1 .. SPDX-License-Identifier: GPL-2.0 1 .. SPDX-License-Identifier: GPL-2.0
2 2
3 ============================ 3 ============================
4 Glock internal locking rules 4 Glock internal locking rules
5 ============================ 5 ============================
6 6
7 This documents the basic principles of the glo 7 This documents the basic principles of the glock state machine
8 internals. Each glock (struct gfs2_glock in fs 8 internals. Each glock (struct gfs2_glock in fs/gfs2/incore.h)
9 has two main (internal) locks: 9 has two main (internal) locks:
10 10
11 1. A spinlock (gl_lockref.lock) which protect 11 1. A spinlock (gl_lockref.lock) which protects the internal state such
12 as gl_state, gl_target and the list of hol 12 as gl_state, gl_target and the list of holders (gl_holders)
13 2. A non-blocking bit lock, GLF_LOCK, which i 13 2. A non-blocking bit lock, GLF_LOCK, which is used to prevent other
14 threads from making calls to the DLM, etc. 14 threads from making calls to the DLM, etc. at the same time. If a
15 thread takes this lock, it must then call 15 thread takes this lock, it must then call run_queue (usually via the
16 workqueue) when it releases it in order to 16 workqueue) when it releases it in order to ensure any pending tasks
17 are completed. 17 are completed.
18 18
19 The gl_holders list contains all the queued lo 19 The gl_holders list contains all the queued lock requests (not
20 just the holders) associated with the glock. I 20 just the holders) associated with the glock. If there are any
21 held locks, then they will be contiguous entri 21 held locks, then they will be contiguous entries at the head
22 of the list. Locks are granted in strictly the 22 of the list. Locks are granted in strictly the order that they
23 are queued. 23 are queued.
24 24
25 There are three lock states that users of the 25 There are three lock states that users of the glock layer can request,
26 namely shared (SH), deferred (DF) and exclusiv 26 namely shared (SH), deferred (DF) and exclusive (EX). Those translate
27 to the following DLM lock modes: 27 to the following DLM lock modes:
28 28
29 ========== ====== ======================= 29 ========== ====== =====================================================
30 Glock mode DLM lock mode 30 Glock mode DLM lock mode
31 ========== ====== ======================= 31 ========== ====== =====================================================
32 UN IV/NL Unlocked (no DLM lock a 32 UN IV/NL Unlocked (no DLM lock associated with glock) or NL
33 SH PR (Protected read) 33 SH PR (Protected read)
34 DF CW (Concurrent write) 34 DF CW (Concurrent write)
35 EX EX (Exclusive) 35 EX EX (Exclusive)
36 ========== ====== ======================= 36 ========== ====== =====================================================
37 37
38 Thus DF is basically a shared mode which is in 38 Thus DF is basically a shared mode which is incompatible with the "normal"
39 shared lock mode, SH. In GFS2 the DF mode is u 39 shared lock mode, SH. In GFS2 the DF mode is used exclusively for direct I/O
40 operations. The glocks are basically a lock pl 40 operations. The glocks are basically a lock plus some routines which deal
41 with cache management. The following rules app 41 with cache management. The following rules apply for the cache:
42 42
43 ========== ============== ========== 43 ========== ============== ========== ========== ==============
44 Glock mode Cache Metadata Cache data 44 Glock mode Cache Metadata Cache data Dirty Data Dirty Metadata
45 ========== ============== ========== 45 ========== ============== ========== ========== ==============
46 UN No No 46 UN No No No No
47 DF Yes No 47 DF Yes No No No
48 SH Yes Yes 48 SH Yes Yes No No
49 EX Yes Yes 49 EX Yes Yes Yes Yes
50 ========== ============== ========== 50 ========== ============== ========== ========== ==============
51 51
52 These rules are implemented using the various 52 These rules are implemented using the various glock operations which
53 are defined for each type of glock. Not all ty 53 are defined for each type of glock. Not all types of glocks use
54 all the modes. Only inode glocks use the DF mo 54 all the modes. Only inode glocks use the DF mode for example.
55 55
56 Table of glock operations and per type constan 56 Table of glock operations and per type constants:
57 57
58 ============== =========================== 58 ============== =============================================================
59 Field Purpose 59 Field Purpose
60 ============== =========================== 60 ============== =============================================================
61 go_sync Called before remote state 61 go_sync Called before remote state change (e.g. to sync dirty data)
62 go_xmote_bh Called after remote state c 62 go_xmote_bh Called after remote state change (e.g. to refill cache)
63 go_inval Called if remote state chan 63 go_inval Called if remote state change requires invalidating the cache
64 go_instantiate Called when a glock has bee 64 go_instantiate Called when a glock has been acquired
65 go_held Called every time a glock h 65 go_held Called every time a glock holder is acquired
66 go_dump Called to print content of 66 go_dump Called to print content of object for debugfs file, or on
67 error to dump glock to the 67 error to dump glock to the log.
