linux_dsm_epyc7002/fs/f2fs/gc.c

1126 lines
28 KiB
C
Raw Normal View History

/*
* fs/f2fs/gc.c
*
* Copyright (c) 2012 Samsung Electronics Co., Ltd.
* http://www.samsung.com/
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 as
* published by the Free Software Foundation.
*/
#include <linux/fs.h>
#include <linux/module.h>
#include <linux/backing-dev.h>
#include <linux/init.h>
#include <linux/f2fs_fs.h>
#include <linux/kthread.h>
#include <linux/delay.h>
#include <linux/freezer.h>
#include "f2fs.h"
#include "node.h"
#include "segment.h"
#include "gc.h"
#include <trace/events/f2fs.h>
static int gc_thread_func(void *data)
{
struct f2fs_sb_info *sbi = data;
struct f2fs_gc_kthread *gc_th = sbi->gc_thread;
wait_queue_head_t *wq = &sbi->gc_thread->gc_wait_queue_head;
unsigned int wait_ms;
wait_ms = gc_th->min_sleep_time;
set_freezable();
do {
wait_event_interruptible_timeout(*wq,
kthread_should_stop() || freezing(current) ||
gc_th->gc_wake,
msecs_to_jiffies(wait_ms));
/* give it a try one time */
if (gc_th->gc_wake)
gc_th->gc_wake = 0;
if (try_to_freeze())
continue;
if (kthread_should_stop())
break;
if (sbi->sb->s_writers.frozen >= SB_FREEZE_WRITE) {
increase_sleep_time(gc_th, &wait_ms);
continue;
}
#ifdef CONFIG_F2FS_FAULT_INJECTION
if (time_to_inject(sbi, FAULT_CHECKPOINT)) {
f2fs_show_injection_info(FAULT_CHECKPOINT);
f2fs_stop_checkpoint(sbi, false);
}
#endif
f2fs: make background threads of f2fs being aware of freezing When ->freeze_fs is called from lvm for doing snapshot, it needs to make sure there will be no more changes in filesystem's data, however, previously, background threads like GC thread wasn't aware of freezing, so in environment with active background threads, data of snapshot becomes unstable. This patch fixes this issue by adding sb_{start,end}_intwrite in below background threads: - GC thread - flush thread - discard thread Note that, don't use sb_start_intwrite() in gc_thread_func() due to: generic/241 reports below bug: ====================================================== WARNING: possible circular locking dependency detected 4.13.0-rc1+ #32 Tainted: G O ------------------------------------------------------ f2fs_gc-250:0/22186 is trying to acquire lock: (&sbi->gc_mutex){+.+...}, at: [<f8fa7f0b>] f2fs_sync_fs+0x7b/0x1b0 [f2fs] but task is already holding lock: (sb_internal#2){++++.-}, at: [<f8fb5609>] gc_thread_func+0x159/0x4a0 [f2fs] which lock already depends on the new lock. the existing dependency chain (in reverse order) is: -> #2 (sb_internal#2){++++.-}: __lock_acquire+0x405/0x7b0 lock_acquire+0xae/0x220 __sb_start_write+0x11d/0x1f0 f2fs_evict_inode+0x2d6/0x4e0 [f2fs] evict+0xa8/0x170 iput+0x1fb/0x2c0 f2fs_sync_inode_meta+0x3f/0xf0 [f2fs] write_checkpoint+0x1b1/0x750 [f2fs] f2fs_sync_fs+0x85/0x1b0 [f2fs] f2fs_do_sync_file.isra.24+0x137/0xa30 [f2fs] f2fs_sync_file+0x34/0x40 [f2fs] vfs_fsync_range+0x4a/0xa0 do_fsync+0x3c/0x60 SyS_fdatasync+0x15/0x20 do_fast_syscall_32+0xa1/0x1b0 entry_SYSENTER_32+0x4c/0x7b -> #1 (&sbi->cp_mutex){+.+...}: __lock_acquire+0x405/0x7b0 lock_acquire+0xae/0x220 __mutex_lock+0x4f/0x830 mutex_lock_nested+0x25/0x30 write_checkpoint+0x2f/0x750 [f2fs] f2fs_sync_fs+0x85/0x1b0 [f2fs] sync_filesystem+0x67/0x80 generic_shutdown_super+0x27/0x100 kill_block_super+0x22/0x50 kill_f2fs_super+0x3a/0x40 [f2fs] deactivate_locked_super+0x3d/0x70 deactivate_super+0x40/0x60 cleanup_mnt+0x39/0x70 __cleanup_mnt+0x10/0x20 task_work_run+0x69/0x80 exit_to_usermode_loop+0x57/0x92 do_fast_syscall_32+0x18c/0x1b0 entry_SYSENTER_32+0x4c/0x7b -> #0 (&sbi->gc_mutex){+.+...}: validate_chain.isra.36+0xc50/0xdb0 __lock_acquire+0x405/0x7b0 lock_acquire+0xae/0x220 __mutex_lock+0x4f/0x830 mutex_lock_nested+0x25/0x30 f2fs_sync_fs+0x7b/0x1b0 [f2fs] f2fs_balance_fs_bg+0xb9/0x200 [f2fs] gc_thread_func+0x302/0x4a0 [f2fs] kthread+0xe9/0x120 ret_from_fork+0x19/0x24 other info that might help us debug this: Chain exists of: &sbi->gc_mutex --> &sbi->cp_mutex --> sb_internal#2 Possible unsafe locking scenario: CPU0 CPU1 ---- ---- lock(sb_internal#2); lock(&sbi->cp_mutex); lock(sb_internal#2); lock(&sbi->gc_mutex); *** DEADLOCK *** 1 lock held by f2fs_gc-250:0/22186: #0: (sb_internal#2){++++.-}, at: [<f8fb5609>] gc_thread_func+0x159/0x4a0 [f2fs] stack backtrace: CPU: 2 PID: 22186 Comm: f2fs_gc-250:0 Tainted: G O 4.13.0-rc1+ #32 Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 Call Trace: dump_stack+0x5f/0x92 print_circular_bug+0x1b3/0x1bd validate_chain.isra.36+0xc50/0xdb0 ? __this_cpu_preempt_check+0xf/0x20 __lock_acquire+0x405/0x7b0 lock_acquire+0xae/0x220 ? f2fs_sync_fs+0x7b/0x1b0 [f2fs] __mutex_lock+0x4f/0x830 ? f2fs_sync_fs+0x7b/0x1b0 [f2fs] mutex_lock_nested+0x25/0x30 ? f2fs_sync_fs+0x7b/0x1b0 [f2fs] f2fs_sync_fs+0x7b/0x1b0 [f2fs] f2fs_balance_fs_bg+0xb9/0x200 [f2fs] gc_thread_func+0x302/0x4a0 [f2fs] ? preempt_schedule_common+0x2f/0x4d ? f2fs_gc+0x540/0x540 [f2fs] kthread+0xe9/0x120 ? f2fs_gc+0x540/0x540 [f2fs] ? kthread_create_on_node+0x30/0x30 ret_from_fork+0x19/0x24 The deadlock occurs in below condition: GC Thread Thread B - sb_start_intwrite - f2fs_sync_file - f2fs_sync_fs - mutex_lock(&sbi->gc_mutex) - write_checkpoint - block_operations - f2fs_sync_inode_meta - iput - sb_start_intwrite - mutex_lock(&sbi->gc_mutex) Fix this by altering sb_start_intwrite to sb_start_write_trylock. Signed-off-by: Chao Yu <yuchao0@huawei.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2017-07-22 07:52:23 +07:00
if (!sb_start_write_trylock(sbi->sb))
continue;
/*
* [GC triggering condition]
* 0. GC is not conducted currently.
* 1. There are enough dirty segments.
* 2. IO subsystem is idle by checking the # of writeback pages.
* 3. IO subsystem is idle by checking the # of requests in
* bdev's request list.
*
* Note) We have to avoid triggering GCs frequently.
* Because it is possible that some segments can be
* invalidated soon after by user update or deletion.
* So, I'd like to wait some time to collect dirty segments.
*/
if (gc_th->gc_urgent) {
wait_ms = gc_th->urgent_sleep_time;
mutex_lock(&sbi->gc_mutex);
goto do_gc;
}
if (!mutex_trylock(&sbi->gc_mutex))
goto next;
if (!is_idle(sbi)) {
increase_sleep_time(gc_th, &wait_ms);
mutex_unlock(&sbi->gc_mutex);
f2fs: make background threads of f2fs being aware of freezing When ->freeze_fs is called from lvm for doing snapshot, it needs to make sure there will be no more changes in filesystem's data, however, previously, background threads like GC thread wasn't aware of freezing, so in environment with active background threads, data of snapshot becomes unstable. This patch fixes this issue by adding sb_{start,end}_intwrite in below background threads: - GC thread - flush thread - discard thread Note that, don't use sb_start_intwrite() in gc_thread_func() due to: generic/241 reports below bug: ====================================================== WARNING: possible circular locking dependency detected 4.13.0-rc1+ #32 Tainted: G O ------------------------------------------------------ f2fs_gc-250:0/22186 is trying to acquire lock: (&sbi->gc_mutex){+.+...}, at: [<f8fa7f0b>] f2fs_sync_fs+0x7b/0x1b0 [f2fs] but task is already holding lock: (sb_internal#2){++++.-}, at: [<f8fb5609>] gc_thread_func+0x159/0x4a0 [f2fs] which lock already depends on the new lock. the existing dependency chain (in reverse order) is: -> #2 (sb_internal#2){++++.-}: __lock_acquire+0x405/0x7b0 lock_acquire+0xae/0x220 __sb_start_write+0x11d/0x1f0 f2fs_evict_inode+0x2d6/0x4e0 [f2fs] evict+0xa8/0x170 iput+0x1fb/0x2c0 f2fs_sync_inode_meta+0x3f/0xf0 [f2fs] write_checkpoint+0x1b1/0x750 [f2fs] f2fs_sync_fs+0x85/0x1b0 [f2fs] f2fs_do_sync_file.isra.24+0x137/0xa30 [f2fs] f2fs_sync_file+0x34/0x40 [f2fs] vfs_fsync_range+0x4a/0xa0 do_fsync+0x3c/0x60 SyS_fdatasync+0x15/0x20 do_fast_syscall_32+0xa1/0x1b0 entry_SYSENTER_32+0x4c/0x7b -> #1 (&sbi->cp_mutex){+.+...}: __lock_acquire+0x405/0x7b0 lock_acquire+0xae/0x220 __mutex_lock+0x4f/0x830 mutex_lock_nested+0x25/0x30 write_checkpoint+0x2f/0x750 [f2fs] f2fs_sync_fs+0x85/0x1b0 [f2fs] sync_filesystem+0x67/0x80 generic_shutdown_super+0x27/0x100 kill_block_super+0x22/0x50 kill_f2fs_super+0x3a/0x40 [f2fs] deactivate_locked_super+0x3d/0x70 deactivate_super+0x40/0x60 cleanup_mnt+0x39/0x70 __cleanup_mnt+0x10/0x20 task_work_run+0x69/0x80 exit_to_usermode_loop+0x57/0x92 do_fast_syscall_32+0x18c/0x1b0 entry_SYSENTER_32+0x4c/0x7b -> #0 (&sbi->gc_mutex){+.+...}: validate_chain.isra.36+0xc50/0xdb0 __lock_acquire+0x405/0x7b0 lock_acquire+0xae/0x220 __mutex_lock+0x4f/0x830 mutex_lock_nested+0x25/0x30 f2fs_sync_fs+0x7b/0x1b0 [f2fs] f2fs_balance_fs_bg+0xb9/0x200 [f2fs] gc_thread_func+0x302/0x4a0 [f2fs] kthread+0xe9/0x120 ret_from_fork+0x19/0x24 other info that might help us debug this: Chain exists of: &sbi->gc_mutex --> &sbi->cp_mutex --> sb_internal#2 Possible unsafe locking scenario: CPU0 CPU1 ---- ---- lock(sb_internal#2); lock(&sbi->cp_mutex); lock(sb_internal#2); lock(&sbi->gc_mutex); *** DEADLOCK *** 1 lock held by f2fs_gc-250:0/22186: #0: (sb_internal#2){++++.-}, at: [<f8fb5609>] gc_thread_func+0x159/0x4a0 [f2fs] stack backtrace: CPU: 2 PID: 22186 Comm: f2fs_gc-250:0 Tainted: G O 4.13.0-rc1+ #32 Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 Call Trace: dump_stack+0x5f/0x92 print_circular_bug+0x1b3/0x1bd validate_chain.isra.36+0xc50/0xdb0 ? __this_cpu_preempt_check+0xf/0x20 __lock_acquire+0x405/0x7b0 lock_acquire+0xae/0x220 ? f2fs_sync_fs+0x7b/0x1b0 [f2fs] __mutex_lock+0x4f/0x830 ? f2fs_sync_fs+0x7b/0x1b0 [f2fs] mutex_lock_nested+0x25/0x30 ? f2fs_sync_fs+0x7b/0x1b0 [f2fs] f2fs_sync_fs+0x7b/0x1b0 [f2fs] f2fs_balance_fs_bg+0xb9/0x200 [f2fs] gc_thread_func+0x302/0x4a0 [f2fs] ? preempt_schedule_common+0x2f/0x4d ? f2fs_gc+0x540/0x540 [f2fs] kthread+0xe9/0x120 ? f2fs_gc+0x540/0x540 [f2fs] ? kthread_create_on_node+0x30/0x30 ret_from_fork+0x19/0x24 The deadlock occurs in below condition: GC Thread Thread B - sb_start_intwrite - f2fs_sync_file - f2fs_sync_fs - mutex_lock(&sbi->gc_mutex) - write_checkpoint - block_operations - f2fs_sync_inode_meta - iput - sb_start_intwrite - mutex_lock(&sbi->gc_mutex) Fix this by altering sb_start_intwrite to sb_start_write_trylock. Signed-off-by: Chao Yu <yuchao0@huawei.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2017-07-22 07:52:23 +07:00
goto next;
}
if (has_enough_invalid_blocks(sbi))
decrease_sleep_time(gc_th, &wait_ms);
else
increase_sleep_time(gc_th, &wait_ms);
do_gc:
stat_inc_bggc_count(sbi);
/* if return value is not zero, no victim was selected */
if (f2fs_gc(sbi, test_opt(sbi, FORCE_FG_GC), true, NULL_SEGNO))
wait_ms = gc_th->no_gc_sleep_time;
trace_f2fs_background_gc(sbi->sb, wait_ms,
prefree_segments(sbi), free_segments(sbi));
/* balancing f2fs's metadata periodically */
f2fs_balance_fs_bg(sbi);
f2fs: make background threads of f2fs being aware of freezing When ->freeze_fs is called from lvm for doing snapshot, it needs to make sure there will be no more changes in filesystem's data, however, previously, background threads like GC thread wasn't aware of freezing, so in environment with active background threads, data of snapshot becomes unstable. This patch fixes this issue by adding sb_{start,end}_intwrite in below background threads: - GC thread - flush thread - discard thread Note that, don't use sb_start_intwrite() in gc_thread_func() due to: generic/241 reports below bug: ====================================================== WARNING: possible circular locking dependency detected 4.13.0-rc1+ #32 Tainted: G O ------------------------------------------------------ f2fs_gc-250:0/22186 is trying to acquire lock: (&sbi->gc_mutex){+.+...}, at: [<f8fa7f0b>] f2fs_sync_fs+0x7b/0x1b0 [f2fs] but task is already holding lock: (sb_internal#2){++++.-}, at: [<f8fb5609>] gc_thread_func+0x159/0x4a0 [f2fs] which lock already depends on the new lock. the existing dependency chain (in reverse order) is: -> #2 (sb_internal#2){++++.-}: __lock_acquire+0x405/0x7b0 lock_acquire+0xae/0x220 __sb_start_write+0x11d/0x1f0 f2fs_evict_inode+0x2d6/0x4e0 [f2fs] evict+0xa8/0x170 iput+0x1fb/0x2c0 f2fs_sync_inode_meta+0x3f/0xf0 [f2fs] write_checkpoint+0x1b1/0x750 [f2fs] f2fs_sync_fs+0x85/0x1b0 [f2fs] f2fs_do_sync_file.isra.24+0x137/0xa30 [f2fs] f2fs_sync_file+0x34/0x40 [f2fs] vfs_fsync_range+0x4a/0xa0 do_fsync+0x3c/0x60 SyS_fdatasync+0x15/0x20 do_fast_syscall_32+0xa1/0x1b0 entry_SYSENTER_32+0x4c/0x7b -> #1 (&sbi->cp_mutex){+.+...}: __lock_acquire+0x405/0x7b0 lock_acquire+0xae/0x220 __mutex_lock+0x4f/0x830 mutex_lock_nested+0x25/0x30 write_checkpoint+0x2f/0x750 [f2fs] f2fs_sync_fs+0x85/0x1b0 [f2fs] sync_filesystem+0x67/0x80 generic_shutdown_super+0x27/0x100 kill_block_super+0x22/0x50 kill_f2fs_super+0x3a/0x40 [f2fs] deactivate_locked_super+0x3d/0x70 deactivate_super+0x40/0x60 cleanup_mnt+0x39/0x70 __cleanup_mnt+0x10/0x20 task_work_run+0x69/0x80 exit_to_usermode_loop+0x57/0x92 do_fast_syscall_32+0x18c/0x1b0 entry_SYSENTER_32+0x4c/0x7b -> #0 (&sbi->gc_mutex){+.+...}: validate_chain.isra.36+0xc50/0xdb0 __lock_acquire+0x405/0x7b0 lock_acquire+0xae/0x220 __mutex_lock+0x4f/0x830 mutex_lock_nested+0x25/0x30 f2fs_sync_fs+0x7b/0x1b0 [f2fs] f2fs_balance_fs_bg+0xb9/0x200 [f2fs] gc_thread_func+0x302/0x4a0 [f2fs] kthread+0xe9/0x120 ret_from_fork+0x19/0x24 other info that might help us debug this: Chain exists of: &sbi->gc_mutex --> &sbi->cp_mutex --> sb_internal#2 Possible unsafe locking scenario: CPU0 CPU1 ---- ---- lock(sb_internal#2); lock(&sbi->cp_mutex); lock(sb_internal#2); lock(&sbi->gc_mutex); *** DEADLOCK *** 1 lock held by f2fs_gc-250:0/22186: #0: (sb_internal#2){++++.-}, at: [<f8fb5609>] gc_thread_func+0x159/0x4a0 [f2fs] stack backtrace: CPU: 2 PID: 22186 Comm: f2fs_gc-250:0 Tainted: G O 4.13.0-rc1+ #32 Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 Call Trace: dump_stack+0x5f/0x92 print_circular_bug+0x1b3/0x1bd validate_chain.isra.36+0xc50/0xdb0 ? __this_cpu_preempt_check+0xf/0x20 __lock_acquire+0x405/0x7b0 lock_acquire+0xae/0x220 ? f2fs_sync_fs+0x7b/0x1b0 [f2fs] __mutex_lock+0x4f/0x830 ? f2fs_sync_fs+0x7b/0x1b0 [f2fs] mutex_lock_nested+0x25/0x30 ? f2fs_sync_fs+0x7b/0x1b0 [f2fs] f2fs_sync_fs+0x7b/0x1b0 [f2fs] f2fs_balance_fs_bg+0xb9/0x200 [f2fs] gc_thread_func+0x302/0x4a0 [f2fs] ? preempt_schedule_common+0x2f/0x4d ? f2fs_gc+0x540/0x540 [f2fs] kthread+0xe9/0x120 ? f2fs_gc+0x540/0x540 [f2fs] ? kthread_create_on_node+0x30/0x30 ret_from_fork+0x19/0x24 The deadlock occurs in below condition: GC Thread Thread B - sb_start_intwrite - f2fs_sync_file - f2fs_sync_fs - mutex_lock(&sbi->gc_mutex) - write_checkpoint - block_operations - f2fs_sync_inode_meta - iput - sb_start_intwrite - mutex_lock(&sbi->gc_mutex) Fix this by altering sb_start_intwrite to sb_start_write_trylock. Signed-off-by: Chao Yu <yuchao0@huawei.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2017-07-22 07:52:23 +07:00
