1 The Linux Journalling API 2 ========================= 3 4 Overview 5 -------- 6 7 Details 8 ~~~~~~~ 9 10 The journalling layer is easy to use. You need to first of all create a 11 journal_t data structure. There are two calls to do this dependent on 12 how you decide to allocate the physical media on which the journal 13 resides. The jbd2_journal_init_inode() call is for journals stored in 14 filesystem inodes, or the jbd2_journal_init_dev() call can be used 15 for journal stored on a raw device (in a continuous range of blocks). A 16 journal_t is a typedef for a struct pointer, so when you are finally 17 finished make sure you call jbd2_journal_destroy() on it to free up 18 any used kernel memory. 19 20 Once you have got your journal_t object you need to 'mount' or load the 21 journal file. The journalling layer expects the space for the journal 22 was already allocated and initialized properly by the userspace tools. 23 When loading the journal you must call jbd2_journal_load() to process 24 journal contents. If the client file system detects the journal contents 25 does not need to be processed (or even need not have valid contents), it 26 may call jbd2_journal_wipe() to clear the journal contents before 27 calling jbd2_journal_load(). 28 29 Note that jbd2_journal_wipe(..,0) calls 30 jbd2_journal_skip_recovery() for you if it detects any outstanding 31 transactions in the journal and similarly jbd2_journal_load() will 32 call jbd2_journal_recover() if necessary. I would advise reading 33 ext4_load_journal() in fs/ext4/super.c for examples on this stage. 34 35 Now you can go ahead and start modifying the underlying filesystem. 36 Almost. 37 38 You still need to actually journal your filesystem changes, this is done 39 by wrapping them into transactions. Additionally you also need to wrap 40 the modification of each of the buffers with calls to the journal layer, 41 so it knows what the modifications you are actually making are. To do 42 this use jbd2_journal_start() which returns a transaction handle. 43 44 jbd2_journal_start() and its counterpart jbd2_journal_stop(), 45 which indicates the end of a transaction are nestable calls, so you can 46 reenter a transaction if necessary, but remember you must call 47 jbd2_journal_stop() the same number of times as 48 jbd2_journal_start() before the transaction is completed (or more 49 accurately leaves the update phase). Ext4/VFS makes use of this feature to 50 simplify handling of inode dirtying, quota support, etc. 51 52 Inside each transaction you need to wrap the modifications to the 53 individual buffers (blocks). Before you start to modify a buffer you 54 need to call jbd2_journal_get_create_access() / 55 jbd2_journal_get_write_access() / 56 jbd2_journal_get_undo_access() as appropriate, this allows the 57 journalling layer to copy the unmodified 58 data if it needs to. After all the buffer may be part of a previously 59 uncommitted transaction. At this point you are at last ready to modify a 60 buffer, and once you are have done so you need to call 61 jbd2_journal_dirty_metadata(). Or if you've asked for access to a 62 buffer you now know is now longer required to be pushed back on the 63 device you can call jbd2_journal_forget() in much the same way as you 64 might have used bforget() in the past. 65 66 A jbd2_journal_flush() may be called at any time to commit and 67 checkpoint all your transactions. 68 69 Then at umount time , in your put_super() you can then call 70 jbd2_journal_destroy() to clean up your in-core journal object. 71 72 Unfortunately there a couple of ways the journal layer can cause a 73 deadlock. The first thing to note is that each task can only have a 74 single outstanding transaction at any one time, remember nothing commits 75 until the outermost jbd2_journal_stop(). This means you must complete 76 the transaction at the end of each file/inode/address etc. operation you 77 perform, so that the journalling system isn't re-entered on another 78 journal. Since transactions can't be nested/batched across differing 79 journals, and another filesystem other than yours (say ext4) may be 80 modified in a later syscall. 81 82 The second case to bear in mind is that jbd2_journal_start() can block 83 if there isn't enough space in the journal for your transaction (based 84 on the passed nblocks param) - when it blocks it merely(!) needs to wait 85 for transactions to complete and be committed from other tasks, so 86 essentially we are waiting for jbd2_journal_stop(). So to avoid 87 deadlocks you must treat jbd2_journal_start() / 88 jbd2_journal_stop() as if they were semaphores and include them in 89 your semaphore ordering rules to prevent 90 deadlocks. Note that jbd2_journal_extend() has similar blocking 91 behaviour to jbd2_journal_start() so you can deadlock here just as 92 easily as on jbd2_journal_start(). 