68 go_callback Called if the DLM sends a c 68 go_callback Called if the DLM sends a callback to drop this lock
69 go_unlocked Called when a glock is unlo 69 go_unlocked Called when a glock is unlocked (dlm_unlock())
70 go_type The type of the glock, ``LM 70 go_type The type of the glock, ``LM_TYPE_*``
71 go_flags GLOF_ASPACE is set, if the 71 go_flags GLOF_ASPACE is set, if the glock has an address space
72 associated with it 72 associated with it
73 ============== =========================== 73 ============== =============================================================
74 74
75 The minimum hold time for each lock is the tim 75 The minimum hold time for each lock is the time after a remote lock
76 grant for which we ignore remote demote reques 76 grant for which we ignore remote demote requests. This is in order to
77 prevent a situation where locks are being boun 77 prevent a situation where locks are being bounced around the cluster
78 from node to node with none of the nodes makin 78 from node to node with none of the nodes making any progress. This
79 tends to show up most with shared mmapped file 79 tends to show up most with shared mmapped files which are being written
80 to by multiple nodes. By delaying the demotion 80 to by multiple nodes. By delaying the demotion in response to a
81 remote callback, that gives the userspace prog 81 remote callback, that gives the userspace program time to make
82 some progress before the pages are unmapped. 82 some progress before the pages are unmapped.
83 83
84 Eventually, we hope to make the glock "EX" mod 84 Eventually, we hope to make the glock "EX" mode locally shared such that any
85 local locking will be done with the i_mutex as 85 local locking will be done with the i_mutex as required rather than via the
86 glock. 86 glock.
87 87
88 Locking rules for glock operations: 88 Locking rules for glock operations:
89 89
90 ============== ====================== === 90 ============== ====================== =============================
91 Operation GLF_LOCK bit lock held gl_ 91 Operation GLF_LOCK bit lock held gl_lockref.lock spinlock held
92 ============== ====================== === 92 ============== ====================== =============================
93 go_sync Yes 93 go_sync Yes No
94 go_xmote_bh Yes 94 go_xmote_bh Yes No
95 go_inval Yes 95 go_inval Yes No
96 go_instantiate No 96 go_instantiate No No
97 go_held No 97 go_held No No
98 go_dump Sometimes 98 go_dump Sometimes Yes
99 go_callback Sometimes (N/A) 99 go_callback Sometimes (N/A) Yes
100 go_unlocked Yes 100 go_unlocked Yes No
101 ============== ====================== === 101 ============== ====================== =============================
102 102
103 .. Note:: 103 .. Note::
104 104
105 Operations must not drop either the bit loc 105 Operations must not drop either the bit lock or the spinlock
106 if its held on entry. go_dump and do_demote 106 if its held on entry. go_dump and do_demote_ok must never block.
107 Note that go_dump will only be called if th 107 Note that go_dump will only be called if the glock's state
108 indicates that it is caching uptodate data. 108 indicates that it is caching uptodate data.
109 109
110 Glock locking order within GFS2: 110 Glock locking order within GFS2:
111 111
112 1. i_rwsem (if required) 112 1. i_rwsem (if required)
113 2. Rename glock (for rename only) 113 2. Rename glock (for rename only)
114 3. Inode glock(s) 114 3. Inode glock(s)
115 (Parents before children, inodes at "same 115 (Parents before children, inodes at "same level" with same parent in
116 lock number order) 116 lock number order)
117 4. Rgrp glock(s) (for (de)allocation operatio 117 4. Rgrp glock(s) (for (de)allocation operations)
118 5. Transaction glock (via gfs2_trans_begin) f 118 5. Transaction glock (via gfs2_trans_begin) for non-read operations
119 6. i_rw_mutex (if required) 119 6. i_rw_mutex (if required)
120 7. Page lock (always last, very important!) 120 7. Page lock (always last, very important!)
121 121
122 There are two glocks per inode. One deals with 122 There are two glocks per inode. One deals with access to the inode
123 itself (locking order as above), and the other 123 itself (locking order as above), and the other, known as the iopen
124 glock is used in conjunction with the i_nlink 124 glock is used in conjunction with the i_nlink field in the inode to
125 determine the lifetime of the inode in questio 125 determine the lifetime of the inode in question. Locking of inodes
126 is on a per-inode basis. Locking of rgrps is o 126 is on a per-inode basis. Locking of rgrps is on a per rgrp basis.
