216 lines · plain
1The Linux Journalling API2=========================3 4Overview5--------6 7Details8~~~~~~~9 10The journalling layer is easy to use. You need to first of all create a11journal_t data structure. There are two calls to do this dependent on12how you decide to allocate the physical media on which the journal13resides. The jbd2_journal_init_inode() call is for journals stored in14filesystem inodes, or the jbd2_journal_init_dev() call can be used15for journal stored on a raw device (in a continuous range of blocks). A16journal_t is a typedef for a struct pointer, so when you are finally17finished make sure you call jbd2_journal_destroy() on it to free up18any used kernel memory.19 20Once you have got your journal_t object you need to 'mount' or load the21journal file. The journalling layer expects the space for the journal22was already allocated and initialized properly by the userspace tools.23When loading the journal you must call jbd2_journal_load() to process24journal contents. If the client file system detects the journal contents25does not need to be processed (or even need not have valid contents), it26may call jbd2_journal_wipe() to clear the journal contents before27calling jbd2_journal_load().28 29Note that jbd2_journal_wipe(..,0) calls30jbd2_journal_skip_recovery() for you if it detects any outstanding31transactions in the journal and similarly jbd2_journal_load() will32call jbd2_journal_recover() if necessary. I would advise reading33ext4_load_journal() in fs/ext4/super.c for examples on this stage.34 35Now you can go ahead and start modifying the underlying filesystem.36Almost.37 38You still need to actually journal your filesystem changes, this is done39by wrapping them into transactions. Additionally you also need to wrap40the modification of each of the buffers with calls to the journal layer,41so it knows what the modifications you are actually making are. To do42this use jbd2_journal_start() which returns a transaction handle.43 44jbd2_journal_start() and its counterpart jbd2_journal_stop(),45which indicates the end of a transaction are nestable calls, so you can46reenter a transaction if necessary, but remember you must call47jbd2_journal_stop() the same number of times as48jbd2_journal_start() before the transaction is completed (or more49accurately leaves the update phase). Ext4/VFS makes use of this feature to50simplify handling of inode dirtying, quota support, etc.51 52Inside each transaction you need to wrap the modifications to the53individual buffers (blocks). Before you start to modify a buffer you54need to call jbd2_journal_get_create_access() /55jbd2_journal_get_write_access() /56jbd2_journal_get_undo_access() as appropriate, this allows the57journalling layer to copy the unmodified58data if it needs to. After all the buffer may be part of a previously59uncommitted transaction. At this point you are at last ready to modify a60buffer, and once you are have done so you need to call61jbd2_journal_dirty_metadata(). Or if you've asked for access to a62buffer you now know is now longer required to be pushed back on the63device you can call jbd2_journal_forget() in much the same way as you64might have used bforget() in the past.65 66A jbd2_journal_flush() may be called at any time to commit and67checkpoint all your transactions.68 69Then at umount time , in your put_super() you can then call70jbd2_journal_destroy() to clean up your in-core journal object.71 72Unfortunately there a couple of ways the journal layer can cause a73deadlock. The first thing to note is that each task can only have a74single outstanding transaction at any one time, remember nothing commits75until the outermost jbd2_journal_stop(). This means you must complete76the transaction at the end of each file/inode/address etc. operation you77perform, so that the journalling system isn't re-entered on another78journal. Since transactions can't be nested/batched across differing79journals, and another filesystem other than yours (say ext4) may be80modified in a later syscall.81 82The second case to bear in mind is that jbd2_journal_start() can block83if there isn't enough space in the journal for your transaction (based84on the passed nblocks param) - when it blocks it merely(!) needs to wait85for transactions to complete and be committed from other tasks, so86essentially we are waiting for jbd2_journal_stop(). So to avoid87deadlocks you must treat jbd2_journal_start() /88jbd2_journal_stop() as if they were semaphores and include them in89your semaphore ordering rules to prevent90deadlocks. Note that jbd2_journal_extend() has similar blocking91behaviour to jbd2_journal_start() so you can deadlock here just