[PATCH v5 0/5] relayfs: misc changes

[PATCH v5 0/5] relayfs: misc changes

[Jason Xing]

From: Jason Xing <kernelxing@tencent.com>

The series mostly focuses on the error counters which helps every user
debug their own kernel module.

---
v5
Link: https://lore.kernel.org/all/20250610004844.66688-1-kerneljasonxing@gmail.com/
1. add Masami's reviewed-by in the first patch
2. fix the wrong printk format specifiers in patch [4/5]

v4
Link: https://lore.kernel.org/all/20250518025734.61479-1-kerneljasonxing@gmail.com/
1. add [1] as the firt/prep/clean-up commit in the series.
2. the rest four patches are not touched, compared to v4.
3. add Masami's reviewed-by tags for last four patches.
[1]: https://lore.kernel.org/all/20250507134225.63248-1-kerneljasonxing@gmail.com/


Jason Xing (5):
  relayfs: abolish prev_padding
  relayfs: support a counter tracking if per-cpu buffers is full
  relayfs: introduce getting relayfs statistics function
  blktrace: use rbuf->stats.full as a drop indicator in relayfs
  relayfs: support a counter tracking if data is too big to write

 drivers/gpu/drm/i915/gt/uc/intel_guc_log.c |  3 +-
 drivers/net/wwan/iosm/iosm_ipc_trace.c     |  3 +-
 drivers/net/wwan/t7xx/t7xx_port_trace.c    |  2 +-
 include/linux/relay.h                      | 24 ++++++--
 kernel/relay.c                             | 66 +++++++++++++++++-----
 kernel/trace/blktrace.c                    | 22 +-------
 6 files changed, 77 insertions(+), 43 deletions(-)

-- 
2.43.5

[PATCH v5 1/5] relayfs: abolish prev_padding

[Jason Xing]

From: Jason Xing <kernelxing@tencent.com>

prev_padding represents the unused space of certain subbuffer. If the
content of a call of relay_write() exceeds the limit of the remainder of
this subbuffer, it will skip storing in the rest space and record the
start point as buf->prev_padding in relay_switch_subbuf(). Since the buf
is a per-cpu big buffer, the point of prev_padding as a global value for
the whole buffer instead of a single subbuffer (whose padding info is
stored in buf->padding[]) seems meaningless from the real use cases, so
we don't bother to record it any more.

Reviewed-by: Yushan Zhou <katrinzhou@tencent.com>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Jason Xing <kernelxing@tencent.com>
---
 drivers/gpu/drm/i915/gt/uc/intel_guc_log.c |  3 +--
 drivers/net/wwan/iosm/iosm_ipc_trace.c     |  3 +--
 drivers/net/wwan/t7xx/t7xx_port_trace.c    |  2 +-
 include/linux/relay.h                      |  5 +----
 kernel/relay.c                             | 14 ++++++++------
 kernel/trace/blktrace.c                    |  2 +-
 6 files changed, 13 insertions(+), 16 deletions(-)

diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_log.c b/drivers/gpu/drm/i915/gt/uc/intel_guc_log.c
index e8a04e476c57..09a64f224c49 100644
--- a/drivers/gpu/drm/i915/gt/uc/intel_guc_log.c
+++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_log.c
@@ -220,8 +220,7 @@ static int guc_action_control_log(struct intel_guc *guc, bool enable,
  */
 static int subbuf_start_callback(struct rchan_buf *buf,
 				 void *subbuf,
-				 void *prev_subbuf,
-				 size_t prev_padding)
+				 void *prev_subbuf)
 {
 	/*
 	 * Use no-overwrite mode by default, where relay will stop accepting
diff --git a/drivers/net/wwan/iosm/iosm_ipc_trace.c b/drivers/net/wwan/iosm/iosm_ipc_trace.c
index eeecfa3d10c5..9656254c1c6c 100644
--- a/drivers/net/wwan/iosm/iosm_ipc_trace.c
+++ b/drivers/net/wwan/iosm/iosm_ipc_trace.c
@@ -51,8 +51,7 @@ static int ipc_trace_remove_buf_file_handler(struct dentry *dentry)
 }
 
