1<html><head><meta http-equiv="Content-Type" content="text/html; charset=ANSI_X3.4-1968"><title>Chapter 6. The Linux Journalling API</title><meta name="generator" content="DocBook XSL Stylesheets V1.78.1"><link rel="home" href="index.html" title="Linux Filesystems API"><link rel="up" href="index.html" title="Linux Filesystems API"><link rel="prev" href="API-debugfs-create-devm-seqfile.html" title="debugfs_create_devm_seqfile"><link rel="next" href="data_types.html" title="Data Types"></head><body bgcolor="white" text="black" link="#0000FF" vlink="#840084" alink="#0000FF"><div class="navheader"><table width="100%" summary="Navigation header"><tr><th colspan="3" align="center">Chapter 6. The Linux Journalling API</th></tr><tr><td width="20%" align="left"><a accesskey="p" href="API-debugfs-create-devm-seqfile.html">Prev</a> </td><th width="60%" align="center"> </th><td width="20%" align="right"> <a accesskey="n" href="data_types.html">Next</a></td></tr></table><hr></div><div class="chapter"><div class="titlepage"><div><div><h1 class="title"><a name="LinuxJDBAPI"></a>Chapter 6. The Linux Journalling API</h1></div><div><div class="authorgroup"><div class="author"><h3 class="author"><span class="firstname">Roger</span> <span class="surname">Gammans</span></h3><div class="affiliation"><div class="address"><p><br> 2      <code class="email"><<a class="email" href="mailto:rgammans@computer-surgery.co.uk">rgammans@computer-surgery.co.uk</a>></code><br> 3     </p></div></div></div></div></div><div><div class="authorgroup"><div class="author"><h3 class="author"><span class="firstname">Stephen</span> <span class="surname">Tweedie</span></h3><div class="affiliation"><div class="address"><p><br> 4      <code class="email"><<a class="email" href="mailto:sct@redhat.com">sct@redhat.com</a>></code><br> 5     </p></div></div></div></div></div><div><p class="copyright">Copyright © 2002 Roger Gammans</p></div></div></div><div class="toc"><p><b>Table of Contents</b></p><dl class="toc"><dt><span class="sect1"><a href="LinuxJDBAPI.html#journaling_overview">Overview</a></span></dt><dd><dl><dt><span class="sect2"><a href="LinuxJDBAPI.html#journaling_details">Details</a></span></dt><dt><span class="sect2"><a href="LinuxJDBAPI.html#jbd_summary">Summary</a></span></dt></dl></dd><dt><span class="sect1"><a href="data_types.html">Data Types</a></span></dt><dd><dl><dt><span class="sect2"><a href="data_types.html#structures">Structures</a></span></dt></dl></dd><dt><span class="sect1"><a href="functions.html">Functions</a></span></dt><dd><dl><dt><span class="sect2"><a href="functions.html#journal_level">Journal Level</a></span></dt><dt><span class="sect2"><a href="functions.html#transaction_level">Transasction Level</a></span></dt></dl></dd><dt><span class="sect1"><a href="see_also.html">See also</a></span></dt></dl></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="journaling_overview"></a>Overview</h2></div></div></div><div class="toc"><dl class="toc"><dt><span class="sect2"><a href="LinuxJDBAPI.html#journaling_details">Details</a></span></dt><dt><span class="sect2"><a href="LinuxJDBAPI.html#jbd_summary">Summary</a></span></dt></dl></div><div class="sect2"><div class="titlepage"><div><div><h3 class="title"><a name="journaling_details"></a>Details</h3></div></div></div><p> 6The journalling layer is easy to use. You need to 7first of all create a journal_t data structure. There are 8two calls to do this dependent on how you decide to allocate the physical 9media on which the journal resides. The journal_init_inode() call 10is for journals stored in filesystem inodes, or the journal_init_dev() 11call can be use for journal stored on a raw device (in a continuous range 12of blocks). A journal_t is a typedef for a struct pointer, so when 13you are finally finished make sure you call journal_destroy() on it 14to free up any used kernel memory. 15</p><p> 16Once you have got your journal_t object you need to 'mount' or load the journal 17file, unless of course you haven't initialised it yet - in which case you 18need to call journal_create(). 