as92easily as on jbd2_journal_start().93 94Try to reserve the right number of blocks the first time. ;-). This will95be the maximum number of blocks you are going to touch in this96transaction. I advise having a look at at least ext4_jbd.h to see the97basis on which ext4 uses to make these decisions.98 99Another wriggle to watch out for is your on-disk block allocation100strategy. Why? Because, if you do a delete, you need to ensure you101haven't reused any of the freed blocks until the transaction freeing102these blocks commits. If you reused these blocks and crash happens,103there is no way to restore the contents of the reallocated blocks at the104end of the last fully committed transaction. One simple way of doing105this is to mark blocks as free in internal in-memory block allocation106structures only after the transaction freeing them commits. Ext4 uses107journal commit callback for this purpose.108 109With journal commit callbacks you can ask the journalling layer to call110a callback function when the transaction is finally committed to disk,111so that you can do some of your own management. You ask the journalling112layer for calling the callback by simply setting113``journal->j_commit_callback`` function pointer and that function is114called after each transaction commit. You can also use115``transaction->t_private_list`` for attaching entries to a transaction116that need processing when the transaction commits.117 118JBD2 also provides a way to block all transaction updates via119jbd2_journal_lock_updates() /120jbd2_journal_unlock_updates(). Ext4 uses this when it wants a121window 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 fs129 jbd2_journal_unlock_updates() // carry on with filesystem use.130 131The opportunities for abuse and DOS attacks with this should be obvious,132if you allow unprivileged userspace to trigger codepaths containing133these calls.134 135Fast commits136~~~~~~~~~~~~137 138JBD2 to also allows you to perform file-system specific delta commits known as139fast commits. In order to use fast commits, you will need to set following140callbacks that perform corresponding work:141 142`journal->j_fc_cleanup_cb`: Cleanup function called after every full commit and143fast commit.144 145`journal->j_fc_replay_cb`: Replay function called for replay of fast commit146blocks.147 148File system is free to perform fast commits as and when it wants as long as it149gets permission from JBD2 to do so by calling the function150:c:func:`jbd2_fc_begin_commit()`. Once a fast commit is done, the client151file system should tell JBD2 about it by calling152:c:func:`jbd2_fc_end_commit()`. If the file system wants JBD2 to perform a full153commit immediately after stopping the fast commit it can do so by calling154:c:func:`jbd2_fc_end_commit_fallback()`. This is useful if fast commit operation155fails for some reason and the only way to guarantee consistency is for JBD2 to156perform the full traditional commit.157 158JBD2 helper functions to manage fast commit buffers. File system can use159:c:func:`jbd2_fc_get_buf()` and :c:func:`jbd2_fc_wait_bufs()` to allocate160and wait on IO completion of fast commit buffers.161 162Currently, only Ext4 implements fast commits. For details of its implementation163of fast commits, please refer to the top level comments in164fs/ext4/fast_commit.c.165 166Summary167~~~~~~~168 169Using the journal is a matter of wrapping the different context changes,170being each mount, each modification (transaction) and each changed171buffer to tell the journalling layer about them.172 173Data Types174----------175 176The journalling layer uses typedefs to 'hide' the concrete definitions177of the structures used. As a client of the JBD2 layer you can just rely178on the using the pointer as a magic cookie of some sort. Obviously the179hiding is not enforced as this is 'C'.180 181Structures182~~~~~~~~~~183 184.. kernel-doc:: include/linux/jbd2.h185 :internal:186 187Functions188---------189 190The functions here are split into two groups those that affect a journal191as a whole, and those which are used to manage transactions192 193Journal Level194~~~~~~~~~~~~~195 196.. kernel-doc:: fs/jbd2/journal.c197 :export:198 199.. kernel-doc:: fs/jbd2/recovery.c200 :internal:201 202Transaction Level203~~~~~~~~~~~~~~~~~~204 205.. kernel-doc:: fs/jbd2/transaction.c206 207See also208--------209 210`Journaling the Linux ext2fs Filesystem, LinuxExpo 98, Stephen211Tweedie <http://kernel.org/pub/linux/kernel/people/sct/ext3/journal-design.ps.gz>`__212 213`Ext3 Journalling FileSystem, OLS 2000, Dr. Stephen214Tweedie <http://olstrans.sourceforge.net/release/OLS2000-ext3/OLS2000-ext3.html>`__215 216