 static int ipc_trace_subbuf_start_handler(struct rchan_buf *buf, void *subbuf,
-					  void *prev_subbuf,
-					  size_t prev_padding)
+					  void *prev_subbuf)
 {
 	if (relay_buf_full(buf)) {
 		pr_err_ratelimited("Relay_buf full dropping traces");
diff --git a/drivers/net/wwan/t7xx/t7xx_port_trace.c b/drivers/net/wwan/t7xx/t7xx_port_trace.c
index 4ed8b4e29bf1..f16d3b01302c 100644
--- a/drivers/net/wwan/t7xx/t7xx_port_trace.c
+++ b/drivers/net/wwan/t7xx/t7xx_port_trace.c
@@ -33,7 +33,7 @@ static int t7xx_trace_remove_buf_file_handler(struct dentry *dentry)
 }
 
 static int t7xx_trace_subbuf_start_handler(struct rchan_buf *buf, void *subbuf,
-					   void *prev_subbuf, size_t prev_padding)
+					   void *prev_subbuf)
 {
 	if (relay_buf_full(buf)) {
 		pr_err_ratelimited("Relay_buf full dropping traces");
diff --git a/include/linux/relay.h b/include/linux/relay.h
index b3224111d074..e10a0fdf4325 100644
--- a/include/linux/relay.h
+++ b/include/linux/relay.h
@@ -47,7 +47,6 @@ struct rchan_buf
 	unsigned int page_count;	/* number of current buffer pages */
 	unsigned int finalized;		/* buffer has been finalized */
 	size_t *padding;		/* padding counts per sub-buffer */
-	size_t prev_padding;		/* temporary variable */
 	size_t bytes_consumed;		/* bytes consumed in cur read subbuf */
 	size_t early_bytes;		/* bytes consumed before VFS inited */
 	unsigned int cpu;		/* this buf's cpu */
@@ -84,7 +83,6 @@ struct rchan_callbacks
 	 * @buf: the channel buffer containing the new sub-buffer
 	 * @subbuf: the start of the new sub-buffer
 	 * @prev_subbuf: the start of the previous sub-buffer
-	 * @prev_padding: unused space at the end of previous sub-buffer
 	 *
 	 * The client should return 1 to continue logging, 0 to stop
 	 * logging.
@@ -100,8 +98,7 @@ struct rchan_callbacks
 	 */
 	int (*subbuf_start) (struct rchan_buf *buf,
 			     void *subbuf,
-			     void *prev_subbuf,
-			     size_t prev_padding);
+			     void *prev_subbuf);
 
 	/*
 	 * create_buf_file - create file to represent a relay channel buffer
diff --git a/kernel/relay.c b/kernel/relay.c
index c0c93a04d4ce..94f79f52d826 100644
--- a/kernel/relay.c
+++ b/kernel/relay.c
@@ -250,13 +250,13 @@ EXPORT_SYMBOL_GPL(relay_buf_full);
  */
 
 static int relay_subbuf_start(struct rchan_buf *buf, void *subbuf,
-			      void *prev_subbuf, size_t prev_padding)
+			      void *prev_subbuf)
 {
 	if (!buf->chan->cb->subbuf_start)
 		return !relay_buf_full(buf);
 
 	return buf->chan->cb->subbuf_start(buf, subbuf,
-					   prev_subbuf, prev_padding);
+					   prev_subbuf);
 }
 
 /**
@@ -302,7 +302,7 @@ static void __relay_reset(struct rchan_buf *buf, unsigned int init)
 	for (i = 0; i < buf->chan->n_subbufs; i++)
 		buf->padding[i] = 0;
 
-	relay_subbuf_start(buf, buf->data, NULL, 0);
+	relay_subbuf_start(buf, buf->data, NULL);
 }
 
