staging: erofs: introduce erofs shrinker

This patch adds a dedicated shrinker targeting to free unneeded
memory consumed by a number of erofs in-memory data structures.

Like F2FS and UBIFS, it also adds:
  - sbi->umount_mutex to avoid races on shrinker and put_super
  - sbi->shrinker_run_no to not revisit recently scaned objects

Signed-off-by: Gao Xiang <gaoxiang25@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

diff --git a/drivers/staging/erofs/internal.h b/drivers/staging/erofs/internal.h
index 18e7b9c..42455f0 100644 (file)
--- a/drivers/staging/erofs/internal.h
+++ b/drivers/staging/erofs/internal.h
@@ -66,6 +66,7 @@ typedef u64 erofs_nid_t;
 struct erofs_sb_info {
        /* list for all registered superblocks, mainly for shrinker */
        struct list_head list;
+       struct mutex umount_mutex;
 
        u32 blocks;
        u32 meta_blkaddr;
@@ -93,6 +94,7 @@ struct erofs_sb_info {
        char *dev_name;
 
        unsigned int mount_opt;
+       unsigned int shrinker_run_no;
 
 #ifdef CONFIG_EROFS_FAULT_INJECTION
        struct erofs_fault_info fault_info;     /* For fault injection */
@@ -416,6 +418,11 @@ extern struct page *erofs_allocpage(struct list_head *pool, gfp_t gfp);
 extern void erofs_register_super(struct super_block *sb);
 extern void erofs_unregister_super(struct super_block *sb);
 
+extern unsigned long erofs_shrink_count(struct shrinker *shrink,
+       struct shrink_control *sc);
+extern unsigned long erofs_shrink_scan(struct shrinker *shrink,
+       struct shrink_control *sc);
+
 #ifndef lru_to_page
 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
 #endif

diff --git a/drivers/staging/erofs/super.c b/drivers/staging/erofs/super.c
index f455d71..ef85884 100644 (file)
--- a/drivers/staging/erofs/super.c
+++ b/drivers/staging/erofs/super.c
@@ -375,7 +375,9 @@ static void erofs_put_super(struct super_block *sb)
        infoln("unmounted for %s", sbi->dev_name);
        __putname(sbi->dev_name);
 
+       mutex_lock(&sbi->umount_mutex);
        erofs_unregister_super(sb);
+       mutex_unlock(&sbi->umount_mutex);
 
        kfree(sbi);
        sb->s_fs_info = NULL;
@@ -415,6 +417,12 @@ static void erofs_kill_sb(struct super_block *sb)
        kill_block_super(sb);
 }
 
+static struct shrinker erofs_shrinker_info = {
+       .scan_objects = erofs_shrink_scan,
+       .count_objects = erofs_shrink_count,
+       .seeks = DEFAULT_SEEKS,
+};
+
 static struct file_system_type erofs_fs_type = {
        .owner          = THIS_MODULE,
        .name           = "erofs",
@@ -435,6 +443,10 @@ static int __init erofs_module_init(void)
        if (err)
                goto icache_err;
 
+       err = register_shrinker(&erofs_shrinker_info);
+       if (err)
+               goto shrinker_err;
+
        err = register_filesystem(&erofs_fs_type);
        if (err)
                goto fs_err;
@@ -443,6 +455,8 @@ static int __init erofs_module_init(void)
        return 0;
 
 fs_err:
+       unregister_shrinker(&erofs_shrinker_info);
+shrinker_err:
        erofs_exit_inode_cache();
 icache_err:
        return err;
@@ -451,6 +465,7 @@ icache_err:
 static void __exit erofs_module_exit(void)
 {
        unregister_filesystem(&erofs_fs_type);
+       unregister_shrinker(&erofs_shrinker_info);
        erofs_exit_inode_cache();
        infoln("successfully finalize erofs");
 }

diff --git a/drivers/staging/erofs/utils.c b/drivers/staging/erofs/utils.c
index 6748def..c1d83ce 100644 (file)
--- a/drivers/staging/erofs/utils.c
+++ b/drivers/staging/erofs/utils.c
@@ -29,20 +29,93 @@ struct page *erofs_allocpage(struct list_head *pool, gfp_t gfp)
        return page;
 }
 
-static DEFINE_MUTEX(erofs_sb_list_lock);
+
+/* protected by 'erofs_sb_list_lock' */
+static unsigned int shrinker_run_no;
+
+/* protects the mounted 'erofs_sb_list' */
+static DEFINE_SPINLOCK(erofs_sb_list_lock);
 static LIST_HEAD(erofs_sb_list);
 
+/* global shrink count (for all mounted EROFS instances) */
+static atomic_long_t erofs_global_shrink_cnt;
+
 void erofs_register_super(struct super_block *sb)
 {
-       mutex_lock(&erofs_sb_list_lock);
-       list_add(&EROFS_SB(sb)->list, &erofs_sb_list);
-       mutex_unlock(&erofs_sb_list_lock);
+       struct erofs_sb_info *sbi = EROFS_SB(sb);
+
+       mutex_init(&sbi->umount_mutex);
+
+       spin_lock(&erofs_sb_list_lock);
+       list_add(&sbi->list, &erofs_sb_list);
+       spin_unlock(&erofs_sb_list_lock);
 }
 
 void erofs_unregister_super(struct super_block *sb)
 {
-       mutex_lock(&erofs_sb_list_lock);
+       spin_lock(&erofs_sb_list_lock);
        list_del(&EROFS_SB(sb)->list);
-       mutex_unlock(&erofs_sb_list_lock);
+       spin_unlock(&erofs_sb_list_lock);
+}
+
+unsigned long erofs_shrink_count(struct shrinker *shrink,
+                                struct shrink_control *sc)
+{
+       return atomic_long_read(&erofs_global_shrink_cnt);
+}
+
+unsigned long erofs_shrink_scan(struct shrinker *shrink,
+                               struct shrink_control *sc)
+{
+       struct erofs_sb_info *sbi;
+       struct list_head *p;
+
+       unsigned long nr = sc->nr_to_scan;
+       unsigned int run_no;
+       unsigned long freed = 0;
+
+       spin_lock(&erofs_sb_list_lock);
+       do
+               run_no = ++shrinker_run_no;
+       while (run_no == 0);
+
+       /* Iterate over all mounted superblocks and try to shrink them */
+       p = erofs_sb_list.next;
+       while (p != &erofs_sb_list) {
+               sbi = list_entry(p, struct erofs_sb_info, list);
+
+               /*
+                * We move the ones we do to the end of the list, so we stop
+                * when we see one we have already done.
+                */
+               if (sbi->shrinker_run_no == run_no)
+                       break;
+
+               if (!mutex_trylock(&sbi->umount_mutex)) {
+                       p = p->next;
+                       continue;
+               }
+
+               spin_unlock(&erofs_sb_list_lock);
+               sbi->shrinker_run_no = run_no;
+
+               /* add scan handlers here */
+
+               spin_lock(&erofs_sb_list_lock);
+               /* Get the next list element before we move this one */
+               p = p->next;
+
+               /*
+                * Move this one to the end of the list to provide some
+                * fairness.
+                */
+               list_move_tail(&sbi->list, &erofs_sb_list);
+               mutex_unlock(&sbi->umount_mutex);
+
+               if (freed >= nr)
+                       break;
+       }
+       spin_unlock(&erofs_sb_list_lock);
+       return freed;
 }