93 94 Try to reserve the right number of blocks the first time. ;-). This will 95 be the maximum number of blocks you are going to touch in this 96 transaction. I advise having a look at at least ext4_jbd.h to see the 97 basis on which ext4 uses to make these decisions. 98 99 Another wriggle to watch out for is your on-disk block allocation 100 strategy. Why? Because, if you do a delete, you need to ensure you 101 haven't reused any of the freed blocks until the transaction freeing 102 these blocks commits. If you reused these blocks and crash happens, 103 there is no way to restore the contents of the reallocated blocks at the 104 end of the last fully committed transaction. One simple way of doing 105 this is to mark blocks as free in internal in-memory block allocation 106 structures only after the transaction freeing them commits. Ext4 uses 107 journal commit callback for this purpose. 108 109 With journal commit callbacks you can ask the journalling layer to call 110 a callback function when the transaction is finally committed to disk, 111 so that you can do some of your own management. You ask the journalling 112 layer for calling the callback by simply setting 113 ``journal->j_commit_callback`` function pointer and that function is 114 called after each transaction commit. You can also use 115 ``transaction->t_private_list`` for attaching entries to a transaction 116 that need processing when the transaction commits. 117 118 JBD2 also provides a way to block all transaction updates via 119 jbd2_journal_lock_updates() / 120 jbd2_journal_unlock_updates(). Ext4 uses this when it wants a 121 window with a clean and stable fs for a moment. E.g. 122 123 :: 124 125 126 jbd2_journal_lock_updates() //stop new stuff happening.. 127 jbd2_journal_flush() // checkpoint everything. 128 ..do stuff on stable fs 129 jbd2_journal_unlock_updates() // carry on with filesystem use. 130 131 The opportunities for abuse and DOS attacks with this should be obvious, 132 if you allow unprivileged userspace to trigger codepaths containing 133 these calls. 134 135 Fast commits 136 ~~~~~~~~~~~~ 137 138 JBD2 to also allows you to perform file-system specific delta commits known as 139 fast commits. In order to use fast commits, you will need to set following 140 callbacks that perform corresponding work: 141 142 `journal->j_fc_cleanup_cb`: Cleanup function called after every full commit and 143 fast commit. 144 145 `journal->j_fc_replay_cb`: Replay function called for replay of fast commit 146 blocks. 147 148 File system is free to perform fast commits as and when it wants as long as it 149 gets permission from JBD2 to do so by calling the function 150 :c:func:`jbd2_fc_begin_commit()`. Once a fast commit is done, the client 151 file system should tell JBD2 about it by calling 152 :c:func:`jbd2_fc_end_commit()`. If the file system wants JBD2 to perform a full 153 commit immediately after stopping the fast commit it can do so by calling 154 :c:func:`jbd2_fc_end_commit_fallback()`. This is useful if fast commit operation 155 fails for some reason and the only way to guarantee consistency is for JBD2 to 156 perform the full traditional commit. 157 158 JBD2 helper functions to manage fast commit buffers. File system can use 159 :c:func:`jbd2_fc_get_buf()` and :c:func:`jbd2_fc_wait_bufs()` to allocate 160 and wait on IO completion of fast commit buffers. 161 162 Currently, only Ext4 implements fast commits. For details of its implementation 163 of fast commits, please refer to the top level comments in 164 fs/ext4/fast_commit.c. 165 166 Summary 167 ~~~~~~~ 168 169 Using the journal is a matter of wrapping the different context changes, 170 being each mount, each modification (transaction) and each changed 171 buffer to tell the journalling layer about them. 172 173 Data Types 174 ---------- 175 176 The journalling layer uses typedefs to 'hide' the concrete definitions 177 of the structures used. As a client of the JBD2 layer you can just rely 178 on the using the pointer as a magic cookie of some sort. Obviously the 179 hiding is not enforced as this is 'C'. 180 181 Structures 182 ~~~~~~~~~~ 183 184 .. kernel-doc:: include/linux/jbd2.h 185 :internal: 186 187 Functions 188 --------- 189 190 The functions here are split into two groups those that affect a journal 191 as a whole, and those which are used to manage transactions 192 193 Journal Level 194 ~~~~~~~~~~~~~ 195 196 .. kernel-doc:: fs/jbd2/journal.c 197 :export: 198 199 .. kernel-doc:: fs/jbd2/recovery.c 200 :internal: 201 202 Transaction Level 203 ~~~~~~~~~~~~~~~~~~ 204 205 .. kernel-doc:: fs/jbd2/transaction.c 206 207 See also 208 -------- 209 210 `Journaling the Linux ext2fs Filesystem, LinuxExpo 98, Stephen 211 Tweedie <http://kernel.org/pub/linux/kernel/people/sct/ext3/journal-design.ps.gz>`__ 212 213 `Ext3 Journalling FileSystem, OLS 2000, Dr. Stephen 214 Tweedie <http://olstrans.sourceforge.net/release/OLS2000-ext3/OLS2000-ext3.html>`__ 215
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.