mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: "Isaac J. Manjarres" <isaacmanjarres@google.com>
To: Petr Mladek <pmladek@suse.com>,
	Steven Rostedt <rostedt@goodmis.org>,
	 John Ogness <john.ogness@linutronix.de>,
	Sergey Senozhatsky <senozhatsky@chromium.org>
Cc: surenb@google.com,
	"Isaac J. Manjarres" <isaacmanjarres@google.com>,
	kernel-team@android.com,  linux-kernel@vger.kernel.org
Subject: [PATCH v2] printk: Improve memory usage logging during boot
Date: Mon, 30 Sep 2024 11:48:24 -0700	[thread overview]
Message-ID: <20240930184826.3595221-1-isaacmanjarres@google.com> (raw)

When the initial printk ring buffer size is updated, setup_log_buf()
allocates a new ring buffer, as well as a set of meta-data structures
for the new ring buffer. The function also emits the new size of the
ring buffer, but not the size of the meta-data structures.

This makes it difficult to assess how changing the log buffer size
impacts memory usage during boot.

For instance, increasing the ring buffer size from 512 KB to 1 MB
through the command line yields an increase of 2304 KB in reserved
memory at boot, while the only obvious change is the 512 KB
difference in the ring buffer sizes:

log_buf_len=512K:

printk: log_buf_len: 524288 bytes
Memory: ... (... 733252K reserved ...)

log_buf_len=1M:

printk: log_buf_len: 1048576 bytes
Memory: ... (... 735556K reserved ...)

This is because of how the size of the meta-data structures scale with
the size of the ring buffer.

Even when there aren't changes to the printk ring buffer size (i.e. the
initial size ==  1 << CONFIG_LOG_BUF_SHIFT), it is impossible to tell
how much memory is consumed by the printk ring buffer during boot.

Therefore, unconditionally log the sizes of the printk ring buffer
and its meta-data structures, so that it's easier to understand
how changing the log buffer size (either through the command line or
by changing CONFIG_LOG_BUF_SHIFT) affects boot time memory usage.

With the new logs, it is much easier to see exactly why the memory
increased by 2304 KB:

log_buf_len=512K:

printk: log_buf_len: 524288 bytes
printk: prb_descs size: 393216 bytes
printk: printk_infos size: 1441792 bytes
Memory: ... (... 733252K reserved ...)

log_buf_len=1M:

printk: log_buf_len: 1048576 bytes
printk: prb_descs size: 786432 bytes
printk: printk_infos size: 2883584 bytes
Memory: ... (... 735556K reserved ...)

Signed-off-by: Isaac J. Manjarres <isaacmanjarres@google.com>
---
 kernel/printk/printk.c | 30 ++++++++++++++++++++++++++----
 1 file changed, 26 insertions(+), 4 deletions(-)

v1 -> v2:
- Consolidated the printk memory usage stats into a single line, and
  summed the printk_info and prb_desc structure sizes into a single
  stat to make the output more user-friendly.
- Fixed an error that caused the memory usage stats to be emitted twice
  when using the default printk buffer.

diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index beb808f4c367..9ba7dd8f352f 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -1156,6 +1156,17 @@ static unsigned int __init add_to_rb(struct printk_ringbuffer *rb,
 
 static char setup_text_buf[PRINTKRB_RECORD_MAX] __initdata;
 
+static void print_log_buf_usage_stats(void)
+{
+	unsigned int descs_count = log_buf_len >> PRB_AVGBITS;
+	size_t meta_data_size;
+
+	meta_data_size = descs_count * (sizeof(struct prb_desc) + sizeof(struct printk_info));
+
+	pr_info("log buffer data + meta data: %u + %zu = %zu bytes\n",
+		log_buf_len, meta_data_size, log_buf_len + meta_data_size);
+}
+
 void __init setup_log_buf(int early)
 {
 	struct printk_info *new_infos;
@@ -1185,20 +1196,29 @@ void __init setup_log_buf(int early)
 	if (!early && !new_log_buf_len)
 		log_buf_add_cpu();
 
-	if (!new_log_buf_len)
+	if (!new_log_buf_len) {
+		/*
+		 * If early == 0 here, then we're using the default buffer,
+		 * but the memory usage stats haven't been printed yet,
+		 * so do that now.
+		 */
+		if (!early)
+			goto out;
+
 		return;
+	}
 
 	new_descs_count = new_log_buf_len >> PRB_AVGBITS;
 	if (new_descs_count == 0) {
 		pr_err("new_log_buf_len: %lu too small\n", new_log_buf_len);
-		return;
+		goto out;
 	}
 
 	new_log_buf = memblock_alloc(new_log_buf_len, LOG_ALIGN);
 	if (unlikely(!new_log_buf)) {
 		pr_err("log_buf_len: %lu text bytes not available\n",
 		       new_log_buf_len);
-		return;
+		goto out;
 	}
 
 	new_descs_size = new_descs_count * sizeof(struct prb_desc);
@@ -1261,7 +1281,7 @@ void __init setup_log_buf(int early)
 		       prb_next_seq(&printk_rb_static) - seq);
 	}
 
-	pr_info("log_buf_len: %u bytes\n", log_buf_len);
+	print_log_buf_usage_stats();
 	pr_info("early log buf free: %u(%u%%)\n",
 		free, (free * 100) / __LOG_BUF_LEN);
 	return;
@@ -1270,6 +1290,8 @@ void __init setup_log_buf(int early)
 	memblock_free(new_descs, new_descs_size);
 err_free_log_buf:
 	memblock_free(new_log_buf, new_log_buf_len);
+out:
+	print_log_buf_usage_stats();
 }
 
 static bool __read_mostly ignore_loglevel;
-- 
2.46.1.824.gd892dcdcdd-goog


             reply	other threads:[~2024-09-30 18:48 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-09-30 18:48 Isaac J. Manjarres [this message]
2024-10-01 15:46 ` Petr Mladek
2024-10-02 18:04   ` Isaac Manjarres
2024-10-15 12:47     ` Isaac Manjarres
2024-10-16 10:24       ` Petr Mladek
2024-10-16 16:38         ` Isaac Manjarres
2024-12-01 11:40   ` Geert Uytterhoeven
2024-12-03  9:42     ` Petr Mladek
2024-12-04 17:17       ` John Ogness

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20240930184826.3595221-1-isaacmanjarres@google.com \
    --to=isaacmanjarres@google.com \
    --cc=john.ogness@linutronix.de \
    --cc=kernel-team@android.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=pmladek@suse.com \
    --cc=rostedt@goodmis.org \
    --cc=senozhatsky@chromium.org \
    --cc=surenb@google.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
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®