zram memory tracking

zRam as swap is useful for small memory device. However, swap means those pages on zram are mostly cold pages due to VM's LRU algorithm. Especially, once init data for application are touched for launching, they tend to be not accessed any more and finally swapped out. zRAM can store such cold pages as compressed form but it's pointless to keep in memory. Better idea is app developers free them directly rather than remaining them on heap.

This patch tell us last access time of each block of zram via "cat /sys/kernel/debug/zram/zram0/block_state".

The output is as follows,

   25064    .wh       100
   25065    s..       9
   25067    ..h       15

First column is zram's block index and second one represents symbol (s: same page w: written page to backing store h: huge page) of the block state. Last one means number of seconds elapsed since the block was last accessed. So above example means the 25065th block is accessed 100 second ago and it was huge so it was written to the backing store.

Admin can leverage this information to catch cold|incompressible pages of process with *pagemap* once part of heaps are swapped out.

+ default n

drivers/block/zram/zram_drv.c +83 -11

@@ -31,10 +31,15 @@
 #include <linux/err.h>
 #include <linux/idr.h>
 #include <linux/sysfs.h>
+#include <linux/debugfs.h>
 #include <linux/cpuhotplug.h>
 
 #include "zram_drv.h"
 
+#ifdef CONFIG_ZRAM_MEMORY_TRACKING
+static struct dentry *zram_debugfs_root;
+#endif
+
 static DEFINE_IDR(zram_index_idr);
 /* idr index must be protected */
 static DEFINE_MUTEX(zram_index_mutex);

@@ -67,6 +72,13 @@ static inline bool init_done(struct zram *zram)
 	return zram->disksize;
 }
 
+static inline bool zram_allocated(struct zram *zram, u32 index)
+{
+
+	return (zram->table[index].value >> (ZRAM_FLAG_SHIFT + 1)) ||
+					zram->table[index].handle;
+}
+
 static inline struct zram *dev_to_zram(struct device *dev)
 {
 	return (struct zram *)dev_to_disk(dev)->private_data;

@@ -83,7 +95,7 @@ static void zram_set_handle(struct zram *zram, u32 index, unsigned long handle)
 }
 
 /* flag operations require table entry bit_spin_lock() being held */
-static int zram_test_flag(struct zram *zram, u32 index,
+static bool zram_test_flag(struct zram *zram, u32 index,
 			enum zram_pageflags flag)
 {
 	return zram->table[index].value & BIT(flag);

@@ -107,16 +119,6 @@ static inline void zram_set_element(struct zram *zram, u32 index,
 	zram->table[index].element = element;
 }
 
-static void zram_accessed(struct zram *zram, u32 index)
-{
-	zram->table[index].ac_time = get_seconds();
-}
-
-static void zram_reset_access(struct zram *zram, u32 index)
-{
-	zram->table[index].ac_time = 0;
-}
-
 static unsigned long zram_get_element(struct zram *zram, u32 index)
 {
 	return zram->table[index].element;

@@ -620,6 +622,98 @@ static int read_from_bdev(struct zram *zram, struct bio_vec *bvec,
 static void zram_wb_clear(struct zram *zram, u32 index) {}
 #endif
 
+#ifdef CONFIG_ZRAM_MEMORY_TRACKING
+static void zram_accessed(struct zram *zram, u32 index)
+{
+	zram->table[index].ac_time = get_seconds();
+}
+
+static void zram_reset_access(struct zram *zram, u32 index)
+{
+	zram->table[index].ac_time = 0;
+}
+
+static ssize_t read_block_state(struct file *file, char __user *buf,
+				size_t count, loff_t *ppos)
+{
+	char *kbuf;
+	ssize_t index, written = 0;
+	struct zram *zram = file->private_data;
+	unsigned long last_access, nr_pages = zram->disksize >> PAGE_SHIFT;
+	char flags[__NR_ZRAM_PAGEFLAGS - ZRAM_FLAG_SHIFT];
+
+	kbuf = kvmalloc(count, GFP_KERNEL);
+	if (!kbuf)
+		return -ENOMEM;
+
+	down_read(&zram->init_lock);
+	if (!init_done(zram)) {
+		up_read(&zram->init_lock);
+		kvfree(kbuf);
+		return -EINVAL;
+	}
+
+	for (index = *ppos; index < nr_pages; index++) {
+		int copied;
+
+		zram_slot_lock(zram, index);
+		if (!zram_allocated(zram, index))
+			goto next;
+
+		snprintf(flags, sizeof(flags), "%c%c%c",
+			zram_test_flag(zram, index, ZRAM_SAME) ? 's' : '.',
+			zram_test_flag(zram, index, ZRAM_WB) ? 'w' : '.',
+			zram_test_flag(zram, index, ZRAM_HUGE) ? 'h' : '.');
+
+		last_access = get_seconds() - zram->table[index].ac_time;
+		copied = snprintf(kbuf + written, count, "%12lu %6s %12lu\n",
+						index, flags, last_access);
+		if (count < copied) {
+			zram_slot_unlock(zram, index);
+			break;
+		}
+		written += copied;
+		count -= copied;
+next:
+		zram_slot_unlock(zram, index);
+		*ppos += 1;
+	}
+
+	up_read(&zram->init_lock);
+	copy_to_user(buf, kbuf, written);
+	kvfree(kbuf);
+
+	return written;
+}
+
+static const struct file_operations proc_zram_block_state_op = {
+	.open = simple_open,
+	.read = read_block_state,
+	.llseek = default_llseek,
+};
+

