From 6986d818e6d2c846c001fc7211b5a4153e5ecd11 Mon Sep 17 00:00:00 2001
From: Shiju Jose <shiju.jose@huawei.com>
Date: Sat, 4 Feb 2023 19:15:55 +0000
Subject: [PATCH] rasdaemon: Fix poll() on per_cpu trace_pipe_raw blocks
 indefinitely

The error events are not received in the rasdaemon since kernel 6.1-rc6.
This issue is firstly detected and reported, when testing the CXL error
events in the rasdaemon.

Debugging showed, poll() on trace_pipe_raw in the ras-events.c do not
return and this issue is seen after the commit
42fb0a1e84ff525ebe560e2baf9451ab69127e2b ("tracing/ring-buffer: Have
polling block on watermark").

This issue is also verified using a test application for poll()
and select() on per_cpu trace_pipe_raw.

There is also a bug reported on this issue,
https://lore.kernel.org/all/31eb3b12-3350-90a4-a0d9-d1494db7cf74@oracle.com/

This issue occurs for the per_cpu case, which calls 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. However poll() does not return
even met the percentage of pages condition.

As a fix, rasdaemon set buffer_percent as 0 through the
/sys/kernel/debug/tracing/instances/rasdaemon/buffer_percent, then the
task will wake up as soon as data is added to any of the specific cpu
buffer and poll() on per_cpu/cpuX/trace_pipe_raw does not block
indefinitely.

Dependency on the kernel fix commit
3e46d910d8acf94e5360126593b68bf4fee4c4a1("tracing: Fix poll() and select()
do not work on per_cpu trace_pipe and trace_pipe_raw")

Signed-off-by: Shiju Jose <shiju.jose@huawei.com>
---
 ras-events.c | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)

diff --git a/ras-events.c b/ras-events.c
index 39f9ce2..49e4f9a 100644
--- a/ras-events.c
+++ b/ras-events.c
@@ -376,6 +376,8 @@ static int read_ras_event_all_cpus(struct pthread_data *pdata,
 	int warnonce[n_cpus];
 	char pipe_raw[PATH_MAX];
 	int legacy_kernel = 0;
+	int fd;
+	char buf[16];
 #if 0
 	int need_sleep = 0;
 #endif
@@ -395,6 +397,26 @@ static int read_ras_event_all_cpus(struct pthread_data *pdata,
 		return -ENOMEM;
 	}
 
+	/* Fix for poll() on the per_cpu trace_pipe and trace_pipe_raw blocks
+	 * indefinitely with the default buffer_percent in the kernel trace system,
+	 * which is introduced by the following change in the kernel.
+	 * https://lore.kernel.org/all/20221020231427.41be3f26@gandalf.local.home/T/#u.
+	 * Set buffer_percent to 0 so that poll() will return immediately
+	 * when the trace data is available in the ras per_cpu trace pipe_raw
+	 */
+	fd = open_trace(pdata[0].ras, "buffer_percent", O_WRONLY);
+	if (fd >= 0) {
+		/* For the backward compatibility to the old kernels, do not return
+		 * if fail to set the buffer_percent.
+		 */
+		snprintf(buf, sizeof(buf), "0");
+		size = write(fd, buf, strlen(buf));
+		if (size <= 0)
+			log(TERM, LOG_WARNING, "can't write to buffer_percent\n");
+		close(fd);
+	} else
+		log(TERM, LOG_WARNING, "Can't open buffer_percent\n");
+
 	for (i = 0; i < (n_cpus + 1); i++)
 		fds[i].fd = -1;
 
-- 
2.25.1