next:
sb_end_write(sbi->sb);
} while (!kthread_should_stop());
return 0;
}
int start_gc_thread(struct f2fs_sb_info *sbi)
{
struct f2fs_gc_kthread *gc_th;
dev_t dev = sbi->sb->s_bdev->bd_dev;
int err = 0;
gc_th = f2fs_kmalloc(sbi, sizeof(struct f2fs_gc_kthread), GFP_KERNEL);
if (!gc_th) {
err = -ENOMEM;
goto out;
}
gc_th->urgent_sleep_time = DEF_GC_THREAD_URGENT_SLEEP_TIME;
gc_th->min_sleep_time = DEF_GC_THREAD_MIN_SLEEP_TIME;
gc_th->max_sleep_time = DEF_GC_THREAD_MAX_SLEEP_TIME;
gc_th->no_gc_sleep_time = DEF_GC_THREAD_NOGC_SLEEP_TIME;
gc_th->gc_idle = 0;
gc_th->gc_urgent = 0;
gc_th->gc_wake= 0;
sbi->gc_thread = gc_th;
init_waitqueue_head(&sbi->gc_thread->gc_wait_queue_head);
sbi->gc_thread->f2fs_gc_task = kthread_run(gc_thread_func, sbi,
"f2fs_gc-%u:%u", MAJOR(dev), MINOR(dev));
if (IS_ERR(gc_th->f2fs_gc_task)) {
err = PTR_ERR(gc_th->f2fs_gc_task);
kfree(gc_th);
sbi->gc_thread = NULL;
}
out:
return err;
}
void stop_gc_thread(struct f2fs_sb_info *sbi)
{
struct f2fs_gc_kthread *gc_th = sbi->gc_thread;
if (!gc_th)
return;
kthread_stop(gc_th->f2fs_gc_task);
kfree(gc_th);
sbi->gc_thread = NULL;
}
static int select_gc_type(struct f2fs_gc_kthread *gc_th, int gc_type)
{
int gc_mode = (gc_type == BG_GC) ? GC_CB : GC_GREEDY;
if (!gc_th)
return gc_mode;
if (gc_th->gc_idle) {
if (gc_th->gc_idle == 1)
gc_mode = GC_CB;
else if (gc_th->gc_idle == 2)
gc_mode = GC_GREEDY;
}
if (gc_th->gc_urgent)
gc_mode = GC_GREEDY;
return gc_mode;
}
static void select_policy(struct f2fs_sb_info *sbi, int gc_type,
int type, struct victim_sel_policy *p)
{
struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
if (p->alloc_mode == SSR) {
p->gc_mode = GC_GREEDY;
p->dirty_segmap = dirty_i->dirty_segmap[type];
f2fs: optimize gc for better performance This patch improves the gc efficiency by optimizing the victim selection policy. With this optimization, the random re-write performance could increase up to 20%. For f2fs, when disk is in shortage of free spaces, gc will selects dirty segments and moves valid blocks around for making more space available. The gc cost of a segment is determined by the valid blocks in the segment. The less the valid blocks, the higher the efficiency. The ideal victim segment is the one that has the most garbage blocks. Currently, it searches up to 20 dirty segments for a victim segment. The selected victim is not likely the best victim for gc when there are much more dirty segments. Why not searching more dirty segments for a better victim? The cost of searching dirty segments is negligible in comparison to moving blocks. In this patch, it enlarges the MAX_VICTIM_SEARCH to 4096 to make the search more aggressively for a possible better victim. Since it also applies to victim selection for SSR, it will likely improve the SSR efficiency as well. The test case is simple. It creates as many files until the disk full. The size for each file is 32KB. Then it writes as many as 100000 records of 4KB size to random offsets of random files in sync mode. The testing was done on a 2GB partition of a SDHC card. Let's see the test result of f2fs without and with the patch. --------------------------------------- 2GB partition, SDHC create 52023 files of size 32768 bytes random re-write 100000 records of 4KB --------------------------------------- | file creation (s) | rewrite time (s) | gc count | gc garbage blocks | [no patch] 341 4227 1174 174840 [patched] 324 2958 645 106682 It's obvious that, with the patch, f2fs finishes the test in 20+% less time than without the patch. And internally it does much less gc with higher efficiency than before. Since the performance improvement is related to gc, it might not be so obvious for other tests that do not trigger gc as often as this one ( This is because f2fs selects dirty segments for SSR use most of the time when free space is in shortage). The well-known iozone test tool was not used for benchmarking the patch becuase it seems do not have a test case that performs random re-write on a full disk. This patch is the revised version based on the suggestion from Jaegeuk Kim. Signed-off-by: Jin Xu <jinuxstyle@gmail.com> [Jaegeuk Kim: suggested simpler solution] Reviewed-by: Jaegeuk Kim <jaegeuk.kim@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2013-09-05 11:45:26 +07:00
p->max_search = dirty_i->nr_dirty[type];
p->ofs_unit = 1;
} else {
p->gc_mode = select_gc_type(sbi->gc_thread, gc_type);
p->dirty_segmap = dirty_i->dirty_segmap[DIRTY];
f2fs: optimize gc for better performance This patch improves the gc efficiency by optimizing the victim selection policy. With this optimization, the random re-write performance could increase up to 20%. For f2fs, when disk is in shortage of free spaces, gc will selects dirty segments and moves valid blocks around for making more space available. The gc cost of a segment is determined by the valid blocks in the segment. The less the valid blocks, the higher the efficiency. The ideal victim segment is the one that has the most garbage blocks. Currently, it searches up to 20 dirty segments for a victim segment. The selected victim is not likely the best victim for gc when there are much more dirty segments. Why not searching more dirty segments for a better victim? The cost of searching dirty segments is negligible in comparison to moving blocks. In this patch, it enlarges the MAX_VICTIM_SEARCH to 4096 to make the search more aggressively for a possible better victim. Since it also applies to victim selection for SSR, it will likely improve the SSR efficiency as well. The test case is simple. It creates as many files until the disk full. The size for each file is 32KB. Then it writes as many as 100000 records of 4KB size to random offsets of random files in sync mode. The testing was done on a 2GB partition of a SDHC card. Let's see the test result of f2fs without and with the patch. --------------------------------------- 2GB partition, SDHC create 52023 files of size 32768 bytes random re-write 100000 records of 4KB --------------------------------------- | file creation (s) | rewrite time (s) | gc count | gc garbage blocks | [no patch] 341 4227 1174 174840 [patched] 324 2958 645 106682 It's obvious that, with the patch, f2fs finishes the test in 20+% less time than without the patch. And internally it does much less gc with higher efficiency than before. Since the performance improvement is related to gc, it might not be so obvious for other tests that do not trigger gc as often as this one ( This is because f2fs selects dirty segments for SSR use most of the time when free space is in shortage). The well-known iozone test tool was not used for benchmarking the patch becuase it seems do not have a test case that performs random re-write on a full disk. This patch is the revised version based on the suggestion from Jaegeuk Kim. Signed-off-by: Jin Xu <jinuxstyle@gmail.com> [Jaegeuk Kim: suggested simpler solution] Reviewed-by: Jaegeuk Kim <jaegeuk.kim@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2013-09-05 11:45:26 +07:00
p->max_search = dirty_i->nr_dirty[DIRTY];
p->ofs_unit = sbi->segs_per_sec;
}
f2fs: optimize gc for better performance This patch improves the gc efficiency by optimizing the victim selection policy. With this optimization, the random re-write performance could increase up to 20%. For f2fs, when disk is in shortage of free spaces, gc will selects dirty segments and moves valid blocks around for making more space available. The gc cost of a segment is determined by the valid blocks in the segment. The less the valid blocks, the higher the efficiency. The ideal victim segment is the one that has the most garbage blocks. Currently, it searches up to 20 dirty segments for a victim segment. The selected victim is not likely the best victim for gc when there are much more dirty segments. Why not searching more dirty segments for a better victim? The cost of searching dirty segments is negligible in comparison to moving blocks. In this patch, it enlarges the MAX_VICTIM_SEARCH to 4096 to make the search more aggressively for a possible better victim. Since it also applies to victim selection for SSR, it will likely improve the SSR efficiency as well. The test case is simple. It creates as many files until the disk full. The size for each file is 32KB. Then it writes as many as 100000 records of 4KB size to random offsets of random files in sync mode. The testing was done on a 2GB partition of a SDHC card. Let's see the test result of f2fs without and with the patch. --------------------------------------- 2GB partition, SDHC create 52023 files of size 32768 bytes random re-write 100000 records of 4KB --------------------------------------- | file creation (s) | rewrite time (s) | gc count | gc garbage blocks | [no patch] 341 4227 1174 174840 [patched] 324 2958 645 106682 It's obvious that, with the patch, f2fs finishes the test in 20+% less time than without the patch. And internally it does much less gc with higher efficiency than before. Since the performance improvement is related to gc, it might not be so obvious for other tests that do not trigger gc as often as this one ( This is because f2fs selects dirty segments for SSR use most of the time when free space is in shortage). The well-known iozone test tool was not used for benchmarking the patch becuase it seems do not have a test case that performs random re-write on a full disk. This patch is the revised version based on the suggestion from Jaegeuk Kim. Signed-off-by: Jin Xu <jinuxstyle@gmail.com> [Jaegeuk Kim: suggested simpler solution] Reviewed-by: Jaegeuk Kim <jaegeuk.kim@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2013-09-05 11:45:26 +07:00
/* we need to check every dirty segments in the FG_GC case */
if (gc_type != FG_GC &&
(sbi->gc_thread && !sbi->gc_thread->gc_urgent) &&
p->max_search > sbi->max_victim_search)
p->max_search = sbi->max_victim_search;
f2fs: optimize gc for better performance This patch improves the gc efficiency by optimizing the victim selection policy. With this optimization, the random re-write performance could increase up to 20%. For f2fs, when disk is in shortage of free spaces, gc will selects dirty segments and moves valid blocks around for making more space available. The gc cost of a segment is determined by the valid blocks in the segment. The less the valid blocks, the higher the efficiency. The ideal victim segment is the one that has the most garbage blocks. Currently, it searches up to 20 dirty segments for a victim segment. The selected victim is not likely the best victim for gc when there are much more dirty segments. Why not searching more dirty segments for a better victim? The cost of searching dirty segments is negligible in comparison to moving blocks. In this patch, it enlarges the MAX_VICTIM_SEARCH to 4096 to make the search more aggressively for a possible better victim. Since it also applies to victim selection for SSR, it will likely improve the SSR efficiency as well. The test case is simple. It creates as many files until the disk full. The size for each file is 32KB. Then it writes as many as 100000 records of 4KB size to random offsets of random files in sync mode. The testing was done on a 2GB partition of a SDHC card. Let's see the test result of f2fs without and with the patch. --------------------------------------- 2GB partition, SDHC create 52023 files of size 32768 bytes random re-write 100000 records of 4KB --------------------------------------- | file creation (s) | rewrite time (s) | gc count | gc garbage blocks | [no patch] 341 4227 1174 174840 [patched] 324 2958 645 106682 It's obvious that, with the patch, f2fs finishes the test in 20+% less time than without the patch. And internally it does much less gc with higher efficiency than before. Since the performance improvement is related to gc, it might not be so obvious for other tests that do not trigger gc as often as this one ( This is because f2fs selects dirty segments for SSR use most of the time when free space is in shortage). The well-known iozone test tool was not used for benchmarking the patch becuase it seems do not have a test case that performs random re-write on a full disk. This patch is the revised version based on the suggestion from Jaegeuk Kim. Signed-off-by: Jin Xu <jinuxstyle@gmail.com> [Jaegeuk Kim: suggested simpler solution] Reviewed-by: Jaegeuk Kim <jaegeuk.kim@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2013-09-05 11:45:26 +07:00
/* let's select beginning hot/small space first in no_heap mode*/
if (test_opt(sbi, NOHEAP) &&
(type == CURSEG_HOT_DATA || IS_NODESEG(type)))
p->offset = 0;
else
p->offset = SIT_I(sbi)->last_victim[p->gc_mode];
}
static unsigned int get_max_cost(struct f2fs_sb_info *sbi,
struct victim_sel_policy *p)
{
/* SSR allocates in a segment unit */
if (p->alloc_mode == SSR)
return sbi->blocks_per_seg;
if (p->gc_mode == GC_GREEDY)
return 2 * sbi->blocks_per_seg * p->ofs_unit;
else if (p->gc_mode == GC_CB)
return UINT_MAX;
else /* No other gc_mode */
return 0;
}
static unsigned int check_bg_victims(struct f2fs_sb_info *sbi)
{
struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
unsigned int secno;
/*
* If the gc_type is FG_GC, we can select victim segments
* selected by background GC before.
* Those segments guarantee they have small valid blocks.