Quoted 23 lines · Minchan Kim

zRam as swap is useful for small memory device. However, swap means those pages on zram are mostly cold pages due to VM's LRU algorithm. Especially, once init data for application are touched for launching, they tend to be not accessed any more and finally swapped out. zRAM can store such cold pages as compressed form but it's pointless to keep in memory. Better idea is app developers free them directly rather than remaining them on heap.

This patch tell us last access time of each block of zram via "cat /sys/kernel/debug/zram/zram0/block_state".

The output is as follows,

   25064    .wh       100
   25065    s..       9
   25067    ..h       15

First column is zram's block index and second one represents symbol (s: same page w: written page to backing store h: huge page) of the block state. Last one means number of seconds elapsed since the block was last accessed. So above example means the 25065th block is accessed 100 second ago and it was huge so it was written to the backing store.

Quoted 2 lines

Admin can leverage this information to catch cold|incompressible pages of process with *pagemap* once part of heaps are swapped out.

For debugging, yes, but remember that no system functionality should ever depend on debugfs :)

drivers/block/zram/Kconfig config ZRAM_WRITEBACK · line 33

Quoted 5 lines

drivers/block/zram/Kconfig +4 -0

+config ZRAM_MEMORY_TRACKING
+       bool "Tracking zram block status"
+       depends on ZRAM
+       select DEBUG_FS

depends on?

Quoted 1 line

+ default n

That's always the default, no need to list it.

drivers/block/zram/Kconfig config ZRAM_WRITEBACK · line 38

Quoted 5 lines

drivers/block/zram/Kconfig +4 -0

+       help
+	 With this feature, admin can track the state of allocated block
+	 of zRAM. Admin could see the information via
+	 /sys/kernel/debug/zram/zramX/block_state.

Odd indentation, please use tabs everywhere here.

Quoted 133 lines

drivers/block/zram/zram_drv.c +83 -11

@@ -31,10 +31,15 @@
 #include <linux/err.h>
 #include <linux/idr.h>
 #include <linux/sysfs.h>
+#include <linux/debugfs.h>
 #include <linux/cpuhotplug.h>
 
 #include "zram_drv.h"
 
+#ifdef CONFIG_ZRAM_MEMORY_TRACKING
+static struct dentry *zram_debugfs_root;
+#endif
+
 static DEFINE_IDR(zram_index_idr);
 /* idr index must be protected */
 static DEFINE_MUTEX(zram_index_mutex);

@@ -67,6 +72,13 @@ static inline bool init_done(struct zram *zram)
 	return zram->disksize;
 }
 
+static inline bool zram_allocated(struct zram *zram, u32 index)
+{
+
+	return (zram->table[index].value >> (ZRAM_FLAG_SHIFT + 1)) ||
+					zram->table[index].handle;
+}
+
 static inline struct zram *dev_to_zram(struct device *dev)
 {
 	return (struct zram *)dev_to_disk(dev)->private_data;

@@ -83,7 +95,7 @@ static void zram_set_handle(struct zram *zram, u32 index, unsigned long handle)
 }
 
 /* flag operations require table entry bit_spin_lock() being held */
-static int zram_test_flag(struct zram *zram, u32 index,
+static bool zram_test_flag(struct zram *zram, u32 index,
 			enum zram_pageflags flag)
 {
 	return zram->table[index].value & BIT(flag);

@@ -107,16 +119,6 @@ static inline void zram_set_element(struct zram *zram, u32 index,
 	zram->table[index].element = element;
 }
 
-static void zram_accessed(struct zram *zram, u32 index)
-{
-	zram->table[index].ac_time = get_seconds();
-}
-
-static void zram_reset_access(struct zram *zram, u32 index)
-{
-	zram->table[index].ac_time = 0;
-}
-
 static unsigned long zram_get_element(struct zram *zram, u32 index)
 {
 	return zram->table[index].element;

