WSL2-Linux-Kernel/fs/ext4/extents_status.c

2173 строки
60 KiB
C
Исходник Обычный вид История

License cleanup: add SPDX GPL-2.0 license identifier to files with no license Many source files in the tree are missing licensing information, which makes it harder for compliance tools to determine the correct license. By default all files without license information are under the default license of the kernel, which is GPL version 2. Update the files which contain no license information with the 'GPL-2.0' SPDX license identifier. The SPDX identifier is a legally binding shorthand, which can be used instead of the full boiler plate text. This patch is based on work done by Thomas Gleixner and Kate Stewart and Philippe Ombredanne. How this work was done: Patches were generated and checked against linux-4.14-rc6 for a subset of the use cases: - file had no licensing information it it. - file was a */uapi/* one with no licensing information in it, - file was a */uapi/* one with existing licensing information, Further patches will be generated in subsequent months to fix up cases where non-standard license headers were used, and references to license had to be inferred by heuristics based on keywords. The analysis to determine which SPDX License Identifier to be applied to a file was done in a spreadsheet of side by side results from of the output of two independent scanners (ScanCode & Windriver) producing SPDX tag:value files created by Philippe Ombredanne. Philippe prepared the base worksheet, and did an initial spot review of a few 1000 files. The 4.13 kernel was the starting point of the analysis with 60,537 files assessed. Kate Stewart did a file by file comparison of the scanner results in the spreadsheet to determine which SPDX license identifier(s) to be applied to the file. She confirmed any determination that was not immediately clear with lawyers working with the Linux Foundation. Criteria used to select files for SPDX license identifier tagging was: - Files considered eligible had to be source code files. - Make and config files were included as candidates if they contained >5 lines of source - File already had some variant of a license header in it (even if <5 lines). All documentation files were explicitly excluded. The following heuristics were used to determine which SPDX license identifiers to apply. - when both scanners couldn't find any license traces, file was considered to have no license information in it, and the top level COPYING file license applied. For non */uapi/* files that summary was: SPDX license identifier # files ---------------------------------------------------|------- GPL-2.0 11139 and resulted in the first patch in this series. If that file was a */uapi/* path one, it was "GPL-2.0 WITH Linux-syscall-note" otherwise it was "GPL-2.0". Results of that was: SPDX license identifier # files ---------------------------------------------------|------- GPL-2.0 WITH Linux-syscall-note 930 and resulted in the second patch in this series. - if a file had some form of licensing information in it, and was one of the */uapi/* ones, it was denoted with the Linux-syscall-note if any GPL family license was found in the file or had no licensing in it (per prior point). Results summary: SPDX license identifier # files ---------------------------------------------------|------ GPL-2.0 WITH Linux-syscall-note 270 GPL-2.0+ WITH Linux-syscall-note 169 ((GPL-2.0 WITH Linux-syscall-note) OR BSD-2-Clause) 21 ((GPL-2.0 WITH Linux-syscall-note) OR BSD-3-Clause) 17 LGPL-2.1+ WITH Linux-syscall-note 15 GPL-1.0+ WITH Linux-syscall-note 14 ((GPL-2.0+ WITH Linux-syscall-note) OR BSD-3-Clause) 5 LGPL-2.0+ WITH Linux-syscall-note 4 LGPL-2.1 WITH Linux-syscall-note 3 ((GPL-2.0 WITH Linux-syscall-note) OR MIT) 3 ((GPL-2.0 WITH Linux-syscall-note) AND MIT) 1 and that resulted in the third patch in this series. - when the two scanners agreed on the detected license(s), that became the concluded license(s). - when there was disagreement between the two scanners (one detected a license but the other didn't, or they both detected different licenses) a manual inspection of the file occurred. - In most cases a manual inspection of the information in the file resulted in a clear resolution of the license that should apply (and which scanner probably needed to revisit its heuristics). - When it was not immediately clear, the license identifier was confirmed with lawyers working with the Linux Foundation. - If there was any question as to the appropriate license identifier, the file was flagged for further research and to be revisited later in time. In total, over 70 hours of logged manual review was done on the spreadsheet to determine the SPDX license identifiers to apply to the source files by Kate, Philippe, Thomas and, in some cases, confirmation by lawyers working with the Linux Foundation. Kate also obtained a third independent scan of the 4.13 code base from FOSSology, and compared selected files where the other two scanners disagreed against that SPDX file, to see if there was new insights. The Windriver scanner is based on an older version of FOSSology in part, so they are related. Thomas did random spot checks in about 500 files from the spreadsheets for the uapi headers and agreed with SPDX license identifier in the files he inspected. For the non-uapi files Thomas did random spot checks in about 15000 files. In initial set of patches against 4.14-rc6, 3 files were found to have copy/paste license identifier errors, and have been fixed to reflect the correct identifier. Additionally Philippe spent 10 hours this week doing a detailed manual inspection and review of the 12,461 patched files from the initial patch version early this week with: - a full scancode scan run, collecting the matched texts, detected license ids and scores - reviewing anything where there was a license detected (about 500+ files) to ensure that the applied SPDX license was correct - reviewing anything where there was no detection but the patch license was not GPL-2.0 WITH Linux-syscall-note to ensure that the applied SPDX license was correct This produced a worksheet with 20 files needing minor correction. This worksheet was then exported into 3 different .csv files for the different types of files to be modified. These .csv files were then reviewed by Greg. Thomas wrote a script to parse the csv files and add the proper SPDX tag to the file, in the format that the file expected. This script was further refined by Greg based on the output to detect more types of files automatically and to distinguish between header and source .c files (which need different comment types.) Finally Greg ran the script using the .csv files to generate the patches. Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org> Reviewed-by: Philippe Ombredanne <pombredanne@nexb.com> Reviewed-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-11-01 17:07:57 +03:00
// SPDX-License-Identifier: GPL-2.0
/*
* fs/ext4/extents_status.c
*
* Written by Yongqiang Yang <xiaoqiangnk@gmail.com>
* Modified by
* Allison Henderson <achender@linux.vnet.ibm.com>
* Hugh Dickins <hughd@google.com>
* Zheng Liu <wenqing.lz@taobao.com>
*
* Ext4 extents status tree core functions.
*/
ext4: improve extent cache shrink mechanism to avoid to burn CPU time Now we maintain an proper in-order LRU list in ext4 to reclaim entries from extent status tree when we are under heavy memory pressure. For keeping this order, a spin lock is used to protect this list. But this lock burns a lot of CPU time. We can use the following steps to trigger it. % cd /dev/shm % dd if=/dev/zero of=ext4-img bs=1M count=2k % mkfs.ext4 ext4-img % mount -t ext4 -o loop ext4-img /mnt % cd /mnt % for ((i=0;i<160;i++)); do truncate -s 64g $i; done % for ((i=0;i<160;i++)); do cp $i /dev/null &; done % perf record -a -g % perf report This commit tries to fix this problem. Now a new member called i_touch_when is added into ext4_inode_info to record the last access time for an inode. Meanwhile we never need to keep a proper in-order LRU list. So this can avoid to burns some CPU time. When we try to reclaim some entries from extent status tree, we use list_sort() to get a proper in-order list. Then we traverse this list to discard some entries. In ext4_sb_info, we use s_es_last_sorted to record the last time of sorting this list. When we traverse the list, we skip the inode that is newer than this time, and move this inode to the tail of LRU list. When the head of the list is newer than s_es_last_sorted, we will sort the LRU list again. In this commit, we break the loop if s_extent_cache_cnt == 0 because that means that all extents in extent status tree have been reclaimed. Meanwhile in this commit, ext4_es_{un}register_shrinker()'s prototype is changed to save a local variable in these functions. Reported-by: Dave Hansen <dave.hansen@intel.com> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2013-07-01 16:12:37 +04:00
#include <linux/list_sort.h>
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
#include <linux/proc_fs.h>
#include <linux/seq_file.h>
#include "ext4.h"
#include <trace/events/ext4.h>
/*
* According to previous discussion in Ext4 Developer Workshop, we
* will introduce a new structure called io tree to track all extent
* status in order to solve some problems that we have met
* (e.g. Reservation space warning), and provide extent-level locking.
* Delay extent tree is the first step to achieve this goal. It is
* original built by Yongqiang Yang. At that time it is called delay
* extent tree, whose goal is only track delayed extents in memory to
* simplify the implementation of fiemap and bigalloc, and introduce
* lseek SEEK_DATA/SEEK_HOLE support. That is why it is still called
* delay extent tree at the first commit. But for better understand
* what it does, it has been rename to extent status tree.
*
* Step1:
* Currently the first step has been done. All delayed extents are
* tracked in the tree. It maintains the delayed extent when a delayed
* allocation is issued, and the delayed extent is written out or
* invalidated. Therefore the implementation of fiemap and bigalloc
* are simplified, and SEEK_DATA/SEEK_HOLE are introduced.
*
* The following comment describes the implemenmtation of extent
* status tree and future works.
*
* Step2:
* In this step all extent status are tracked by extent status tree.
* Thus, we can first try to lookup a block mapping in this tree before
* finding it in extent tree. Hence, single extent cache can be removed
* because extent status tree can do a better job. Extents in status
* tree are loaded on-demand. Therefore, the extent status tree may not
* contain all of the extents in a file. Meanwhile we define a shrinker
* to reclaim memory from extent status tree because fragmented extent
* tree will make status tree cost too much memory. written/unwritten/-
* hole extents in the tree will be reclaimed by this shrinker when we
* are under high memory pressure. Delayed extents will not be
* reclimed because fiemap, bigalloc, and seek_data/hole need it.
*/
/*
* Extent status tree implementation for ext4.
*
*
* ==========================================================================
* Extent status tree tracks all extent status.
*
* 1. Why we need to implement extent status tree?
*
* Without extent status tree, ext4 identifies a delayed extent by looking
* up page cache, this has several deficiencies - complicated, buggy,
* and inefficient code.
*
* FIEMAP, SEEK_HOLE/DATA, bigalloc, and writeout all need to know if a
* block or a range of blocks are belonged to a delayed extent.
*
* Let us have a look at how they do without extent status tree.
* -- FIEMAP
* FIEMAP looks up page cache to identify delayed allocations from holes.
*
* -- SEEK_HOLE/DATA
* SEEK_HOLE/DATA has the same problem as FIEMAP.
*
* -- bigalloc
* bigalloc looks up page cache to figure out if a block is
* already under delayed allocation or not to determine whether
* quota reserving is needed for the cluster.
*
* -- writeout
* Writeout looks up whole page cache to see if a buffer is
* mapped, If there are not very many delayed buffers, then it is
* time consuming.
*
* With extent status tree implementation, FIEMAP, SEEK_HOLE/DATA,
* bigalloc and writeout can figure out if a block or a range of
* blocks is under delayed allocation(belonged to a delayed extent) or
* not by searching the extent tree.
*
*
* ==========================================================================
* 2. Ext4 extent status tree impelmentation
*
* -- extent
* A extent is a range of blocks which are contiguous logically and
* physically. Unlike extent in extent tree, this extent in ext4 is
* a in-memory struct, there is no corresponding on-disk data. There
* is no limit on length of extent, so an extent can contain as many
* blocks as they are contiguous logically and physically.
*
* -- extent status tree
* Every inode has an extent status tree and all allocation blocks
* are added to the tree with different status. The extent in the
* tree are ordered by logical block no.
*
* -- operations on a extent status tree
* There are three important operations on a delayed extent tree: find
* next extent, adding a extent(a range of blocks) and removing a extent.
*
* -- race on a extent status tree
* Extent status tree is protected by inode->i_es_lock.
*
* -- memory consumption
* Fragmented extent tree will make extent status tree cost too much
* memory. Hence, we will reclaim written/unwritten/hole extents from
* the tree under a heavy memory pressure.
*
*
* ==========================================================================
* 3. Performance analysis
*
* -- overhead
* 1. There is a cache extent for write access, so if writes are
* not very random, adding space operaions are in O(1) time.
*
* -- gain
* 2. Code is much simpler, more readable, more maintainable and
* more efficient.