*/
for_each_set_bit(secno, dirty_i->victim_secmap, MAIN_SECS(sbi)) {
if (sec_usage_check(sbi, secno))
continue;
if (no_fggc_candidate(sbi, secno))
continue;
clear_bit(secno, dirty_i->victim_secmap);
return GET_SEG_FROM_SEC(sbi, secno);
}
return NULL_SEGNO;
}
static unsigned int get_cb_cost(struct f2fs_sb_info *sbi, unsigned int segno)
{
struct sit_info *sit_i = SIT_I(sbi);
unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
unsigned int start = GET_SEG_FROM_SEC(sbi, secno);
unsigned long long mtime = 0;
unsigned int vblocks;
unsigned char age = 0;
unsigned char u;
unsigned int i;
for (i = 0; i < sbi->segs_per_sec; i++)
mtime += get_seg_entry(sbi, start + i)->mtime;
vblocks = get_valid_blocks(sbi, segno, true);
mtime = div_u64(mtime, sbi->segs_per_sec);
vblocks = div_u64(vblocks, sbi->segs_per_sec);
u = (vblocks * 100) >> sbi->log_blocks_per_seg;
/* Handle if the system time has changed by the user */
if (mtime < sit_i->min_mtime)
sit_i->min_mtime = mtime;
if (mtime > sit_i->max_mtime)
sit_i->max_mtime = mtime;
if (sit_i->max_mtime != sit_i->min_mtime)
age = 100 - div64_u64(100 * (mtime - sit_i->min_mtime),
sit_i->max_mtime - sit_i->min_mtime);
return UINT_MAX - ((100 * (100 - u) * age) / (100 + u));
}
static inline unsigned int get_gc_cost(struct f2fs_sb_info *sbi,
unsigned int segno, struct victim_sel_policy *p)
{
if (p->alloc_mode == SSR)
return get_seg_entry(sbi, segno)->ckpt_valid_blocks;
/* alloc_mode == LFS */
if (p->gc_mode == GC_GREEDY)
return get_valid_blocks(sbi, segno, true);
else
return get_cb_cost(sbi, segno);
}
static unsigned int count_bits(const unsigned long *addr,
unsigned int offset, unsigned int len)
{
unsigned int end = offset + len, sum = 0;
while (offset < end) {
if (test_bit(offset++, addr))
++sum;
}
return sum;
}
/*
* This function is called from two paths.
* One is garbage collection and the other is SSR segment selection.
* When it is called during GC, it just gets a victim segment
* and it does not remove it from dirty seglist.
* When it is called from SSR segment selection, it finds a segment
* which has minimum valid blocks and removes it from dirty seglist.
*/
static int get_victim_by_default(struct f2fs_sb_info *sbi,
unsigned int *result, int gc_type, int type, char alloc_mode)
{
struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
struct sit_info *sm = SIT_I(sbi);
struct victim_sel_policy p;
unsigned int secno, last_victim;
unsigned int last_segment = MAIN_SEGS(sbi);
unsigned int nsearched = 0;
mutex_lock(&dirty_i->seglist_lock);
p.alloc_mode = alloc_mode;
select_policy(sbi, gc_type, type, &p);
p.min_segno = NULL_SEGNO;
p.min_cost = get_max_cost(sbi, &p);
if (*result != NULL_SEGNO) {
if (IS_DATASEG(get_seg_entry(sbi, *result)->type) &&
get_valid_blocks(sbi, *result, false) &&
!sec_usage_check(sbi, GET_SEC_FROM_SEG(sbi, *result)))
p.min_segno = *result;
goto out;
}
if (p.max_search == 0)
goto out;
last_victim = sm->last_victim[p.gc_mode];
if (p.alloc_mode == LFS && gc_type == FG_GC) {
p.min_segno = check_bg_victims(sbi);
if (p.min_segno != NULL_SEGNO)
goto got_it;
}
while (1) {
unsigned long cost;
unsigned int segno;
segno = find_next_bit(p.dirty_segmap, last_segment, p.offset);
if (segno >= last_segment) {
if (sm->last_victim[p.gc_mode]) {
last_segment =
sm->last_victim[p.gc_mode];
sm->last_victim[p.gc_mode] = 0;
p.offset = 0;
continue;
}
break;
}
p.offset = segno + p.ofs_unit;
if (p.ofs_unit > 1) {
p.offset -= segno % p.ofs_unit;
nsearched += count_bits(p.dirty_segmap,
p.offset - p.ofs_unit,
p.ofs_unit);
} else {
nsearched++;
}
secno = GET_SEC_FROM_SEG(sbi, segno);
if (sec_usage_check(sbi, secno))
goto next;
if (gc_type == BG_GC && test_bit(secno, dirty_i->victim_secmap))
goto next;
if (gc_type == FG_GC && p.alloc_mode == LFS &&
no_fggc_candidate(sbi, secno))
goto next;
cost = get_gc_cost(sbi, segno, &p);
if (p.min_cost > cost) {
p.min_segno = segno;
p.min_cost = cost;
}
next:
if (nsearched >= p.max_search) {
if (!sm->last_victim[p.gc_mode] && segno <= last_victim)
sm->last_victim[p.gc_mode] = last_victim + 1;
else
sm->last_victim[p.gc_mode] = segno + 1;
sm->last_victim[p.gc_mode] %= MAIN_SEGS(sbi);
break;
}
}
if (p.min_segno != NULL_SEGNO) {
got_it:
if (p.alloc_mode == LFS) {
secno = GET_SEC_FROM_SEG(sbi, p.min_segno);
if (gc_type == FG_GC)
sbi->cur_victim_sec = secno;
else
set_bit(secno, dirty_i->victim_secmap);
}
*result = (p.min_segno / p.ofs_unit) * p.ofs_unit;
trace_f2fs_get_victim(sbi->sb, type, gc_type, &p,
sbi->cur_victim_sec,
prefree_segments(sbi), free_segments(sbi));
}
out:
mutex_unlock(&dirty_i->seglist_lock);
return (p.min_segno == NULL_SEGNO) ? 0 : 1;
}
static const struct victim_selection default_v_ops = {
.get_victim = get_victim_by_default,
};
static struct inode *find_gc_inode(struct gc_inode_list *gc_list, nid_t ino)
{
struct inode_entry *ie;
ie = radix_tree_lookup(&gc_list->iroot, ino);
if (ie)
return ie->inode;
return NULL;
}
static void add_gc_inode(struct gc_inode_list *gc_list, struct inode *inode)
{
struct inode_entry *new_ie;
if (inode == find_gc_inode(gc_list, inode->i_ino)) {
iput(inode);
return;
}
new_ie = f2fs_kmem_cache_alloc(inode_entry_slab, GFP_NOFS);
new_ie->inode = inode;
f2fs_radix_tree_insert(&gc_list->iroot, inode->i_ino, new_ie);
list_add_tail(&new_ie->list, &gc_list->ilist);
}
static void put_gc_inode(struct gc_inode_list *gc_list)
{
struct inode_entry *ie, *next_ie;
list_for_each_entry_safe(ie, next_ie, &gc_list->ilist, list) {
radix_tree_delete(&gc_list->iroot, ie->inode->i_ino);
iput(ie->inode);
list_del(&ie->list);
kmem_cache_free(inode_entry_slab, ie);
}
}
static int check_valid_map(struct f2fs_sb_info *sbi,
unsigned int segno, int offset)
{
struct sit_info *sit_i = SIT_I(sbi);
struct seg_entry *sentry;
int ret;
down_read(&sit_i->sentry_lock);
sentry = get_seg_entry(sbi, segno);
ret = f2fs_test_bit(offset, sentry->cur_valid_map);
up_read(&sit_i->sentry_lock);
return ret;
}
/*
* This function compares node address got in summary with that in NAT.
* On validity, copy that node with cold status, otherwise (invalid node)
* ignore that.
*/
f2fs: enhance foreground GC If we configure section consist of multiple segments, foreground GC will do the garbage collection with following approach: for each segment in victim section blk_start_plug for each valid block in segment write out by OPU method submit bio cache <--- blk_finish_plug <--- There are two issue: 1) for most of the time, 'submit bio cache' will break the merging in current bio buffer from writes of next segments, making a smaller bio submitting. 2) block plug only cover IO submitting in one segment, which reduce opportunity of merging IOs in plug with multiple segments. So refactor the code as below structure to strive for biggest opportunity of merging IOs: blk_start_plug for each segment in victim section for each valid block in segment write out by OPU method submit bio cache blk_finish_plug Test method: 1. mkfs.f2fs -s 8 /dev/sdX 2. touch 32 files 3. write 2M data into each file 4. punch 1.5M data from offset 0 for each file 5. trigger foreground gc through ioctl Before patch, there are totoally 40 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65776, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66016, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66256, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66496, size = 32768 ----repeat for 8 times After patch, there are totally 35 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 ----repeat 34 times f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 73696, size = 16384 Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-23 15:23:55 +07:00
static void gc_node_segment(struct f2fs_sb_info *sbi,
struct f2fs_summary *sum, unsigned int segno, int gc_type)
{
struct f2fs_summary *entry;
block_t start_addr;
int off;
f2fs: do in batch synchronously readahead during GC In order to enhance performance, we try to readahead node page during GC, but before loading node page we should get block address of node page which is stored in NAT table, so synchronously read of single NAT page block our readahead flow. f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0xa1e, oldaddr = 0xa1e, newaddr = 0xa1e, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x35e9, oldaddr = 0x72d7a, newaddr = 0x72d7a, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0xc1f, oldaddr = 0xc1f, newaddr = 0xc1f, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x389d, oldaddr = 0x72d7d, newaddr = 0x72d7d, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x3a82, oldaddr = 0x72d7f, newaddr = 0x72d7f, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x3bfa, oldaddr = 0x72d86, newaddr = 0x72d86, rw = READAHEAD ^H, type = NODE This patch adds one phase that do readahead NAT pages in batch before readahead node page for more effeciently. f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0x1952, oldaddr = 0x1952, newaddr = 0x1952, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc34, oldaddr = 0xc34, newaddr = 0xc34, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa33, oldaddr = 0xa33, newaddr = 0xa33, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc30, oldaddr = 0xc30, newaddr = 0xc30, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc32, oldaddr = 0xc32, newaddr = 0xc32, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc26, oldaddr = 0xc26, newaddr = 0xc26, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa2b, oldaddr = 0xa2b, newaddr = 0xa2b, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc23, oldaddr = 0xc23, newaddr = 0xc23, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc24, oldaddr = 0xc24, newaddr = 0xc24, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa10, oldaddr = 0xa10, newaddr = 0xa10, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc2c, oldaddr = 0xc2c, newaddr = 0xc2c, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5db7, oldaddr = 0x6be00, newaddr = 0x6be00, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5db9, oldaddr = 0x6be17, newaddr = 0x6be17, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dbc, oldaddr = 0x6be1a, newaddr = 0x6be1a, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc3, oldaddr = 0x6be20, newaddr = 0x6be20, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc7, oldaddr = 0x6be24, newaddr = 0x6be24, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc9, oldaddr = 0x6be25, newaddr = 0x6be25, rw = READAHEAD ^H, type = NODE Signed-off-by: Chao Yu <yuchao0@huawei.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-08-26 23:14:31 +07:00
int phase = 0;
start_addr = START_BLOCK(sbi, segno);
next_step:
entry = sum;
f2fs: give a chance to merge IOs by IO scheduler Previously, background GC submits many 4KB read requests to load victim blocks and/or its (i)node blocks. ... f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb61, blkaddr = 0x3b964ed f2fs_gc : block_rq_complete: 8,16 R () 499854968 + 8 [0] f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb6f, blkaddr = 0x3b964ee f2fs_gc : block_rq_complete: 8,16 R () 499854976 + 8 [0] f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb79, blkaddr = 0x3b964ef f2fs_gc : block_rq_complete: 8,16 R () 499854984 + 8 [0] ... However, by the fact that many IOs are sequential, we can give a chance to merge the IOs by IO scheduler. In order to do that, let's use blk_plug. ... f2fs_gc : f2fs_iget: ino = 143 f2fs_gc : f2fs_readpage: ino = 143, page_index = 0x1c6, blkaddr = 0x2e6ee f2fs_gc : f2fs_iget: ino = 143 f2fs_gc : f2fs_readpage: ino = 143, page_index = 0x1c7, blkaddr = 0x2e6ef <idle> : block_rq_complete: 8,16 R () 1519616 + 8 [0] <idle> : block_rq_complete: 8,16 R () 1519848 + 8 [0] <idle> : block_rq_complete: 8,16 R () 1520432 + 96 [0] <idle> : block_rq_complete: 8,16 R () 1520536 + 104 [0] <idle> : block_rq_complete: 8,16 R () 1521008 + 112 [0] <idle> : block_rq_complete: 8,16 R () 1521440 + 152 [0] <idle> : block_rq_complete: 8,16 R () 1521688 + 144 [0] <idle> : block_rq_complete: 8,16 R () 1522128 + 192 [0] <idle> : block_rq_complete: 8,16 R () 1523256 + 328 [0] ... Note that this issue should be addressed in checkpoint, and some readahead flows too. Reviewed-by: Namjae Jeon <namjae.jeon@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2013-04-24 11:19:56 +07:00
for (off = 0; off < sbi->blocks_per_seg; off++, entry++) {
nid_t nid = le32_to_cpu(entry->nid);
struct page *node_page;
struct node_info ni;
/* stop BG_GC if there is not enough free sections. */
if (gc_type == BG_GC && has_not_enough_free_secs(sbi, 0, 0))
f2fs: enhance foreground GC If we configure section consist of multiple segments, foreground GC will do the garbage collection with following approach: for each segment in victim section blk_start_plug for each valid block in segment write out by OPU method submit bio cache <--- blk_finish_plug <--- There are two issue: 1) for most of the time, 'submit bio cache' will break the merging in current bio buffer from writes of next segments, making a smaller bio submitting. 2) block plug only cover IO submitting in one segment, which reduce opportunity of merging IOs in plug with multiple segments. So refactor the code as below structure to strive for biggest opportunity of merging IOs: blk_start_plug for each segment in victim section for each valid block in segment write out by OPU method submit bio cache blk_finish_plug Test method: 1. mkfs.f2fs -s 8 /dev/sdX 2. touch 32 files 3. write 2M data into each file 4. punch 1.5M data from offset 0 for each file 5. trigger foreground gc through ioctl Before patch, there are totoally 40 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65776, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66016, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66256, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66496, size = 32768 ----repeat for 8 times After patch, there are totally 35 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 ----repeat 34 times f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 73696, size = 16384 Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-23 15:23:55 +07:00
return;
if (check_valid_map(sbi, segno, off) == 0)
continue;
f2fs: do in batch synchronously readahead during GC In order to enhance performance, we try to readahead node page during GC, but before loading node page we should get block address of node page which is stored in NAT table, so synchronously read of single NAT page block our readahead flow. f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0xa1e, oldaddr = 0xa1e, newaddr = 0xa1e, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x35e9, oldaddr = 0x72d7a, newaddr = 0x72d7a, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0xc1f, oldaddr = 0xc1f, newaddr = 0xc1f, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x389d, oldaddr = 0x72d7d, newaddr = 0x72d7d, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x3a82, oldaddr = 0x72d7f, newaddr = 0x72d7f, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x3bfa, oldaddr = 0x72d86, newaddr = 0x72d86, rw = READAHEAD ^H, type = NODE This patch adds one phase that do readahead NAT pages in batch before readahead node page for more effeciently. f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0x1952, oldaddr = 0x1952, newaddr = 0x1952, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc34, oldaddr = 0xc34, newaddr = 0xc34, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa33, oldaddr = 0xa33, newaddr = 0xa33, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc30, oldaddr = 0xc30, newaddr = 0xc30, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc32, oldaddr = 0xc32, newaddr = 0xc32, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc26, oldaddr = 0xc26, newaddr = 0xc26, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa2b, oldaddr = 0xa2b, newaddr = 0xa2b, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc23, oldaddr = 0xc23, newaddr = 0xc23, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc24, oldaddr = 0xc24, newaddr = 0xc24, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa10, oldaddr = 0xa10, newaddr = 0xa10, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc2c, oldaddr = 0xc2c, newaddr = 0xc2c, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5db7, oldaddr = 0x6be00, newaddr = 0x6be00, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5db9, oldaddr = 0x6be17, newaddr = 0x6be17, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dbc, oldaddr = 0x6be1a, newaddr = 0x6be1a, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc3, oldaddr = 0x6be20, newaddr = 0x6be20, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc7, oldaddr = 0x6be24, newaddr = 0x6be24, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc9, oldaddr = 0x6be25, newaddr = 0x6be25, rw = READAHEAD ^H, type = NODE Signed-off-by: Chao Yu <yuchao0@huawei.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-08-26 23:14:31 +07:00