@@ -620,6 +622,98 @@ static int read_from_bdev(struct zram *zram, struct bio_vec *bvec,
 static void zram_wb_clear(struct zram *zram, u32 index) {}
 #endif
 
+#ifdef CONFIG_ZRAM_MEMORY_TRACKING
+static void zram_accessed(struct zram *zram, u32 index)
+{
+	zram->table[index].ac_time = get_seconds();
+}
+
+static void zram_reset_access(struct zram *zram, u32 index)
+{
+	zram->table[index].ac_time = 0;
+}
+
+static ssize_t read_block_state(struct file *file, char __user *buf,
+				size_t count, loff_t *ppos)
+{
+	char *kbuf;
+	ssize_t index, written = 0;
+	struct zram *zram = file->private_data;
+	unsigned long last_access, nr_pages = zram->disksize >> PAGE_SHIFT;
+	char flags[__NR_ZRAM_PAGEFLAGS - ZRAM_FLAG_SHIFT];
+
+	kbuf = kvmalloc(count, GFP_KERNEL);
+	if (!kbuf)
+		return -ENOMEM;
+
+	down_read(&zram->init_lock);
+	if (!init_done(zram)) {
+		up_read(&zram->init_lock);
+		kvfree(kbuf);
+		return -EINVAL;
+	}
+
+	for (index = *ppos; index < nr_pages; index++) {
+		int copied;
+
+		zram_slot_lock(zram, index);
+		if (!zram_allocated(zram, index))
+			goto next;
+
+		snprintf(flags, sizeof(flags), "%c%c%c",
+			zram_test_flag(zram, index, ZRAM_SAME) ? 's' : '.',
+			zram_test_flag(zram, index, ZRAM_WB) ? 'w' : '.',
+			zram_test_flag(zram, index, ZRAM_HUGE) ? 'h' : '.');
+
+		last_access = get_seconds() - zram->table[index].ac_time;
+		copied = snprintf(kbuf + written, count, "%12lu %6s %12lu\n",
+						index, flags, last_access);
+		if (count < copied) {
+			zram_slot_unlock(zram, index);
+			break;
+		}
+		written += copied;
+		count -= copied;
+next:
+		zram_slot_unlock(zram, index);
+		*ppos += 1;
+	}
+
+	up_read(&zram->init_lock);
+	copy_to_user(buf, kbuf, written);
+	kvfree(kbuf);
+
+	return written;
+}
+
+static const struct file_operations proc_zram_block_state_op = {
+	.open = simple_open,
+	.read = read_block_state,
+	.llseek = default_llseek,
+};
+

drivers/block/zram/zram_drv.c static int read_from_bdev(struct zram *zram, struct bio_vec *bvec, · line 704

Quoted 10 lines

drivers/block/zram/zram_drv.c +10 -0

+static void zram_debugfs_register(struct zram *zram)
+{
+	if (!zram_debugfs_root)
+		return;
+
+	zram->debugfs_dir = debugfs_create_dir(zram->disk->disk_name,
+						zram_debugfs_root);
+	debugfs_create_file("block_state", 0400, zram->debugfs_dir,
+				zram, &proc_zram_block_state_op);
+}

Much nicer code, thanks for the changes.

drivers/block/zram/zram_drv.c static int zram_remove_cb(int id, void *ptr, void *data) · line 1839

Quoted 6 lines

drivers/block/zram/zram_drv.c +3 -0

 static void destroy_devices(void)
 {
 	class_unregister(&zram_control_class);
+#ifdef CONFIG_ZRAM_MEMORY_TRACKING
+	debugfs_remove_recursive(zram_debugfs_root);
+#endif

No need for #ifdefs in .c code if your .h files are correct.

Quoted 7 lines

drivers/block/zram/zram_drv.c +0 -0

 	idr_for_each(&zram_index_idr, &zram_remove_cb, NULL);
 	idr_destroy(&zram_index_idr);
 	unregister_blkdev(zram_major, "zram");

@@ -1760,6 +1859,9 @@ static int __init zram_init(void)
 		return ret;
 	}
 

drivers/block/zram/zram_drv.c static int __init zram_init(void) · line 1864

Quoted 3 lines

drivers/block/zram/zram_drv.c +3 -0

+#ifdef CONFIG_ZRAM_MEMORY_TRACKING
+	zram_debugfs_root = debugfs_create_dir("zram", NULL);
+#endif

Same here. Put this in a wrapper function somewhere.