*
*
* ==========================================================================
* 4. TODO list
*
* -- Refactor delayed space reservation
*
* -- Extent-level locking
*/
static struct kmem_cache *ext4_es_cachep;
static struct kmem_cache *ext4_pending_cachep;
static int __es_insert_extent(struct inode *inode, struct extent_status *newes);
static int __es_remove_extent(struct inode *inode, ext4_lblk_t lblk,
ext4_lblk_t end, int *reserved);
static int es_reclaim_extents(struct ext4_inode_info *ei, int *nr_to_scan);
static int __es_shrink(struct ext4_sb_info *sbi, int nr_to_scan,
struct ext4_inode_info *locked_ei);
static void __revise_pending(struct inode *inode, ext4_lblk_t lblk,
ext4_lblk_t len);
int __init ext4_init_es(void)
{
ext4_es_cachep = kmem_cache_create("ext4_extent_status",
sizeof(struct extent_status),
0, (SLAB_RECLAIM_ACCOUNT), NULL);
if (ext4_es_cachep == NULL)
return -ENOMEM;
return 0;
}
void ext4_exit_es(void)
{
kmem_cache_destroy(ext4_es_cachep);
}
void ext4_es_init_tree(struct ext4_es_tree *tree)
{
tree->root = RB_ROOT;
tree->cache_es = NULL;
}
#ifdef ES_DEBUG__
static void ext4_es_print_tree(struct inode *inode)
{
struct ext4_es_tree *tree;
struct rb_node *node;
printk(KERN_DEBUG "status extents for inode %lu:", inode->i_ino);
tree = &EXT4_I(inode)->i_es_tree;
node = rb_first(&tree->root);
while (node) {
struct extent_status *es;
es = rb_entry(node, struct extent_status, rb_node);
printk(KERN_DEBUG " [%u/%u) %llu %x",
es->es_lblk, es->es_len,
ext4_es_pblock(es), ext4_es_status(es));
node = rb_next(node);
}
printk(KERN_DEBUG "\n");
}
#else
#define ext4_es_print_tree(inode)
#endif
static inline ext4_lblk_t ext4_es_end(struct extent_status *es)
{
BUG_ON(es->es_lblk + es->es_len < es->es_lblk);
return es->es_lblk + es->es_len - 1;
}
/*
* search through the tree for an delayed extent with a given offset. If
* it can't be found, try to find next extent.
*/
static struct extent_status *__es_tree_search(struct rb_root *root,
ext4_lblk_t lblk)
{
struct rb_node *node = root->rb_node;
struct extent_status *es = NULL;
while (node) {
es = rb_entry(node, struct extent_status, rb_node);
if (lblk < es->es_lblk)
node = node->rb_left;
else if (lblk > ext4_es_end(es))
node = node->rb_right;
else
return es;
}
if (es && lblk < es->es_lblk)
return es;
if (es && lblk > ext4_es_end(es)) {
node = rb_next(&es->rb_node);
return node ? rb_entry(node, struct extent_status, rb_node) :
NULL;
}
return NULL;
}
/*
* ext4_es_find_extent_range - find extent with specified status within block
* range or next extent following block range in
* extents status tree
*
* @inode - file containing the range
* @matching_fn - pointer to function that matches extents with desired status
* @lblk - logical block defining start of range
* @end - logical block defining end of range
* @es - extent found, if any
*
* Find the first extent within the block range specified by @lblk and @end
* in the extents status tree that satisfies @matching_fn. If a match
* is found, it's returned in @es. If not, and a matching extent is found
* beyond the block range, it's returned in @es. If no match is found, an
* extent is returned in @es whose es_lblk, es_len, and es_pblk components
* are 0.
*/
static void __es_find_extent_range(struct inode *inode,
int (*matching_fn)(struct extent_status *es),
ext4_lblk_t lblk, ext4_lblk_t end,
struct extent_status *es)
{
struct ext4_es_tree *tree = NULL;
struct extent_status *es1 = NULL;
struct rb_node *node;
WARN_ON(es == NULL);
WARN_ON(end < lblk);
tree = &EXT4_I(inode)->i_es_tree;
/* see if the extent has been cached */
es->es_lblk = es->es_len = es->es_pblk = 0;
if (tree->cache_es) {
es1 = tree->cache_es;
if (in_range(lblk, es1->es_lblk, es1->es_len)) {
es_debug("%u cached by [%u/%u) %llu %x\n",
lblk, es1->es_lblk, es1->es_len,
ext4_es_pblock(es1), ext4_es_status(es1));
goto out;
}
}
es1 = __es_tree_search(&tree->root, lblk);
out:
if (es1 && !matching_fn(es1)) {
while ((node = rb_next(&es1->rb_node)) != NULL) {
es1 = rb_entry(node, struct extent_status, rb_node);
if (es1->es_lblk > end) {
es1 = NULL;
break;
}
if (matching_fn(es1))
break;
}
}
if (es1 && matching_fn(es1)) {
tree->cache_es = es1;
es->es_lblk = es1->es_lblk;
es->es_len = es1->es_len;
es->es_pblk = es1->es_pblk;
}
}
/*
* Locking for __es_find_extent_range() for external use
*/
void ext4_es_find_extent_range(struct inode *inode,
int (*matching_fn)(struct extent_status *es),
ext4_lblk_t lblk, ext4_lblk_t end,
struct extent_status *es)
{
trace_ext4_es_find_extent_range_enter(inode, lblk);
read_lock(&EXT4_I(inode)->i_es_lock);
__es_find_extent_range(inode, matching_fn, lblk, end, es);
read_unlock(&EXT4_I(inode)->i_es_lock);
trace_ext4_es_find_extent_range_exit(inode, es);
}
/*
* __es_scan_range - search block range for block with specified status
* in extents status tree
*
* @inode - file containing the range
* @matching_fn - pointer to function that matches extents with desired status
* @lblk - logical block defining start of range
* @end - logical block defining end of range
*
* Returns true if at least one block in the specified block range satisfies
* the criterion specified by @matching_fn, and false if not. If at least
* one extent has the specified status, then there is at least one block
* in the cluster with that status. Should only be called by code that has
* taken i_es_lock.
*/
static bool __es_scan_range(struct inode *inode,
int (*matching_fn)(struct extent_status *es),
ext4_lblk_t start, ext4_lblk_t end)
{
struct extent_status es;
__es_find_extent_range(inode, matching_fn, start, end, &es);
if (es.es_len == 0)
return false; /* no matching extent in the tree */
else if (es.es_lblk <= start &&
start < es.es_lblk + es.es_len)
return true;
else if (start <= es.es_lblk && es.es_lblk <= end)
return true;
else
return false;
}
/*
* Locking for __es_scan_range() for external use
*/
bool ext4_es_scan_range(struct inode *inode,
int (*matching_fn)(struct extent_status *es),
ext4_lblk_t lblk, ext4_lblk_t end)
{
bool ret;
read_lock(&EXT4_I(inode)->i_es_lock);
ret = __es_scan_range(inode, matching_fn, lblk, end);
read_unlock(&EXT4_I(inode)->i_es_lock);
return ret;
}
/*
* __es_scan_clu - search cluster for block with specified status in
* extents status tree
*
* @inode - file containing the cluster
* @matching_fn - pointer to function that matches extents with desired status
* @lblk - logical block in cluster to be searched
*
* Returns true if at least one extent in the cluster containing @lblk
* satisfies the criterion specified by @matching_fn, and false if not. If at
* least one extent has the specified status, then there is at least one block
* in the cluster with that status. Should only be called by code that has
* taken i_es_lock.
*/
static bool __es_scan_clu(struct inode *inode,
int (*matching_fn)(struct extent_status *es),
ext4_lblk_t lblk)
{
struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
ext4_lblk_t lblk_start, lblk_end;
lblk_start = EXT4_LBLK_CMASK(sbi, lblk);
lblk_end = lblk_start + sbi->s_cluster_ratio - 1;
return __es_scan_range(inode, matching_fn, lblk_start, lblk_end);
}
/*
* Locking for __es_scan_clu() for external use
*/
bool ext4_es_scan_clu(struct inode *inode,
int (*matching_fn)(struct extent_status *es),
ext4_lblk_t lblk)
{
bool ret;
read_lock(&EXT4_I(inode)->i_es_lock);
ret = __es_scan_clu(inode, matching_fn, lblk);
read_unlock(&EXT4_I(inode)->i_es_lock);
return ret;
}
static void ext4_es_list_add(struct inode *inode)
{
struct ext4_inode_info *ei = EXT4_I(inode);
struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
if (!list_empty(&ei->i_es_list))
return;
spin_lock(&sbi->s_es_lock);
if (list_empty(&ei->i_es_list)) {
list_add_tail(&ei->i_es_list, &sbi->s_es_list);
sbi->s_es_nr_inode++;
}
spin_unlock(&sbi->s_es_lock);
}
static void ext4_es_list_del(struct inode *inode)
{
struct ext4_inode_info *ei = EXT4_I(inode);
struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
spin_lock(&sbi->s_es_lock);
if (!list_empty(&ei->i_es_list)) {
list_del_init(&ei->i_es_list);
sbi->s_es_nr_inode--;
WARN_ON_ONCE(sbi->s_es_nr_inode < 0);
}
spin_unlock(&sbi->s_es_lock);
}
static struct extent_status *
ext4_es_alloc_extent(struct inode *inode, ext4_lblk_t lblk, ext4_lblk_t len,
ext4_fsblk_t pblk)
{
struct extent_status *es;
es = kmem_cache_alloc(ext4_es_cachep, GFP_ATOMIC);
if (es == NULL)
return NULL;
es->es_lblk = lblk;
es->es_len = len;
es->es_pblk = pblk;
/*
* We don't count delayed extent because we never try to reclaim them
*/
if (!ext4_es_is_delayed(es)) {
if (!EXT4_I(inode)->i_es_shk_nr++)
ext4_es_list_add(inode);
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
percpu_counter_inc(&EXT4_SB(inode->i_sb)->
s_es_stats.es_stats_shk_cnt);
}
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
EXT4_I(inode)->i_es_all_nr++;
percpu_counter_inc(&EXT4_SB(inode->i_sb)->s_es_stats.es_stats_all_cnt);
return es;
}
static void ext4_es_free_extent(struct inode *inode, struct extent_status *es)
{
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
EXT4_I(inode)->i_es_all_nr--;
percpu_counter_dec(&EXT4_SB(inode->i_sb)->s_es_stats.es_stats_all_cnt);
/* Decrease the shrink counter when this es is not delayed */
if (!ext4_es_is_delayed(es)) {
BUG_ON(EXT4_I(inode)->i_es_shk_nr == 0);
if (!--EXT4_I(inode)->i_es_shk_nr)
ext4_es_list_del(inode);
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
percpu_counter_dec(&EXT4_SB(inode->i_sb)->
s_es_stats.es_stats_shk_cnt);
}
kmem_cache_free(ext4_es_cachep, es);
}
/*
* Check whether or not two extents can be merged
* Condition:
* - logical block number is contiguous
* - physical block number is contiguous
* - status is equal
*/
static int ext4_es_can_be_merged(struct extent_status *es1,
struct extent_status *es2)
{
if (ext4_es_type(es1) != ext4_es_type(es2))
return 0;
if (((__u64) es1->es_len) + es2->es_len > EXT_MAX_BLOCKS) {
pr_warn("ES assertion failed when merging extents. "
"The sum of lengths of es1 (%d) and es2 (%d) "
"is bigger than allowed file size (%d)\n",
es1->es_len, es2->es_len, EXT_MAX_BLOCKS);
WARN_ON(1);
return 0;
}
if (((__u64) es1->es_lblk) + es1->es_len != es2->es_lblk)
return 0;
if ((ext4_es_is_written(es1) || ext4_es_is_unwritten(es1)) &&
(ext4_es_pblock(es1) + es1->es_len == ext4_es_pblock(es2)))
return 1;
if (ext4_es_is_hole(es1))
return 1;
/* we need to check delayed extent is without unwritten status */
if (ext4_es_is_delayed(es1) && !ext4_es_is_unwritten(es1))
return 1;
return 0;
}
static struct extent_status *
ext4_es_try_to_merge_left(struct inode *inode, struct extent_status *es)
{
struct ext4_es_tree *tree = &EXT4_I(inode)->i_es_tree;
struct extent_status *es1;
struct rb_node *node;
node = rb_prev(&es->rb_node);
if (!node)
return es;
es1 = rb_entry(node, struct extent_status, rb_node);
if (ext4_es_can_be_merged(es1, es)) {
es1->es_len += es->es_len;
if (ext4_es_is_referenced(es))
ext4_es_set_referenced(es1);
rb_erase(&es->rb_node, &tree->root);
ext4_es_free_extent(inode, es);
es = es1;
}
return es;
}
static struct extent_status *
ext4_es_try_to_merge_right(struct inode *inode, struct extent_status *es)
{
struct ext4_es_tree *tree = &EXT4_I(inode)->i_es_tree;
struct extent_status *es1;
struct rb_node *node;
node = rb_next(&es->rb_node);
if (!node)
return es;
es1 = rb_entry(node, struct extent_status, rb_node);
if (ext4_es_can_be_merged(es, es1)) {
es->es_len += es1->es_len;
if (ext4_es_is_referenced(es1))
ext4_es_set_referenced(es);
rb_erase(node, &tree->root);
ext4_es_free_extent(inode, es1);
}
return es;
}
#ifdef ES_AGGRESSIVE_TEST
#include "ext4_extents.h" /* Needed when ES_AGGRESSIVE_TEST is defined */
static void ext4_es_insert_extent_ext_check(struct inode *inode,
struct extent_status *es)
{
struct ext4_ext_path *path = NULL;
struct ext4_extent *ex;
ext4_lblk_t ee_block;
ext4_fsblk_t ee_start;
unsigned short ee_len;
int depth, ee_status, es_status;
path = ext4_find_extent(inode, es->es_lblk, NULL, EXT4_EX_NOCACHE);
if (IS_ERR(path))
return;
depth = ext_depth(inode);
ex = path[depth].p_ext;
if (ex) {
ee_block = le32_to_cpu(ex->ee_block);
ee_start = ext4_ext_pblock(ex);
ee_len = ext4_ext_get_actual_len(ex);
ee_status = ext4_ext_is_unwritten(ex) ? 1 : 0;
es_status = ext4_es_is_unwritten(es) ? 1 : 0;
/*
* Make sure ex and es are not overlap when we try to insert
* a delayed/hole extent.