if (phase == 0) {
ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nid), 1,
META_NAT, true);
continue;
}
if (phase == 1) {
ra_node_page(sbi, nid);
continue;
}
f2fs: do in batch synchronously readahead during GC In order to enhance performance, we try to readahead node page during GC, but before loading node page we should get block address of node page which is stored in NAT table, so synchronously read of single NAT page block our readahead flow. f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0xa1e, oldaddr = 0xa1e, newaddr = 0xa1e, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x35e9, oldaddr = 0x72d7a, newaddr = 0x72d7a, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0xc1f, oldaddr = 0xc1f, newaddr = 0xc1f, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x389d, oldaddr = 0x72d7d, newaddr = 0x72d7d, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x3a82, oldaddr = 0x72d7f, newaddr = 0x72d7f, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x3bfa, oldaddr = 0x72d86, newaddr = 0x72d86, rw = READAHEAD ^H, type = NODE This patch adds one phase that do readahead NAT pages in batch before readahead node page for more effeciently. f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0x1952, oldaddr = 0x1952, newaddr = 0x1952, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc34, oldaddr = 0xc34, newaddr = 0xc34, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa33, oldaddr = 0xa33, newaddr = 0xa33, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc30, oldaddr = 0xc30, newaddr = 0xc30, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc32, oldaddr = 0xc32, newaddr = 0xc32, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc26, oldaddr = 0xc26, newaddr = 0xc26, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa2b, oldaddr = 0xa2b, newaddr = 0xa2b, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc23, oldaddr = 0xc23, newaddr = 0xc23, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc24, oldaddr = 0xc24, newaddr = 0xc24, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa10, oldaddr = 0xa10, newaddr = 0xa10, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc2c, oldaddr = 0xc2c, newaddr = 0xc2c, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5db7, oldaddr = 0x6be00, newaddr = 0x6be00, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5db9, oldaddr = 0x6be17, newaddr = 0x6be17, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dbc, oldaddr = 0x6be1a, newaddr = 0x6be1a, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc3, oldaddr = 0x6be20, newaddr = 0x6be20, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc7, oldaddr = 0x6be24, newaddr = 0x6be24, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc9, oldaddr = 0x6be25, newaddr = 0x6be25, rw = READAHEAD ^H, type = NODE Signed-off-by: Chao Yu <yuchao0@huawei.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-08-26 23:14:31 +07:00
/* phase == 2 */
node_page = get_node_page(sbi, nid);
if (IS_ERR(node_page))
continue;
/* block may become invalid during get_node_page */
if (check_valid_map(sbi, segno, off) == 0) {
f2fs_put_page(node_page, 1);
continue;
}
get_node_info(sbi, nid, &ni);
if (ni.blk_addr != start_addr + off) {
f2fs_put_page(node_page, 1);
continue;
}
move_node_page(node_page, gc_type);
stat_inc_node_blk_count(sbi, 1, gc_type);
}
f2fs: give a chance to merge IOs by IO scheduler Previously, background GC submits many 4KB read requests to load victim blocks and/or its (i)node blocks. ... f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb61, blkaddr = 0x3b964ed f2fs_gc : block_rq_complete: 8,16 R () 499854968 + 8 [0] f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb6f, blkaddr = 0x3b964ee f2fs_gc : block_rq_complete: 8,16 R () 499854976 + 8 [0] f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb79, blkaddr = 0x3b964ef f2fs_gc : block_rq_complete: 8,16 R () 499854984 + 8 [0] ... However, by the fact that many IOs are sequential, we can give a chance to merge the IOs by IO scheduler. In order to do that, let's use blk_plug. ... f2fs_gc : f2fs_iget: ino = 143 f2fs_gc : f2fs_readpage: ino = 143, page_index = 0x1c6, blkaddr = 0x2e6ee f2fs_gc : f2fs_iget: ino = 143 f2fs_gc : f2fs_readpage: ino = 143, page_index = 0x1c7, blkaddr = 0x2e6ef <idle> : block_rq_complete: 8,16 R () 1519616 + 8 [0] <idle> : block_rq_complete: 8,16 R () 1519848 + 8 [0] <idle> : block_rq_complete: 8,16 R () 1520432 + 96 [0] <idle> : block_rq_complete: 8,16 R () 1520536 + 104 [0] <idle> : block_rq_complete: 8,16 R () 1521008 + 112 [0] <idle> : block_rq_complete: 8,16 R () 1521440 + 152 [0] <idle> : block_rq_complete: 8,16 R () 1521688 + 144 [0] <idle> : block_rq_complete: 8,16 R () 1522128 + 192 [0] <idle> : block_rq_complete: 8,16 R () 1523256 + 328 [0] ... Note that this issue should be addressed in checkpoint, and some readahead flows too. Reviewed-by: Namjae Jeon <namjae.jeon@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2013-04-24 11:19:56 +07:00
f2fs: do in batch synchronously readahead during GC In order to enhance performance, we try to readahead node page during GC, but before loading node page we should get block address of node page which is stored in NAT table, so synchronously read of single NAT page block our readahead flow. f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0xa1e, oldaddr = 0xa1e, newaddr = 0xa1e, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x35e9, oldaddr = 0x72d7a, newaddr = 0x72d7a, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0xc1f, oldaddr = 0xc1f, newaddr = 0xc1f, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x389d, oldaddr = 0x72d7d, newaddr = 0x72d7d, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x3a82, oldaddr = 0x72d7f, newaddr = 0x72d7f, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x3bfa, oldaddr = 0x72d86, newaddr = 0x72d86, rw = READAHEAD ^H, type = NODE This patch adds one phase that do readahead NAT pages in batch before readahead node page for more effeciently. f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0x1952, oldaddr = 0x1952, newaddr = 0x1952, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc34, oldaddr = 0xc34, newaddr = 0xc34, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa33, oldaddr = 0xa33, newaddr = 0xa33, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc30, oldaddr = 0xc30, newaddr = 0xc30, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc32, oldaddr = 0xc32, newaddr = 0xc32, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc26, oldaddr = 0xc26, newaddr = 0xc26, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa2b, oldaddr = 0xa2b, newaddr = 0xa2b, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc23, oldaddr = 0xc23, newaddr = 0xc23, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc24, oldaddr = 0xc24, newaddr = 0xc24, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa10, oldaddr = 0xa10, newaddr = 0xa10, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc2c, oldaddr = 0xc2c, newaddr = 0xc2c, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5db7, oldaddr = 0x6be00, newaddr = 0x6be00, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5db9, oldaddr = 0x6be17, newaddr = 0x6be17, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dbc, oldaddr = 0x6be1a, newaddr = 0x6be1a, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc3, oldaddr = 0x6be20, newaddr = 0x6be20, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc7, oldaddr = 0x6be24, newaddr = 0x6be24, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc9, oldaddr = 0x6be25, newaddr = 0x6be25, rw = READAHEAD ^H, type = NODE Signed-off-by: Chao Yu <yuchao0@huawei.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-08-26 23:14:31 +07:00
if (++phase < 3)
goto next_step;
}
/*
* Calculate start block index indicating the given node offset.
* Be careful, caller should give this node offset only indicating direct node
* blocks. If any node offsets, which point the other types of node blocks such
* as indirect or double indirect node blocks, are given, it must be a caller's
* bug.
*/
block_t start_bidx_of_node(unsigned int node_ofs, struct inode *inode)
{
unsigned int indirect_blks = 2 * NIDS_PER_BLOCK + 4;
unsigned int bidx;
if (node_ofs == 0)
return 0;
if (node_ofs <= 2) {
bidx = node_ofs - 1;
} else if (node_ofs <= indirect_blks) {
int dec = (node_ofs - 4) / (NIDS_PER_BLOCK + 1);
bidx = node_ofs - 2 - dec;
} else {
int dec = (node_ofs - indirect_blks - 3) / (NIDS_PER_BLOCK + 1);
bidx = node_ofs - 5 - dec;
}
return bidx * ADDRS_PER_BLOCK + ADDRS_PER_INODE(inode);
}
static bool is_alive(struct f2fs_sb_info *sbi, struct f2fs_summary *sum,
struct node_info *dni, block_t blkaddr, unsigned int *nofs)
{
struct page *node_page;
nid_t nid;
unsigned int ofs_in_node;
block_t source_blkaddr;
nid = le32_to_cpu(sum->nid);
ofs_in_node = le16_to_cpu(sum->ofs_in_node);
node_page = get_node_page(sbi, nid);
if (IS_ERR(node_page))
return false;
get_node_info(sbi, nid, dni);
if (sum->version != dni->version) {
f2fs_msg(sbi->sb, KERN_WARNING,
"%s: valid data with mismatched node version.",
__func__);
set_sbi_flag(sbi, SBI_NEED_FSCK);
}
*nofs = ofs_of_node(node_page);
f2fs: enhance on-disk inode structure scalability This patch add new flag F2FS_EXTRA_ATTR storing in inode.i_inline to indicate that on-disk structure of current inode is extended. In order to extend, we changed the inode structure a bit: Original one: struct f2fs_inode { ... struct f2fs_extent i_ext; __le32 i_addr[DEF_ADDRS_PER_INODE]; __le32 i_nid[DEF_NIDS_PER_INODE]; } Extended one: struct f2fs_inode { ... struct f2fs_extent i_ext; union { struct { __le16 i_extra_isize; __le16 i_padding; __le32 i_extra_end[0]; }; __le32 i_addr[DEF_ADDRS_PER_INODE]; }; __le32 i_nid[DEF_NIDS_PER_INODE]; } Once F2FS_EXTRA_ATTR is set, we will steal four bytes in the head of i_addr field for storing i_extra_isize and i_padding. with i_extra_isize, we can calculate actual size of reserved space in i_addr, available attribute fields included in total extra attribute fields for current inode can be described as below: +--------------------+ | .i_mode | | ... | | .i_ext | +--------------------+ | .i_extra_isize |-----+ | .i_padding | | | .i_prjid | | | .i_atime_extra | | | .i_ctime_extra | | | .i_mtime_extra |<----+ | .i_inode_cs |<----- store blkaddr/inline from here | .i_xattr_cs | | ... | +--------------------+ | | | block address | | | +--------------------+ | .i_nid | +--------------------+ | node_footer | | (nid, ino, offset) | +--------------------+ Hence, with this patch, we would enhance scalability of f2fs inode for storing more newly added attribute. Signed-off-by: Chao Yu <yuchao0@huawei.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2017-07-18 23:19:06 +07:00
source_blkaddr = datablock_addr(NULL, node_page, ofs_in_node);
f2fs_put_page(node_page, 1);
if (source_blkaddr != blkaddr)
return false;
return true;
}
/*
* Move data block via META_MAPPING while keeping locked data page.
* This can be used to move blocks, aka LBAs, directly on disk.
*/
static void move_data_block(struct inode *inode, block_t bidx,
unsigned int segno, int off)
{
struct f2fs_io_info fio = {
.sbi = F2FS_I_SB(inode),
.ino = inode->i_ino,
.type = DATA,
.temp = COLD,
.op = REQ_OP_READ,
.op_flags = 0,
.encrypted_page = NULL,
.in_list = false,
};
struct dnode_of_data dn;
struct f2fs_summary sum;
struct node_info ni;
struct page *page;
block_t newaddr;
int err;
/* do not read out */
page = f2fs_grab_cache_page(inode->i_mapping, bidx, false);
if (!page)
return;
if (!check_valid_map(F2FS_I_SB(inode), segno, off))
goto out;
if (f2fs_is_atomic_file(inode))
goto out;
if (f2fs_is_pinned_file(inode)) {
f2fs_pin_file_control(inode, true);
goto out;
}
set_new_dnode(&dn, inode, NULL, NULL, 0);
err = get_dnode_of_data(&dn, bidx, LOOKUP_NODE);
if (err)
goto out;
f2fs crypto: fix racing of accessing encrypted page among different competitors Since we use different page cache (normally inode's page cache for R/W and meta inode's page cache for GC) to cache the same physical block which is belong to an encrypted inode. Writeback of these two page cache should be exclusive, but now we didn't handle writeback state well, so there may be potential racing problem: a) kworker: f2fs_gc: - f2fs_write_data_pages - f2fs_write_data_page - do_write_data_page - write_data_page - f2fs_submit_page_mbio (page#1 in inode's page cache was queued in f2fs bio cache, and be ready to write to new blkaddr) - gc_data_segment - move_encrypted_block - pagecache_get_page (page#2 in meta inode's page cache was cached with the invalid datas of physical block located in new blkaddr) - f2fs_submit_page_mbio (page#1 was submitted, later, page#2 with invalid data will be submitted) b) f2fs_gc: - gc_data_segment - move_encrypted_block - f2fs_submit_page_mbio (page#1 in meta inode's page cache was queued in f2fs bio cache, and be ready to write to new blkaddr) user thread: - f2fs_write_begin - f2fs_submit_page_bio (we submit the request to block layer to update page#2 in inode's page cache with physical block located in new blkaddr, so here we may read gabbage data from new blkaddr since GC hasn't writebacked the page#1 yet) This patch fixes above potential racing problem for encrypted inode. Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-10-08 12:27:34 +07:00
if (unlikely(dn.data_blkaddr == NULL_ADDR)) {
ClearPageUptodate(page);
goto put_out;
f2fs crypto: fix racing of accessing encrypted page among different competitors Since we use different page cache (normally inode's page cache for R/W and meta inode's page cache for GC) to cache the same physical block which is belong to an encrypted inode. Writeback of these two page cache should be exclusive, but now we didn't handle writeback state well, so there may be potential racing problem: a) kworker: f2fs_gc: - f2fs_write_data_pages - f2fs_write_data_page - do_write_data_page - write_data_page - f2fs_submit_page_mbio (page#1 in inode's page cache was queued in f2fs bio cache, and be ready to write to new blkaddr) - gc_data_segment - move_encrypted_block - pagecache_get_page (page#2 in meta inode's page cache was cached with the invalid datas of physical block located in new blkaddr) - f2fs_submit_page_mbio (page#1 was submitted, later, page#2 with invalid data will be submitted) b) f2fs_gc: - gc_data_segment - move_encrypted_block - f2fs_submit_page_mbio (page#1 in meta inode's page cache was queued in f2fs bio cache, and be ready to write to new blkaddr) user thread: - f2fs_write_begin - f2fs_submit_page_bio (we submit the request to block layer to update page#2 in inode's page cache with physical block located in new blkaddr, so here we may read gabbage data from new blkaddr since GC hasn't writebacked the page#1 yet) This patch fixes above potential racing problem for encrypted inode. Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-10-08 12:27:34 +07:00
}
/*
* don't cache encrypted data into meta inode until previous dirty
* data were writebacked to avoid racing between GC and flush.
