mm: introduce statistic for inode's gen&tier

Commit Message

zhaoyang.huang June 27, 2023, 9:17 a.m. UTC

From: Zhaoyang Huang <[e-mail]>

As mglru scale page's activity more presiced than before, I would like to
introduce statistics over these two properties on all pages of the inode, which
could help some mechanisms have ability to judge the inode's activity, etc madivse.

Signed-off-by: Zhaoyang Huang <[e-mail]>
---
 fs/proc/task_mmu.c        |  9 +++++++++
 include/linux/fs.h        |  2 ++
 include/linux/mm_inline.h | 14 ++++++++++++++
 mm/filemap.c              | 11 +++++++++++
 mm/swap.c                 |  1 +
 5 files changed, 37 insertions(+)

Comments

Christoph Hellwig June 28, 2023, 6:11 a.m. UTC | #1

On Tue, Jun 27, 2023 at 05:17:18PM +0800, zhaoyang.huang wrote:
> --- a/include/linux/fs.h
> +++ b/include/linux/fs.h
> @@ -434,6 +434,8 @@ struct address_space {
>  	struct rb_root_cached	i_mmap;
>  	struct rw_semaphore	i_mmap_rwsem;
>  	unsigned long		nrpages;
> +	atomic_long_t		gen;
> +	atomic_long_t		tier;

This increases the size of the inode by 16 byes, and better have really
good data supporting it, as increases of the inode size impact a lot
of workloads.

Patch

diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
index e35a039..3ed30ef 100644
--- a/fs/proc/task_mmu.c
+++ b/fs/proc/task_mmu.c
@@ -283,17 +283,26 @@  static void show_vma_header_prefix(struct seq_file *m,
 	unsigned long start, end;
 	dev_t dev = 0;
 	const char *name = NULL;
+	long nrpages = 0, gen = 0, tier = 0;
 
 	if (file) {
 		struct inode *inode = file_inode(vma->vm_file);
 		dev = inode->i_sb->s_dev;
 		ino = inode->i_ino;
 		pgoff = ((loff_t)vma->vm_pgoff) << PAGE_SHIFT;
+		nrpages = inode->i_mapping->nrpages;
+		gen = atomic_long_read(&inode->i_mapping->gen);
+		tier = atomic_long_read(&inode->i_mapping->tier);
 	}
 
 	start = vma->vm_start;
 	end = vma->vm_end;
 	show_vma_header_prefix(m, start, end, flags, pgoff, dev, ino);
+
+	seq_put_hex_ll(m, NULL, nrpages, 8);
+	seq_put_hex_ll(m, ":", gen, 8);
+	seq_put_hex_ll(m, ":", tier, 8);
+
 	if (mm)
 		anon_name = anon_vma_name(vma);
 
diff --git a/include/linux/fs.h b/include/linux/fs.h
index c1769a2..4f4c3a2 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -434,6 +434,8 @@  struct address_space {
 	struct rb_root_cached	i_mmap;
 	struct rw_semaphore	i_mmap_rwsem;
 	unsigned long		nrpages;
+	atomic_long_t		gen;
+	atomic_long_t		tier;
 	pgoff_t			writeback_index;
 	const struct address_space_operations *a_ops;
 	unsigned long		flags;
diff --git a/include/linux/mm_inline.h b/include/linux/mm_inline.h
index ff3f3f2..f68bd06 100644
--- a/include/linux/mm_inline.h
+++ b/include/linux/mm_inline.h
@@ -307,6 +307,20 @@  static inline bool lru_gen_del_folio(struct lruvec *lruvec, struct folio *folio,
 	return false;
 }
 
+static inline int lru_tier_from_refs(int refs)
+{
+	return 0;
+}
+
+static inline int folio_lru_refs(struct folio *folio)
+{
+	return 0;
+}
+
+static inline int folio_lru_gen(struct folio *folio)
+{
+	return 0;
+}
 #endif /* CONFIG_LRU_GEN */
 
 static __always_inline
diff --git a/mm/filemap.c b/mm/filemap.c
index c4d4ace..a1c68a9 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -45,6 +45,7 @@ 
 #include <asm/pgalloc.h>
 #include <asm/tlbflush.h>
 #include "internal.h"
+#include <linux/mm_inline.h>
 
 #define CREATE_TRACE_POINTS
 #include <trace/events/filemap.h>
@@ -126,6 +127,9 @@  static void page_cache_delete(struct address_space *mapping,
 {
 	XA_STATE(xas, &mapping->i_pages, folio->index);
 	long nr = 1;
+	int refs = folio_lru_refs(folio);
+	int tier = lru_tier_from_refs(refs);
+	int gen = folio_lru_gen(folio);
 
 	mapping_set_update(&xas, mapping);
 
@@ -143,6 +147,8 @@  static void page_cache_delete(struct address_space *mapping,
 	folio->mapping = NULL;
 	/* Leave page->index set: truncation lookup relies upon it */
 	mapping->nrpages -= nr;
+	atomic_long_sub(gen, &mapping->gen);
+	atomic_long_sub(tier, &mapping->tier);
 }
 
 static void filemap_unaccount_folio(struct address_space *mapping,
@@ -844,6 +850,9 @@  noinline int __filemap_add_folio(struct address_space *mapping,
 	int huge = folio_test_hugetlb(folio);
 	bool charged = false;
 	long nr = 1;
+	int refs = folio_lru_refs(folio);
+	int tier = lru_tier_from_refs(refs);
+	int gen = folio_lru_gen(folio);
 
 	VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
 	VM_BUG_ON_FOLIO(folio_test_swapbacked(folio), folio);
@@ -898,6 +907,8 @@  noinline int __filemap_add_folio(struct address_space *mapping,
 			goto unlock;
 
 		mapping->nrpages += nr;
+		atomic_long_add(gen, &mapping->gen);
+		atomic_long_add(tier, &mapping->tier);
 
 		/* hugetlb pages do not participate in page cache accounting */
 		if (!huge) {
diff --git a/mm/swap.c b/mm/swap.c
index 70e2063..6322c1c 100644
--- a/mm/swap.c
+++ b/mm/swap.c
@@ -468,6 +468,7 @@  static void folio_inc_refs(struct folio *folio)
 		new_flags += BIT(LRU_REFS_PGOFF);
 		new_flags |= old_flags & ~LRU_REFS_MASK;
 	} while (!try_cmpxchg(&folio->flags, &old_flags, new_flags));
+	atomic_long_inc(&folio->mapping->tier);
 }
 #else
 static void folio_inc_refs(struct folio *folio)