*/
if (!ext4_es_is_written(es) && !ext4_es_is_unwritten(es)) {
if (in_range(es->es_lblk, ee_block, ee_len)) {
pr_warn("ES insert assertion failed for "
"inode: %lu we can find an extent "
"at block [%d/%d/%llu/%c], but we "
"want to add a delayed/hole extent "
"[%d/%d/%llu/%x]\n",
inode->i_ino, ee_block, ee_len,
ee_start, ee_status ? 'u' : 'w',
es->es_lblk, es->es_len,
ext4_es_pblock(es), ext4_es_status(es));
}
goto out;
}
/*
* We don't check ee_block == es->es_lblk, etc. because es
* might be a part of whole extent, vice versa.
*/
if (es->es_lblk < ee_block ||
ext4_es_pblock(es) != ee_start + es->es_lblk - ee_block) {
pr_warn("ES insert assertion failed for inode: %lu "
"ex_status [%d/%d/%llu/%c] != "
"es_status [%d/%d/%llu/%c]\n", inode->i_ino,
ee_block, ee_len, ee_start,
ee_status ? 'u' : 'w', es->es_lblk, es->es_len,
ext4_es_pblock(es), es_status ? 'u' : 'w');
goto out;
}
if (ee_status ^ es_status) {
pr_warn("ES insert assertion failed for inode: %lu "
"ex_status [%d/%d/%llu/%c] != "
"es_status [%d/%d/%llu/%c]\n", inode->i_ino,
ee_block, ee_len, ee_start,
ee_status ? 'u' : 'w', es->es_lblk, es->es_len,
ext4_es_pblock(es), es_status ? 'u' : 'w');
}
} else {
/*
* We can't find an extent on disk. So we need to make sure
* that we don't want to add an written/unwritten extent.
*/
if (!ext4_es_is_delayed(es) && !ext4_es_is_hole(es)) {
pr_warn("ES insert assertion failed for inode: %lu "
"can't find an extent at block %d but we want "
"to add a written/unwritten extent "
"[%d/%d/%llu/%x]\n", inode->i_ino,
es->es_lblk, es->es_lblk, es->es_len,
ext4_es_pblock(es), ext4_es_status(es));
}
}
out:
ext4_ext_drop_refs(path);
kfree(path);
}
static void ext4_es_insert_extent_ind_check(struct inode *inode,
struct extent_status *es)
{
struct ext4_map_blocks map;
int retval;
/*
* Here we call ext4_ind_map_blocks to lookup a block mapping because
* 'Indirect' structure is defined in indirect.c. So we couldn't
* access direct/indirect tree from outside. It is too dirty to define
* this function in indirect.c file.
*/
map.m_lblk = es->es_lblk;
map.m_len = es->es_len;
retval = ext4_ind_map_blocks(NULL, inode, &map, 0);
if (retval > 0) {
if (ext4_es_is_delayed(es) || ext4_es_is_hole(es)) {
/*
* We want to add a delayed/hole extent but this
* block has been allocated.
*/
pr_warn("ES insert assertion failed for inode: %lu "
"We can find blocks but we want to add a "
"delayed/hole extent [%d/%d/%llu/%x]\n",
inode->i_ino, es->es_lblk, es->es_len,
ext4_es_pblock(es), ext4_es_status(es));
return;
} else if (ext4_es_is_written(es)) {
if (retval != es->es_len) {
pr_warn("ES insert assertion failed for "
"inode: %lu retval %d != es_len %d\n",
inode->i_ino, retval, es->es_len);
return;
}
if (map.m_pblk != ext4_es_pblock(es)) {
pr_warn("ES insert assertion failed for "
"inode: %lu m_pblk %llu != "
"es_pblk %llu\n",
inode->i_ino, map.m_pblk,
ext4_es_pblock(es));
return;
}
} else {
/*
* We don't need to check unwritten extent because
* indirect-based file doesn't have it.
*/
BUG();
}
} else if (retval == 0) {
if (ext4_es_is_written(es)) {
pr_warn("ES insert assertion failed for inode: %lu "
"We can't find the block but we want to add "
"a written extent [%d/%d/%llu/%x]\n",
inode->i_ino, es->es_lblk, es->es_len,
ext4_es_pblock(es), ext4_es_status(es));
return;
}
}
}
static inline void ext4_es_insert_extent_check(struct inode *inode,
struct extent_status *es)
{
/*
* We don't need to worry about the race condition because
* caller takes i_data_sem locking.
*/
BUG_ON(!rwsem_is_locked(&EXT4_I(inode)->i_data_sem));
if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
ext4_es_insert_extent_ext_check(inode, es);
else
ext4_es_insert_extent_ind_check(inode, es);
}
#else
static inline void ext4_es_insert_extent_check(struct inode *inode,
struct extent_status *es)
{
}
#endif
static int __es_insert_extent(struct inode *inode, struct extent_status *newes)
{
struct ext4_es_tree *tree = &EXT4_I(inode)->i_es_tree;
struct rb_node **p = &tree->root.rb_node;
struct rb_node *parent = NULL;
struct extent_status *es;
while (*p) {
parent = *p;
es = rb_entry(parent, struct extent_status, rb_node);
if (newes->es_lblk < es->es_lblk) {
if (ext4_es_can_be_merged(newes, es)) {
/*
* Here we can modify es_lblk directly
* because it isn't overlapped.
*/
es->es_lblk = newes->es_lblk;
es->es_len += newes->es_len;
if (ext4_es_is_written(es) ||
ext4_es_is_unwritten(es))
ext4_es_store_pblock(es,
newes->es_pblk);
es = ext4_es_try_to_merge_left(inode, es);
goto out;
}
p = &(*p)->rb_left;
} else if (newes->es_lblk > ext4_es_end(es)) {
if (ext4_es_can_be_merged(es, newes)) {
es->es_len += newes->es_len;
es = ext4_es_try_to_merge_right(inode, es);
goto out;
}
p = &(*p)->rb_right;
} else {
BUG();
return -EINVAL;
}
}
es = ext4_es_alloc_extent(inode, newes->es_lblk, newes->es_len,
newes->es_pblk);
if (!es)
return -ENOMEM;
rb_link_node(&es->rb_node, parent, p);
rb_insert_color(&es->rb_node, &tree->root);
out:
tree->cache_es = es;
return 0;
}
/*
* ext4_es_insert_extent() adds information to an inode's extent
* status tree.
*
* Return 0 on success, error code on failure.
*/
int ext4_es_insert_extent(struct inode *inode, ext4_lblk_t lblk,
ext4_lblk_t len, ext4_fsblk_t pblk,
unsigned int status)
{
struct extent_status newes;
ext4_lblk_t end = lblk + len - 1;
int err = 0;
struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
es_debug("add [%u/%u) %llu %x to extent status tree of inode %lu\n",
lblk, len, pblk, status, inode->i_ino);
if (!len)
return 0;
BUG_ON(end < lblk);
ext4: fix data corruption caused by unwritten and delayed extents Currently it is possible to lose whole file system block worth of data when we hit the specific interaction with unwritten and delayed extents in status extent tree. The problem is that when we insert delayed extent into extent status tree the only way to get rid of it is when we write out delayed buffer. However there is a limitation in the extent status tree implementation so that when inserting unwritten extent should there be even a single delayed block the whole unwritten extent would be marked as delayed. At this point, there is no way to get rid of the delayed extents, because there are no delayed buffers to write out. So when a we write into said unwritten extent we will convert it to written, but it still remains delayed. When we try to write into that block later ext4_da_map_blocks() will set the buffer new and delayed and map it to invalid block which causes the rest of the block to be zeroed loosing already written data. For now we can fix this by simply not allowing to set delayed status on written extent in the extent status tree. Also add WARN_ON() to make sure that we notice if this happens in the future. This problem can be easily reproduced by running the following xfs_io. xfs_io -f -c "pwrite -S 0xaa 4096 2048" \ -c "falloc 0 131072" \ -c "pwrite -S 0xbb 65536 2048" \ -c "fsync" /mnt/test/fff echo 3 > /proc/sys/vm/drop_caches xfs_io -c "pwrite -S 0xdd 67584 2048" /mnt/test/fff This can be theoretically also reproduced by at random by running fsx, but it's not very reliable, though on machines with bigger page size (like ppc) this can be seen more often (especially xfstest generic/127) Signed-off-by: Lukas Czerner <lczerner@redhat.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu> Cc: stable@vger.kernel.org
2015-05-03 04:36:55 +03:00
if ((status & EXTENT_STATUS_DELAYED) &&
(status & EXTENT_STATUS_WRITTEN)) {
ext4_warning(inode->i_sb, "Inserting extent [%u/%u] as "
" delayed and written which can potentially "
" cause data loss.", lblk, len);
ext4: fix data corruption caused by unwritten and delayed extents Currently it is possible to lose whole file system block worth of data when we hit the specific interaction with unwritten and delayed extents in status extent tree. The problem is that when we insert delayed extent into extent status tree the only way to get rid of it is when we write out delayed buffer. However there is a limitation in the extent status tree implementation so that when inserting unwritten extent should there be even a single delayed block the whole unwritten extent would be marked as delayed. At this point, there is no way to get rid of the delayed extents, because there are no delayed buffers to write out. So when a we write into said unwritten extent we will convert it to written, but it still remains delayed. When we try to write into that block later ext4_da_map_blocks() will set the buffer new and delayed and map it to invalid block which causes the rest of the block to be zeroed loosing already written data. For now we can fix this by simply not allowing to set delayed status on written extent in the extent status tree. Also add WARN_ON() to make sure that we notice if this happens in the future. This problem can be easily reproduced by running the following xfs_io. xfs_io -f -c "pwrite -S 0xaa 4096 2048" \ -c "falloc 0 131072" \ -c "pwrite -S 0xbb 65536 2048" \ -c "fsync" /mnt/test/fff echo 3 > /proc/sys/vm/drop_caches xfs_io -c "pwrite -S 0xdd 67584 2048" /mnt/test/fff This can be theoretically also reproduced by at random by running fsx, but it's not very reliable, though on machines with bigger page size (like ppc) this can be seen more often (especially xfstest generic/127) Signed-off-by: Lukas Czerner <lczerner@redhat.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu> Cc: stable@vger.kernel.org
2015-05-03 04:36:55 +03:00
WARN_ON(1);
}
newes.es_lblk = lblk;
newes.es_len = len;
ext4_es_store_pblock_status(&newes, pblk, status);
trace_ext4_es_insert_extent(inode, &newes);
ext4_es_insert_extent_check(inode, &newes);
write_lock(&EXT4_I(inode)->i_es_lock);
err = __es_remove_extent(inode, lblk, end, NULL);
if (err != 0)
goto error;
retry:
err = __es_insert_extent(inode, &newes);
if (err == -ENOMEM && __es_shrink(EXT4_SB(inode->i_sb),
128, EXT4_I(inode)))
goto retry;
if (err == -ENOMEM && !ext4_es_is_delayed(&newes))
err = 0;
if (sbi->s_cluster_ratio > 1 && test_opt(inode->i_sb, DELALLOC) &&
(status & EXTENT_STATUS_WRITTEN ||
status & EXTENT_STATUS_UNWRITTEN))
__revise_pending(inode, lblk, len);
error:
write_unlock(&EXT4_I(inode)->i_es_lock);
ext4_es_print_tree(inode);
return err;
}
/*
* ext4_es_cache_extent() inserts information into the extent status
* tree if and only if there isn't information about the range in
* question already.
