- 07 Apr, 2014 1 commit
-
-
Jaegeuk Kim authored
Some storage devices show relatively high latencies to complete cache_flush commands, even though their normal IO speed is prettry much high. In such the case, it needs to merge cache_flush commands as much as possible to avoid issuing them redundantly. So, this patch introduces a mount option, "-o flush_merge", to mitigate such the overhead. If this option is enabled by user, F2FS merges the cache_flush commands and then issues just one cache_flush on behalf of them. Once the single command is finished, F2FS sends a completion signal to all the pending threads. Note that, this option can be used under a workload consisting of very intensive concurrent fsync calls, while the storage handles cache_flush commands slowly. Signed-off-by:Jaegeuk Kim <jaegeuk.kim@samsung.com>
-
- 02 Apr, 2014 4 commits
-
-
Jaegeuk Kim authored
In the f2fs_wait_on_page_writeback, io->bio should be covered by io_rwsem. Otherwise, the bio pointer can become a dangling pointer due to data races. Signed-off-by: -
Chao Yu authored
We should unlock page in ->readpage() path and also should unlock & release page in error path of ->write_begin() to avoid deadlock or memory leak. So let's add release code to fix the problem when we fail to read inline data. Signed-off-by:
Chao Yu <chao2.yu@samsung.com> Signed-off-by:
-
Chao Yu authored
This patch use list_for_each_entry{_safe} instead of list_for_each{_safe} for simplfying code. Signed-off-by: -
Chao Yu authored
Move kmem_cache_free out of spinlock protection region for better performance. Change log from v1: o remove spinlock protection for kmem_cache_free in destroy_node_manager suggested by Jaegeuk Kim. Signed-off-by:
-
- 01 Apr, 2014 3 commits
-
-
Chao Yu authored
In f2fs_setxattr we have limit this attribute name length, so we should also check it in f2fs_getxattr to avoid useless lookup caused by invalid name length. Signed-off-by:
-
Chao Yu authored
This patch introduce is_merged_page() to check whether current page is merged in f2fs bio cache. When page is not in cache, we can avoid submitting bio cache, resulting in having more chance to merge pages. Signed-off-by:
-
Jaegeuk Kim authored
During the cleaing of node segments, F2FS can get errored node blocks due to data race between node page lock and its valid bitmap operations. In that case, it needs to return an error to skip such the obsolete block copy. Signed-off-by:
-
- 20 Mar, 2014 9 commits
-
-
Jaegeuk Kim authored
This patch should resolve the following possible bug. RECLAIM_FS-ON-W at: mark_held_locks+0xb9/0x140 lockdep_trace_alloc+0x85/0xf0 __kmalloc+0x53/0x1d0 read_all_xattrs+0x3d1/0x3f0 [f2fs] f2fs_getxattr+0x4f/0x100 [f2fs] f2fs_get_acl+0x4c/0x290 [f2fs] get_acl+0x4f/0x80 posix_acl_create+0x72/0x180 f2fs_init_acl+0x29/0xcc [f2fs] __f2fs_add_link+0x259/0x710 [f2fs] f2fs_create+0xad/0x1c0 [f2fs] vfs_create+0xed/0x150 do_last+0xd36/0xed0 path_openat+0xc5/0x680 do_filp_open+0x43/0xa0 do_sys_open+0x13c/0x230 SyS_creat+0x1e/0x20 system_call_fastpath+0x16/0x1b Signed-off-by: -
Jaegeuk Kim authored
If multiple redundant fsync calls are triggered, we don't need to write its node pages with fsync mark continuously. So, this patch adds FI_NEED_FSYNC to track whether the latest node block is written with the fsync mark or not. If the mark was set, a new fsync doesn't need to write a node block. Otherwise, we should do a new node block with the mark for roll-forward recovery. Signed-off-by: -
Jaegeuk Kim authored
This patch introduces fi->i_sem to protect fi's info that includes xattr_ver, pino, i_nlink. This enables to remove i_mutex during f2fs_sync_file, resulting in performance improvement when a number of fsync calls are triggered from many concurrent threads. Signed-off-by: -
Jaegeuk Kim authored
It is more reasonable to determine the reclaiming rate of prefree segments according to the volume size, which is set to 5% by default. For example, if the volume is 128GB, the prefree segments are reclaimed when the number reaches to 6.4GB. Signed-off-by: -
Jaegeuk Kim authored
This patch adds missing dir_level documentation. Signed-off-by: -
Jaegeuk Kim authored
The NM_WOUT_THRESHOLD is now obsolete since f2fs starts to control on a basis of the memory footprint. Signed-off-by: -
Jaegeuk Kim authored