*/
f2fs_wait_on_page_writeback(page, DATA, true);
get_node_info(fio.sbi, dn.nid, &ni);
set_summary(&sum, dn.nid, dn.ofs_in_node, ni.version);
/* read page */
fio.page = page;
fio.new_blkaddr = fio.old_blkaddr = dn.data_blkaddr;
allocate_data_block(fio.sbi, NULL, fio.old_blkaddr, &newaddr,
&sum, CURSEG_COLD_DATA, NULL, false);
fio.encrypted_page = f2fs_pagecache_get_page(META_MAPPING(fio.sbi),
newaddr, FGP_LOCK | FGP_CREAT, GFP_NOFS);
if (!fio.encrypted_page) {
err = -ENOMEM;
goto recover_block;
}
err = f2fs_submit_page_bio(&fio);
if (err)
goto put_page_out;
/* write page */
lock_page(fio.encrypted_page);
if (unlikely(fio.encrypted_page->mapping != META_MAPPING(fio.sbi))) {
err = -EIO;
goto put_page_out;
}
if (unlikely(!PageUptodate(fio.encrypted_page))) {
err = -EIO;
goto put_page_out;
}
f2fs: call set_page_dirty to attach i_wb for cgroup The cgroup attaches inode->i_wb via mark_inode_dirty and when set_page_writeback is called, __inc_wb_stat() updates i_wb's stat. So, we need to explicitly call set_page_dirty->__mark_inode_dirty in prior to any writebacking pages. This patch should resolve the following kernel panic reported by Andreas Reis. https://bugzilla.kernel.org/show_bug.cgi?id=101801 --- Comment #2 from Andreas Reis <andreas.reis@gmail.com> --- BUG: unable to handle kernel NULL pointer dereference at 00000000000000a8 IP: [<ffffffff8149deea>] __percpu_counter_add+0x1a/0x90 PGD 2951ff067 PUD 2df43f067 PMD 0 Oops: 0000 [#1] PREEMPT SMP Modules linked in: CPU: 7 PID: 10356 Comm: gcc Tainted: G W 4.2.0-1-cu #1 Hardware name: Gigabyte Technology Co., Ltd. G1.Sniper M5/G1.Sniper M5, BIOS T01 02/03/2015 task: ffff880295044f80 ti: ffff880295140000 task.ti: ffff880295140000 RIP: 0010:[<ffffffff8149deea>] [<ffffffff8149deea>] __percpu_counter_add+0x1a/0x90 RSP: 0018:ffff880295143ac8 EFLAGS: 00010082 RAX: 0000000000000003 RBX: ffffea000a526d40 RCX: 0000000000000001 RDX: 0000000000000020 RSI: 0000000000000001 RDI: 0000000000000088 RBP: ffff880295143ae8 R08: 0000000000000000 R09: ffff88008f69bb30 R10: 00000000fffffffa R11: 0000000000000000 R12: 0000000000000088 R13: 0000000000000001 R14: ffff88041d099000 R15: ffff880084a205d0 FS: 00007f8549374700(0000) GS:ffff88042f3c0000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00000000000000a8 CR3: 000000033e1d5000 CR4: 00000000001406e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Stack: 0000000000000000 ffffea000a526d40 ffff880084a20738 ffff880084a20750 ffff880295143b48 ffffffff811cc91e ffff880000000000 0000000000000296 0000000000000000 ffff880417090198 0000000000000000 ffffea000a526d40 Call Trace: [<ffffffff811cc91e>] __test_set_page_writeback+0xde/0x1d0 [<ffffffff813fee87>] do_write_data_page+0xe7/0x3a0 [<ffffffff813faeea>] gc_data_segment+0x5aa/0x640 [<ffffffff813fb0b8>] do_garbage_collect+0x138/0x150 [<ffffffff813fb3fe>] f2fs_gc+0x1be/0x3e0 [<ffffffff81405541>] f2fs_balance_fs+0x81/0x90 [<ffffffff813ee357>] f2fs_unlink+0x47/0x1d0 [<ffffffff81239329>] vfs_unlink+0x109/0x1b0 [<ffffffff8123e3d7>] do_unlinkat+0x287/0x2c0 [<ffffffff8123ebc6>] SyS_unlink+0x16/0x20 [<ffffffff81942e2e>] entry_SYSCALL_64_fastpath+0x12/0x71 Code: 41 5e 5d c3 0f 1f 00 66 2e 0f 1f 84 00 00 00 00 00 55 48 89 e5 41 55 49 89 f5 41 54 49 89 fc 53 48 83 ec 08 65 ff 05 e6 d9 b6 7e <48> 8b 47 20 48 63 ca 65 8b 18 48 63 db 48 01 f3 48 39 cb 7d 0a RIP [<ffffffff8149deea>] __percpu_counter_add+0x1a/0x90 RSP <ffff880295143ac8> CR2: 00000000000000a8 ---[ end trace 5132449a58ed93a3 ]--- note: gcc[10356] exited with preempt_count 2 Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-07-25 14:29:17 +07:00
set_page_dirty(fio.encrypted_page);
f2fs_wait_on_page_writeback(fio.encrypted_page, DATA, true);
f2fs: call set_page_dirty to attach i_wb for cgroup The cgroup attaches inode->i_wb via mark_inode_dirty and when set_page_writeback is called, __inc_wb_stat() updates i_wb's stat. So, we need to explicitly call set_page_dirty->__mark_inode_dirty in prior to any writebacking pages. This patch should resolve the following kernel panic reported by Andreas Reis. https://bugzilla.kernel.org/show_bug.cgi?id=101801 --- Comment #2 from Andreas Reis <andreas.reis@gmail.com> --- BUG: unable to handle kernel NULL pointer dereference at 00000000000000a8 IP: [<ffffffff8149deea>] __percpu_counter_add+0x1a/0x90 PGD 2951ff067 PUD 2df43f067 PMD 0 Oops: 0000 [#1] PREEMPT SMP Modules linked in: CPU: 7 PID: 10356 Comm: gcc Tainted: G W 4.2.0-1-cu #1 Hardware name: Gigabyte Technology Co., Ltd. G1.Sniper M5/G1.Sniper M5, BIOS T01 02/03/2015 task: ffff880295044f80 ti: ffff880295140000 task.ti: ffff880295140000 RIP: 0010:[<ffffffff8149deea>] [<ffffffff8149deea>] __percpu_counter_add+0x1a/0x90 RSP: 0018:ffff880295143ac8 EFLAGS: 00010082 RAX: 0000000000000003 RBX: ffffea000a526d40 RCX: 0000000000000001 RDX: 0000000000000020 RSI: 0000000000000001 RDI: 0000000000000088 RBP: ffff880295143ae8 R08: 0000000000000000 R09: ffff88008f69bb30 R10: 00000000fffffffa R11: 0000000000000000 R12: 0000000000000088 R13: 0000000000000001 R14: ffff88041d099000 R15: ffff880084a205d0 FS: 00007f8549374700(0000) GS:ffff88042f3c0000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00000000000000a8 CR3: 000000033e1d5000 CR4: 00000000001406e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Stack: 0000000000000000 ffffea000a526d40 ffff880084a20738 ffff880084a20750 ffff880295143b48 ffffffff811cc91e ffff880000000000 0000000000000296 0000000000000000 ffff880417090198 0000000000000000 ffffea000a526d40 Call Trace: [<ffffffff811cc91e>] __test_set_page_writeback+0xde/0x1d0 [<ffffffff813fee87>] do_write_data_page+0xe7/0x3a0 [<ffffffff813faeea>] gc_data_segment+0x5aa/0x640 [<ffffffff813fb0b8>] do_garbage_collect+0x138/0x150 [<ffffffff813fb3fe>] f2fs_gc+0x1be/0x3e0 [<ffffffff81405541>] f2fs_balance_fs+0x81/0x90 [<ffffffff813ee357>] f2fs_unlink+0x47/0x1d0 [<ffffffff81239329>] vfs_unlink+0x109/0x1b0 [<ffffffff8123e3d7>] do_unlinkat+0x287/0x2c0 [<ffffffff8123ebc6>] SyS_unlink+0x16/0x20 [<ffffffff81942e2e>] entry_SYSCALL_64_fastpath+0x12/0x71 Code: 41 5e 5d c3 0f 1f 00 66 2e 0f 1f 84 00 00 00 00 00 55 48 89 e5 41 55 49 89 f5 41 54 49 89 fc 53 48 83 ec 08 65 ff 05 e6 d9 b6 7e <48> 8b 47 20 48 63 ca 65 8b 18 48 63 db 48 01 f3 48 39 cb 7d 0a RIP [<ffffffff8149deea>] __percpu_counter_add+0x1a/0x90 RSP <ffff880295143ac8> CR2: 00000000000000a8 ---[ end trace 5132449a58ed93a3 ]--- note: gcc[10356] exited with preempt_count 2 Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-07-25 14:29:17 +07:00
if (clear_page_dirty_for_io(fio.encrypted_page))
dec_page_count(fio.sbi, F2FS_DIRTY_META);
set_page_writeback(fio.encrypted_page);
ClearPageError(page);
/* allocate block address */
f2fs_wait_on_page_writeback(dn.node_page, NODE, true);
fio.op = REQ_OP_WRITE;
fio.op_flags = REQ_SYNC;
fio.new_blkaddr = newaddr;
f2fs: avoid hungtask when GC encrypted block if io_bits is set When io_bits is set, GCing encrypted block may hit the following hungtask. Since io_bits requires aligned block address, f2fs_submit_page_write may return -EAGAIN if new_blkaddr does not satisify io_bits alignment. As a result, the encrypted page will never be writtenback. This patch makes move_data_block aware the EAGAIN error and cancel the writeback. [ 246.751371] INFO: task kworker/u4:4:797 blocked for more than 90 seconds. [ 246.752423] Not tainted 4.15.0-rc4+ #11 [ 246.754176] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. [ 246.755336] kworker/u4:4 D25448 797 2 0x80000000 [ 246.755597] Workqueue: writeback wb_workfn (flush-7:0) [ 246.755616] Call Trace: [ 246.755695] ? __schedule+0x322/0xa90 [ 246.755761] ? blk_init_request_from_bio+0x120/0x120 [ 246.755773] ? pci_mmcfg_check_reserved+0xb0/0xb0 [ 246.755801] ? __radix_tree_create+0x19e/0x200 [ 246.755813] ? delete_node+0x136/0x370 [ 246.755838] schedule+0x43/0xc0 [ 246.755904] io_schedule+0x17/0x40 [ 246.755939] wait_on_page_bit_common+0x17b/0x240 [ 246.755950] ? wake_page_function+0xa0/0xa0 [ 246.755961] ? add_to_page_cache_lru+0x160/0x160 [ 246.755972] ? page_cache_tree_insert+0x170/0x170 [ 246.755983] ? __lru_cache_add+0x96/0xb0 [ 246.756086] __filemap_fdatawait_range+0x14f/0x1c0 [ 246.756097] ? wait_on_page_bit_common+0x240/0x240 [ 246.756120] ? __wake_up_locked_key_bookmark+0x20/0x20 [ 246.756167] ? wait_on_all_pages_writeback+0xc9/0x100 [ 246.756179] ? __remove_ino_entry+0x120/0x120 [ 246.756192] ? wait_woken+0x100/0x100 [ 246.756204] filemap_fdatawait_range+0x9/0x20 [ 246.756216] write_checkpoint+0x18a1/0x1f00 [ 246.756254] ? blk_get_request+0x10/0x10 [ 246.756265] ? cpumask_next_and+0x43/0x60 [ 246.756279] ? f2fs_sync_inode_meta+0x160/0x160 [ 246.756289] ? remove_element.isra.4+0xa0/0xa0 [ 246.756300] ? __put_compound_page+0x40/0x40 [ 246.756310] ? f2fs_sync_fs+0xec/0x1c0 [ 246.756320] ? f2fs_sync_fs+0x120/0x1c0 [ 246.756329] f2fs_sync_fs+0x120/0x1c0 [ 246.756357] ? trace_event_raw_event_f2fs__page+0x260/0x260 [ 246.756393] ? ata_build_rw_tf+0x173/0x410 [ 246.756397] f2fs_balance_fs_bg+0x198/0x390 [ 246.756405] ? drop_inmem_page+0x230/0x230 [ 246.756415] ? ahci_qc_prep+0x1bb/0x2e0 [ 246.756418] ? ahci_qc_issue+0x1df/0x290 [ 246.756422] ? __accumulate_pelt_segments+0x42/0xd0 [ 246.756426] ? f2fs_write_node_pages+0xd1/0x380 [ 246.756429] f2fs_write_node_pages+0xd1/0x380 [ 246.756437] ? sync_node_pages+0x8f0/0x8f0 [ 246.756440] ? update_curr+0x53/0x220 [ 246.756444] ? __accumulate_pelt_segments+0xa2/0xd0 [ 246.756448] ? __update_load_avg_se.isra.39+0x349/0x360 [ 246.756452] ? do_writepages+0x2a/0xa0 [ 246.756456] do_writepages+0x2a/0xa0 [ 246.756460] __writeback_single_inode+0x70/0x490 [ 246.756463] ? check_preempt_wakeup+0x199/0x310 [ 246.756467] writeback_sb_inodes+0x2a2/0x660 [ 246.756471] ? is_empty_dir_inode+0x40/0x40 [ 246.756474] ? __writeback_single_inode+0x490/0x490 [ 246.756477] ? string+0xbf/0xf0 [ 246.756480] ? down_read_trylock+0x35/0x60 [ 246.756484] __writeback_inodes_wb+0x9f/0xf0 [ 246.756488] wb_writeback+0x41d/0x4b0 [ 246.756492] ? writeback_inodes_wb.constprop.55+0x150/0x150 [ 246.756498] ? set_worker_desc+0xf7/0x130 [ 246.756502] ? current_is_workqueue_rescuer+0x60/0x60 [ 246.756511] ? _find_next_bit+0x2c/0xa0 [ 246.756514] ? wb_workfn+0x400/0x5d0 [ 246.756518] wb_workfn+0x400/0x5d0 [ 246.756521] ? finish_task_switch+0xdf/0x2a0 [ 246.756525] ? inode_wait_for_writeback+0x30/0x30 [ 246.756529] process_one_work+0x3a7/0x6f0 [ 246.756533] worker_thread+0x82/0x750 [ 246.756537] kthread+0x16f/0x1c0 [ 246.756541] ? trace_event_raw_event_workqueue_work+0x110/0x110 [ 246.756544] ? kthread_create_worker_on_cpu+0xb0/0xb0 [ 246.756548] ret_from_fork+0x1f/0x30 Signed-off-by: Sheng Yong <shengyong1@huawei.com> Reviewed-by: Chao Yu <yuchao0@huawei.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-01-17 11:11:31 +07:00
err = f2fs_submit_page_write(&fio);
if (err) {
if (PageWriteback(fio.encrypted_page))
end_page_writeback(fio.encrypted_page);
goto put_page_out;
}
f2fs_update_iostat(fio.sbi, FS_GC_DATA_IO, F2FS_BLKSIZE);
f2fs_update_data_blkaddr(&dn, newaddr);
set_inode_flag(inode, FI_APPEND_WRITE);
if (page->index == 0)
set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
put_page_out:
f2fs_put_page(fio.encrypted_page, 1);
recover_block:
if (err)
__f2fs_replace_block(fio.sbi, &sum, newaddr, fio.old_blkaddr,
true, true);
put_out:
f2fs_put_dnode(&dn);
out:
f2fs_put_page(page, 1);
}
static void move_data_page(struct inode *inode, block_t bidx, int gc_type,
unsigned int segno, int off)
{
struct page *page;
page = get_lock_data_page(inode, bidx, true);
if (IS_ERR(page))
return;
if (!check_valid_map(F2FS_I_SB(inode), segno, off))
goto out;
if (f2fs_is_atomic_file(inode))
goto out;
if (f2fs_is_pinned_file(inode)) {
if (gc_type == FG_GC)
f2fs_pin_file_control(inode, true);
goto out;
}
if (gc_type == BG_GC) {
if (PageWriteback(page))
goto out;
set_page_dirty(page);
set_cold_data(page);
} else {
struct f2fs_io_info fio = {
.sbi = F2FS_I_SB(inode),
.ino = inode->i_ino,
.type = DATA,
.temp = COLD,
.op = REQ_OP_WRITE,
.op_flags = REQ_SYNC,
.old_blkaddr = NULL_ADDR,
.page = page,
.encrypted_page = NULL,
.need_lock = LOCK_REQ,
.io_type = FS_GC_DATA_IO,
};
bool is_dirty = PageDirty(page);
int err;
retry:
f2fs: call set_page_dirty to attach i_wb for cgroup The cgroup attaches inode->i_wb via mark_inode_dirty and when set_page_writeback is called, __inc_wb_stat() updates i_wb's stat. So, we need to explicitly call set_page_dirty->__mark_inode_dirty in prior to any writebacking pages. This patch should resolve the following kernel panic reported by Andreas Reis. https://bugzilla.kernel.org/show_bug.cgi?id=101801 --- Comment #2 from Andreas Reis <andreas.reis@gmail.com> --- BUG: unable to handle kernel NULL pointer dereference at 00000000000000a8 IP: [<ffffffff8149deea>] __percpu_counter_add+0x1a/0x90 PGD 2951ff067 PUD 2df43f067 PMD 0 Oops: 0000 [#1] PREEMPT SMP Modules linked in: CPU: 7 PID: 10356 Comm: gcc Tainted: G W 4.2.0-1-cu #1 Hardware name: Gigabyte Technology Co., Ltd. G1.Sniper M5/G1.Sniper M5, BIOS T01 02/03/2015 task: ffff880295044f80 ti: ffff880295140000 task.ti: ffff880295140000 RIP: 0010:[<ffffffff8149deea>] [<ffffffff8149deea>] __percpu_counter_add+0x1a/0x90 RSP: 0018:ffff880295143ac8 EFLAGS: 00010082 RAX: 0000000000000003 RBX: ffffea000a526d40 RCX: 0000000000000001 RDX: 0000000000000020 RSI: 0000000000000001 RDI: 0000000000000088 RBP: ffff880295143ae8 R08: 0000000000000000 R09: ffff88008f69bb30 R10: 00000000fffffffa R11: 0000000000000000 R12: 0000000000000088 R13: 0000000000000001 R14: ffff88041d099000 R15: ffff880084a205d0 FS: 00007f8549374700(0000) GS:ffff88042f3c0000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00000000000000a8 CR3: 000000033e1d5000 CR4: 00000000001406e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Stack: 0000000000000000 ffffea000a526d40 ffff880084a20738 ffff880084a20750 ffff880295143b48 ffffffff811cc91e ffff880000000000 0000000000000296 0000000000000000 ffff880417090198 0000000000000000 ffffea000a526d40 Call Trace: [<ffffffff811cc91e>] __test_set_page_writeback+0xde/0x1d0 [<ffffffff813fee87>] do_write_data_page+0xe7/0x3a0 [<ffffffff813faeea>] gc_data_segment+0x5aa/0x640 [<ffffffff813fb0b8>] do_garbage_collect+0x138/0x150 [<ffffffff813fb3fe>] f2fs_gc+0x1be/0x3e0 [<ffffffff81405541>] f2fs_balance_fs+0x81/0x90 [<ffffffff813ee357>] f2fs_unlink+0x47/0x1d0 [<ffffffff81239329>] vfs_unlink+0x109/0x1b0 [<ffffffff8123e3d7>] do_unlinkat+0x287/0x2c0 [<ffffffff8123ebc6>] SyS_unlink+0x16/0x20 [<ffffffff81942e2e>] entry_SYSCALL_64_fastpath+0x12/0x71 Code: 41 5e 5d c3 0f 1f 00 66 2e 0f 1f 84 00 00 00 00 00 55 48 89 e5 41 55 49 89 f5 41 54 49 89 fc 53 48 83 ec 08 65 ff 05 e6 d9 b6 7e <48> 8b 47 20 48 63 ca 65 8b 18 48 63 db 48 01 f3 48 39 cb 7d 0a RIP [<ffffffff8149deea>] __percpu_counter_add+0x1a/0x90 RSP <ffff880295143ac8> CR2: 00000000000000a8 ---[ end trace 5132449a58ed93a3 ]--- note: gcc[10356] exited with preempt_count 2 Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-07-25 14:29:17 +07:00
set_page_dirty(page);
f2fs_wait_on_page_writeback(page, DATA, true);
if (clear_page_dirty_for_io(page)) {
inode_dec_dirty_pages(inode);
remove_dirty_inode(inode);
}
set_cold_data(page);
err = do_write_data_page(&fio);
if (err == -ENOMEM && is_dirty) {
congestion_wait(BLK_RW_ASYNC, HZ/50);
goto retry;
}
}
out:
f2fs_put_page(page, 1);
}
/*
* This function tries to get parent node of victim data block, and identifies
* data block validity. If the block is valid, copy that with cold status and
* modify parent node.