*/
void ext4_es_cache_extent(struct inode *inode, ext4_lblk_t lblk,
ext4_lblk_t len, ext4_fsblk_t pblk,
unsigned int status)
{
struct extent_status *es;
struct extent_status newes;
ext4_lblk_t end = lblk + len - 1;
newes.es_lblk = lblk;
newes.es_len = len;
ext4_es_store_pblock_status(&newes, pblk, status);
trace_ext4_es_cache_extent(inode, &newes);
if (!len)
return;
BUG_ON(end < lblk);
write_lock(&EXT4_I(inode)->i_es_lock);
es = __es_tree_search(&EXT4_I(inode)->i_es_tree.root, lblk);
if (!es || es->es_lblk > end)
__es_insert_extent(inode, &newes);
write_unlock(&EXT4_I(inode)->i_es_lock);
}
/*
* ext4_es_lookup_extent() looks up an extent in extent status tree.
*
* ext4_es_lookup_extent is called by ext4_map_blocks/ext4_da_map_blocks.
*
* Return: 1 on found, 0 on not
*/
int ext4_es_lookup_extent(struct inode *inode, ext4_lblk_t lblk,
ext4_lblk_t *next_lblk,
struct extent_status *es)
{
struct ext4_es_tree *tree;
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
struct ext4_es_stats *stats;
struct extent_status *es1 = NULL;
struct rb_node *node;
int found = 0;
trace_ext4_es_lookup_extent_enter(inode, lblk);
es_debug("lookup extent in block %u\n", lblk);
tree = &EXT4_I(inode)->i_es_tree;
read_lock(&EXT4_I(inode)->i_es_lock);
/* find extent in cache firstly */
es->es_lblk = es->es_len = es->es_pblk = 0;
if (tree->cache_es) {
es1 = tree->cache_es;
if (in_range(lblk, es1->es_lblk, es1->es_len)) {
es_debug("%u cached by [%u/%u)\n",
lblk, es1->es_lblk, es1->es_len);
found = 1;
goto out;
}
}
node = tree->root.rb_node;
while (node) {
es1 = rb_entry(node, struct extent_status, rb_node);
if (lblk < es1->es_lblk)
node = node->rb_left;
else if (lblk > ext4_es_end(es1))
node = node->rb_right;
else {
found = 1;
break;
}
}
out:
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
stats = &EXT4_SB(inode->i_sb)->s_es_stats;
if (found) {
BUG_ON(!es1);
es->es_lblk = es1->es_lblk;
es->es_len = es1->es_len;
es->es_pblk = es1->es_pblk;
if (!ext4_es_is_referenced(es1))
ext4_es_set_referenced(es1);
percpu_counter_inc(&stats->es_stats_cache_hits);
if (next_lblk) {
node = rb_next(&es1->rb_node);
if (node) {
es1 = rb_entry(node, struct extent_status,
rb_node);
*next_lblk = es1->es_lblk;
} else
*next_lblk = 0;
}
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
} else {
percpu_counter_inc(&stats->es_stats_cache_misses);
}
read_unlock(&EXT4_I(inode)->i_es_lock);
trace_ext4_es_lookup_extent_exit(inode, es, found);
return found;
}
struct rsvd_count {
int ndelonly;
bool first_do_lblk_found;
ext4_lblk_t first_do_lblk;
ext4_lblk_t last_do_lblk;
struct extent_status *left_es;
bool partial;
ext4_lblk_t lclu;
};
/*
* init_rsvd - initialize reserved count data before removing block range
* in file from extent status tree
*
* @inode - file containing range
* @lblk - first block in range
* @es - pointer to first extent in range
* @rc - pointer to reserved count data
*
* Assumes es is not NULL
*/
static void init_rsvd(struct inode *inode, ext4_lblk_t lblk,
struct extent_status *es, struct rsvd_count *rc)
{
struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
struct rb_node *node;
rc->ndelonly = 0;
/*
* for bigalloc, note the first delonly block in the range has not
* been found, record the extent containing the block to the left of
* the region to be removed, if any, and note that there's no partial
* cluster to track
*/
if (sbi->s_cluster_ratio > 1) {
rc->first_do_lblk_found = false;
if (lblk > es->es_lblk) {
rc->left_es = es;
} else {
node = rb_prev(&es->rb_node);
rc->left_es = node ? rb_entry(node,
struct extent_status,
rb_node) : NULL;
}
rc->partial = false;
}
}
/*
* count_rsvd - count the clusters containing delayed and not unwritten
* (delonly) blocks in a range within an extent and add to
* the running tally in rsvd_count
*
* @inode - file containing extent
* @lblk - first block in range
* @len - length of range in blocks
* @es - pointer to extent containing clusters to be counted
* @rc - pointer to reserved count data
*
* Tracks partial clusters found at the beginning and end of extents so
* they aren't overcounted when they span adjacent extents
*/
static void count_rsvd(struct inode *inode, ext4_lblk_t lblk, long len,
struct extent_status *es, struct rsvd_count *rc)
{
struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
ext4_lblk_t i, end, nclu;
if (!ext4_es_is_delonly(es))
return;
WARN_ON(len <= 0);
if (sbi->s_cluster_ratio == 1) {
rc->ndelonly += (int) len;
return;
}
/* bigalloc */
i = (lblk < es->es_lblk) ? es->es_lblk : lblk;
end = lblk + (ext4_lblk_t) len - 1;
end = (end > ext4_es_end(es)) ? ext4_es_end(es) : end;
/* record the first block of the first delonly extent seen */
if (!rc->first_do_lblk_found) {
rc->first_do_lblk = i;
rc->first_do_lblk_found = true;
}
/* update the last lblk in the region seen so far */
rc->last_do_lblk = end;
/*
* if we're tracking a partial cluster and the current extent
* doesn't start with it, count it and stop tracking
*/
if (rc->partial && (rc->lclu != EXT4_B2C(sbi, i))) {
rc->ndelonly++;
rc->partial = false;
}
/*
* if the first cluster doesn't start on a cluster boundary but
* ends on one, count it
*/
if (EXT4_LBLK_COFF(sbi, i) != 0) {
if (end >= EXT4_LBLK_CFILL(sbi, i)) {
rc->ndelonly++;
rc->partial = false;
i = EXT4_LBLK_CFILL(sbi, i) + 1;
}
}
/*
* if the current cluster starts on a cluster boundary, count the
* number of whole delonly clusters in the extent
*/
if ((i + sbi->s_cluster_ratio - 1) <= end) {
nclu = (end - i + 1) >> sbi->s_cluster_bits;
rc->ndelonly += nclu;
i += nclu << sbi->s_cluster_bits;
}
/*
* start tracking a partial cluster if there's a partial at the end
* of the current extent and we're not already tracking one
*/
if (!rc->partial && i <= end) {
rc->partial = true;
rc->lclu = EXT4_B2C(sbi, i);
}
}
/*
* __pr_tree_search - search for a pending cluster reservation
*
* @root - root of pending reservation tree
* @lclu - logical cluster to search for
*
* Returns the pending reservation for the cluster identified by @lclu
* if found. If not, returns a reservation for the next cluster if any,
* and if not, returns NULL.
*/
static struct pending_reservation *__pr_tree_search(struct rb_root *root,
ext4_lblk_t lclu)
{
struct rb_node *node = root->rb_node;
struct pending_reservation *pr = NULL;
while (node) {
pr = rb_entry(node, struct pending_reservation, rb_node);
if (lclu < pr->lclu)
node = node->rb_left;
else if (lclu > pr->lclu)
node = node->rb_right;
else
return pr;
}
if (pr && lclu < pr->lclu)
return pr;
if (pr && lclu > pr->lclu) {
node = rb_next(&pr->rb_node);
return node ? rb_entry(node, struct pending_reservation,
rb_node) : NULL;
}
return NULL;
}
/*
* get_rsvd - calculates and returns the number of cluster reservations to be
* released when removing a block range from the extent status tree
* and releases any pending reservations within the range
*
* @inode - file containing block range
* @end - last block in range
* @right_es - pointer to extent containing next block beyond end or NULL
* @rc - pointer to reserved count data
*
* The number of reservations to be released is equal to the number of
* clusters containing delayed and not unwritten (delonly) blocks within
* the range, minus the number of clusters still containing delonly blocks
* at the ends of the range, and minus the number of pending reservations
* within the range.
*/
static unsigned int get_rsvd(struct inode *inode, ext4_lblk_t end,
struct extent_status *right_es,
struct rsvd_count *rc)
{
struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
struct pending_reservation *pr;
struct ext4_pending_tree *tree = &EXT4_I(inode)->i_pending_tree;
struct rb_node *node;
ext4_lblk_t first_lclu, last_lclu;
bool left_delonly, right_delonly, count_pending;
struct extent_status *es;
if (sbi->s_cluster_ratio > 1) {
/* count any remaining partial cluster */
if (rc->partial)
rc->ndelonly++;
if (rc->ndelonly == 0)
return 0;
first_lclu = EXT4_B2C(sbi, rc->first_do_lblk);
last_lclu = EXT4_B2C(sbi, rc->last_do_lblk);
/*
* decrease the delonly count by the number of clusters at the
* ends of the range that still contain delonly blocks -
* these clusters still need to be reserved
*/
left_delonly = right_delonly = false;
es = rc->left_es;
while (es && ext4_es_end(es) >=
EXT4_LBLK_CMASK(sbi, rc->first_do_lblk)) {
if (ext4_es_is_delonly(es)) {
rc->ndelonly--;
left_delonly = true;
break;
}
node = rb_prev(&es->rb_node);
if (!node)
break;
es = rb_entry(node, struct extent_status, rb_node);
}
if (right_es && (!left_delonly || first_lclu != last_lclu)) {
if (end < ext4_es_end(right_es)) {
es = right_es;
} else {
node = rb_next(&right_es->rb_node);
es = node ? rb_entry(node, struct extent_status,
rb_node) : NULL;
}
while (es && es->es_lblk <=
EXT4_LBLK_CFILL(sbi, rc->last_do_lblk)) {
if (ext4_es_is_delonly(es)) {
rc->ndelonly--;
right_delonly = true;
break;
}
node = rb_next(&es->rb_node);
if (!node)
break;
es = rb_entry(node, struct extent_status,
rb_node);
}
}
/*
* Determine the block range that should be searched for
* pending reservations, if any. Clusters on the ends of the
* original removed range containing delonly blocks are
* excluded. They've already been accounted for and it's not
* possible to determine if an associated pending reservation
* should be released with the information available in the
* extents status tree.
*/
if (first_lclu == last_lclu) {
if (left_delonly | right_delonly)
count_pending = false;
else
count_pending = true;
} else {
if (left_delonly)
first_lclu++;
if (right_delonly)
last_lclu--;
if (first_lclu <= last_lclu)
count_pending = true;
else
count_pending = false;
}
/*
* a pending reservation found between first_lclu and last_lclu
* represents an allocated cluster that contained at least one
* delonly block, so the delonly total must be reduced by one
* for each pending reservation found and released
*/
if (count_pending) {
pr = __pr_tree_search(&tree->root, first_lclu);
while (pr && pr->lclu <= last_lclu) {
rc->ndelonly--;
node = rb_next(&pr->rb_node);
rb_erase(&pr->rb_node, &tree->root);
kmem_cache_free(ext4_pending_cachep, pr);
if (!node)
break;
pr = rb_entry(node, struct pending_reservation,
rb_node);
}
}
}
return rc->ndelonly;
}
/*
* __es_remove_extent - removes block range from extent status tree
*
* @inode - file containing range
* @lblk - first block in range
* @end - last block in range
* @reserved - number of cluster reservations released
*
* If @reserved is not NULL and delayed allocation is enabled, counts
* block/cluster reservations freed by removing range and if bigalloc
* enabled cancels pending reservations as needed. Returns 0 on success,
* error code on failure.