 /**
@@ -555,9 +555,11 @@ size_t relay_switch_subbuf(struct rchan_buf *buf, size_t length)
 		goto toobig;
 
 	if (buf->offset != buf->chan->subbuf_size + 1) {
-		buf->prev_padding = buf->chan->subbuf_size - buf->offset;
+		size_t prev_padding;
+
+		prev_padding = buf->chan->subbuf_size - buf->offset;
 		old_subbuf = buf->subbufs_produced % buf->chan->n_subbufs;
-		buf->padding[old_subbuf] = buf->prev_padding;
+		buf->padding[old_subbuf] = prev_padding;
 		buf->subbufs_produced++;
 		if (buf->dentry)
 			d_inode(buf->dentry)->i_size +=
@@ -582,7 +584,7 @@ size_t relay_switch_subbuf(struct rchan_buf *buf, size_t length)
 	new_subbuf = buf->subbufs_produced % buf->chan->n_subbufs;
 	new = buf->start + new_subbuf * buf->chan->subbuf_size;
 	buf->offset = 0;
-	if (!relay_subbuf_start(buf, new, old, buf->prev_padding)) {
+	if (!relay_subbuf_start(buf, new, old)) {
 		buf->offset = buf->chan->subbuf_size + 1;
 		return 0;
 	}
diff --git a/kernel/trace/blktrace.c b/kernel/trace/blktrace.c
index 3f6a7bdc6edf..d3083c88474e 100644
--- a/kernel/trace/blktrace.c
+++ b/kernel/trace/blktrace.c
@@ -461,7 +461,7 @@ static const struct file_operations blk_msg_fops = {
  * the user space app in telling how many lost events there were.
  */
 static int blk_subbuf_start_callback(struct rchan_buf *buf, void *subbuf,
-				     void *prev_subbuf, size_t prev_padding)
+				     void *prev_subbuf)
 {
 	struct blk_trace *bt;
 
-- 
2.43.5

[PATCH v5 2/5] relayfs: support a counter tracking if per-cpu buffers is full

[Jason Xing]

From: Jason Xing <kernelxing@tencent.com>

When using relay mechanism, we often encounter the case where new data
are lost or old unconsumed data are overwritten because of slow reader.

Add 'full' field in per-cpu buffer structure to detect if the above case
is happening.  Relay has two modes: 1) non-overwrite mode, 2) overwrite
mode. So buffer being full here respectively means: 1) relayfs doesn't
intend to accept new data and then simply drop them, or 2) relayfs is
going to start over again and overwrite old unread data with new data.

Note: this counter doesn't need any explicit lock to protect from being
modified by different threads for the better performance consideration.
Writers calling __relay_write/relay_write should consider how to use
the lock and ensure it performs under the lock protection, thus it's
not necessary to add a new small lock here.

Reviewed-by: Yushan Zhou <katrinzhou@tencent.com>
Reviewed-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Jason Xing <kernelxing@tencent.com>
---
 include/linux/relay.h | 9 +++++++++
 kernel/relay.c        | 8 +++++++-
 2 files changed, 16 insertions(+), 1 deletion(-)

diff --git a/include/linux/relay.h b/include/linux/relay.h
index e10a0fdf4325..cd77eb285a48 100644
--- a/include/linux/relay.h
+++ b/include/linux/relay.h
@@ -28,6 +28,14 @@
  */
 #define RELAYFS_CHANNEL_VERSION		7
 