127 In general we prefer to lock local locks prior 127 In general we prefer to lock local locks prior to cluster locks.
128 128
129 Glock Statistics 129 Glock Statistics
130 ---------------- 130 ----------------
131 131
132 The stats are divided into two sets: those rel 132 The stats are divided into two sets: those relating to the
133 super block and those relating to an individua 133 super block and those relating to an individual glock. The
134 super block stats are done on a per cpu basis 134 super block stats are done on a per cpu basis in order to
135 try and reduce the overhead of gathering them. 135 try and reduce the overhead of gathering them. They are also
136 further divided by glock type. All timings are 136 further divided by glock type. All timings are in nanoseconds.
137 137
138 In the case of both the super block and glock 138 In the case of both the super block and glock statistics,
139 the same information is gathered in each case. 139 the same information is gathered in each case. The super
140 block timing statistics are used to provide de 140 block timing statistics are used to provide default values for
141 the glock timing statistics, so that newly cre 141 the glock timing statistics, so that newly created glocks
142 should have, as far as possible, a sensible st 142 should have, as far as possible, a sensible starting point.
143 The per-glock counters are initialised to zero 143 The per-glock counters are initialised to zero when the
144 glock is created. The per-glock statistics are 144 glock is created. The per-glock statistics are lost when
145 the glock is ejected from memory. 145 the glock is ejected from memory.
146 146
147 The statistics are divided into three pairs of 147 The statistics are divided into three pairs of mean and
148 variance, plus two counters. The mean/variance 148 variance, plus two counters. The mean/variance pairs are
149 smoothed exponential estimates and the algorit 149 smoothed exponential estimates and the algorithm used is
150 one which will be very familiar to those used 150 one which will be very familiar to those used to calculation
151 of round trip times in network code. See "TCP/ 151 of round trip times in network code. See "TCP/IP Illustrated,
152 Volume 1", W. Richard Stevens, sect 21.3, "Rou 152 Volume 1", W. Richard Stevens, sect 21.3, "Round-Trip Time Measurement",
153 p. 299 and onwards. Also, Volume 2, Sect. 25.1 153 p. 299 and onwards. Also, Volume 2, Sect. 25.10, p. 838 and onwards.
154 Unlike the TCP/IP Illustrated case, the mean a 154 Unlike the TCP/IP Illustrated case, the mean and variance are
155 not scaled, but are in units of integer nanose 155 not scaled, but are in units of integer nanoseconds.
156 156
157 The three pairs of mean/variance measure the f 157 The three pairs of mean/variance measure the following
158 things: 158 things:
159 159
160 1. DLM lock time (non-blocking requests) 160 1. DLM lock time (non-blocking requests)
161 2. DLM lock time (blocking requests) 161 2. DLM lock time (blocking requests)
162 3. Inter-request time (again to the DLM) 162 3. Inter-request time (again to the DLM)
163 163
164 A non-blocking request is one which will compl 164 A non-blocking request is one which will complete right
165 away, whatever the state of the DLM lock in qu 165 away, whatever the state of the DLM lock in question. That
166 currently means any requests when (a) the curr 166 currently means any requests when (a) the current state of
167 the lock is exclusive, i.e. a lock demotion (b 167 the lock is exclusive, i.e. a lock demotion (b) the requested
168 state is either null or unlocked (again, a dem 168 state is either null or unlocked (again, a demotion) or (c) the
169 "try lock" flag is set. A blocking request cov 169 "try lock" flag is set. A blocking request covers all the other
170 lock requests. 170 lock requests.
171 171
172 There are two counters. The first is there pri 172 There are two counters. The first is there primarily to show
173 how many lock requests have been made, and thu 173 how many lock requests have been made, and thus how much data
174 has gone into the mean/variance calculations. 174 has gone into the mean/variance calculations. The other counter
175 is counting queuing of holders at the top laye 175 is counting queuing of holders at the top layer of the glock
176 code. Hopefully that number will be a lot larg 176 code. Hopefully that number will be a lot larger than the number
177 of dlm lock requests issued. 177 of dlm lock requests issued.