*/
static int __es_remove_extent(struct inode *inode, ext4_lblk_t lblk,
ext4_lblk_t end, int *reserved)
{
struct ext4_es_tree *tree = &EXT4_I(inode)->i_es_tree;
struct rb_node *node;
struct extent_status *es;
struct extent_status orig_es;
ext4_lblk_t len1, len2;
ext4_fsblk_t block;
int err;
bool count_reserved = true;
struct rsvd_count rc;
if (reserved == NULL || !test_opt(inode->i_sb, DELALLOC))
count_reserved = false;
retry:
err = 0;
es = __es_tree_search(&tree->root, lblk);
if (!es)
goto out;
if (es->es_lblk > end)
goto out;
/* Simply invalidate cache_es. */
tree->cache_es = NULL;
if (count_reserved)
init_rsvd(inode, lblk, es, &rc);
orig_es.es_lblk = es->es_lblk;
orig_es.es_len = es->es_len;
orig_es.es_pblk = es->es_pblk;
len1 = lblk > es->es_lblk ? lblk - es->es_lblk : 0;
len2 = ext4_es_end(es) > end ? ext4_es_end(es) - end : 0;
if (len1 > 0)
es->es_len = len1;
if (len2 > 0) {
if (len1 > 0) {
struct extent_status newes;
newes.es_lblk = end + 1;
newes.es_len = len2;
block = 0x7FDEADBEEFULL;
if (ext4_es_is_written(&orig_es) ||
ext4_es_is_unwritten(&orig_es))
block = ext4_es_pblock(&orig_es) +
orig_es.es_len - len2;
ext4_es_store_pblock_status(&newes, block,
ext4_es_status(&orig_es));
err = __es_insert_extent(inode, &newes);
if (err) {
es->es_lblk = orig_es.es_lblk;
es->es_len = orig_es.es_len;
if ((err == -ENOMEM) &&
__es_shrink(EXT4_SB(inode->i_sb),
128, EXT4_I(inode)))
goto retry;
goto out;
}
} else {
es->es_lblk = end + 1;
es->es_len = len2;
if (ext4_es_is_written(es) ||
ext4_es_is_unwritten(es)) {
block = orig_es.es_pblk + orig_es.es_len - len2;
ext4_es_store_pblock(es, block);
}
}
if (count_reserved)
count_rsvd(inode, lblk, orig_es.es_len - len1 - len2,
&orig_es, &rc);
goto out;
}
if (len1 > 0) {
if (count_reserved)
count_rsvd(inode, lblk, orig_es.es_len - len1,
&orig_es, &rc);
node = rb_next(&es->rb_node);
if (node)
es = rb_entry(node, struct extent_status, rb_node);
else
es = NULL;
}
while (es && ext4_es_end(es) <= end) {
if (count_reserved)
count_rsvd(inode, es->es_lblk, es->es_len, es, &rc);
node = rb_next(&es->rb_node);
rb_erase(&es->rb_node, &tree->root);
ext4_es_free_extent(inode, es);
if (!node) {
es = NULL;
break;
}
es = rb_entry(node, struct extent_status, rb_node);
}
if (es && es->es_lblk < end + 1) {
ext4_lblk_t orig_len = es->es_len;
len1 = ext4_es_end(es) - end;
if (count_reserved)
count_rsvd(inode, es->es_lblk, orig_len - len1,
es, &rc);
es->es_lblk = end + 1;
es->es_len = len1;
if (ext4_es_is_written(es) || ext4_es_is_unwritten(es)) {
block = es->es_pblk + orig_len - len1;
ext4_es_store_pblock(es, block);
}
}
if (count_reserved)
*reserved = get_rsvd(inode, end, es, &rc);
out:
return err;
}
/*
* ext4_es_remove_extent - removes block range from extent status tree
*
* @inode - file containing range
* @lblk - first block in range
* @len - number of blocks to remove
*
* Reduces block/cluster reservation count and for bigalloc cancels pending
* reservations as needed. Returns 0 on success, error code on failure.
*/
int ext4_es_remove_extent(struct inode *inode, ext4_lblk_t lblk,
ext4_lblk_t len)
{
ext4_lblk_t end;
int err = 0;
int reserved = 0;
trace_ext4_es_remove_extent(inode, lblk, len);
es_debug("remove [%u/%u) from extent status tree of inode %lu\n",
lblk, len, inode->i_ino);
if (!len)
return err;
end = lblk + len - 1;
BUG_ON(end < lblk);
/*
* ext4_clear_inode() depends on us taking i_es_lock unconditionally
* so that we are sure __es_shrink() is done with the inode before it
* is reclaimed.
*/
write_lock(&EXT4_I(inode)->i_es_lock);
err = __es_remove_extent(inode, lblk, end, &reserved);
write_unlock(&EXT4_I(inode)->i_es_lock);
ext4_es_print_tree(inode);
ext4_da_release_space(inode, reserved);
return err;
}
static int __es_shrink(struct ext4_sb_info *sbi, int nr_to_scan,
struct ext4_inode_info *locked_ei)
{
struct ext4_inode_info *ei;
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
struct ext4_es_stats *es_stats;
ktime_t start_time;
u64 scan_time;
int nr_to_walk;
fs: convert fs shrinkers to new scan/count API Convert the filesystem shrinkers to use the new API, and standardise some of the behaviours of the shrinkers at the same time. For example, nr_to_scan means the number of objects to scan, not the number of objects to free. I refactored the CIFS idmap shrinker a little - it really needs to be broken up into a shrinker per tree and keep an item count with the tree root so that we don't need to walk the tree every time the shrinker needs to count the number of objects in the tree (i.e. all the time under memory pressure). [glommer@openvz.org: fixes for ext4, ubifs, nfs, cifs and glock. Fixes are needed mainly due to new code merged in the tree] [assorted fixes folded in] Signed-off-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Glauber Costa <glommer@openvz.org> Acked-by: Mel Gorman <mgorman@suse.de> Acked-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Acked-by: Jan Kara <jack@suse.cz> Acked-by: Steven Whitehouse <swhiteho@redhat.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Cc: Arve Hjønnevåg <arve@android.com> Cc: Carlos Maiolino <cmaiolino@redhat.com> Cc: Christoph Hellwig <hch@lst.de> Cc: Chuck Lever <chuck.lever@oracle.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: David Rientjes <rientjes@google.com> Cc: Gleb Natapov <gleb@redhat.com> Cc: Greg Thelen <gthelen@google.com> Cc: J. Bruce Fields <bfields@redhat.com> Cc: Jan Kara <jack@suse.cz> Cc: Jerome Glisse <jglisse@redhat.com> Cc: John Stultz <john.stultz@linaro.org> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Cc: Kent Overstreet <koverstreet@google.com> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Mel Gorman <mgorman@suse.de> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Thomas Hellstrom <thellstrom@vmware.com> Cc: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2013-08-28 04:18:09 +04:00
int nr_shrunk = 0;
int retried = 0, nr_skipped = 0;
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
es_stats = &sbi->s_es_stats;
start_time = ktime_get();
ext4: improve extent cache shrink mechanism to avoid to burn CPU time Now we maintain an proper in-order LRU list in ext4 to reclaim entries from extent status tree when we are under heavy memory pressure. For keeping this order, a spin lock is used to protect this list. But this lock burns a lot of CPU time. We can use the following steps to trigger it. % cd /dev/shm % dd if=/dev/zero of=ext4-img bs=1M count=2k % mkfs.ext4 ext4-img % mount -t ext4 -o loop ext4-img /mnt % cd /mnt % for ((i=0;i<160;i++)); do truncate -s 64g $i; done % for ((i=0;i<160;i++)); do cp $i /dev/null &; done % perf record -a -g % perf report This commit tries to fix this problem. Now a new member called i_touch_when is added into ext4_inode_info to record the last access time for an inode. Meanwhile we never need to keep a proper in-order LRU list. So this can avoid to burns some CPU time. When we try to reclaim some entries from extent status tree, we use list_sort() to get a proper in-order list. Then we traverse this list to discard some entries. In ext4_sb_info, we use s_es_last_sorted to record the last time of sorting this list. When we traverse the list, we skip the inode that is newer than this time, and move this inode to the tail of LRU list. When the head of the list is newer than s_es_last_sorted, we will sort the LRU list again. In this commit, we break the loop if s_extent_cache_cnt == 0 because that means that all extents in extent status tree have been reclaimed. Meanwhile in this commit, ext4_es_{un}register_shrinker()'s prototype is changed to save a local variable in these functions. Reported-by: Dave Hansen <dave.hansen@intel.com> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2013-07-01 16:12:37 +04:00
retry:
spin_lock(&sbi->s_es_lock);
nr_to_walk = sbi->s_es_nr_inode;
while (nr_to_walk-- > 0) {
if (list_empty(&sbi->s_es_list)) {
spin_unlock(&sbi->s_es_lock);
goto out;
}
ei = list_first_entry(&sbi->s_es_list, struct ext4_inode_info,
i_es_list);
/* Move the inode to the tail */
list_move_tail(&ei->i_es_list, &sbi->s_es_list);
/*
* Normally we try hard to avoid shrinking precached inodes,
* but we will as a last resort.
*/
if (!retried && ext4_test_inode_state(&ei->vfs_inode,
EXT4_STATE_EXT_PRECACHED)) {
nr_skipped++;
continue;
}
ext4: improve extent cache shrink mechanism to avoid to burn CPU time Now we maintain an proper in-order LRU list in ext4 to reclaim entries from extent status tree when we are under heavy memory pressure. For keeping this order, a spin lock is used to protect this list. But this lock burns a lot of CPU time. We can use the following steps to trigger it. % cd /dev/shm % dd if=/dev/zero of=ext4-img bs=1M count=2k % mkfs.ext4 ext4-img % mount -t ext4 -o loop ext4-img /mnt % cd /mnt % for ((i=0;i<160;i++)); do truncate -s 64g $i; done % for ((i=0;i<160;i++)); do cp $i /dev/null &; done % perf record -a -g % perf report This commit tries to fix this problem. Now a new member called i_touch_when is added into ext4_inode_info to record the last access time for an inode. Meanwhile we never need to keep a proper in-order LRU list. So this can avoid to burns some CPU time. When we try to reclaim some entries from extent status tree, we use list_sort() to get a proper in-order list. Then we traverse this list to discard some entries. In ext4_sb_info, we use s_es_last_sorted to record the last time of sorting this list. When we traverse the list, we skip the inode that is newer than this time, and move this inode to the tail of LRU list. When the head of the list is newer than s_es_last_sorted, we will sort the LRU list again. In this commit, we break the loop if s_extent_cache_cnt == 0 because that means that all extents in extent status tree have been reclaimed. Meanwhile in this commit, ext4_es_{un}register_shrinker()'s prototype is changed to save a local variable in these functions. Reported-by: Dave Hansen <dave.hansen@intel.com> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2013-07-01 16:12:37 +04:00
if (ei == locked_ei || !write_trylock(&ei->i_es_lock)) {
nr_skipped++;
ext4: improve extent cache shrink mechanism to avoid to burn CPU time Now we maintain an proper in-order LRU list in ext4 to reclaim entries from extent status tree when we are under heavy memory pressure. For keeping this order, a spin lock is used to protect this list. But this lock burns a lot of CPU time. We can use the following steps to trigger it. % cd /dev/shm % dd if=/dev/zero of=ext4-img bs=1M count=2k % mkfs.ext4 ext4-img % mount -t ext4 -o loop ext4-img /mnt % cd /mnt % for ((i=0;i<160;i++)); do truncate -s 64g $i; done % for ((i=0;i<160;i++)); do cp $i /dev/null &; done % perf record -a -g % perf report This commit tries to fix this problem. Now a new member called i_touch_when is added into ext4_inode_info to record the last access time for an inode. Meanwhile we never need to keep a proper in-order LRU list. So this can avoid to burns some CPU time. When we try to reclaim some entries from extent status tree, we use list_sort() to get a proper in-order list. Then we traverse this list to discard some entries. In ext4_sb_info, we use s_es_last_sorted to record the last time of sorting this list. When we traverse the list, we skip the inode that is newer than this time, and move this inode to the tail of LRU list. When the head of the list is newer than s_es_last_sorted, we will sort the LRU list again. In this commit, we break the loop if s_extent_cache_cnt == 0 because that means that all extents in extent status tree have been reclaimed. Meanwhile in this commit, ext4_es_{un}register_shrinker()'s prototype is changed to save a local variable in these functions. Reported-by: Dave Hansen <dave.hansen@intel.com> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2013-07-01 16:12:37 +04:00
continue;
}
/*
* Now we hold i_es_lock which protects us from inode reclaim
* freeing inode under us
*/
spin_unlock(&sbi->s_es_lock);
nr_shrunk += es_reclaim_extents(ei, &nr_to_scan);
write_unlock(&ei->i_es_lock);
if (nr_to_scan <= 0)
goto out;
spin_lock(&sbi->s_es_lock);
}
spin_unlock(&sbi->s_es_lock);
/*
* If we skipped any inodes, and we weren't able to make any
* forward progress, try again to scan precached inodes.
