From: Alex Markuze <amarkuze@redhat.com>
Date: Wed, 26 Feb 2025 15:04:52 +0000 (+0000)
Subject: tls: adding an allocation histogram
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=764e50ba296da76e7332d2d08223d3e484aea2b3;p=ceph-client.git

tls: adding an allocation histogram
---

diff --git a/fs/ceph/debugfs.c b/fs/ceph/debugfs.c
index 9fa6203bb741..2f2bd49cc482 100644
--- a/fs/ceph/debugfs.c
+++ b/fs/ceph/debugfs.c
@@ -23,6 +23,46 @@
 #include "mds_client.h"
 #include "metric.h"
 
+static int histogram_show(struct seq_file *s, void *p)
+{
+	struct ceph_san_tls_logger *tls;
+	size_t cpu;
+	u64 total_histogram[16] = {0};
+	int i;
+
+	seq_printf(s, "Ceph TLS Histogram Summary:\n");
+	seq_printf(s, "%-10s %-16s\n", "Bucket", "Count");
+	seq_printf(s, "-------------------------\n");
+
+	/* Sum up histograms from all CPUs */
+	for_each_possible_cpu(cpu) {
+		tls = &per_cpu(ceph_san_tls, cpu);
+
+		/* Add each CPU's histogram data to the total */
+		for (i = 0; i < 16; i++) {
+				total_histogram[i] += tls->histogram.counters[i];
+		}
+	}
+
+	/* Calculate total sum for normalization */
+	u64 sum = 0;
+	for (i = 0; i < 16; i++) {
+		sum += total_histogram[i];
+	}
+
+	/* Display normalized histogram with stars and percentages */
+	for (i = 0; i < 16; i++) {
+		int stars = sum ? (total_histogram[i] * 128) / sum : 0;
+		u64 percent = sum ? (total_histogram[i] * 100) / sum : 0;
+		seq_printf(s, "%-4d [%3.1f%%] ", i, percent);
+		while (stars-- > 0)
+			seq_printf(s, "*");
+		seq_printf(s, "\n");
+	}
+
+	return 0;
+}
+
 static int ceph_san_show(struct seq_file *s, void *p)
 {
 	struct ceph_san_tls_logger *tls;
@@ -413,6 +453,7 @@ DEFINE_SHOW_ATTRIBUTE(metrics_latency);
 DEFINE_SHOW_ATTRIBUTE(metrics_size);
 DEFINE_SHOW_ATTRIBUTE(metrics_caps);
 DEFINE_SHOW_ATTRIBUTE(ceph_san);
+DEFINE_SHOW_ATTRIBUTE(histogram);
 
 
 /*
@@ -449,6 +490,7 @@ void ceph_fs_debugfs_cleanup(struct ceph_fs_client *fsc)
 	debugfs_remove(fsc->debugfs_status);
 	debugfs_remove(fsc->debugfs_mdsc);
 	debugfs_remove(fsc->debugfs_cephsan);
+	debugfs_remove(fsc->debugfs_histogram);
 	debugfs_remove_recursive(fsc->debugfs_metrics_dir);
 	doutc(fsc->client, "done\n");
 }
@@ -507,6 +549,12 @@ void ceph_fs_debugfs_init(struct ceph_fs_client *fsc)
 							fsc,
 							&ceph_san_fops);
 
+	fsc->debugfs_histogram = debugfs_create_file("histogram",
+							0444,
+							fsc->client->debugfs_dir,
+							fsc,
+							&histogram_fops);
+
 	fsc->debugfs_metrics_dir = debugfs_create_dir("metrics",
 						      fsc->client->debugfs_dir);
 
diff --git a/fs/ceph/super.h b/fs/ceph/super.h
index d7e52f0c79a4..4f901006cb48 100644
--- a/fs/ceph/super.h
+++ b/fs/ceph/super.h
@@ -154,6 +154,7 @@ struct ceph_fs_client {
 	struct dentry *debugfs_mds_sessions;
 	struct dentry *debugfs_metrics_dir;
 	struct dentry *debugfs_cephsan;
+	struct dentry *debugfs_histogram;
 #endif
 
 #ifdef CONFIG_CEPH_FSCACHE
diff --git a/include/linux/ceph/ceph_san.h b/include/linux/ceph/ceph_san.h
index e6d3c34ce6ca..ee1445a2f37b 100644
--- a/include/linux/ceph/ceph_san.h
+++ b/include/linux/ceph/ceph_san.h
@@ -5,7 +5,7 @@
 #include <linux/list.h>
 #include <linux/spinlock.h>
 #include <linux/sched.h>
-
+#include <linux/slab.h>
 
 DECLARE_PER_CPU(struct ceph_san_tls_logger, ceph_san_tls);
 DECLARE_PER_CPU(struct cephsan_pagefrag, ceph_san_pagefrag);
@@ -94,11 +94,14 @@ struct ceph_san_log_entry {
     pid_t pid;
     u32 len;
 };
-
+struct histogram {
+	u64 counters[16];
+};
 struct ceph_san_tls_logger {
     size_t head_idx;
     struct page *pages;
     struct ceph_san_log_entry *logs;
+    struct histogram histogram;
 };
 #else /* CONFIG_DEBUG_FS */
 
diff --git a/net/ceph/ceph_san.c b/net/ceph/ceph_san.c
index ed6592b5b0ea..5b41ce8b0f12 100644
--- a/net/ceph/ceph_san.c
+++ b/net/ceph/ceph_san.c
@@ -26,7 +26,7 @@ void log_cephsan(char *buf) {
     struct ceph_san_tls_logger *tls = this_cpu_ptr(&ceph_san_tls);
     struct cephsan_pagefrag *pf = this_cpu_ptr(&ceph_san_pagefrag);
 
-    int head_idx = tls->head_idx++ & (CEPH_SAN_MAX_LOGS - 1);
+    int head_idx = tls->head_idx + 1 & (CEPH_SAN_MAX_LOGS - 1);
     int pre_len = tls->logs[head_idx].len;
 
     buf[len-1] = '\0';
@@ -35,9 +35,12 @@ void log_cephsan(char *buf) {
     memcpy(tls->logs[head_idx].comm, current->comm, TASK_COMM_LEN);
 
     cephsan_pagefrag_free(pf, pre_len);
+    tls->logs[head_idx].len = 0;
 
     buf_idx = cephsan_pagefrag_alloc(pf, len);
     if (buf_idx) {
+		tls->head_idx = head_idx;
+		tls->histogram.counters[len >> 3]++;
 		tls->logs[head_idx].len = len;
         tls->logs[head_idx].buf = cephsan_pagefrag_get_ptr(pf, buf_idx);
 		memcpy(tls->logs[head_idx].buf, buf, len);
@@ -71,7 +74,6 @@ int cephsan_init(void)
 	struct ceph_san_tls_logger *tls;
 	struct cephsan_pagefrag *pf;
 
-	struct task_struct *task = current;
 	for_each_possible_cpu(cpu) {
 		tls = per_cpu_ptr(&ceph_san_tls, cpu);
 		tls->pages = alloc_pages(GFP_KERNEL, get_order(CEPH_SAN_MAX_LOGS * sizeof(struct ceph_san_log_entry)));