* [PATCH] tracing: Keep copying markers when one instance is stopped
@ 2026-10-31 7:35 Jaewon Jung
2026-11-02 14:17 ` Steven Rostedt
0 siblings, 1 reply; 2+ messages in thread
From: Jaewon Jung @ 2026-10-31 7:35 UTC (permalink / raw)
To: Steven Rostedt, Masami Hiramatsu, Mathieu Desnoyers
Cc: stable, linux-kernel, linux-trace-kernel
When several instances enable copy_trace_marker, stopping one instance
can prevent the remaining buffers from receiving markers.
Continue copying text and raw markers even if one copy fails, and
return the last error after trying all buffers.
Fixes: 7b382efd5e8a ("tracing: Allow the top level trace_marker to write into another instances")
Cc: stable@vger.kernel.org
Signed-off-by: Jaewon Jung <jaewon8868@gmail.com>
---
The current code stops at the first error without trying the remaining
buffers. This patch tries all buffers even if a write fails, and returns
the last error.
Returning either the first or last error cannot report all failures.
I chose the last error to keep the code simple. Per-buffer error
reporting is outside the scope of this patch.
Please let me know if you prefer a different error return policy.
Tested on x86_64 QEMU:
- Setup: Enable copy_trace_marker in multiple instances and set
tracing_on=0 in one instance before a live buffer in the copy order.
- Before: No markers in live buffers after the stopped instance;
write() returns an error.
- After: Markers in all live buffers; write() returns an error.
- Existing text/raw marker and instance selftests: 6 PASS.
kernel/trace/trace.c | 12 ++++++------
1 file changed, 6 insertions(+), 6 deletions(-)
diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
index e8c4efa8107b..bddcb35835be 100644
--- a/kernel/trace/trace.c
+++ b/kernel/trace/trace.c
@@ -6316,6 +6316,7 @@ tracing_mark_write(struct file *filp, const char __user *ubuf,
{
struct trace_array *tr = filp->private_data;
ssize_t written = -ENODEV;
+ ssize_t error = 0;
unsigned long ip;
char *buf;
@@ -6346,14 +6347,13 @@ tracing_mark_write(struct file *filp, const char __user *ubuf,
guard(rcu)();
list_for_each_entry_rcu(tr, &marker_copies, marker_list) {
written = write_marker_to_buffer(tr, buf, cnt, ip);
- if (written < 0)
- break;
+ error = written < 0 ? written : error;
}
} else {
written = write_marker_to_buffer(tr, buf, cnt, ip);
}
- return written;
+ return error ? error : written;
}
static ssize_t write_raw_marker_to_buffer(struct trace_array *tr,
@@ -6397,6 +6397,7 @@ tracing_mark_raw_write(struct file *filp, const char __user *ubuf,
{
struct trace_array *tr = filp->private_data;
ssize_t written = -ENODEV;
+ ssize_t error = 0;
char *buf;
if (unlikely(tracing_disabled))
@@ -6425,14 +6426,13 @@ tracing_mark_raw_write(struct file *filp, const char __user *ubuf,
guard(rcu)();
list_for_each_entry_rcu(tr, &marker_copies, marker_list) {
written = write_raw_marker_to_buffer(tr, buf, cnt);
- if (written < 0)
- break;
+ error = written < 0 ? written : error;
}
} else {
written = write_raw_marker_to_buffer(tr, buf, cnt);
}
- return written;
+ return error ? error : written;
}
static int tracing_mark_open(struct inode *inode, struct file *filp)
base-commit: c339f4010dea8b0d754f84ace70cd1b089878ef8
--
2.43.0
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: [PATCH] tracing: Keep copying markers when one instance is stopped
2026-10-31 7:35 [PATCH] tracing: Keep copying markers when one instance is stopped Jaewon Jung
@ 2026-11-02 14:17 ` Steven Rostedt
0 siblings, 0 replies; 2+ messages in thread
From: Steven Rostedt @ 2026-11-02 14:17 UTC (permalink / raw)
To: Jaewon Jung
Cc: Masami Hiramatsu, Mathieu Desnoyers, stable, linux-kernel,
linux-trace-kernel
On Sat, 31 Oct 2026 16:35:45 +0900
Jaewon Jung <jaewon8868@gmail.com> wrote:
> When several instances enable copy_trace_marker, stopping one instance
> can prevent the remaining buffers from receiving markers.
>
> Continue copying text and raw markers even if one copy fails, and
> return the last error after trying all buffers.
>
> Fixes: 7b382efd5e8a ("tracing: Allow the top level trace_marker to write into another instances")
> Cc: stable@vger.kernel.org
Please remove the stable tag, this is more of an enhancement than a bug fix.
You can keep the fixes tag though.
> Signed-off-by: Jaewon Jung <jaewon8868@gmail.com>
> ---
> The current code stops at the first error without trying the remaining
> buffers. This patch tries all buffers even if a write fails, and returns
> the last error.
>
> Returning either the first or last error cannot report all failures.
> I chose the last error to keep the code simple. Per-buffer error
> reporting is outside the scope of this patch.
>
> Please let me know if you prefer a different error return policy.
>
> Tested on x86_64 QEMU:
> - Setup: Enable copy_trace_marker in multiple instances and set
> tracing_on=0 in one instance before a live buffer in the copy order.
> - Before: No markers in live buffers after the stopped instance;
> write() returns an error.
> - After: Markers in all live buffers; write() returns an error.
> - Existing text/raw marker and instance selftests: 6 PASS.
Honestly, just skip the warnings of the copies and don't even bother
returning its value.
-- Steve
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2026-11-02 14:18 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-10-31 7:35 [PATCH] tracing: Keep copying markers when one instance is stopped Jaewon Jung
2026-11-02 14:17 ` Steven Rostedt
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®