*/
if ((nr_shrunk == 0) && nr_skipped && !retried) {
retried++;
goto retry;
}
if (locked_ei && nr_shrunk == 0)
nr_shrunk = es_reclaim_extents(locked_ei, &nr_to_scan);
out:
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
scan_time = ktime_to_ns(ktime_sub(ktime_get(), start_time));
if (likely(es_stats->es_stats_scan_time))
es_stats->es_stats_scan_time = (scan_time +
es_stats->es_stats_scan_time*3) / 4;
else
es_stats->es_stats_scan_time = scan_time;
if (scan_time > es_stats->es_stats_max_scan_time)
es_stats->es_stats_max_scan_time = scan_time;
if (likely(es_stats->es_stats_shrunk))
es_stats->es_stats_shrunk = (nr_shrunk +
es_stats->es_stats_shrunk*3) / 4;
else
es_stats->es_stats_shrunk = nr_shrunk;
trace_ext4_es_shrink(sbi->s_sb, nr_shrunk, scan_time,
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
nr_skipped, retried);
return nr_shrunk;
}
fs: convert fs shrinkers to new scan/count API Convert the filesystem shrinkers to use the new API, and standardise some of the behaviours of the shrinkers at the same time. For example, nr_to_scan means the number of objects to scan, not the number of objects to free. I refactored the CIFS idmap shrinker a little - it really needs to be broken up into a shrinker per tree and keep an item count with the tree root so that we don't need to walk the tree every time the shrinker needs to count the number of objects in the tree (i.e. all the time under memory pressure). [glommer@openvz.org: fixes for ext4, ubifs, nfs, cifs and glock. Fixes are needed mainly due to new code merged in the tree] [assorted fixes folded in] Signed-off-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Glauber Costa <glommer@openvz.org> Acked-by: Mel Gorman <mgorman@suse.de> Acked-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Acked-by: Jan Kara <jack@suse.cz> Acked-by: Steven Whitehouse <swhiteho@redhat.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Cc: Arve Hjønnevåg <arve@android.com> Cc: Carlos Maiolino <cmaiolino@redhat.com> Cc: Christoph Hellwig <hch@lst.de> Cc: Chuck Lever <chuck.lever@oracle.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: David Rientjes <rientjes@google.com> Cc: Gleb Natapov <gleb@redhat.com> Cc: Greg Thelen <gthelen@google.com> Cc: J. Bruce Fields <bfields@redhat.com> Cc: Jan Kara <jack@suse.cz> Cc: Jerome Glisse <jglisse@redhat.com> Cc: John Stultz <john.stultz@linaro.org> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Cc: Kent Overstreet <koverstreet@google.com> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Mel Gorman <mgorman@suse.de> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Thomas Hellstrom <thellstrom@vmware.com> Cc: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2013-08-28 04:18:09 +04:00
static unsigned long ext4_es_count(struct shrinker *shrink,
struct shrink_control *sc)
{
unsigned long nr;
struct ext4_sb_info *sbi;
sbi = container_of(shrink, struct ext4_sb_info, s_es_shrinker);
nr = percpu_counter_read_positive(&sbi->s_es_stats.es_stats_shk_cnt);
trace_ext4_es_shrink_count(sbi->s_sb, sc->nr_to_scan, nr);
fs: convert fs shrinkers to new scan/count API Convert the filesystem shrinkers to use the new API, and standardise some of the behaviours of the shrinkers at the same time. For example, nr_to_scan means the number of objects to scan, not the number of objects to free. I refactored the CIFS idmap shrinker a little - it really needs to be broken up into a shrinker per tree and keep an item count with the tree root so that we don't need to walk the tree every time the shrinker needs to count the number of objects in the tree (i.e. all the time under memory pressure). [glommer@openvz.org: fixes for ext4, ubifs, nfs, cifs and glock. Fixes are needed mainly due to new code merged in the tree] [assorted fixes folded in] Signed-off-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Glauber Costa <glommer@openvz.org> Acked-by: Mel Gorman <mgorman@suse.de> Acked-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Acked-by: Jan Kara <jack@suse.cz> Acked-by: Steven Whitehouse <swhiteho@redhat.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Cc: Arve Hjønnevåg <arve@android.com> Cc: Carlos Maiolino <cmaiolino@redhat.com> Cc: Christoph Hellwig <hch@lst.de> Cc: Chuck Lever <chuck.lever@oracle.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: David Rientjes <rientjes@google.com> Cc: Gleb Natapov <gleb@redhat.com> Cc: Greg Thelen <gthelen@google.com> Cc: J. Bruce Fields <bfields@redhat.com> Cc: Jan Kara <jack@suse.cz> Cc: Jerome Glisse <jglisse@redhat.com> Cc: John Stultz <john.stultz@linaro.org> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Cc: Kent Overstreet <koverstreet@google.com> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Mel Gorman <mgorman@suse.de> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Thomas Hellstrom <thellstrom@vmware.com> Cc: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2013-08-28 04:18:09 +04:00
return nr;
}
static unsigned long ext4_es_scan(struct shrinker *shrink,
struct shrink_control *sc)
{
struct ext4_sb_info *sbi = container_of(shrink,
struct ext4_sb_info, s_es_shrinker);
int nr_to_scan = sc->nr_to_scan;
int ret, nr_shrunk;
ret = percpu_counter_read_positive(&sbi->s_es_stats.es_stats_shk_cnt);
trace_ext4_es_shrink_scan_enter(sbi->s_sb, nr_to_scan, ret);
if (!nr_to_scan)
return ret;
nr_shrunk = __es_shrink(sbi, nr_to_scan, NULL);
trace_ext4_es_shrink_scan_exit(sbi->s_sb, nr_shrunk, ret);
fs: convert fs shrinkers to new scan/count API Convert the filesystem shrinkers to use the new API, and standardise some of the behaviours of the shrinkers at the same time. For example, nr_to_scan means the number of objects to scan, not the number of objects to free. I refactored the CIFS idmap shrinker a little - it really needs to be broken up into a shrinker per tree and keep an item count with the tree root so that we don't need to walk the tree every time the shrinker needs to count the number of objects in the tree (i.e. all the time under memory pressure). [glommer@openvz.org: fixes for ext4, ubifs, nfs, cifs and glock. Fixes are needed mainly due to new code merged in the tree] [assorted fixes folded in] Signed-off-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Glauber Costa <glommer@openvz.org> Acked-by: Mel Gorman <mgorman@suse.de> Acked-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Acked-by: Jan Kara <jack@suse.cz> Acked-by: Steven Whitehouse <swhiteho@redhat.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Cc: Arve Hjønnevåg <arve@android.com> Cc: Carlos Maiolino <cmaiolino@redhat.com> Cc: Christoph Hellwig <hch@lst.de> Cc: Chuck Lever <chuck.lever@oracle.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: David Rientjes <rientjes@google.com> Cc: Gleb Natapov <gleb@redhat.com> Cc: Greg Thelen <gthelen@google.com> Cc: J. Bruce Fields <bfields@redhat.com> Cc: Jan Kara <jack@suse.cz> Cc: Jerome Glisse <jglisse@redhat.com> Cc: John Stultz <john.stultz@linaro.org> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Cc: Kent Overstreet <koverstreet@google.com> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Mel Gorman <mgorman@suse.de> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Thomas Hellstrom <thellstrom@vmware.com> Cc: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2013-08-28 04:18:09 +04:00
return nr_shrunk;
}
int ext4_seq_es_shrinker_info_show(struct seq_file *seq, void *v)
{
struct ext4_sb_info *sbi = EXT4_SB((struct super_block *) seq->private);
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
struct ext4_es_stats *es_stats = &sbi->s_es_stats;
struct ext4_inode_info *ei, *max = NULL;
unsigned int inode_cnt = 0;
if (v != SEQ_START_TOKEN)
return 0;
/* here we just find an inode that has the max nr. of objects */
spin_lock(&sbi->s_es_lock);
list_for_each_entry(ei, &sbi->s_es_list, i_es_list) {
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
inode_cnt++;
if (max && max->i_es_all_nr < ei->i_es_all_nr)
max = ei;
else if (!max)
max = ei;
}
spin_unlock(&sbi->s_es_lock);
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
seq_printf(seq, "stats:\n %lld objects\n %lld reclaimable objects\n",
percpu_counter_sum_positive(&es_stats->es_stats_all_cnt),
percpu_counter_sum_positive(&es_stats->es_stats_shk_cnt));
seq_printf(seq, " %lld/%lld cache hits/misses\n",
percpu_counter_sum_positive(&es_stats->es_stats_cache_hits),
percpu_counter_sum_positive(&es_stats->es_stats_cache_misses));
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
if (inode_cnt)
seq_printf(seq, " %d inodes on list\n", inode_cnt);
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
seq_printf(seq, "average:\n %llu us scan time\n",
div_u64(es_stats->es_stats_scan_time, 1000));
seq_printf(seq, " %lu shrunk objects\n", es_stats->es_stats_shrunk);
if (inode_cnt)
seq_printf(seq,
"maximum:\n %lu inode (%u objects, %u reclaimable)\n"
" %llu us max scan time\n",
max->vfs_inode.i_ino, max->i_es_all_nr, max->i_es_shk_nr,
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
div_u64(es_stats->es_stats_max_scan_time, 1000));
return 0;
}
int ext4_es_register_shrinker(struct ext4_sb_info *sbi)
{
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
int err;
/* Make sure we have enough bits for physical block number */
BUILD_BUG_ON(ES_SHIFT < 48);
INIT_LIST_HEAD(&sbi->s_es_list);
sbi->s_es_nr_inode = 0;
spin_lock_init(&sbi->s_es_lock);
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
sbi->s_es_stats.es_stats_shrunk = 0;
err = percpu_counter_init(&sbi->s_es_stats.es_stats_cache_hits, 0,
GFP_KERNEL);
if (err)
return err;
err = percpu_counter_init(&sbi->s_es_stats.es_stats_cache_misses, 0,
GFP_KERNEL);
if (err)
goto err1;
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
sbi->s_es_stats.es_stats_scan_time = 0;
sbi->s_es_stats.es_stats_max_scan_time = 0;
A large number of cleanups and bug fixes, with some (minor) journal optimizations. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2 iQIcBAABCAAGBQJUPlLCAAoJENNvdpvBGATwpN8P/jnbDL1RqM9ZEAWfbDhvYumR Fi59b3IDzSJHuuJeP0nTblVbbWclpO9ljCd18ttsHr8gBXA0ViaEU0XvWbpHIwPN 1fr1/Ovd0wvBdIVdLlaLXTR9skH4lbkiXxv/tkfjVCOSpzqiKID98Z72e/gUjB7Z 8xjAn/mTCnXKnhqMGzi8RC2MP1wgY//ErR21bj6so/8RC8zu4P6JuVj/hI6s0y5i IPtAmjhdM7nxnS0wJwj7dLT0yNDftDh69qE6CgIwyK+Xn/SZFgYwE6+l02dj3DET ZcAzTT9ToTMJdWtMu+5Y4LY8ObJ5xqMPbMoUclQ3DWe6nZicvtcBVCjfG/J8pFlY IFD0nfh/OpX9cQMwJ+5Y8P4TrMiqM+FfuLfu+X83gLyrAyIazwoaZls2lxlEyC0w M25oAqeKGUeVakVlmDZlVyBf05cu5m62x1rRvpcwMXMNhJl8/xwsSdhdYGeJfbO0 0MfL1n6GmvHvouMXKNsXlat/w3QVaQWVRzqdF9x7Q730fSHC/zxVGO+Po3jz2fBd fBdfE14BIIU7nkyBVy0CZG5SDmQW4YACocOv/ATmII9j76F9eZQ3zsA8J1x+dLmJ dP1Uxvsn1C3HW8Ua239j0XUJncglb06iEId0ywdkmWcc1rbzsyZ/NzXN/QBdZmqB 9g4GKAXAyh15PeBTJ5K/ =vWic -----END PGP SIGNATURE----- Merge tag 'ext4_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4 Pull ext4 updates from Ted Ts'o: "A large number of cleanups and bug fixes, with some (minor) journal optimizations" [ This got sent to me before -rc1, but was stuck in my spam folder. - Linus ] * tag 'ext4_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4: (67 commits) ext4: check s_chksum_driver when looking for bg csum presence ext4: move error report out of atomic context in ext4_init_block_bitmap() ext4: Replace open coded mdata csum feature to helper function ext4: delete useless comments about ext4_move_extents ext4: fix reservation overflow in ext4_da_write_begin ext4: add ext4_iget_normal() which is to be used for dir tree lookups ext4: don't orphan or truncate the boot loader inode ext4: grab missed write_count for EXT4_IOC_SWAP_BOOT ext4: optimize block allocation on grow indepth ext4: get rid of code duplication ext4: fix over-defensive complaint after journal abort ext4: fix return value of ext4_do_update_inode ext4: fix mmap data corruption when blocksize < pagesize vfs: fix data corruption when blocksize < pagesize for mmaped data ext4: fold ext4_nojournal_sops into ext4_sops ext4: support freezing ext2 (nojournal) file systems ext4: fold ext4_sync_fs_nojournal() into ext4_sync_fs() ext4: don't check quota format when there are no quota files jbd2: simplify calling convention around __jbd2_journal_clean_checkpoint_list jbd2: avoid pointless scanning of checkpoint lists ...