* If the parent node is not valid or the data block address is different,
* the victim data block is ignored.
*/
f2fs: enhance foreground GC If we configure section consist of multiple segments, foreground GC will do the garbage collection with following approach: for each segment in victim section blk_start_plug for each valid block in segment write out by OPU method submit bio cache <--- blk_finish_plug <--- There are two issue: 1) for most of the time, 'submit bio cache' will break the merging in current bio buffer from writes of next segments, making a smaller bio submitting. 2) block plug only cover IO submitting in one segment, which reduce opportunity of merging IOs in plug with multiple segments. So refactor the code as below structure to strive for biggest opportunity of merging IOs: blk_start_plug for each segment in victim section for each valid block in segment write out by OPU method submit bio cache blk_finish_plug Test method: 1. mkfs.f2fs -s 8 /dev/sdX 2. touch 32 files 3. write 2M data into each file 4. punch 1.5M data from offset 0 for each file 5. trigger foreground gc through ioctl Before patch, there are totoally 40 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65776, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66016, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66256, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66496, size = 32768 ----repeat for 8 times After patch, there are totally 35 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 ----repeat 34 times f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 73696, size = 16384 Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-23 15:23:55 +07:00
static void gc_data_segment(struct f2fs_sb_info *sbi, struct f2fs_summary *sum,
struct gc_inode_list *gc_list, unsigned int segno, int gc_type)
{
struct super_block *sb = sbi->sb;
struct f2fs_summary *entry;
block_t start_addr;
int off;
int phase = 0;
start_addr = START_BLOCK(sbi, segno);
next_step:
entry = sum;
f2fs: give a chance to merge IOs by IO scheduler Previously, background GC submits many 4KB read requests to load victim blocks and/or its (i)node blocks. ... f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb61, blkaddr = 0x3b964ed f2fs_gc : block_rq_complete: 8,16 R () 499854968 + 8 [0] f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb6f, blkaddr = 0x3b964ee f2fs_gc : block_rq_complete: 8,16 R () 499854976 + 8 [0] f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb79, blkaddr = 0x3b964ef f2fs_gc : block_rq_complete: 8,16 R () 499854984 + 8 [0] ... However, by the fact that many IOs are sequential, we can give a chance to merge the IOs by IO scheduler. In order to do that, let's use blk_plug. ... f2fs_gc : f2fs_iget: ino = 143 f2fs_gc : f2fs_readpage: ino = 143, page_index = 0x1c6, blkaddr = 0x2e6ee f2fs_gc : f2fs_iget: ino = 143 f2fs_gc : f2fs_readpage: ino = 143, page_index = 0x1c7, blkaddr = 0x2e6ef <idle> : block_rq_complete: 8,16 R () 1519616 + 8 [0] <idle> : block_rq_complete: 8,16 R () 1519848 + 8 [0] <idle> : block_rq_complete: 8,16 R () 1520432 + 96 [0] <idle> : block_rq_complete: 8,16 R () 1520536 + 104 [0] <idle> : block_rq_complete: 8,16 R () 1521008 + 112 [0] <idle> : block_rq_complete: 8,16 R () 1521440 + 152 [0] <idle> : block_rq_complete: 8,16 R () 1521688 + 144 [0] <idle> : block_rq_complete: 8,16 R () 1522128 + 192 [0] <idle> : block_rq_complete: 8,16 R () 1523256 + 328 [0] ... Note that this issue should be addressed in checkpoint, and some readahead flows too. Reviewed-by: Namjae Jeon <namjae.jeon@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2013-04-24 11:19:56 +07:00
for (off = 0; off < sbi->blocks_per_seg; off++, entry++) {
struct page *data_page;
struct inode *inode;
struct node_info dni; /* dnode info for the data */
unsigned int ofs_in_node, nofs;
block_t start_bidx;
f2fs: do in batch synchronously readahead during GC In order to enhance performance, we try to readahead node page during GC, but before loading node page we should get block address of node page which is stored in NAT table, so synchronously read of single NAT page block our readahead flow. f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0xa1e, oldaddr = 0xa1e, newaddr = 0xa1e, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x35e9, oldaddr = 0x72d7a, newaddr = 0x72d7a, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0xc1f, oldaddr = 0xc1f, newaddr = 0xc1f, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x389d, oldaddr = 0x72d7d, newaddr = 0x72d7d, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x3a82, oldaddr = 0x72d7f, newaddr = 0x72d7f, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x3bfa, oldaddr = 0x72d86, newaddr = 0x72d86, rw = READAHEAD ^H, type = NODE This patch adds one phase that do readahead NAT pages in batch before readahead node page for more effeciently. f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0x1952, oldaddr = 0x1952, newaddr = 0x1952, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc34, oldaddr = 0xc34, newaddr = 0xc34, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa33, oldaddr = 0xa33, newaddr = 0xa33, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc30, oldaddr = 0xc30, newaddr = 0xc30, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc32, oldaddr = 0xc32, newaddr = 0xc32, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc26, oldaddr = 0xc26, newaddr = 0xc26, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa2b, oldaddr = 0xa2b, newaddr = 0xa2b, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc23, oldaddr = 0xc23, newaddr = 0xc23, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc24, oldaddr = 0xc24, newaddr = 0xc24, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa10, oldaddr = 0xa10, newaddr = 0xa10, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc2c, oldaddr = 0xc2c, newaddr = 0xc2c, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5db7, oldaddr = 0x6be00, newaddr = 0x6be00, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5db9, oldaddr = 0x6be17, newaddr = 0x6be17, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dbc, oldaddr = 0x6be1a, newaddr = 0x6be1a, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc3, oldaddr = 0x6be20, newaddr = 0x6be20, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc7, oldaddr = 0x6be24, newaddr = 0x6be24, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc9, oldaddr = 0x6be25, newaddr = 0x6be25, rw = READAHEAD ^H, type = NODE Signed-off-by: Chao Yu <yuchao0@huawei.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-08-26 23:14:31 +07:00
nid_t nid = le32_to_cpu(entry->nid);
/* stop BG_GC if there is not enough free sections. */
if (gc_type == BG_GC && has_not_enough_free_secs(sbi, 0, 0))
f2fs: enhance foreground GC If we configure section consist of multiple segments, foreground GC will do the garbage collection with following approach: for each segment in victim section blk_start_plug for each valid block in segment write out by OPU method submit bio cache <--- blk_finish_plug <--- There are two issue: 1) for most of the time, 'submit bio cache' will break the merging in current bio buffer from writes of next segments, making a smaller bio submitting. 2) block plug only cover IO submitting in one segment, which reduce opportunity of merging IOs in plug with multiple segments. So refactor the code as below structure to strive for biggest opportunity of merging IOs: blk_start_plug for each segment in victim section for each valid block in segment write out by OPU method submit bio cache blk_finish_plug Test method: 1. mkfs.f2fs -s 8 /dev/sdX 2. touch 32 files 3. write 2M data into each file 4. punch 1.5M data from offset 0 for each file 5. trigger foreground gc through ioctl Before patch, there are totoally 40 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65776, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66016, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66256, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66496, size = 32768 ----repeat for 8 times After patch, there are totally 35 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 ----repeat 34 times f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 73696, size = 16384 Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-23 15:23:55 +07:00
return;
if (check_valid_map(sbi, segno, off) == 0)
continue;
if (phase == 0) {
f2fs: do in batch synchronously readahead during GC In order to enhance performance, we try to readahead node page during GC, but before loading node page we should get block address of node page which is stored in NAT table, so synchronously read of single NAT page block our readahead flow. f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0xa1e, oldaddr = 0xa1e, newaddr = 0xa1e, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x35e9, oldaddr = 0x72d7a, newaddr = 0x72d7a, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0xc1f, oldaddr = 0xc1f, newaddr = 0xc1f, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x389d, oldaddr = 0x72d7d, newaddr = 0x72d7d, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x3a82, oldaddr = 0x72d7f, newaddr = 0x72d7f, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x3bfa, oldaddr = 0x72d86, newaddr = 0x72d86, rw = READAHEAD ^H, type = NODE This patch adds one phase that do readahead NAT pages in batch before readahead node page for more effeciently. f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0x1952, oldaddr = 0x1952, newaddr = 0x1952, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc34, oldaddr = 0xc34, newaddr = 0xc34, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa33, oldaddr = 0xa33, newaddr = 0xa33, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc30, oldaddr = 0xc30, newaddr = 0xc30, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc32, oldaddr = 0xc32, newaddr = 0xc32, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc26, oldaddr = 0xc26, newaddr = 0xc26, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa2b, oldaddr = 0xa2b, newaddr = 0xa2b, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc23, oldaddr = 0xc23, newaddr = 0xc23, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc24, oldaddr = 0xc24, newaddr = 0xc24, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa10, oldaddr = 0xa10, newaddr = 0xa10, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc2c, oldaddr = 0xc2c, newaddr = 0xc2c, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5db7, oldaddr = 0x6be00, newaddr = 0x6be00, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5db9, oldaddr = 0x6be17, newaddr = 0x6be17, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dbc, oldaddr = 0x6be1a, newaddr = 0x6be1a, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc3, oldaddr = 0x6be20, newaddr = 0x6be20, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc7, oldaddr = 0x6be24, newaddr = 0x6be24, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc9, oldaddr = 0x6be25, newaddr = 0x6be25, rw = READAHEAD ^H, type = NODE Signed-off-by: Chao Yu <yuchao0@huawei.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-08-26 23:14:31 +07:00
ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nid), 1,
META_NAT, true);
continue;
}
if (phase == 1) {
ra_node_page(sbi, nid);
continue;
}
/* Get an inode by ino with checking validity */
if (!is_alive(sbi, entry, &dni, start_addr + off, &nofs))
continue;
f2fs: do in batch synchronously readahead during GC In order to enhance performance, we try to readahead node page during GC, but before loading node page we should get block address of node page which is stored in NAT table, so synchronously read of single NAT page block our readahead flow. f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0xa1e, oldaddr = 0xa1e, newaddr = 0xa1e, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x35e9, oldaddr = 0x72d7a, newaddr = 0x72d7a, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0xc1f, oldaddr = 0xc1f, newaddr = 0xc1f, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x389d, oldaddr = 0x72d7d, newaddr = 0x72d7d, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x3a82, oldaddr = 0x72d7f, newaddr = 0x72d7f, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x3bfa, oldaddr = 0x72d86, newaddr = 0x72d86, rw = READAHEAD ^H, type = NODE This patch adds one phase that do readahead NAT pages in batch before readahead node page for more effeciently. f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0x1952, oldaddr = 0x1952, newaddr = 0x1952, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc34, oldaddr = 0xc34, newaddr = 0xc34, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa33, oldaddr = 0xa33, newaddr = 0xa33, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc30, oldaddr = 0xc30, newaddr = 0xc30, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc32, oldaddr = 0xc32, newaddr = 0xc32, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc26, oldaddr = 0xc26, newaddr = 0xc26, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa2b, oldaddr = 0xa2b, newaddr = 0xa2b, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc23, oldaddr = 0xc23, newaddr = 0xc23, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc24, oldaddr = 0xc24, newaddr = 0xc24, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa10, oldaddr = 0xa10, newaddr = 0xa10, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc2c, oldaddr = 0xc2c, newaddr = 0xc2c, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5db7, oldaddr = 0x6be00, newaddr = 0x6be00, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5db9, oldaddr = 0x6be17, newaddr = 0x6be17, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dbc, oldaddr = 0x6be1a, newaddr = 0x6be1a, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc3, oldaddr = 0x6be20, newaddr = 0x6be20, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc7, oldaddr = 0x6be24, newaddr = 0x6be24, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc9, oldaddr = 0x6be25, newaddr = 0x6be25, rw = READAHEAD ^H, type = NODE Signed-off-by: Chao Yu <yuchao0@huawei.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-08-26 23:14:31 +07:00
if (phase == 2) {
ra_node_page(sbi, dni.ino);
continue;
}
ofs_in_node = le16_to_cpu(entry->ofs_in_node);
f2fs: do in batch synchronously readahead during GC In order to enhance performance, we try to readahead node page during GC, but before loading node page we should get block address of node page which is stored in NAT table, so synchronously read of single NAT page block our readahead flow. f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0xa1e, oldaddr = 0xa1e, newaddr = 0xa1e, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x35e9, oldaddr = 0x72d7a, newaddr = 0x72d7a, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0xc1f, oldaddr = 0xc1f, newaddr = 0xc1f, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x389d, oldaddr = 0x72d7d, newaddr = 0x72d7d, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x3a82, oldaddr = 0x72d7f, newaddr = 0x72d7f, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x3bfa, oldaddr = 0x72d86, newaddr = 0x72d86, rw = READAHEAD ^H, type = NODE This patch adds one phase that do readahead NAT pages in batch before readahead node page for more effeciently. f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0x1952, oldaddr = 0x1952, newaddr = 0x1952, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc34, oldaddr = 0xc34, newaddr = 0xc34, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa33, oldaddr = 0xa33, newaddr = 0xa33, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc30, oldaddr = 0xc30, newaddr = 0xc30, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc32, oldaddr = 0xc32, newaddr = 0xc32, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc26, oldaddr = 0xc26, newaddr = 0xc26, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa2b, oldaddr = 0xa2b, newaddr = 0xa2b, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc23, oldaddr = 0xc23, newaddr = 0xc23, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc24, oldaddr = 0xc24, newaddr = 0xc24, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa10, oldaddr = 0xa10, newaddr = 0xa10, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc2c, oldaddr = 0xc2c, newaddr = 0xc2c, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5db7, oldaddr = 0x6be00, newaddr = 0x6be00, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5db9, oldaddr = 0x6be17, newaddr = 0x6be17, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dbc, oldaddr = 0x6be1a, newaddr = 0x6be1a, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc3, oldaddr = 0x6be20, newaddr = 0x6be20, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc7, oldaddr = 0x6be24, newaddr = 0x6be24, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc9, oldaddr = 0x6be25, newaddr = 0x6be25, rw = READAHEAD ^H, type = NODE Signed-off-by: Chao Yu <yuchao0@huawei.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-08-26 23:14:31 +07:00
if (phase == 3) {