178 178
179 So why gather these statistics? There are seve 179 So why gather these statistics? There are several reasons
180 we'd like to get a better idea of these timing 180 we'd like to get a better idea of these timings:
181 181
182 1. To be able to better set the glock "min hol 182 1. To be able to better set the glock "min hold time"
183 2. To spot performance issues more easily 183 2. To spot performance issues more easily
184 3. To improve the algorithm for selecting reso 184 3. To improve the algorithm for selecting resource groups for
185 allocation (to base it on lock wait time, r 185 allocation (to base it on lock wait time, rather than blindly
186 using a "try lock") 186 using a "try lock")
187 187
188 Due to the smoothing action of the updates, a 188 Due to the smoothing action of the updates, a step change in
189 some input quantity being sampled will only fu 189 some input quantity being sampled will only fully be taken
190 into account after 8 samples (or 4 for the var 190 into account after 8 samples (or 4 for the variance) and this
191 needs to be carefully considered when interpre 191 needs to be carefully considered when interpreting the
192 results. 192 results.
193 193
194 Knowing both the time it takes a lock request 194 Knowing both the time it takes a lock request to complete and
195 the average time between lock requests for a g 195 the average time between lock requests for a glock means we
196 can compute the total percentage of the time f 196 can compute the total percentage of the time for which the
197 node is able to use a glock vs. time that the 197 node is able to use a glock vs. time that the rest of the
198 cluster has its share. That will be very usefu 198 cluster has its share. That will be very useful when setting
199 the lock min hold time. 199 the lock min hold time.
200 200
201 Great care has been taken to ensure that we 201 Great care has been taken to ensure that we
202 measure exactly the quantities that we want, a 202 measure exactly the quantities that we want, as accurately
203 as possible. There are always inaccuracies in 203 as possible. There are always inaccuracies in any
204 measuring system, but I hope this is as accura 204 measuring system, but I hope this is as accurate as we
205 can reasonably make it. 205 can reasonably make it.
206 206
207 Per sb stats can be found here:: 207 Per sb stats can be found here::
208 208
209 /sys/kernel/debug/gfs2/<fsname>/sbstats 209 /sys/kernel/debug/gfs2/<fsname>/sbstats
210 210
211 Per glock stats can be found here:: 211 Per glock stats can be found here::
212 212
213 /sys/kernel/debug/gfs2/<fsname>/glstats 213 /sys/kernel/debug/gfs2/<fsname>/glstats
214 214
215 Assuming that debugfs is mounted on /sys/kerne 215 Assuming that debugfs is mounted on /sys/kernel/debug and also
216 that <fsname> is replaced with the name of the 216 that <fsname> is replaced with the name of the gfs2 filesystem
217 in question. 217 in question.
218 218
219 The abbreviations used in the output as are fo 219 The abbreviations used in the output as are follows:
220 220
221 ========= =================================== 221 ========= ================================================================
222 srtt Smoothed round trip time for non bl 222 srtt Smoothed round trip time for non blocking dlm requests
223 srttvar Variance estimate for srtt 223 srttvar Variance estimate for srtt
224 srttb Smoothed round trip time for (poten 224 srttb Smoothed round trip time for (potentially) blocking dlm requests
225 srttvarb Variance estimate for srttb 225 srttvarb Variance estimate for srttb
226 sirt Smoothed inter request time (for dl 226 sirt Smoothed inter request time (for dlm requests)
227 sirtvar Variance estimate for sirt 227 sirtvar Variance estimate for sirt
228 dlm Number of dlm requests made (dcnt i 228 dlm Number of dlm requests made (dcnt in glstats file)
229 queue Number of glock requests queued (qc 229 queue Number of glock requests queued (qcnt in glstats file)
230 ========= =================================== 230 ========= ================================================================
231 231
232 The sbstats file contains a set of these stats 232 The sbstats file contains a set of these stats for each glock type (so 8 lines
233 for each type) and for each cpu (one column pe 233 for each type) and for each cpu (one column per cpu). The glstats file contains
234 a set of these stats for each glock in a simil 234 a set of these stats for each glock in a similar format to the glocks file, but
235 using the format mean/variance for each of the 235 using the format mean/variance for each of the timing stats.
236 236
237 The gfs2_glock_lock_time tracepoint prints out 237 The gfs2_glock_lock_time tracepoint prints out the current values of the stats
238 for the glock in question, along with some add 238 for the glock in question, along with some addition information on each dlm
239 reply that is received: 239 reply that is received:
240 240
241 ====== ===================================== 241 ====== =======================================
242 status The status of the dlm request 242 status The status of the dlm request
243 flags The dlm request flags 243 flags The dlm request flags
244 tdiff The time taken by this specific reque 244 tdiff The time taken by this specific request
245 ====== ===================================== 245 ====== =======================================
246 246
247 (remaining fields as per above list) 247 (remaining fields as per above list)
248 248
249 249
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.