This patch introduces ram_thresh, a sysfs entry, which controls the memory footprint used by the free nid list and the nat cache. Previously, the free nid list was controlled by MAX_FREE_NIDS, while the nat cache was managed by NM_WOUT_THRESHOLD. However, this approach cannot be applied dynamically according to the system. So, this patch adds ram_thresh that users can specify the threshold, which is in order of 1 / 1024. For example, if the total ram size is 4GB and the value is set to 10 by default, f2fs tries to control the number of free nids and nat caches not to consume over 10 * (4GB / 1024) = 10MB. Signed-off-by: -
Jaegeuk Kim authored
The try_to_free_nats should not receive the negative nr_shrink. Otherwise, it can drop all the nat entries by the while loop. Signed-off-by: -
Jaegeuk Kim authored
If a page is on writeback, f2fs can face with deadlock due to under writepages. This is caused by merging IOs inside f2fs, so if it comes to detect, let's throw merged IOs, which is implemented by f2fs_wait_on_page_writeback. Signed-off-by:
-
- 18 Mar, 2014 8 commits
-
-
Jaegeuk Kim authored
This patch introduces nr_pages_to_write to align page writes to the segment or other operational unit size, which can be tuned according to the system environment. Signed-off-by: -
Jaegeuk Kim authored
This patch increases pages_skipped when skipping writepages. Signed-off-by: -
Jaegeuk Kim authored
This patch introduces nr_pages_to_skip(sbi, type) to determine writepages can be skipped. The dentry, node, and meta pages can be conrolled by F2FS without breaking the FS consistency. Signed-off-by: -
Jaegeuk Kim authored
The get_dirty_dents gives us the number of dirty dentry pages. Signed-off-by: -
Chao Yu authored
Previously 'background_gc={on***,off***}' is being parsed as correct option, with this patch we cloud fix the trivial bug in mount process. Change log from v1: o need to check length of parameter suggested by Jaegeuk Kim. Signed-off-by: -
Chao Yu authored
We should return error number of read_normal_summaries instead of -EINVAL when read_normal_summaries failed. Signed-off-by:
-
Chao Yu authored
This patch introduces a help function f2fs_has_xattr_block for better readability. Signed-off-by:
-
Chao Yu authored
The original segment_info's show looks out-of-format: cat /proc/fs/f2fs/loop0/segment_info 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 512 512 512 512 512 512 512 512 0 0 512 348 0 263 0 0 512 0 0 512 512 512 512 0 512 512 512 512 512 512 512 512 512 511 328 512 512 512 512 512 512 512 512 512 512 512 512 512 0 0 175 Let's fix this and show type for each segment. cat /proc/fs/f2fs/loop0/segment_info format: segment_type|valid_blocks segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN) 0 2|0 1|0 0|0 0|0 0|0 0|0 0|0 0|0 0|0 0|0 10 0|0 0|0 0|0 0|0 0|0 0|0 0|0 0|0 0|0 0|0 20 0|0 0|0 0|0 0|0 0|0 0|0 0|0 0|0 0|0 0|0 30 0|0 0|0 0|0 0|0 0|0 0|0 0|0 0|0 0|0 0|0 40 0|0 0|0 0|0 0|0 0|0 0|0 0|0 0|0 0|0 0|0 50 3|0 3|0 3|0 3|0 3|0 3|0 3|0 0|0 3|0 3|0 60 3|0 3|0 3|0 3|0 3|0 3|0 3|0 3|0 3|0 3|512 70 3|512 3|512 3|512 3|512 3|512 3|512 3|512 3|0 3|0 3|512 80 3|0 3|0 3|0 3|0 3|0 3|512 3|0 3|0 3|512 3|512 90 3|512 0|512 3|274 0|512 0|512 0|512 0|512 0|512 0|512 3|512 100 3|512 0|512 3|511 0|328 3|512 0|512 0|512 3|512 0|512 0|512 110 0|512 0|512 0|512 0|512 0|512 0|512 0|512 5|0 4|0 3|512 Signed-off-by:
-
- 12 Mar, 2014 2 commits
-
-
Chao Yu authored
Upper bound checking of ino should be added to f2fs_nfs_get_inode, so unneeded process before do_read_inode in f2fs_iget could be avoided when ino is invalid. Signed-off-by:
-
Chao Yu authored
This patch introduces a help function f2fs_has_inline_xattr for better readability. Signed-off-by:
-
- 11 Mar, 2014 1 commit
-
-
Chao Yu authored
Previously we do not recover inline xattr data of inode after power-cut, so inline xattr data may be lost. We should recover the data during the roll-forward process. Signed-off-by:
-
- 10 Mar, 2014 4 commits
-
-
Gu Zheng authored
Previously, we ra_sum_pages to pre-read contiguous pages as more as possible, and if we fail to alloc more pages, an ENOMEM error will be reported upstream, even though we have alloced some pages yet. In fact, we can use the available pages to do the job partly, and continue the rest in the following circle. Only reporting ENOMEM upstream if we really can not alloc any available page. And another fix is ignoring dealing with the following pages if an EIO occurs when reading page from page_list. Signed-off-by:
Gu Zheng <guz.fnst@cn.fujitsu.com> Reviewed-by:
-
Gu Zheng authored
The original segment_info's show is a bit out-of-format: [root@guz Demoes]# cat /proc/fs/f2fs/loop0/segment_info 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ...... 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 [root@guz Demoes]# so we fix it here for better legibility. [root@guz Demoes]# cat /proc/fs/f2fs/loop0/segment_info 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ...... 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 [root@guz Demoes]# Signed-off-by:
-
Gu Zheng authored
Signed-off-by:
-
Gu Zheng authored
Integrated a couple of minor changes for better readability suggested by Chao Yu. Signed-off-by:
-
- 05 Mar, 2014 1 commit
-
-
Jaegeuk Kim authored
This patch includes a simple fix to adjust coding style. Signed-off-by:
-
- 03 Mar, 2014 1 commit
-
-
Jaegeuk Kim authored
This patch fixes performance regression of dbench reported by Alex <hbx7d@yandex.com>. This issue was revealed by Phoronix tests results: http://www.phoronix.com/scan.php?page=article&item=linux_314_ssdfs&num=2 It turns out that we need to assign WRITE_SYNC to the node writes, if fsync is triggered. The performance numbers are like below, which is measured by Alex. 1. 355MB/s ext4 2. 225MB/s f2fs : WRITE for node writes 3. 525MB/s f2fs : WRITE_SYNC for node writes Reported-And-Tested-by: Alex <hbx7d@yandex.com>. Signed-off-by:
-
- 28 Feb, 2014 1 commit
-
-
Chao Yu authored
We should de-account dirty counters for page when redirty in ->writepage(). Wu Fengguang described in 'commit 971767ca': "writeback: fix dirtied pages accounting on redirty De-account the accumulative dirty counters on page redirty. Page redirties (very common in ext4) will introduce mismatch between counters (a) and (b) a) NR_DIRTIED, BDI_DIRTIED, tsk->nr_dirtied b) NR_WRITTEN, BDI_WRITTEN This will introduce systematic errors in balanced_rate and result in dirty page position errors (ie. the dirty pages are no longer balanced around the global/bdi setpoints)." Signed-off-by:
-
- 27 Feb, 2014 5 commits
-
-
Chao Yu authored
This patch use existing macro F2FS_INODE/NEXT_FREE_BLKADDR to clean up some codes. Signed-off-by:
-
Chao Yu authored
If there are multi segments in one section, we will read those SSA blocks which have contiguous address one by one in f2fs_gc. It may lost performance, let's read ahead SSA blocks by merge multi read request. Signed-off-by:
-
Jaegeuk Kim authored
This patch adds an sysfs entry to control dir_level used by the large directory. The description of this entry is: dir_level This parameter controls the directory level to support large directory. If a directory has a number of files, it can reduce the file lookup latency by increasing this dir_level value. Otherwise, it needs to decrease this value to reduce the space overhead. The default value is 0. Signed-off-by: -
Jaegeuk Kim authored
This patch introduces an i_dir_level field to support large directory. Previously, f2fs maintains multi-level hash tables to find a dentry quickly from a bunch of chiild dentries in a directory, and the hash tables consist of the following tree structure as below. In Documentation/filesystems/f2fs.txt, ---------------------- A : bucket B : block N : MAX_DIR_HASH_DEPTH ---------------------- level #0 | A(2B) | level #1 | A(2B) - A(2B) | level #2 | A(2B) - A(2B) - A(2B) - A(2B) . | . . . . level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B) But, if we can guess that a directory will handle a number of child files, we don't need to traverse the tree from level #0 to #N all the time. Since the lower level tables contain relatively small number of dentries, the miss ratio of the target dentry is likely to be high. In order to avoid that, we can configure the hash tables sparsely from level #0 like this. level #0 | A(2B) - A(2B) - A(2B) - A(2B) level #1 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B) With this structure, we can skip the ineffective tree searches in lower level hash tables. This patch adds just a facility for this by introducing i_dir_level in f2fs_inode. Signed-off-by: -
Jaegeuk Kim authored
It turns out that a bit operation like find_next_bit is not always fast enough for f2fs_find_entry. Instead, it is pretty much simple and fast to traverse each dentries. Signed-off-by:
-