f2fs: avoid balanc_fs during evict_inode 1. Background Previously, if f2fs tries to move data blocks of an *evicting* inode during the cleaning process, it stops the process incompletely and then restarts the whole process, since it needs a locked inode to grab victim data pages in its address space. In order to get a locked inode, iget_locked() by f2fs_iget() is normally used, but, it waits if the inode is on freeing. So, here is a deadlock scenario. 1. f2fs_evict_inode() <- inode "A" 2. f2fs_balance_fs() 3. f2fs_gc() 4. gc_data_segment() 5. f2fs_iget() <- inode "A" too! If step #1 and #5 treat a same inode "A", step #5 would fall into deadlock since the inode "A" is on freeing. In order to resolve this, f2fs_iget_nowait() which skips __wait_on_freeing_inode() was introduced in step #5, and stops f2fs_gc() to complete f2fs_evict_inode(). 1. f2fs_evict_inode() <- inode "A" 2. f2fs_balance_fs() 3. f2fs_gc() 4. gc_data_segment() 5. f2fs_iget_nowait() <- inode "A", then stop f2fs_gc() w/ -ENOENT 2. Problem and Solution In the above scenario, however, f2fs cannot finish f2fs_evict_inode() only if: o there are not enough free sections, and o f2fs_gc() tries to move data blocks of the *evicting* inode repeatedly. So, the final solution is to use f2fs_iget() and remove f2fs_balance_fs() in f2fs_evict_inode(). The f2fs_evict_inode() actually truncates all the data and node blocks, which means that it doesn't produce any dirty node pages accordingly. So, we don't need to do f2fs_balance_fs() in practical. Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2013-01-31 13:36:04 +07:00
inode = f2fs_iget(sb, dni.ino);
if (IS_ERR(inode) || is_bad_inode(inode))
continue;
/* if inode uses special I/O path, let's go phase 3 */
if (f2fs_post_read_required(inode)) {
add_gc_inode(gc_list, inode);
continue;
}
if (!down_write_trylock(
&F2FS_I(inode)->dio_rwsem[WRITE])) {
iput(inode);
continue;
}
start_bidx = start_bidx_of_node(nofs, inode);
data_page = get_read_data_page(inode,
start_bidx + ofs_in_node, REQ_RAHEAD,
true);
up_write(&F2FS_I(inode)->dio_rwsem[WRITE]);
if (IS_ERR(data_page)) {
iput(inode);
continue;
}
f2fs_put_page(data_page, 0);
add_gc_inode(gc_list, inode);
continue;
}
f2fs: do in batch synchronously readahead during GC In order to enhance performance, we try to readahead node page during GC, but before loading node page we should get block address of node page which is stored in NAT table, so synchronously read of single NAT page block our readahead flow. f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0xa1e, oldaddr = 0xa1e, newaddr = 0xa1e, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x35e9, oldaddr = 0x72d7a, newaddr = 0x72d7a, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0xc1f, oldaddr = 0xc1f, newaddr = 0xc1f, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x389d, oldaddr = 0x72d7d, newaddr = 0x72d7d, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x3a82, oldaddr = 0x72d7f, newaddr = 0x72d7f, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x3bfa, oldaddr = 0x72d86, newaddr = 0x72d86, rw = READAHEAD ^H, type = NODE This patch adds one phase that do readahead NAT pages in batch before readahead node page for more effeciently. f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0x1952, oldaddr = 0x1952, newaddr = 0x1952, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc34, oldaddr = 0xc34, newaddr = 0xc34, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa33, oldaddr = 0xa33, newaddr = 0xa33, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc30, oldaddr = 0xc30, newaddr = 0xc30, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc32, oldaddr = 0xc32, newaddr = 0xc32, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc26, oldaddr = 0xc26, newaddr = 0xc26, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa2b, oldaddr = 0xa2b, newaddr = 0xa2b, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc23, oldaddr = 0xc23, newaddr = 0xc23, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc24, oldaddr = 0xc24, newaddr = 0xc24, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa10, oldaddr = 0xa10, newaddr = 0xa10, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc2c, oldaddr = 0xc2c, newaddr = 0xc2c, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5db7, oldaddr = 0x6be00, newaddr = 0x6be00, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5db9, oldaddr = 0x6be17, newaddr = 0x6be17, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dbc, oldaddr = 0x6be1a, newaddr = 0x6be1a, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc3, oldaddr = 0x6be20, newaddr = 0x6be20, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc7, oldaddr = 0x6be24, newaddr = 0x6be24, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc9, oldaddr = 0x6be25, newaddr = 0x6be25, rw = READAHEAD ^H, type = NODE Signed-off-by: Chao Yu <yuchao0@huawei.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-08-26 23:14:31 +07:00
/* phase 4 */
inode = find_gc_inode(gc_list, dni.ino);
if (inode) {
struct f2fs_inode_info *fi = F2FS_I(inode);
bool locked = false;
if (S_ISREG(inode->i_mode)) {
if (!down_write_trylock(&fi->dio_rwsem[READ]))
continue;
if (!down_write_trylock(
&fi->dio_rwsem[WRITE])) {
up_write(&fi->dio_rwsem[READ]);
continue;
}
locked = true;
/* wait for all inflight aio data */
inode_dio_wait(inode);
}
start_bidx = start_bidx_of_node(nofs, inode)
+ ofs_in_node;
if (f2fs_post_read_required(inode))
move_data_block(inode, start_bidx, segno, off);
else
move_data_page(inode, start_bidx, gc_type,
segno, off);
if (locked) {
up_write(&fi->dio_rwsem[WRITE]);
up_write(&fi->dio_rwsem[READ]);
}
stat_inc_data_blk_count(sbi, 1, gc_type);
}
}
f2fs: give a chance to merge IOs by IO scheduler Previously, background GC submits many 4KB read requests to load victim blocks and/or its (i)node blocks. ... f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb61, blkaddr = 0x3b964ed f2fs_gc : block_rq_complete: 8,16 R () 499854968 + 8 [0] f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb6f, blkaddr = 0x3b964ee f2fs_gc : block_rq_complete: 8,16 R () 499854976 + 8 [0] f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb79, blkaddr = 0x3b964ef f2fs_gc : block_rq_complete: 8,16 R () 499854984 + 8 [0] ... However, by the fact that many IOs are sequential, we can give a chance to merge the IOs by IO scheduler. In order to do that, let's use blk_plug. ... f2fs_gc : f2fs_iget: ino = 143 f2fs_gc : f2fs_readpage: ino = 143, page_index = 0x1c6, blkaddr = 0x2e6ee f2fs_gc : f2fs_iget: ino = 143 f2fs_gc : f2fs_readpage: ino = 143, page_index = 0x1c7, blkaddr = 0x2e6ef <idle> : block_rq_complete: 8,16 R () 1519616 + 8 [0] <idle> : block_rq_complete: 8,16 R () 1519848 + 8 [0] <idle> : block_rq_complete: 8,16 R () 1520432 + 96 [0] <idle> : block_rq_complete: 8,16 R () 1520536 + 104 [0] <idle> : block_rq_complete: 8,16 R () 1521008 + 112 [0] <idle> : block_rq_complete: 8,16 R () 1521440 + 152 [0] <idle> : block_rq_complete: 8,16 R () 1521688 + 144 [0] <idle> : block_rq_complete: 8,16 R () 1522128 + 192 [0] <idle> : block_rq_complete: 8,16 R () 1523256 + 328 [0] ... Note that this issue should be addressed in checkpoint, and some readahead flows too. Reviewed-by: Namjae Jeon <namjae.jeon@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2013-04-24 11:19:56 +07:00
f2fs: do in batch synchronously readahead during GC In order to enhance performance, we try to readahead node page during GC, but before loading node page we should get block address of node page which is stored in NAT table, so synchronously read of single NAT page block our readahead flow. f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0xa1e, oldaddr = 0xa1e, newaddr = 0xa1e, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x35e9, oldaddr = 0x72d7a, newaddr = 0x72d7a, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0xc1f, oldaddr = 0xc1f, newaddr = 0xc1f, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x389d, oldaddr = 0x72d7d, newaddr = 0x72d7d, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x3a82, oldaddr = 0x72d7f, newaddr = 0x72d7f, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x3bfa, oldaddr = 0x72d86, newaddr = 0x72d86, rw = READAHEAD ^H, type = NODE This patch adds one phase that do readahead NAT pages in batch before readahead node page for more effeciently. f2fs_submit_page_bio: dev = (251,0), ino = 2, page_index = 0x1952, oldaddr = 0x1952, newaddr = 0x1952, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc34, oldaddr = 0xc34, newaddr = 0xc34, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa33, oldaddr = 0xa33, newaddr = 0xa33, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc30, oldaddr = 0xc30, newaddr = 0xc30, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc32, oldaddr = 0xc32, newaddr = 0xc32, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc26, oldaddr = 0xc26, newaddr = 0xc26, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa2b, oldaddr = 0xa2b, newaddr = 0xa2b, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc23, oldaddr = 0xc23, newaddr = 0xc23, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc24, oldaddr = 0xc24, newaddr = 0xc24, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xa10, oldaddr = 0xa10, newaddr = 0xa10, rw = READ_SYNC(MP), type = META f2fs_submit_page_mbio: dev = (251,0), ino = 2, page_index = 0xc2c, oldaddr = 0xc2c, newaddr = 0xc2c, rw = READ_SYNC(MP), type = META f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5db7, oldaddr = 0x6be00, newaddr = 0x6be00, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5db9, oldaddr = 0x6be17, newaddr = 0x6be17, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dbc, oldaddr = 0x6be1a, newaddr = 0x6be1a, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc3, oldaddr = 0x6be20, newaddr = 0x6be20, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc7, oldaddr = 0x6be24, newaddr = 0x6be24, rw = READAHEAD ^H, type = NODE f2fs_submit_page_bio: dev = (251,0), ino = 1, page_index = 0x5dc9, oldaddr = 0x6be25, newaddr = 0x6be25, rw = READAHEAD ^H, type = NODE Signed-off-by: Chao Yu <yuchao0@huawei.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-08-26 23:14:31 +07:00
if (++phase < 5)
goto next_step;
}
static int __get_victim(struct f2fs_sb_info *sbi, unsigned int *victim,
int gc_type)
{
struct sit_info *sit_i = SIT_I(sbi);
int ret;
down_write(&sit_i->sentry_lock);
ret = DIRTY_I(sbi)->v_ops->get_victim(sbi, victim, gc_type,
NO_CHECK_TYPE, LFS);
up_write(&sit_i->sentry_lock);
return ret;
}
f2fs: enhance foreground GC If we configure section consist of multiple segments, foreground GC will do the garbage collection with following approach: for each segment in victim section blk_start_plug for each valid block in segment write out by OPU method submit bio cache <--- blk_finish_plug <--- There are two issue: 1) for most of the time, 'submit bio cache' will break the merging in current bio buffer from writes of next segments, making a smaller bio submitting. 2) block plug only cover IO submitting in one segment, which reduce opportunity of merging IOs in plug with multiple segments. So refactor the code as below structure to strive for biggest opportunity of merging IOs: blk_start_plug for each segment in victim section for each valid block in segment write out by OPU method submit bio cache blk_finish_plug Test method: 1. mkfs.f2fs -s 8 /dev/sdX 2. touch 32 files 3. write 2M data into each file 4. punch 1.5M data from offset 0 for each file 5. trigger foreground gc through ioctl Before patch, there are totoally 40 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65776, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66016, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66256, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66496, size = 32768 ----repeat for 8 times After patch, there are totally 35 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 ----repeat 34 times f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 73696, size = 16384 Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-23 15:23:55 +07:00
static int do_garbage_collect(struct f2fs_sb_info *sbi,
unsigned int start_segno,
struct gc_inode_list *gc_list, int gc_type)
{
struct page *sum_page;
struct f2fs_summary_block *sum;
f2fs: give a chance to merge IOs by IO scheduler Previously, background GC submits many 4KB read requests to load victim blocks and/or its (i)node blocks. ... f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb61, blkaddr = 0x3b964ed f2fs_gc : block_rq_complete: 8,16 R () 499854968 + 8 [0] f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb6f, blkaddr = 0x3b964ee f2fs_gc : block_rq_complete: 8,16 R () 499854976 + 8 [0] f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb79, blkaddr = 0x3b964ef f2fs_gc : block_rq_complete: 8,16 R () 499854984 + 8 [0] ... However, by the fact that many IOs are sequential, we can give a chance to merge the IOs by IO scheduler. In order to do that, let's use blk_plug. ... f2fs_gc : f2fs_iget: ino = 143 f2fs_gc : f2fs_readpage: ino = 143, page_index = 0x1c6, blkaddr = 0x2e6ee f2fs_gc : f2fs_iget: ino = 143 f2fs_gc : f2fs_readpage: ino = 143, page_index = 0x1c7, blkaddr = 0x2e6ef <idle> : block_rq_complete: 8,16 R () 1519616 + 8 [0] <idle> : block_rq_complete: 8,16 R () 1519848 + 8 [0] <idle> : block_rq_complete: 8,16 R () 1520432 + 96 [0] <idle> : block_rq_complete: 8,16 R () 1520536 + 104 [0] <idle> : block_rq_complete: 8,16 R () 1521008 + 112 [0] <idle> : block_rq_complete: 8,16 R () 1521440 + 152 [0] <idle> : block_rq_complete: 8,16 R () 1521688 + 144 [0] <idle> : block_rq_complete: 8,16 R () 1522128 + 192 [0] <idle> : block_rq_complete: 8,16 R () 1523256 + 328 [0] ... Note that this issue should be addressed in checkpoint, and some readahead flows too. Reviewed-by: Namjae Jeon <namjae.jeon@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2013-04-24 11:19:56 +07:00
struct blk_plug plug;
f2fs: enhance foreground GC If we configure section consist of multiple segments, foreground GC will do the garbage collection with following approach: for each segment in victim section blk_start_plug for each valid block in segment write out by OPU method submit bio cache <--- blk_finish_plug <--- There are two issue: 1) for most of the time, 'submit bio cache' will break the merging in current bio buffer from writes of next segments, making a smaller bio submitting. 2) block plug only cover IO submitting in one segment, which reduce opportunity of merging IOs in plug with multiple segments. So refactor the code as below structure to strive for biggest opportunity of merging IOs: blk_start_plug for each segment in victim section for each valid block in segment write out by OPU method submit bio cache blk_finish_plug Test method: 1. mkfs.f2fs -s 8 /dev/sdX 2. touch 32 files 3. write 2M data into each file 4. punch 1.5M data from offset 0 for each file 5. trigger foreground gc through ioctl Before patch, there are totoally 40 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65776, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66016, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66256, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66496, size = 32768 ----repeat for 8 times After patch, there are totally 35 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 ----repeat 34 times f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 73696, size = 16384 Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-23 15:23:55 +07:00
unsigned int segno = start_segno;
unsigned int end_segno = start_segno + sbi->segs_per_sec;
int seg_freed = 0;
f2fs: enhance foreground GC If we configure section consist of multiple segments, foreground GC will do the garbage collection with following approach: for each segment in victim section blk_start_plug for each valid block in segment write out by OPU method submit bio cache <--- blk_finish_plug <--- There are two issue: 1) for most of the time, 'submit bio cache' will break the merging in current bio buffer from writes of next segments, making a smaller bio submitting. 2) block plug only cover IO submitting in one segment, which reduce opportunity of merging IOs in plug with multiple segments. So refactor the code as below structure to strive for biggest opportunity of merging IOs: blk_start_plug for each segment in victim section for each valid block in segment write out by OPU method submit bio cache blk_finish_plug Test method: 1. mkfs.f2fs -s 8 /dev/sdX 2. touch 32 files 3. write 2M data into each file 4. punch 1.5M data from offset 0 for each file 5. trigger foreground gc through ioctl Before patch, there are totoally 40 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65776, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66016, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66256, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66496, size = 32768 ----repeat for 8 times After patch, there are totally 35 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 ----repeat 34 times f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 73696, size = 16384 Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-23 15:23:55 +07:00
unsigned char type = IS_DATASEG(get_seg_entry(sbi, segno)->type) ?