2014-10-20 20:50:11 +04:00
err = percpu_counter_init(&sbi->s_es_stats.es_stats_all_cnt, 0, GFP_KERNEL);
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
if (err)
goto err2;
err = percpu_counter_init(&sbi->s_es_stats.es_stats_shk_cnt, 0, GFP_KERNEL);
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
if (err)
goto err3;
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
fs: convert fs shrinkers to new scan/count API Convert the filesystem shrinkers to use the new API, and standardise some of the behaviours of the shrinkers at the same time. For example, nr_to_scan means the number of objects to scan, not the number of objects to free. I refactored the CIFS idmap shrinker a little - it really needs to be broken up into a shrinker per tree and keep an item count with the tree root so that we don't need to walk the tree every time the shrinker needs to count the number of objects in the tree (i.e. all the time under memory pressure). [glommer@openvz.org: fixes for ext4, ubifs, nfs, cifs and glock. Fixes are needed mainly due to new code merged in the tree] [assorted fixes folded in] Signed-off-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Glauber Costa <glommer@openvz.org> Acked-by: Mel Gorman <mgorman@suse.de> Acked-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Acked-by: Jan Kara <jack@suse.cz> Acked-by: Steven Whitehouse <swhiteho@redhat.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Cc: Arve Hjønnevåg <arve@android.com> Cc: Carlos Maiolino <cmaiolino@redhat.com> Cc: Christoph Hellwig <hch@lst.de> Cc: Chuck Lever <chuck.lever@oracle.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: David Rientjes <rientjes@google.com> Cc: Gleb Natapov <gleb@redhat.com> Cc: Greg Thelen <gthelen@google.com> Cc: J. Bruce Fields <bfields@redhat.com> Cc: Jan Kara <jack@suse.cz> Cc: Jerome Glisse <jglisse@redhat.com> Cc: John Stultz <john.stultz@linaro.org> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Cc: Kent Overstreet <koverstreet@google.com> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Mel Gorman <mgorman@suse.de> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Thomas Hellstrom <thellstrom@vmware.com> Cc: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2013-08-28 04:18:09 +04:00
sbi->s_es_shrinker.scan_objects = ext4_es_scan;
sbi->s_es_shrinker.count_objects = ext4_es_count;
sbi->s_es_shrinker.seeks = DEFAULT_SEEKS;
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
err = register_shrinker(&sbi->s_es_shrinker);
if (err)
goto err4;
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
return 0;
err4:
percpu_counter_destroy(&sbi->s_es_stats.es_stats_shk_cnt);
err3:
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
percpu_counter_destroy(&sbi->s_es_stats.es_stats_all_cnt);
err2:
percpu_counter_destroy(&sbi->s_es_stats.es_stats_cache_misses);
err1:
percpu_counter_destroy(&sbi->s_es_stats.es_stats_cache_hits);
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
return err;
}
ext4: improve extent cache shrink mechanism to avoid to burn CPU time Now we maintain an proper in-order LRU list in ext4 to reclaim entries from extent status tree when we are under heavy memory pressure. For keeping this order, a spin lock is used to protect this list. But this lock burns a lot of CPU time. We can use the following steps to trigger it. % cd /dev/shm % dd if=/dev/zero of=ext4-img bs=1M count=2k % mkfs.ext4 ext4-img % mount -t ext4 -o loop ext4-img /mnt % cd /mnt % for ((i=0;i<160;i++)); do truncate -s 64g $i; done % for ((i=0;i<160;i++)); do cp $i /dev/null &; done % perf record -a -g % perf report This commit tries to fix this problem. Now a new member called i_touch_when is added into ext4_inode_info to record the last access time for an inode. Meanwhile we never need to keep a proper in-order LRU list. So this can avoid to burns some CPU time. When we try to reclaim some entries from extent status tree, we use list_sort() to get a proper in-order list. Then we traverse this list to discard some entries. In ext4_sb_info, we use s_es_last_sorted to record the last time of sorting this list. When we traverse the list, we skip the inode that is newer than this time, and move this inode to the tail of LRU list. When the head of the list is newer than s_es_last_sorted, we will sort the LRU list again. In this commit, we break the loop if s_extent_cache_cnt == 0 because that means that all extents in extent status tree have been reclaimed. Meanwhile in this commit, ext4_es_{un}register_shrinker()'s prototype is changed to save a local variable in these functions. Reported-by: Dave Hansen <dave.hansen@intel.com> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2013-07-01 16:12:37 +04:00
void ext4_es_unregister_shrinker(struct ext4_sb_info *sbi)
{
percpu_counter_destroy(&sbi->s_es_stats.es_stats_cache_hits);
percpu_counter_destroy(&sbi->s_es_stats.es_stats_cache_misses);
ext4: track extent status tree shrinker delay statictics This commit adds some statictics in extent status tree shrinker. The purpose to add these is that we want to collect more details when we encounter a stall caused by extent status tree shrinker. Here we count the following statictics: stats: the number of all objects on all extent status trees the number of reclaimable objects on lru list cache hits/misses the last sorted interval the number of inodes on lru list average: scan time for shrinking some objects the number of shrunk objects maximum: the inode that has max nr. of objects on lru list the maximum scan time for shrinking some objects The output looks like below: $ cat /proc/fs/ext4/sda1/es_shrinker_info stats: 28228 objects 6341 reclaimable objects 5281/631 cache hits/misses 586 ms last sorted interval 250 inodes on lru list average: 153 us scan time 128 shrunk objects maximum: 255 inode (255 objects, 198 reclaimable) 125723 us max scan time If the lru list has never been sorted, the following line will not be printed: 586ms last sorted interval If there is an empty lru list, the following lines also will not be printed: 250 inodes on lru list ... maximum: 255 inode (255 objects, 198 reclaimable) 0 us max scan time Meanwhile in this commit a new trace point is defined to print some details in __ext4_es_shrink(). Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2014-09-02 06:26:49 +04:00
percpu_counter_destroy(&sbi->s_es_stats.es_stats_all_cnt);
percpu_counter_destroy(&sbi->s_es_stats.es_stats_shk_cnt);
ext4: improve extent cache shrink mechanism to avoid to burn CPU time Now we maintain an proper in-order LRU list in ext4 to reclaim entries from extent status tree when we are under heavy memory pressure. For keeping this order, a spin lock is used to protect this list. But this lock burns a lot of CPU time. We can use the following steps to trigger it. % cd /dev/shm % dd if=/dev/zero of=ext4-img bs=1M count=2k % mkfs.ext4 ext4-img % mount -t ext4 -o loop ext4-img /mnt % cd /mnt % for ((i=0;i<160;i++)); do truncate -s 64g $i; done % for ((i=0;i<160;i++)); do cp $i /dev/null &; done % perf record -a -g % perf report This commit tries to fix this problem. Now a new member called i_touch_when is added into ext4_inode_info to record the last access time for an inode. Meanwhile we never need to keep a proper in-order LRU list. So this can avoid to burns some CPU time. When we try to reclaim some entries from extent status tree, we use list_sort() to get a proper in-order list. Then we traverse this list to discard some entries. In ext4_sb_info, we use s_es_last_sorted to record the last time of sorting this list. When we traverse the list, we skip the inode that is newer than this time, and move this inode to the tail of LRU list. When the head of the list is newer than s_es_last_sorted, we will sort the LRU list again. In this commit, we break the loop if s_extent_cache_cnt == 0 because that means that all extents in extent status tree have been reclaimed. Meanwhile in this commit, ext4_es_{un}register_shrinker()'s prototype is changed to save a local variable in these functions. Reported-by: Dave Hansen <dave.hansen@intel.com> Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2013-07-01 16:12:37 +04:00
unregister_shrinker(&sbi->s_es_shrinker);
}
/*
* Shrink extents in given inode from ei->i_es_shrink_lblk till end. Scan at
* most *nr_to_scan extents, update *nr_to_scan accordingly.
*
* Return 0 if we hit end of tree / interval, 1 if we exhausted nr_to_scan.
* Increment *nr_shrunk by the number of reclaimed extents. Also update
* ei->i_es_shrink_lblk to where we should continue scanning.
*/
static int es_do_reclaim_extents(struct ext4_inode_info *ei, ext4_lblk_t end,
int *nr_to_scan, int *nr_shrunk)
{
struct inode *inode = &ei->vfs_inode;
struct ext4_es_tree *tree = &ei->i_es_tree;
struct extent_status *es;
struct rb_node *node;
es = __es_tree_search(&tree->root, ei->i_es_shrink_lblk);
if (!es)
goto out_wrap;
while (*nr_to_scan > 0) {
if (es->es_lblk > end) {
ei->i_es_shrink_lblk = end + 1;
return 0;
}
(*nr_to_scan)--;
node = rb_next(&es->rb_node);
/*
* We can't reclaim delayed extent from status tree because
* fiemap, bigallic, and seek_data/hole need to use it.
*/
if (ext4_es_is_delayed(es))
goto next;
if (ext4_es_is_referenced(es)) {
ext4_es_clear_referenced(es);
goto next;
}
rb_erase(&es->rb_node, &tree->root);
ext4_es_free_extent(inode, es);
(*nr_shrunk)++;
next:
if (!node)
goto out_wrap;
es = rb_entry(node, struct extent_status, rb_node);
}
ei->i_es_shrink_lblk = es->es_lblk;
return 1;
out_wrap:
ei->i_es_shrink_lblk = 0;
return 0;
}
static int es_reclaim_extents(struct ext4_inode_info *ei, int *nr_to_scan)
{
struct inode *inode = &ei->vfs_inode;
int nr_shrunk = 0;
ext4_lblk_t start = ei->i_es_shrink_lblk;
static DEFINE_RATELIMIT_STATE(_rs, DEFAULT_RATELIMIT_INTERVAL,
DEFAULT_RATELIMIT_BURST);
if (ei->i_es_shk_nr == 0)
return 0;
if (ext4_test_inode_state(inode, EXT4_STATE_EXT_PRECACHED) &&
__ratelimit(&_rs))
ext4_warning(inode->i_sb, "forced shrink of precached extents");
if (!es_do_reclaim_extents(ei, EXT_MAX_BLOCKS, nr_to_scan, &nr_shrunk) &&
start != 0)
es_do_reclaim_extents(ei, start - 1, nr_to_scan, &nr_shrunk);
ei->i_es_tree.cache_es = NULL;
return nr_shrunk;
}
/*
* Called to support EXT4_IOC_CLEAR_ES_CACHE. We can only remove
* discretionary entries from the extent status cache. (Some entries
* must be present for proper operations.)
