ -6931,6 +6938,8 @@ static void perf_output_read_group(struct perf_output_handle *handle,
__output_copy(handle, values, n * sizeof(u64));
}
+
+ local_irq_restore(flags);
}
#define PERF_FORMAT_TOTAL_TIMES (PERF_FORMAT_TOTAL_TIME_ENABLED|\