| Message ID | 20230409024616.31099-1-Tze-nan.Wu@mediatek.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | [v2] ring-buffer: Prevent inconsistent operation on cpu_buffer->resize_disabled | expand |
On 4/9/23 09:46, Tze-nan Wu wrote: > This issue can be reproduced by "echo 0 > trace" and hotplug cpu at the > same time. After reproducing success, we can find out buffer_size_kb > will not be functional anymore. > Do you mean disabling tracing while hotplugging CPU? Or disabling both tracing and hotplug CPU? > This patch uses cpus_read_lock() to prevent cpu_online_mask being changed > between two different "for_each_online_buffer_cpu". > "Use cpu_read_lock() to prevent ..." > Changes in v2: > Use cpus_read_lock() instead of copying cpu_online_mask at the entry of > function > To resolve kernel test robot warnings ([1] and [2])? Or have they been fixed? [1]: https://lore.kernel.org/stable/202304081615.eiaqpbV8-lkp@intel.com/ [2]: https://lore.kernel.org/stable/202304082051.Dp50upfS-lkp@intel.com/ Thanks.
Hi Bagas, On Sun, 2023-04-09 at 19:30 +0700, Bagas Sanjaya wrote: > > On 4/9/23 09:46, Tze-nan Wu wrote: > > This issue can be reproduced by "echo 0 > trace" and hotplug cpu at > > the > > same time. After reproducing success, we can find out > > buffer_size_kb > > will not be functional anymore. > > > > Do you mean disabling tracing while hotplugging CPU? Or disabling > both > tracing and hotplug CPU? > I mean disabling tracing while hotplugging CPU, sorry for the confusion here. > > This patch uses cpus_read_lock() to prevent cpu_online_mask being > > changed > > between two different "for_each_online_buffer_cpu". > > > > "Use cpu_read_lock() to prevent ..." > Thanks for the suggestion > > Changes in v2: > > Use cpus_read_lock() instead of copying cpu_online_mask at the > > entry of > > function > > > Since I change to fix the issue by using cpus_read_lock(), we don't need a copy of cpu_online_mask anymore, hence the two warnings will not meet in the v2 patch. > To resolve kernel test robot warnings ([1] and [2])? Or have they > been fixed? > > [1]: > https://lore.kernel.org/stable/202304081615.eiaqpbV8-lkp@intel.com/ > [2]: > https://lore.kernel.org/stable/202304082051.Dp50upfS-lkp@intel.com/ > > Thanks. > > -- > An old man doll... just what I always wanted! - Clara >
diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c index 76a2d91eecad..44d833252fb0 100644 --- a/kernel/trace/ring_buffer.c +++ b/kernel/trace/ring_buffer.c @@ -5357,6 +5357,7 @@ void ring_buffer_reset_online_cpus(struct trace_buffer *buffer) /* prevent another thread from changing buffer sizes */ mutex_lock(&buffer->mutex); + cpus_read_lock(); for_each_online_buffer_cpu(buffer, cpu) { cpu_buffer = buffer->buffers[cpu]; @@ -5377,6 +5378,7 @@ void ring_buffer_reset_online_cpus(struct trace_buffer *buffer) atomic_dec(&cpu_buffer->resize_disabled); } + cpus_read_unlock(); mutex_unlock(&buffer->mutex); }
Write to buffer_size_kb can permanently fail due to cpu_online_mask changed between two for_each_online_buffer_cpu loops. The number of increasing and decreasing on cpu_buffer->resize_disable may be inconsistent, leading that the resize_disabled in some CPUs becoming none zero after ring_buffer_reset_online_cpus return. This issue can be reproduced by "echo 0 > trace" and hotplug cpu at the same time. After reproducing success, we can find out buffer_size_kb will not be functional anymore. This patch uses cpus_read_lock() to prevent cpu_online_mask being changed between two different "for_each_online_buffer_cpu". Changes in v2: Use cpus_read_lock() instead of copying cpu_online_mask at the entry of function Link: https://lore.kernel.org/lkml/20230408052226.25268-1-Tze-nan.Wu@mediatek.com/ Cc: stable@vger.kernel.org Signed-off-by: Tze-nan Wu <Tze-nan.Wu@mediatek.com> --- kernel/trace/ring_buffer.c | 2 ++ 1 file changed, 2 insertions(+)