+/*
+ * Relay buffer statistics
+ */
+struct rchan_buf_stats
+{
+	unsigned int full_count;	/* counter for buffer full */
+};
+
 /*
  * Per-cpu relay channel buffer
  */
@@ -43,6 +51,7 @@ struct rchan_buf
 	struct irq_work wakeup_work;	/* reader wakeup */
 	struct dentry *dentry;		/* channel file dentry */
 	struct kref kref;		/* channel buffer refcount */
+	struct rchan_buf_stats stats;	/* buffer stats */
 	struct page **page_array;	/* array of current buffer pages */
 	unsigned int page_count;	/* number of current buffer pages */
 	unsigned int finalized;		/* buffer has been finalized */
diff --git a/kernel/relay.c b/kernel/relay.c
index 94f79f52d826..eb3f630f3896 100644
--- a/kernel/relay.c
+++ b/kernel/relay.c
@@ -252,8 +252,13 @@ EXPORT_SYMBOL_GPL(relay_buf_full);
 static int relay_subbuf_start(struct rchan_buf *buf, void *subbuf,
 			      void *prev_subbuf)
 {
+	int full = relay_buf_full(buf);
+
+	if (full)
+		buf->stats.full_count++;
+
 	if (!buf->chan->cb->subbuf_start)
-		return !relay_buf_full(buf);
+		return !full;
 
 	return buf->chan->cb->subbuf_start(buf, subbuf,
 					   prev_subbuf);
@@ -298,6 +303,7 @@ static void __relay_reset(struct rchan_buf *buf, unsigned int init)
 	buf->finalized = 0;
 	buf->data = buf->start;
 	buf->offset = 0;
+	buf->stats.full_count = 0;
 
 	for (i = 0; i < buf->chan->n_subbufs; i++)
 		buf->padding[i] = 0;
-- 
2.43.5

[PATCH v5 3/5] relayfs: introduce getting relayfs statistics function

[Jason Xing]

From: Jason Xing <kernelxing@tencent.com>

In this version, only support getting the counter for buffer full and
implement the framework of how it works.

Users can pass certain flag to fetch what field/statistics they expect
to know. Each time it only returns one result. So do not pass multiple
flags.

Reviewed-by: Yushan Zhou <katrinzhou@tencent.com>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Jason Xing <kernelxing@tencent.com>
---
 include/linux/relay.h |  7 +++++++
 kernel/relay.c        | 30 ++++++++++++++++++++++++++++++
 2 files changed, 37 insertions(+)

diff --git a/include/linux/relay.h b/include/linux/relay.h
index cd77eb285a48..5310967f9d74 100644
--- a/include/linux/relay.h
+++ b/include/linux/relay.h
@@ -31,6 +31,12 @@
 /*
  * Relay buffer statistics
  */
+enum {
+	RELAY_STATS_BUF_FULL = (1 << 0),
+
+	RELAY_STATS_LAST = RELAY_STATS_BUF_FULL,
+};
+
 struct rchan_buf_stats
 {
 	unsigned int full_count;	/* counter for buffer full */
@@ -167,6 +173,7 @@ struct rchan *relay_open(const char *base_filename,
 			 void *private_data);
 extern void relay_close(struct rchan *chan);
 extern void relay_flush(struct rchan *chan);
+size_t relay_stats(struct rchan *chan, int flags);
 extern void relay_subbufs_consumed(struct rchan *chan,
 				   unsigned int cpu,
 				   size_t consumed);
diff --git a/kernel/relay.c b/kernel/relay.c
index eb3f630f3896..fd70d0e03216 100644
--- a/kernel/relay.c
+++ b/kernel/relay.c
@@ -701,6 +701,36 @@ void relay_flush(struct rchan *chan)
 }
 EXPORT_SYMBOL_GPL(relay_flush);
 
+/**
+ *	relay_stats - get channel buffer statistics
+ *	@chan: the channel
+ *	@flags: select particular information to get
+ *
+ *	Returns the count of certain field that caller specifies.
+ */
+size_t relay_stats(struct rchan *chan, int flags)
+{
+	unsigned int i, count = 0;
+	struct rchan_buf *rbuf;
+
+	if (!chan || flags > RELAY_STATS_LAST)
+		return 0;
+
+	if (chan->is_global) {
+		rbuf = *per_cpu_ptr(chan->buf, 0);
+		if (flags & RELAY_STATS_BUF_FULL)
+			count = rbuf->stats.full_count;
+	} else {
+		for_each_online_cpu(i) {
+			rbuf = *per_cpu_ptr(chan->buf, i);
+			if (rbuf && flags & RELAY_STATS_BUF_FULL)
+				count += rbuf->stats.full_count;
+		}
+	}
+
+	return count;
+}
+
 /**
  *	relay_file_open - open file op for relay files
  *	@inode: the inode
-- 
2.43.5

[PATCH v5 4/5] blktrace: use rbuf->stats.full as a drop indicator in relayfs

[Jason Xing]

From: Jason Xing <kernelxing@tencent.com>

Replace internal subbuf_start in blktrace with the default policy
in relayfs.

