]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/ext4/mballoc.h
ext4: convert instrumentation from markers to tracepoints
[mirror_ubuntu-artful-kernel.git] / fs / ext4 / mballoc.h
CommitLineData
8f6e39a7
MC
1/*
2 * fs/ext4/mballoc.h
3 *
4 * Written by: Alex Tomas <alex@clusterfs.com>
5 *
6 */
7#ifndef _EXT4_MBALLOC_H
8#define _EXT4_MBALLOC_H
9
10#include <linux/time.h>
11#include <linux/fs.h>
12#include <linux/namei.h>
13#include <linux/quotaops.h>
14#include <linux/buffer_head.h>
15#include <linux/module.h>
16#include <linux/swap.h>
17#include <linux/proc_fs.h>
18#include <linux/pagemap.h>
19#include <linux/seq_file.h>
20#include <linux/version.h>
8a0aba73 21#include <linux/blkdev.h>
920313a7 22#include <linux/mutex.h>
8f6e39a7
MC
23#include "ext4_jbd2.h"
24#include "ext4.h"
8f6e39a7
MC
25
26/*
27 * with AGGRESSIVE_CHECK allocator runs consistency checks over
28 * structures. these checks slow things down a lot
29 */
30#define AGGRESSIVE_CHECK__
31
32/*
33 * with DOUBLE_CHECK defined mballoc creates persistent in-core
34 * bitmaps, maintains and uses them to check for double allocations
35 */
36#define DOUBLE_CHECK__
37
38/*
39 */
40#define MB_DEBUG__
41#ifdef MB_DEBUG
42#define mb_debug(fmt, a...) printk(fmt, ##a)
43#else
44#define mb_debug(fmt, a...)
45#endif
46
47/*
48 * with EXT4_MB_HISTORY mballoc stores last N allocations in memory
49 * and you can monitor it in /proc/fs/ext4/<dev>/mb_history
50 */
51#define EXT4_MB_HISTORY
52#define EXT4_MB_HISTORY_ALLOC 1 /* allocation */
53#define EXT4_MB_HISTORY_PREALLOC 2 /* preallocated blocks used */
54#define EXT4_MB_HISTORY_DISCARD 4 /* preallocation discarded */
55#define EXT4_MB_HISTORY_FREE 8 /* free */
56
57#define EXT4_MB_HISTORY_DEFAULT (EXT4_MB_HISTORY_ALLOC | \
58 EXT4_MB_HISTORY_PREALLOC)
59
60/*
61 * How long mballoc can look for a best extent (in found extents)
62 */
63#define MB_DEFAULT_MAX_TO_SCAN 200
64
65/*
66 * How long mballoc must look for a best extent
67 */
68#define MB_DEFAULT_MIN_TO_SCAN 10
69
70/*
71 * How many groups mballoc will scan looking for the best chunk
72 */
73#define MB_DEFAULT_MAX_GROUPS_TO_SCAN 5
74
75/*
76 * with 'ext4_mb_stats' allocator will collect stats that will be
77 * shown at umount. The collecting costs though!
78 */
79#define MB_DEFAULT_STATS 1
80
81/*
82 * files smaller than MB_DEFAULT_STREAM_THRESHOLD are served
83 * by the stream allocator, which purpose is to pack requests
84 * as close each to other as possible to produce smooth I/O traffic
85 * We use locality group prealloc space for stream request.
86 * We can tune the same via /proc/fs/ext4/<parition>/stream_req
87 */
88#define MB_DEFAULT_STREAM_THRESHOLD 16 /* 64K */
89
90/*
91 * for which requests use 2^N search using buddies
92 */
93#define MB_DEFAULT_ORDER2_REQS 2
94
95/*
96 * default group prealloc size 512 blocks
97 */
98#define MB_DEFAULT_GROUP_PREALLOC 512
99
8f6e39a7 100
c894058d
AK
101struct ext4_free_data {
102 /* this links the free block information from group_info */
103 struct rb_node node;
8f6e39a7 104
c894058d 105 /* this links the free block information from ext4_sb_info */
8f6e39a7 106 struct list_head list;
c894058d
AK
107
108 /* group which free block extent belongs */
109 ext4_group_t group;
110
111 /* free block extent */
112 ext4_grpblk_t start_blk;
113 ext4_grpblk_t count;
114
115 /* transaction which freed this extent */
116 tid_t t_tid;
8f6e39a7
MC
117};
118
8f6e39a7
MC
119struct ext4_prealloc_space {
120 struct list_head pa_inode_list;
121 struct list_head pa_group_list;
122 union {
123 struct list_head pa_tmp_list;
124 struct rcu_head pa_rcu;
125 } u;
126 spinlock_t pa_lock;
127 atomic_t pa_count;
128 unsigned pa_deleted;
129 ext4_fsblk_t pa_pstart; /* phys. block */
130 ext4_lblk_t pa_lstart; /* log. block */
131 unsigned short pa_len; /* len of preallocated chunk */
132 unsigned short pa_free; /* how many blocks are free */
