diff mbox series

[RFC,1/1] tracing: Fix poll() and select() do not work on per_cpu trace_pipe and trace_pipe_raw

Message ID 20230202182309.742-2-shiju.jose@huawei.com (mailing list archive)
State Accepted
Commit 3e46d910d8acf94e5360126593b68bf4fee4c4a1
Headers show
Series [RFC,1/1] tracing: Fix poll() and select() do not work on per_cpu trace_pipe and trace_pipe_raw | expand

Commit Message

Shiju Jose Feb. 2, 2023, 6:23 p.m. UTC
From: Shiju Jose <shiju.jose@huawei.com>

poll() and select() on per_cpu trace_pipe and trace_pipe_raw do not work
since kernel 6.1-rc6. This issue is seen after the commit
42fb0a1e84ff525ebe560e2baf9451ab69127e2b ("tracing/ring-buffer: Have
polling block on watermark").

This issue is firstly detected and reported, when testing the CXL error
events in the rasdaemon and also erified using the test application for poll()
and select().

This issue occurs for the per_cpu case, when calling the ring_buffer_poll_wait(),
in kernel/trace/ring_buffer.c, with the buffer_percent > 0 and then wait until the
percentage of pages are available. The default value set for the buffer_percent is 50
in the kernel/trace/trace.c.

As a fix, allow userspace application could set buffer_percent as 0 through
the buffer_percent_fops, so that the task will wake up as soon as data is added
to any of the specific cpu buffer.

Fixes: 42fb0a1e84ff5 ("tracing/ring-buffer: Have polling block on watermark")
Signed-off-by: Shiju Jose <shiju.jose@huawei.com>
---
 kernel/trace/trace.c | 3 ---
 1 file changed, 3 deletions(-)

Comments

Steven Rostedt Feb. 2, 2023, 9:18 p.m. UTC | #1
On Thu, 2 Feb 2023 18:23:09 +0000
<shiju.jose@huawei.com> wrote:

> From: Shiju Jose <shiju.jose@huawei.com>
> 
> poll() and select() on per_cpu trace_pipe and trace_pipe_raw do not work
> since kernel 6.1-rc6. This issue is seen after the commit
> 42fb0a1e84ff525ebe560e2baf9451ab69127e2b ("tracing/ring-buffer: Have
> polling block on watermark").
> 
> This issue is firstly detected and reported, when testing the CXL error
> events in the rasdaemon and also erified using the test application for poll()
> and select().
> 
> This issue occurs for the per_cpu case, when calling the ring_buffer_poll_wait(),
> in kernel/trace/ring_buffer.c, with the buffer_percent > 0 and then wait until the
> percentage of pages are available. The default value set for the buffer_percent is 50
> in the kernel/trace/trace.c.
> 
> As a fix, allow userspace application could set buffer_percent as 0 through
> the buffer_percent_fops, so that the task will wake up as soon as data is added
> to any of the specific cpu buffer.
> 
> Fixes: 42fb0a1e84ff5 ("tracing/ring-buffer: Have polling block on watermark")
> Signed-off-by: Shiju Jose <shiju.jose@huawei.com>
> ---

This makes sense to me. I'm going to run a bunch of tests on this and
see if anything else breaks.

Thanks!

-- Steve


>  kernel/trace/trace.c | 3 ---
>  1 file changed, 3 deletions(-)
> 
> diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
> index a555a861b978..01164c78483a 100644
> --- a/kernel/trace/trace.c
> +++ b/kernel/trace/trace.c
> @@ -9148,9 +9148,6 @@ buffer_percent_write(struct file *filp, const char __user *ubuf,
>  	if (val > 100)
>  		return -EINVAL;
>  
> -	if (!val)
> -		val = 1;
> -
>  	tr->buffer_percent = val;
>  
>  	(*ppos)++;
diff mbox series

Patch

diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
index a555a861b978..01164c78483a 100644
--- a/kernel/trace/trace.c
+++ b/kernel/trace/trace.c
@@ -9148,9 +9148,6 @@  buffer_percent_write(struct file *filp, const char __user *ubuf,
 	if (val > 100)
 		return -EINVAL;
 
-	if (!val)
-		val = 1;
-
 	tr->buffer_percent = val;
 
 	(*ppos)++;