Remove dropped field from struct blktrace. Correspondingly, call the
common helper in relay. By incrementing full_count to keep track of how
many times we encountered a full buffer issue, user space will know how
many events were lost.

Reviewed-by: Yushan Zhou <katrinzhou@tencent.com>
Reviewed-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Jason Xing <kernelxing@tencent.com>
---
 kernel/trace/blktrace.c | 22 ++--------------------
 1 file changed, 2 insertions(+), 20 deletions(-)

diff --git a/kernel/trace/blktrace.c b/kernel/trace/blktrace.c
index d3083c88474e..5401b9006135 100644
--- a/kernel/trace/blktrace.c
+++ b/kernel/trace/blktrace.c
@@ -415,9 +415,10 @@ static ssize_t blk_dropped_read(struct file *filp, char __user *buffer,
 				size_t count, loff_t *ppos)
 {
 	struct blk_trace *bt = filp->private_data;
+	size_t dropped = relay_stats(bt->rchan, RELAY_STATS_BUF_FULL);
 	char buf[16];
 
-	snprintf(buf, sizeof(buf), "%u\n", atomic_read(&bt->dropped));
+	snprintf(buf, sizeof(buf), "%zu\n", dropped);
 
 	return simple_read_from_buffer(buffer, count, ppos, buf, strlen(buf));
 }
@@ -456,23 +457,6 @@ static const struct file_operations blk_msg_fops = {
 	.llseek =	noop_llseek,
 };
 
-/*
- * Keep track of how many times we encountered a full subbuffer, to aid
- * the user space app in telling how many lost events there were.
- */
-static int blk_subbuf_start_callback(struct rchan_buf *buf, void *subbuf,
-				     void *prev_subbuf)
-{
-	struct blk_trace *bt;
-
-	if (!relay_buf_full(buf))
-		return 1;
-
-	bt = buf->chan->private_data;
-	atomic_inc(&bt->dropped);
-	return 0;
-}
-
 static int blk_remove_buf_file_callback(struct dentry *dentry)
 {
 	debugfs_remove(dentry);
@@ -491,7 +475,6 @@ static struct dentry *blk_create_buf_file_callback(const char *filename,
 }
 
 static const struct rchan_callbacks blk_relay_callbacks = {
-	.subbuf_start		= blk_subbuf_start_callback,
 	.create_buf_file	= blk_create_buf_file_callback,
 	.remove_buf_file	= blk_remove_buf_file_callback,
 };
@@ -580,7 +563,6 @@ static int do_blk_trace_setup(struct request_queue *q, char *name, dev_t dev,
 	}
 
 	bt->dev = dev;
-	atomic_set(&bt->dropped, 0);
 	INIT_LIST_HEAD(&bt->running_list);
 
 	ret = -EIO;
-- 
2.43.5

[PATCH v5 5/5] relayfs: support a counter tracking if data is too big to write

[Jason Xing]

From: Jason Xing <kernelxing@tencent.com>

It really doesn't matter if the user/admin knows what the last too
big value is. Record how many times this case is triggered would be
helpful.

Solve the existing issue where relay_reset() doesn't restore
the value.

Store the counter in the per-cpu buffer structure instead of the global
buffer structure. It also solves the racy condition which is likely
to happen when a few of per-cpu buffers encounter the too big data case
and then access the global field last_toobig without lock protection.