*/
void ext4_clear_inode_es(struct inode *inode)
{
struct ext4_inode_info *ei = EXT4_I(inode);
struct extent_status *es;
struct ext4_es_tree *tree;
struct rb_node *node;
write_lock(&ei->i_es_lock);
tree = &EXT4_I(inode)->i_es_tree;
tree->cache_es = NULL;
node = rb_first(&tree->root);
while (node) {
es = rb_entry(node, struct extent_status, rb_node);
node = rb_next(node);
if (!ext4_es_is_delayed(es)) {
rb_erase(&es->rb_node, &tree->root);
ext4_es_free_extent(inode, es);
}
}
ext4_clear_inode_state(inode, EXT4_STATE_EXT_PRECACHED);
write_unlock(&ei->i_es_lock);
}
#ifdef ES_DEBUG__
static void ext4_print_pending_tree(struct inode *inode)
{
struct ext4_pending_tree *tree;
struct rb_node *node;
struct pending_reservation *pr;
printk(KERN_DEBUG "pending reservations for inode %lu:", inode->i_ino);
tree = &EXT4_I(inode)->i_pending_tree;
node = rb_first(&tree->root);
while (node) {
pr = rb_entry(node, struct pending_reservation, rb_node);
printk(KERN_DEBUG " %u", pr->lclu);
node = rb_next(node);
}
printk(KERN_DEBUG "\n");
}
#else
#define ext4_print_pending_tree(inode)
#endif
int __init ext4_init_pending(void)
{
ext4_pending_cachep = kmem_cache_create("ext4_pending_reservation",
sizeof(struct pending_reservation),
0, (SLAB_RECLAIM_ACCOUNT), NULL);
if (ext4_pending_cachep == NULL)
return -ENOMEM;
return 0;
}
void ext4_exit_pending(void)
{
kmem_cache_destroy(ext4_pending_cachep);
}
void ext4_init_pending_tree(struct ext4_pending_tree *tree)
{
tree->root = RB_ROOT;
}
/*
* __get_pending - retrieve a pointer to a pending reservation
*
* @inode - file containing the pending cluster reservation
* @lclu - logical cluster of interest
*
* Returns a pointer to a pending reservation if it's a member of
* the set, and NULL if not. Must be called holding i_es_lock.
*/
static struct pending_reservation *__get_pending(struct inode *inode,
ext4_lblk_t lclu)
{
struct ext4_pending_tree *tree;
struct rb_node *node;
struct pending_reservation *pr = NULL;
tree = &EXT4_I(inode)->i_pending_tree;
node = (&tree->root)->rb_node;
while (node) {
pr = rb_entry(node, struct pending_reservation, rb_node);
if (lclu < pr->lclu)
node = node->rb_left;
else if (lclu > pr->lclu)
node = node->rb_right;
else if (lclu == pr->lclu)
return pr;
}
return NULL;
}
/*
* __insert_pending - adds a pending cluster reservation to the set of
* pending reservations
*
* @inode - file containing the cluster
* @lblk - logical block in the cluster to be added
*
* Returns 0 on successful insertion and -ENOMEM on failure. If the
* pending reservation is already in the set, returns successfully.
*/
static int __insert_pending(struct inode *inode, ext4_lblk_t lblk)
{
struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
struct ext4_pending_tree *tree = &EXT4_I(inode)->i_pending_tree;
struct rb_node **p = &tree->root.rb_node;
struct rb_node *parent = NULL;
struct pending_reservation *pr;
ext4_lblk_t lclu;
int ret = 0;
lclu = EXT4_B2C(sbi, lblk);
/* search to find parent for insertion */
while (*p) {
parent = *p;
pr = rb_entry(parent, struct pending_reservation, rb_node);
if (lclu < pr->lclu) {
p = &(*p)->rb_left;
} else if (lclu > pr->lclu) {
p = &(*p)->rb_right;
} else {
/* pending reservation already inserted */
goto out;
}
}
pr = kmem_cache_alloc(ext4_pending_cachep, GFP_ATOMIC);
if (pr == NULL) {
ret = -ENOMEM;
goto out;
}
pr->lclu = lclu;
rb_link_node(&pr->rb_node, parent, p);
rb_insert_color(&pr->rb_node, &tree->root);
out:
return ret;
}
/*
* __remove_pending - removes a pending cluster reservation from the set
* of pending reservations
*
* @inode - file containing the cluster
* @lblk - logical block in the pending cluster reservation to be removed
*
* Returns successfully if pending reservation is not a member of the set.
*/
static void __remove_pending(struct inode *inode, ext4_lblk_t lblk)
{
struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
struct pending_reservation *pr;
struct ext4_pending_tree *tree;
pr = __get_pending(inode, EXT4_B2C(sbi, lblk));
if (pr != NULL) {
tree = &EXT4_I(inode)->i_pending_tree;
rb_erase(&pr->rb_node, &tree->root);
kmem_cache_free(ext4_pending_cachep, pr);
}
}
/*
* ext4_remove_pending - removes a pending cluster reservation from the set
* of pending reservations
*
* @inode - file containing the cluster
* @lblk - logical block in the pending cluster reservation to be removed
*
* Locking for external use of __remove_pending.
*/
void ext4_remove_pending(struct inode *inode, ext4_lblk_t lblk)
{
struct ext4_inode_info *ei = EXT4_I(inode);
write_lock(&ei->i_es_lock);
__remove_pending(inode, lblk);
write_unlock(&ei->i_es_lock);
}
/*
* ext4_is_pending - determine whether a cluster has a pending reservation
* on it
*
* @inode - file containing the cluster
* @lblk - logical block in the cluster
*
* Returns true if there's a pending reservation for the cluster in the
* set of pending reservations, and false if not.
*/
bool ext4_is_pending(struct inode *inode, ext4_lblk_t lblk)
{
struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
struct ext4_inode_info *ei = EXT4_I(inode);
bool ret;
read_lock(&ei->i_es_lock);
ret = (bool)(__get_pending(inode, EXT4_B2C(sbi, lblk)) != NULL);
read_unlock(&ei->i_es_lock);
return ret;
}
/*
* ext4_es_insert_delayed_block - adds a delayed block to the extents status
* tree, adding a pending reservation where
* needed
*
* @inode - file containing the newly added block
* @lblk - logical block to be added
* @allocated - indicates whether a physical cluster has been allocated for
* the logical cluster that contains the block
*
* Returns 0 on success, negative error code on failure.
*/
int ext4_es_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk,
bool allocated)
{
struct extent_status newes;
int err = 0;
es_debug("add [%u/1) delayed to extent status tree of inode %lu\n",
lblk, inode->i_ino);
newes.es_lblk = lblk;
newes.es_len = 1;
ext4_es_store_pblock_status(&newes, ~0, EXTENT_STATUS_DELAYED);
trace_ext4_es_insert_delayed_block(inode, &newes, allocated);
ext4_es_insert_extent_check(inode, &newes);
write_lock(&EXT4_I(inode)->i_es_lock);
err = __es_remove_extent(inode, lblk, lblk, NULL);
if (err != 0)
goto error;
retry:
err = __es_insert_extent(inode, &newes);
if (err == -ENOMEM && __es_shrink(EXT4_SB(inode->i_sb),
128, EXT4_I(inode)))
goto retry;
if (err != 0)
goto error;
if (allocated)
__insert_pending(inode, lblk);
error:
write_unlock(&EXT4_I(inode)->i_es_lock);
ext4_es_print_tree(inode);
ext4_print_pending_tree(inode);
return err;
}
/*
* __es_delayed_clu - count number of clusters containing blocks that
* are delayed only
*
* @inode - file containing block range
* @start - logical block defining start of range
* @end - logical block defining end of range
*
* Returns the number of clusters containing only delayed (not delayed
* and unwritten) blocks in the range specified by @start and @end. Any
* cluster or part of a cluster within the range and containing a delayed
* and not unwritten block within the range is counted as a whole cluster.
*/
static unsigned int __es_delayed_clu(struct inode *inode, ext4_lblk_t start,
ext4_lblk_t end)
{
struct ext4_es_tree *tree = &EXT4_I(inode)->i_es_tree;
struct extent_status *es;
struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
struct rb_node *node;
ext4_lblk_t first_lclu, last_lclu;
unsigned long long last_counted_lclu;
unsigned int n = 0;
/* guaranteed to be unequal to any ext4_lblk_t value */
last_counted_lclu = ~0ULL;
es = __es_tree_search(&tree->root, start);
while (es && (es->es_lblk <= end)) {
if (ext4_es_is_delonly(es)) {
if (es->es_lblk <= start)
first_lclu = EXT4_B2C(sbi, start);
else
first_lclu = EXT4_B2C(sbi, es->es_lblk);
if (ext4_es_end(es) >= end)
last_lclu = EXT4_B2C(sbi, end);
else
last_lclu = EXT4_B2C(sbi, ext4_es_end(es));
if (first_lclu == last_counted_lclu)
n += last_lclu - first_lclu;
else
n += last_lclu - first_lclu + 1;
last_counted_lclu = last_lclu;
}
node = rb_next(&es->rb_node);
if (!node)
break;
es = rb_entry(node, struct extent_status, rb_node);
}
return n;
}
/*
* ext4_es_delayed_clu - count number of clusters containing blocks that
* are both delayed and unwritten
*
* @inode - file containing block range
* @lblk - logical block defining start of range
* @len - number of blocks in range
*
* Locking for external use of __es_delayed_clu().
*/
unsigned int ext4_es_delayed_clu(struct inode *inode, ext4_lblk_t lblk,
ext4_lblk_t len)
{
struct ext4_inode_info *ei = EXT4_I(inode);
ext4_lblk_t end;
unsigned int n;
if (len == 0)
return 0;
end = lblk + len - 1;
WARN_ON(end < lblk);
read_lock(&ei->i_es_lock);
n = __es_delayed_clu(inode, lblk, end);
read_unlock(&ei->i_es_lock);
return n;
}
/*
* __revise_pending - makes, cancels, or leaves unchanged pending cluster
* reservations for a specified block range depending
* upon the presence or absence of delayed blocks
* outside the range within clusters at the ends of the
* range
*
* @inode - file containing the range
* @lblk - logical block defining the start of range
* @len - length of range in blocks
*
* Used after a newly allocated extent is added to the extents status tree.
* Requires that the extents in the range have either written or unwritten
* status. Must be called while holding i_es_lock.
*/
static void __revise_pending(struct inode *inode, ext4_lblk_t lblk,
ext4_lblk_t len)
{
struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
ext4_lblk_t end = lblk + len - 1;
ext4_lblk_t first, last;
bool f_del = false, l_del = false;
if (len == 0)
return;
/*
* Two cases - block range within single cluster and block range
* spanning two or more clusters. Note that a cluster belonging
* to a range starting and/or ending on a cluster boundary is treated
* as if it does not contain a delayed extent. The new range may
* have allocated space for previously delayed blocks out to the
* cluster boundary, requiring that any pre-existing pending
* reservation be canceled. Because this code only looks at blocks
* outside the range, it should revise pending reservations
* correctly even if the extent represented by the range can't be
* inserted in the extents status tree due to ENOSPC.
*/
if (EXT4_B2C(sbi, lblk) == EXT4_B2C(sbi, end)) {
first = EXT4_LBLK_CMASK(sbi, lblk);
if (first != lblk)
f_del = __es_scan_range(inode, &ext4_es_is_delonly,
first, lblk - 1);
if (f_del) {
__insert_pending(inode, first);
} else {
last = EXT4_LBLK_CMASK(sbi, end) +
sbi->s_cluster_ratio - 1;
if (last != end)
l_del = __es_scan_range(inode,
&ext4_es_is_delonly,
end + 1, last);
if (l_del)
__insert_pending(inode, last);
else
__remove_pending(inode, last);
}
} else {
first = EXT4_LBLK_CMASK(sbi, lblk);
if (first != lblk)
f_del = __es_scan_range(inode, &ext4_es_is_delonly,
first, lblk - 1);
if (f_del)
__insert_pending(inode, first);
else
__remove_pending(inode, first);
last = EXT4_LBLK_CMASK(sbi, end) + sbi->s_cluster_ratio - 1;
if (last != end)
l_del = __es_scan_range(inode, &ext4_es_is_delonly,
end + 1, last);
if (l_del)
__insert_pending(inode, last);
else
__remove_pending(inode, last);
}
}