cc0fb9ad 133 unsigned short pa_type; /* pa type. inode or group */
8f6e39a7
MC
134 spinlock_t *pa_obj_lock;
135 struct inode *pa_inode; /* hack, for history only */
136};
137
cc0fb9ad
AK
138enum {
139 MB_INODE_PA = 0,
140 MB_GROUP_PA = 1
141};
8f6e39a7
MC
142
143struct ext4_free_extent {
144 ext4_lblk_t fe_logical;
145 ext4_grpblk_t fe_start;
146 ext4_group_t fe_group;
147 int fe_len;
148};
149
150/*
151 * Locality group:
152 * we try to group all related changes together
153 * so that writeback can flush/allocate them together as well
6be2ded1
AK
154 * Size of lg_prealloc_list hash is determined by MB_DEFAULT_GROUP_PREALLOC
155 * (512). We store prealloc space into the hash based on the pa_free blocks
156 * order value.ie, fls(pa_free)-1;
8f6e39a7 157 */
6be2ded1 158#define PREALLOC_TB_SIZE 10
8f6e39a7
MC
159struct ext4_locality_group {
160 /* for allocator */
6be2ded1
AK
161 /* to serialize allocates */
162 struct mutex lg_mutex;
163 /* list of preallocations */
164 struct list_head lg_prealloc_list[PREALLOC_TB_SIZE];
8f6e39a7
MC
165 spinlock_t lg_prealloc_lock;
166};
167
168struct ext4_allocation_context {
169 struct inode *ac_inode;
170 struct super_block *ac_sb;
171
172 /* original request */
173 struct ext4_free_extent ac_o_ex;
174
175 /* goal request (after normalization) */
176 struct ext4_free_extent ac_g_ex;
177
178 /* the best found extent */
179 struct ext4_free_extent ac_b_ex;
180
181 /* copy of the bext found extent taken before preallocation efforts */
182 struct ext4_free_extent ac_f_ex;
183
184 /* number of iterations done. we have to track to limit searching */
185 unsigned long ac_ex_scanned;
186 __u16 ac_groups_scanned;
187 __u16 ac_found;
188 __u16 ac_tail;
189 __u16 ac_buddy;
190 __u16 ac_flags; /* allocation hints */
191 __u8 ac_status;
192 __u8 ac_criteria;
193 __u8 ac_repeats;
194 __u8 ac_2order; /* if request is to allocate 2^N blocks and
195 * N > 0, the field stores N, otherwise 0 */
196 __u8 ac_op; /* operation, for history only */
197 struct page *ac_bitmap_page;
198 struct page *ac_buddy_page;
8556e8f3
AK
199 /*
200 * pointer to the held semaphore upon successful
201 * block allocation
202 */
203 struct rw_semaphore *alloc_semp;
8f6e39a7
MC
204 struct ext4_prealloc_space *ac_pa;
205 struct ext4_locality_group *ac_lg;
206};
207
208#define AC_STATUS_CONTINUE 1
209#define AC_STATUS_FOUND 2
210#define AC_STATUS_BREAK 3
211
212struct ext4_mb_history {
213 struct ext4_free_extent orig; /* orig allocation */
214 struct ext4_free_extent goal; /* goal allocation */
215 struct ext4_free_extent result; /* result allocation */
216 unsigned pid;
217 unsigned ino;
218 __u16 found; /* how many extents have been found */
219 __u16 groups; /* how many groups have been scanned */
220 __u16 tail; /* what tail broke some buddy */
221 __u16 buddy; /* buddy the tail ^^^ broke */
222 __u16 flags;
223 __u8 cr:3; /* which phase the result extent was found at */
224 __u8 op:4;
225 __u8 merged:1;
226};
227
228struct ext4_buddy {
229 struct page *bd_buddy_page;
230 void *bd_buddy;
231 struct page *bd_bitmap_page;
232 void *bd_bitmap;
233 struct ext4_group_info *bd_info;
234 struct super_block *bd_sb;
235 __u16 bd_blkbits;
236 ext4_group_t bd_group;
920313a7 237 struct rw_semaphore *alloc_semp;
8f6e39a7
MC
238};
239#define EXT4_MB_BITMAP(e4b) ((e4b)->bd_bitmap)
240#define EXT4_MB_BUDDY(e4b) ((e4b)->bd_buddy)
241
242#ifndef EXT4_MB_HISTORY
243static inline void ext4_mb_store_history(struct ext4_allocation_context *ac)
244{
245 return;
246}
8f6e39a7
MC
247#endif
248
249#define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
250
c3a326a6 251static inline ext4_fsblk_t ext4_grp_offs_to_block(struct super_block *sb,
8f6e39a7
MC
252 struct ext4_free_extent *fex)
253{
254 ext4_fsblk_t block;
255
256 block = (ext4_fsblk_t) fex->fe_group * EXT4_BLOCKS_PER_GROUP(sb)
257 + fex->fe_start
258 + le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
259 return block;
260}
261#endif