Remove the printk in relay_close() since kernel module can directly call
relay_stats() as they want.

Reviewed-by: Yushan Zhou <katrinzhou@tencent.com>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Jason Xing <kernelxing@tencent.com>
---
 include/linux/relay.h |  5 +++--
 kernel/relay.c        | 18 ++++++++++--------
 2 files changed, 13 insertions(+), 10 deletions(-)

diff --git a/include/linux/relay.h b/include/linux/relay.h
index 5310967f9d74..6772a7075840 100644
--- a/include/linux/relay.h
+++ b/include/linux/relay.h
@@ -33,13 +33,15 @@
  */
 enum {
 	RELAY_STATS_BUF_FULL = (1 << 0),
+	RELAY_STATS_WRT_BIG = (1 << 1),
 
-	RELAY_STATS_LAST = RELAY_STATS_BUF_FULL,
+	RELAY_STATS_LAST = RELAY_STATS_WRT_BIG,
 };
 
 struct rchan_buf_stats
 {
 	unsigned int full_count;	/* counter for buffer full */
+	unsigned int big_count;		/* counter for too big to write */
 };
 
 /*
@@ -79,7 +81,6 @@ struct rchan
 	const struct rchan_callbacks *cb; /* client callbacks */
 	struct kref kref;		/* channel refcount */
 	void *private_data;		/* for user-defined data */
-	size_t last_toobig;		/* tried to log event > subbuf size */
 	struct rchan_buf * __percpu *buf; /* per-cpu channel buffers */
 	int is_global;			/* One global buffer ? */
 	struct list_head list;		/* for channel list */
diff --git a/kernel/relay.c b/kernel/relay.c
index fd70d0e03216..bcc889dc0970 100644
--- a/kernel/relay.c
+++ b/kernel/relay.c
@@ -304,6 +304,7 @@ static void __relay_reset(struct rchan_buf *buf, unsigned int init)
 	buf->data = buf->start;
 	buf->offset = 0;
 	buf->stats.full_count = 0;
+	buf->stats.big_count = 0;
 
 	for (i = 0; i < buf->chan->n_subbufs; i++)
 		buf->padding[i] = 0;
@@ -603,7 +604,7 @@ size_t relay_switch_subbuf(struct rchan_buf *buf, size_t length)
 	return length;
 
 toobig:
-	buf->chan->last_toobig = length;
+	buf->stats.big_count++;
 	return 0;
 }
 EXPORT_SYMBOL_GPL(relay_switch_subbuf);
@@ -663,11 +664,6 @@ void relay_close(struct rchan *chan)
 			if ((buf = *per_cpu_ptr(chan->buf, i)))
 				relay_close_buf(buf);
 
-	if (chan->last_toobig)
-		printk(KERN_WARNING "relay: one or more items not logged "
-		       "[item size (%zd) > sub-buffer size (%zd)]\n",
-		       chan->last_toobig, chan->subbuf_size);
-
 	list_del(&chan->list);
 	kref_put(&chan->kref, relay_destroy_channel);
 	mutex_unlock(&relay_channels_mutex);
@@ -720,11 +716,17 @@ size_t relay_stats(struct rchan *chan, int flags)
 		rbuf = *per_cpu_ptr(chan->buf, 0);
 		if (flags & RELAY_STATS_BUF_FULL)
 			count = rbuf->stats.full_count;
+		else if (flags & RELAY_STATS_WRT_BIG)
+			count = rbuf->stats.big_count;
 	} else {
 		for_each_online_cpu(i) {
 			rbuf = *per_cpu_ptr(chan->buf, i);
-			if (rbuf && flags & RELAY_STATS_BUF_FULL)
-				count += rbuf->stats.full_count;
+			if (rbuf) {
+				if (flags & RELAY_STATS_BUF_FULL)
+					count += rbuf->stats.full_count;
+				else if (flags & RELAY_STATS_WRT_BIG)
+					count += rbuf->stats.big_count;
+			}
 		}
 	}
 
-- 
2.43.5