SUM_TYPE_DATA : SUM_TYPE_NODE;
f2fs: enhance foreground GC If we configure section consist of multiple segments, foreground GC will do the garbage collection with following approach: for each segment in victim section blk_start_plug for each valid block in segment write out by OPU method submit bio cache <--- blk_finish_plug <--- There are two issue: 1) for most of the time, 'submit bio cache' will break the merging in current bio buffer from writes of next segments, making a smaller bio submitting. 2) block plug only cover IO submitting in one segment, which reduce opportunity of merging IOs in plug with multiple segments. So refactor the code as below structure to strive for biggest opportunity of merging IOs: blk_start_plug for each segment in victim section for each valid block in segment write out by OPU method submit bio cache blk_finish_plug Test method: 1. mkfs.f2fs -s 8 /dev/sdX 2. touch 32 files 3. write 2M data into each file 4. punch 1.5M data from offset 0 for each file 5. trigger foreground gc through ioctl Before patch, there are totoally 40 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65776, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66016, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66256, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66496, size = 32768 ----repeat for 8 times After patch, there are totally 35 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 ----repeat 34 times f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 73696, size = 16384 Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-23 15:23:55 +07:00
/* readahead multi ssa blocks those have contiguous address */
if (sbi->segs_per_sec > 1)
ra_meta_pages(sbi, GET_SUM_BLOCK(sbi, segno),
sbi->segs_per_sec, META_SSA, true);
/* reference all summary page */
while (segno < end_segno) {
sum_page = get_sum_page(sbi, segno++);
unlock_page(sum_page);
}
f2fs: give a chance to merge IOs by IO scheduler Previously, background GC submits many 4KB read requests to load victim blocks and/or its (i)node blocks. ... f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb61, blkaddr = 0x3b964ed f2fs_gc : block_rq_complete: 8,16 R () 499854968 + 8 [0] f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb6f, blkaddr = 0x3b964ee f2fs_gc : block_rq_complete: 8,16 R () 499854976 + 8 [0] f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb79, blkaddr = 0x3b964ef f2fs_gc : block_rq_complete: 8,16 R () 499854984 + 8 [0] ... However, by the fact that many IOs are sequential, we can give a chance to merge the IOs by IO scheduler. In order to do that, let's use blk_plug. ... f2fs_gc : f2fs_iget: ino = 143 f2fs_gc : f2fs_readpage: ino = 143, page_index = 0x1c6, blkaddr = 0x2e6ee f2fs_gc : f2fs_iget: ino = 143 f2fs_gc : f2fs_readpage: ino = 143, page_index = 0x1c7, blkaddr = 0x2e6ef <idle> : block_rq_complete: 8,16 R () 1519616 + 8 [0] <idle> : block_rq_complete: 8,16 R () 1519848 + 8 [0] <idle> : block_rq_complete: 8,16 R () 1520432 + 96 [0] <idle> : block_rq_complete: 8,16 R () 1520536 + 104 [0] <idle> : block_rq_complete: 8,16 R () 1521008 + 112 [0] <idle> : block_rq_complete: 8,16 R () 1521440 + 152 [0] <idle> : block_rq_complete: 8,16 R () 1521688 + 144 [0] <idle> : block_rq_complete: 8,16 R () 1522128 + 192 [0] <idle> : block_rq_complete: 8,16 R () 1523256 + 328 [0] ... Note that this issue should be addressed in checkpoint, and some readahead flows too. Reviewed-by: Namjae Jeon <namjae.jeon@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2013-04-24 11:19:56 +07:00
blk_start_plug(&plug);
f2fs: enhance foreground GC If we configure section consist of multiple segments, foreground GC will do the garbage collection with following approach: for each segment in victim section blk_start_plug for each valid block in segment write out by OPU method submit bio cache <--- blk_finish_plug <--- There are two issue: 1) for most of the time, 'submit bio cache' will break the merging in current bio buffer from writes of next segments, making a smaller bio submitting. 2) block plug only cover IO submitting in one segment, which reduce opportunity of merging IOs in plug with multiple segments. So refactor the code as below structure to strive for biggest opportunity of merging IOs: blk_start_plug for each segment in victim section for each valid block in segment write out by OPU method submit bio cache blk_finish_plug Test method: 1. mkfs.f2fs -s 8 /dev/sdX 2. touch 32 files 3. write 2M data into each file 4. punch 1.5M data from offset 0 for each file 5. trigger foreground gc through ioctl Before patch, there are totoally 40 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65776, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66016, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66256, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66496, size = 32768 ----repeat for 8 times After patch, there are totally 35 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 ----repeat 34 times f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 73696, size = 16384 Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-23 15:23:55 +07:00
for (segno = start_segno; segno < end_segno; segno++) {
f2fs: enhance foreground GC If we configure section consist of multiple segments, foreground GC will do the garbage collection with following approach: for each segment in victim section blk_start_plug for each valid block in segment write out by OPU method submit bio cache <--- blk_finish_plug <--- There are two issue: 1) for most of the time, 'submit bio cache' will break the merging in current bio buffer from writes of next segments, making a smaller bio submitting. 2) block plug only cover IO submitting in one segment, which reduce opportunity of merging IOs in plug with multiple segments. So refactor the code as below structure to strive for biggest opportunity of merging IOs: blk_start_plug for each segment in victim section for each valid block in segment write out by OPU method submit bio cache blk_finish_plug Test method: 1. mkfs.f2fs -s 8 /dev/sdX 2. touch 32 files 3. write 2M data into each file 4. punch 1.5M data from offset 0 for each file 5. trigger foreground gc through ioctl Before patch, there are totoally 40 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65776, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66016, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66256, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66496, size = 32768 ----repeat for 8 times After patch, there are totally 35 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 ----repeat 34 times f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 73696, size = 16384 Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-23 15:23:55 +07:00
/* find segment summary of victim */
sum_page = find_get_page(META_MAPPING(sbi),
GET_SUM_BLOCK(sbi, segno));
f2fs_put_page(sum_page, 0);
if (get_valid_blocks(sbi, segno, false) == 0 ||
!PageUptodate(sum_page) ||
unlikely(f2fs_cp_error(sbi)))
goto next;
f2fs: enhance foreground GC If we configure section consist of multiple segments, foreground GC will do the garbage collection with following approach: for each segment in victim section blk_start_plug for each valid block in segment write out by OPU method submit bio cache <--- blk_finish_plug <--- There are two issue: 1) for most of the time, 'submit bio cache' will break the merging in current bio buffer from writes of next segments, making a smaller bio submitting. 2) block plug only cover IO submitting in one segment, which reduce opportunity of merging IOs in plug with multiple segments. So refactor the code as below structure to strive for biggest opportunity of merging IOs: blk_start_plug for each segment in victim section for each valid block in segment write out by OPU method submit bio cache blk_finish_plug Test method: 1. mkfs.f2fs -s 8 /dev/sdX 2. touch 32 files 3. write 2M data into each file 4. punch 1.5M data from offset 0 for each file 5. trigger foreground gc through ioctl Before patch, there are totoally 40 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65776, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66016, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66256, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66496, size = 32768 ----repeat for 8 times After patch, there are totally 35 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 ----repeat 34 times f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 73696, size = 16384 Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-23 15:23:55 +07:00
sum = page_address(sum_page);
f2fs_bug_on(sbi, type != GET_SUM_TYPE((&sum->footer)));
/*
* this is to avoid deadlock:
* - lock_page(sum_page) - f2fs_replace_block
* - check_valid_map() - down_write(sentry_lock)
* - down_read(sentry_lock) - change_curseg()
f2fs: enhance foreground GC If we configure section consist of multiple segments, foreground GC will do the garbage collection with following approach: for each segment in victim section blk_start_plug for each valid block in segment write out by OPU method submit bio cache <--- blk_finish_plug <--- There are two issue: 1) for most of the time, 'submit bio cache' will break the merging in current bio buffer from writes of next segments, making a smaller bio submitting. 2) block plug only cover IO submitting in one segment, which reduce opportunity of merging IOs in plug with multiple segments. So refactor the code as below structure to strive for biggest opportunity of merging IOs: blk_start_plug for each segment in victim section for each valid block in segment write out by OPU method submit bio cache blk_finish_plug Test method: 1. mkfs.f2fs -s 8 /dev/sdX 2. touch 32 files 3. write 2M data into each file 4. punch 1.5M data from offset 0 for each file 5. trigger foreground gc through ioctl Before patch, there are totoally 40 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65776, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66016, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66256, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66496, size = 32768 ----repeat for 8 times After patch, there are totally 35 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 ----repeat 34 times f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 73696, size = 16384 Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-23 15:23:55 +07:00
* - lock_page(sum_page)
*/
if (type == SUM_TYPE_NODE)
gc_node_segment(sbi, sum->entries, segno, gc_type);
else
gc_data_segment(sbi, sum->entries, gc_list, segno,
gc_type);
stat_inc_seg_count(sbi, type, gc_type);
if (gc_type == FG_GC &&
get_valid_blocks(sbi, segno, false) == 0)
seg_freed++;
next:
f2fs: enhance foreground GC If we configure section consist of multiple segments, foreground GC will do the garbage collection with following approach: for each segment in victim section blk_start_plug for each valid block in segment write out by OPU method submit bio cache <--- blk_finish_plug <--- There are two issue: 1) for most of the time, 'submit bio cache' will break the merging in current bio buffer from writes of next segments, making a smaller bio submitting. 2) block plug only cover IO submitting in one segment, which reduce opportunity of merging IOs in plug with multiple segments. So refactor the code as below structure to strive for biggest opportunity of merging IOs: blk_start_plug for each segment in victim section for each valid block in segment write out by OPU method submit bio cache blk_finish_plug Test method: 1. mkfs.f2fs -s 8 /dev/sdX 2. touch 32 files 3. write 2M data into each file 4. punch 1.5M data from offset 0 for each file 5. trigger foreground gc through ioctl Before patch, there are totoally 40 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65776, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66016, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66256, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66496, size = 32768 ----repeat for 8 times After patch, there are totally 35 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 ----repeat 34 times f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 73696, size = 16384 Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-23 15:23:55 +07:00
f2fs_put_page(sum_page, 0);
}
if (gc_type == FG_GC)
f2fs_submit_merged_write(sbi,
(type == SUM_TYPE_NODE) ? NODE : DATA);
f2fs: give a chance to merge IOs by IO scheduler Previously, background GC submits many 4KB read requests to load victim blocks and/or its (i)node blocks. ... f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb61, blkaddr = 0x3b964ed f2fs_gc : block_rq_complete: 8,16 R () 499854968 + 8 [0] f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb6f, blkaddr = 0x3b964ee f2fs_gc : block_rq_complete: 8,16 R () 499854976 + 8 [0] f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb79, blkaddr = 0x3b964ef f2fs_gc : block_rq_complete: 8,16 R () 499854984 + 8 [0] ... However, by the fact that many IOs are sequential, we can give a chance to merge the IOs by IO scheduler. In order to do that, let's use blk_plug. ... f2fs_gc : f2fs_iget: ino = 143 f2fs_gc : f2fs_readpage: ino = 143, page_index = 0x1c6, blkaddr = 0x2e6ee f2fs_gc : f2fs_iget: ino = 143 f2fs_gc : f2fs_readpage: ino = 143, page_index = 0x1c7, blkaddr = 0x2e6ef <idle> : block_rq_complete: 8,16 R () 1519616 + 8 [0] <idle> : block_rq_complete: 8,16 R () 1519848 + 8 [0] <idle> : block_rq_complete: 8,16 R () 1520432 + 96 [0] <idle> : block_rq_complete: 8,16 R () 1520536 + 104 [0] <idle> : block_rq_complete: 8,16 R () 1521008 + 112 [0] <idle> : block_rq_complete: 8,16 R () 1521440 + 152 [0] <idle> : block_rq_complete: 8,16 R () 1521688 + 144 [0] <idle> : block_rq_complete: 8,16 R () 1522128 + 192 [0] <idle> : block_rq_complete: 8,16 R () 1523256 + 328 [0] ... Note that this issue should be addressed in checkpoint, and some readahead flows too. Reviewed-by: Namjae Jeon <namjae.jeon@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2013-04-24 11:19:56 +07:00
f2fs: enhance foreground GC If we configure section consist of multiple segments, foreground GC will do the garbage collection with following approach: for each segment in victim section blk_start_plug for each valid block in segment write out by OPU method submit bio cache <--- blk_finish_plug <--- There are two issue: 1) for most of the time, 'submit bio cache' will break the merging in current bio buffer from writes of next segments, making a smaller bio submitting. 2) block plug only cover IO submitting in one segment, which reduce opportunity of merging IOs in plug with multiple segments. So refactor the code as below structure to strive for biggest opportunity of merging IOs: blk_start_plug for each segment in victim section for each valid block in segment write out by OPU method submit bio cache blk_finish_plug Test method: 1. mkfs.f2fs -s 8 /dev/sdX 2. touch 32 files 3. write 2M data into each file 4. punch 1.5M data from offset 0 for each file 5. trigger foreground gc through ioctl Before patch, there are totoally 40 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65776, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66016, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66256, size = 122880 f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 66496, size = 32768 ----repeat for 8 times After patch, there are totally 35 bios submitted. f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 65536, size = 122880 ----repeat 34 times f2fs_submit_write_bio: dev = (8,32), WRITE_SYNC, DATA, sector = 73696, size = 16384 Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-01-23 15:23:55 +07:00
blk_finish_plug(&plug);
stat_inc_call_count(sbi->stat_info);
return seg_freed;
}
int f2fs_gc(struct f2fs_sb_info *sbi, bool sync,
bool background, unsigned int segno)
{
int gc_type = sync ? FG_GC : BG_GC;
int sec_freed = 0, seg_freed = 0, total_freed = 0;
int ret = 0;
struct cp_control cpc;
unsigned int init_segno = segno;
struct gc_inode_list gc_list = {
.ilist = LIST_HEAD_INIT(gc_list.ilist),
.iroot = RADIX_TREE_INIT(gc_list.iroot, GFP_NOFS),
};
trace_f2fs_gc_begin(sbi->sb, sync, background,
get_pages(sbi, F2FS_DIRTY_NODES),
get_pages(sbi, F2FS_DIRTY_DENTS),
get_pages(sbi, F2FS_DIRTY_IMETA),
free_sections(sbi),
free_segments(sbi),
reserved_segments(sbi),
prefree_segments(sbi));
cpc.reason = __get_cp_reason(sbi);
gc_more:
Rename superblock flags (MS_xyz -> SB_xyz) This is a pure automated search-and-replace of the internal kernel superblock flags. The s_flags are now called SB_*, with the names and the values for the moment mirroring the MS_* flags that they're equivalent to. Note how the MS_xyz flags are the ones passed to the mount system call, while the SB_xyz flags are what we then use in sb->s_flags. The script to do this was: # places to look in; re security/*: it generally should *not* be # touched (that stuff parses mount(2) arguments directly), but # there are two places where we really deal with superblock flags. FILES="drivers/mtd drivers/staging/lustre fs ipc mm \ include/linux/fs.h include/uapi/linux/bfs_fs.h \ security/apparmor/apparmorfs.c security/apparmor/include/lib.h" # the list of MS_... constants SYMS="RDONLY NOSUID NODEV NOEXEC SYNCHRONOUS REMOUNT MANDLOCK \ DIRSYNC NOATIME NODIRATIME BIND MOVE REC VERBOSE SILENT \ POSIXACL UNBINDABLE PRIVATE SLAVE SHARED RELATIME KERNMOUNT \ I_VERSION STRICTATIME LAZYTIME SUBMOUNT NOREMOTELOCK NOSEC BORN \ ACTIVE NOUSER" SED_PROG= for i in $SYMS; do SED_PROG="$SED_PROG -e s/MS_$i/SB_$i/g"; done # we want files that contain at least one of MS_..., # with fs/namespace.c and fs/pnode.c excluded. L=$(for i in $SYMS; do git grep -w -l MS_$i $FILES; done| sort|uniq|grep -v '^fs/namespace.c'|grep -v '^fs/pnode.c') for f in $L; do sed -i $f $SED_PROG; done Requested-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-11-28 04:05:09 +07:00
if (unlikely(!(sbi->sb->s_flags & SB_ACTIVE))) {
ret = -EINVAL;
goto stop;
}
if (unlikely(f2fs_cp_error(sbi))) {
ret = -EIO;
goto stop;
}
if (gc_type == BG_GC && has_not_enough_free_secs(sbi, 0, 0)) {
/*
* For example, if there are many prefree_segments below given
* threshold, we can make them free by checkpoint. Then, we
* secure free segments which doesn't need fggc any more.
*/
if (prefree_segments(sbi)) {
ret = write_checkpoint(sbi, &cpc);
if (ret)
goto stop;
}
if (has_not_enough_free_secs(sbi, 0, 0))
gc_type = FG_GC;
}
/* f2fs_balance_fs doesn't need to do BG_GC in critical path. */
if (gc_type == BG_GC && !background) {
ret = -EINVAL;
goto stop;
}
if (!__get_victim(sbi, &segno, gc_type)) {
ret = -ENODATA;
goto stop;
}
seg_freed = do_garbage_collect(sbi, segno, &gc_list, gc_type);
if (gc_type == FG_GC && seg_freed == sbi->segs_per_sec)
sec_freed++;
total_freed += seg_freed;
if (gc_type == FG_GC)
sbi->cur_victim_sec = NULL_SEGNO;
if (!sync) {
if (has_not_enough_free_secs(sbi, sec_freed, 0)) {
segno = NULL_SEGNO;
goto gc_more;
}
if (gc_type == FG_GC)
ret = write_checkpoint(sbi, &cpc);
}
stop:
SIT_I(sbi)->last_victim[ALLOC_NEXT] = 0;
SIT_I(sbi)->last_victim[FLUSH_DEVICE] = init_segno;
trace_f2fs_gc_end(sbi->sb, ret, total_freed, sec_freed,
get_pages(sbi, F2FS_DIRTY_NODES),
get_pages(sbi, F2FS_DIRTY_DENTS),
get_pages(sbi, F2FS_DIRTY_IMETA),
free_sections(sbi),
free_segments(sbi),
reserved_segments(sbi),
prefree_segments(sbi));
mutex_unlock(&sbi->gc_mutex);
put_gc_inode(&gc_list);
if (sync)
ret = sec_freed ? 0 : -EAGAIN;
return ret;
}
void build_gc_manager(struct f2fs_sb_info *sbi)
{
u64 main_count, resv_count, ovp_count;
DIRTY_I(sbi)->v_ops = &default_v_ops;
/* threshold of # of valid blocks in a section for victims of FG_GC */
main_count = SM_I(sbi)->main_segments << sbi->log_blocks_per_seg;
resv_count = SM_I(sbi)->reserved_segments << sbi->log_blocks_per_seg;
ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
sbi->fggc_threshold = div64_u64((main_count - ovp_count) *
BLKS_PER_SEC(sbi), (main_count - resv_count));
sbi->gc_pin_file_threshold = DEF_GC_FAILED_PINNED_FILES;
/* give warm/cold data area from slower device */
if (sbi->s_ndevs && sbi->segs_per_sec == 1)
SIT_I(sbi)->last_victim[ALLOC_NEXT] =
GET_SEGNO(sbi, FDEV(0).end_blk) + 1;
}