Push the asynchronous function call concept a little further
cover Push the asynchronous function call concept a little further
Hi,
the 2 patches in this series try to push the asynchronous function call concept a little further;
the first patch allows for creating multiple queues so that you can do synchronization on your own queue without creating congestion in the one global queue (for example, as used in the 2nd patch, a per superblock queue).
the second patch uses a per superblock queue to make the last part of file deletion (the actual removal of the file content from disk) using the async function call concept.
The cut is done after the inode is removed from the VFS (and where the code can and does sleep), but before calling into the filesystem to do the actual removal. The locking etc around this should be safe due to the sleep-ability and the fact that the inode is already out of the VFS namespaces (eg it's in I_FREEING).
On my testbox, rm -rf of a kernel tree went from 11.6 to 8.6 seconds with this; with smaller rm's I expect the gain to be more (to the point where applications that only occasionally delete files no longer will stutter around this as much).
Signature · 3 lines
Arjan van de Ven Intel Open Source Technology Centre For development, discussion and tips for power savings, visit http://www.lesswatts.org
PATCH 1/2 async: allow for multiple independent synchronization domains
Quoted 1 line
From
6353d3c8c9e11127961c2f366abc8ea0c7cb0e92Mon Sep 17 00:00:00 2001
From: Arjan van de Ven <[email protected]>
Date: Tue, 6 Jan 2009 07:19:33 -0800
Subject: [PATCH] async: allow for multiple independent synchronization domains
for a wider use of async function calls it's nice to have multiple domains for synchronization, for example, in the VFS world each superblock would be its own domain.
--- include/linux/async.h | 4 ++++ kernel/async.c | 49 +++++++++++++++++++++++++++++++++++-------------- 2 files changed, 39 insertions(+), 14 deletions(-)
include/linux/async.h +4 -0
@@ -11,11 +11,15 @@*/ #include <linux/types.h> +#include <linux/list.h> typedef u64 async_cookie_t ;typedef void (async_func_ptr ) (void *data,async_cookie_t cookie);extern async_cookie_t async_schedule (async_func_ptr *ptr,void *data); +extern async_cookie_t async_schedule_special (async_func_ptr *ptr,void *data,struct list_head *list);extern void async_synchronize_full (void ); +extern void async_synchronize_full_special (struct list_head *list);extern void async_synchronize_cookie (async_cookie_t cookie); +extern void async_synchronize_cookie_special (async_cookie_t cookie,struct list_head *list);
kernel/async.c +35 -14
121 lines not on this page · raw message
Signature · 8 lines
1.5.5.1 -- Arjan van de Ven Intel Open Source Technology Centre For development, discussion and tips for power savings, visit http://www.lesswatts.org
PATCH 2/2 async: make the final inode deletion an asynchronous event
In reply to Arjan van de VenQuoted 1 line
From
fc952f480bc4aec90e189b28196ecd7079d0b778Mon Sep 17 00:00:00 2001
From: Arjan van de Ven <[email protected]>
Date: Tue, 6 Jan 2009 07:20:54 -0800
Subject: [PATCH] async: make the final inode deletion an asynchronous event
this makes "rm -rf" on a (names cached) kernel tree go from 11.6 to 8.6 seconds on an ext3 filesystem
--- fs/inode.c | 20 +++++++++++++------- fs/super.c | 10 ++++++++++ include/linux/fs.h | 5 +++++ 3 files changed, 28 insertions(+), 7 deletions(-)
fs/inode.c +13 -7
@@ -22,6 +22,7 @@#include <linux/bootmem.h> #include <linux/inotify.h> #include <linux/mount.h> +#include <linux/async.h> /* * This is needed for the following functions:
@@ -1136,16 +1137,11 @@ EXPORT_SYMBOL(remove_inode_hash);* I_FREEING is set so that no-one will take a new reference to the inode while * it is being deleted. */ -void generic_delete_inode (struct inode *inode) +static void generic_delete_inode_async (void *data,async_cookie_t cookie) { +struct inode *inode = data;const struct super_operations *op = inode->i_sb->s_op; -list_del_init (&inode->i_list); -list_del_init (&inode->i_sb_list); - inode->i_state |=I_FREEING ; - inodes_stat.nr_inodes--; -spin_unlock (&inode_lock); -security_inode_delete (inode);if (op->delete_inode) {
@@ -1169,6 +1165,16 @@ void generic_delete_inode(struct inode *inode)destroy_inode (inode); } +void generic_delete_inode (struct inode *inode) +{ +list_del_init (&inode->i_list); +list_del_init (&inode->i_sb_list); + inode->i_state |=I_FREEING ; + inodes_stat.nr_inodes--; +spin_unlock (&inode_lock); +async_schedule_special (generic_delete_inode_async, inode, &inode->i_sb->s_async_list); +} +EXPORT_SYMBOL (generic_delete_inode);static void generic_forget_inode (struct inode *inode)
fs/super.c +10 -0
@@ -38,6 +38,7 @@#include <linux/kobject.h> #include <linux/mutex.h> #include <linux/file.h> +#include <linux/async.h> #include <asm/uaccess.h> #include "internal.h"
@@ -71,6 +72,7 @@ static struct super_block *alloc_super(struct file_system_type *type)INIT_HLIST_HEAD (&s->s_anon);INIT_LIST_HEAD (&s->s_inodes);INIT_LIST_HEAD (&s->s_dentry_lru); +INIT_LIST_HEAD (&s->s_async_list);init_rwsem (&s->s_umount);mutex_init (&s->s_lock);lockdep_set_class (&s->s_umount, &type->s_umount_key);
@@ -289,11 +291,18 @@ void generic_shutdown_super(struct super_block *sb) {const struct super_operations *sop = sb->s_op; +if (sb->s_root) {shrink_dcache_for_umount (sb);fsync_super (sb);lock_super (sb); sb->s_flags &= ~MS_ACTIVE ; + +/* +* wait for asynchronous fs operations to finish before going further +*/ +async_synchronize_full_special (&sb->s_async_list); +/* bad name - it should be evict_inodes() */ invalidate_inodes (sb);lock_kernel ();
@@ -449,6 +458,7 @@ void sync_filesystems(int wait)if (sb->s_flags &MS_RDONLY )continue ; sb->s_need_sync_fs =1 ; +async_synchronize_full_special (&sb->s_async_list); } restart:
include/linux/fs.h +5 -0
@@ -1185,6 +1185,11 @@ struct super_block {* generic_show_options() */ char *s_options; + +/* +* storage for asynchronous operations +*/ +struct list_head s_async_list; };extern struct timespec current_fs_time (struct super_block *sb);
Signature · 9 lines
1.5.5.1 -- Arjan van de Ven Intel Open Source Technology Centre For development, discussion and tips for power savings, visit http://www.lesswatts.org