In ring_buffer_reset_online_cpus, the buffer_size_kb write operation
may permanently fail if the cpu_online_mask changes between two
for_each_online_buffer_cpu loops. The number of increases and decreases
on both cpu_buffer->resize_disabled and cpu_buffer->record_disabled may be
inconsistent, causing some CPUs to have non-zero values for these atomic
variables after the function returns.
This issue can be reproduced by "echo 0 > trace" while hotplugging cpu.
After reproducing success, we can find out buffer_size_kb will not be
functional anymore.
To prevent leaving 'resize_disabled' and 'record_disabled' non-zero after
ring_buffer_reset_online_cpus returns, we ensure that each atomic variable
has been set up before atomic_sub() to it.
Cc: stable@xxxxxxxxxxxxxxx
Cc: npiggin@xxxxxxxxx
Fixes: b23d7a5f4a07 ("ring-buffer: speed up buffer resets by avoiding synchronize_rcu for each CPU")
Reviewed-by: Cheng-Jui Wang <cheng-jui.wang@xxxxxxxxxxxx>
Signed-off-by: Tze-nan Wu <Tze-nan.Wu@xxxxxxxxxxxx>
---
Changes from v1 to v3: https://lore.kernel.org/all/20230408052226.25268-1-Tze-nan.Wu@xxxxxxxxxxxx/
- Declare the cpumask variable statically rather than dynamically.
Changes from v2 to v3: https://lore.kernel.org/all/20230409024616.31099-1-Tze-nan.Wu@xxxxxxxxxxxx/
- Considering holding cpu_hotplug_lock too long because of the
synchronize_rcu(), maybe it's better to prevent the issue by copying
cpu_online_mask at the entry of the function as V1 does, instead of
using cpus_read_lock().
Changes from v3 to v4: https://lore.kernel.org/all/20230410073512.13362-1-Tze-nan.Wu@xxxxxxxxxxxx/
- Considering that the size of cpumask may not be too big on some machines
We no longer adopt the approach of copying cpumask at the beginning of
the function. Instead, we ensure that atomic variables have been set up
before atomic_sub() is called.
- Change the title of the patch.
---
kernel/trace/ring_buffer.c | 14 +++++++++++---
1 file changed, 11 insertions(+), 3 deletions(-)
diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c
index 76a2d91eecad..8c647d8b5bb4 100644
--- a/kernel/trace/ring_buffer.c
+++ b/kernel/trace/ring_buffer.c
@@ -5361,20 +5361,28 @@ void ring_buffer_reset_online_cpus(struct trace_buffer *buffer)
for_each_online_buffer_cpu(buffer, cpu) {
cpu_buffer = buffer->buffers[cpu];
- atomic_inc(&cpu_buffer->resize_disabled);
+#define RESET_BIT (1 << 30)
+ atomic_add(RESET_BIT, &cpu_buffer->resize_disabled);
atomic_inc(&cpu_buffer->record_disabled);
}
/* Make sure all commits have finished */
synchronize_rcu();
- for_each_online_buffer_cpu(buffer, cpu) {
+ for_each_buffer_cpu(buffer, cpu) {
cpu_buffer = buffer->buffers[cpu];
+ /*
+ * If a CPU came online during the synchronize_rcu(), then
+ * ignore it.
+ */
+ if (!(atomic_read(&cpu_buffer->resize_disabled) & RESET_BIT))
+ continue;
+
reset_disabled_cpu_buffer(cpu_buffer);
atomic_dec(&cpu_buffer->record_disabled);
- atomic_dec(&cpu_buffer->resize_disabled);
+ atomic_sub(RESET_BIT, &cpu_buffer->resize_disabled);
}
mutex_unlock(&buffer->mutex);
--
2.18.0