19</p><p> 20Most of the time however your journal file will already have been created, but 21before you load it you must call journal_wipe() to empty the journal file. 22Hang on, you say , what if the filesystem wasn't cleanly umount()'d . Well, it is the 23job of the client file system to detect this and skip the call to journal_wipe(). 24</p><p> 25In either case the next call should be to journal_load() which prepares the 26journal file for use. Note that journal_wipe(..,0) calls journal_skip_recovery() 27for you if it detects any outstanding transactions in the journal and similarly 28journal_load() will call journal_recover() if necessary. 29I would advise reading fs/ext3/super.c for examples on this stage. 30[RGG: Why is the journal_wipe() call necessary - doesn't this needlessly 31complicate the API. Or isn't a good idea for the journal layer to hide 32dirty mounts from the client fs] 33</p><p> 34Now you can go ahead and start modifying the underlying 35filesystem. Almost. 36</p><p> 37 38You still need to actually journal your filesystem changes, this 39is done by wrapping them into transactions. Additionally you 40also need to wrap the modification of each of the buffers 41with calls to the journal layer, so it knows what the modifications 42you are actually making are. To do this use journal_start() which 43returns a transaction handle. 44</p><p> 45journal_start() 46and its counterpart journal_stop(), which indicates the end of a transaction 47are nestable calls, so you can reenter a transaction if necessary, 48but remember you must call journal_stop() the same number of times as 49journal_start() before the transaction is completed (or more accurately 50leaves the update phase). Ext3/VFS makes use of this feature to simplify 51quota support. 52</p><p> 53Inside each transaction you need to wrap the modifications to the 54individual buffers (blocks). Before you start to modify a buffer you 55need to call journal_get_{create,write,undo}_access() as appropriate, 56this allows the journalling layer to copy the unmodified data if it 57needs to. After all the buffer may be part of a previously uncommitted 58transaction. 59At this point you are at last ready to modify a buffer, and once 60you are have done so you need to call journal_dirty_{meta,}data(). 61Or if you've asked for access to a buffer you now know is now longer 62required to be pushed back on the device you can call journal_forget() 63in much the same way as you might have used bforget() in the past. 64</p><p> 65A journal_flush() may be called at any time to commit and checkpoint 66all your transactions. 67</p><p> 68Then at umount time , in your put_super() you can then call journal_destroy() 69to clean up your in-core journal object. 70</p><p> 71Unfortunately there a couple of ways the journal layer can cause a deadlock. 72The first thing to note is that each task can only have 73a single outstanding transaction at any one time, remember nothing 74commits until the outermost journal_stop(). This means 75you must complete the transaction at the end of each file/inode/address 76etc. operation you perform, so that the journalling system isn't re-entered 77on another journal. Since transactions can't be nested/batched 78across differing journals, and another filesystem other than 79yours (say ext3) may be modified in a later syscall. 80</p><p> 81The second case to bear in mind is that journal_start() can 82block if there isn't enough space in the journal for your transaction 83(based on the passed nblocks param) - when it blocks it merely(!) needs to 84wait for transactions to complete and be committed from other tasks, 85so essentially we are waiting for journal_stop(). So to avoid 86deadlocks you must treat journal_start/stop() as if they 87were semaphores and include them in your semaphore ordering rules to prevent 88deadlocks. Note that journal_extend() has similar blocking behaviour to 89journal_start() so you can deadlock here just as easily as on journal_start(). 90</p><p> 91Try to reserve the right number of blocks the first time. ;-). This will 92be the maximum number of blocks you are going to touch in this transaction. 93I advise having a look at at least ext3_jbd.h to see the basis on which 94ext3 uses to make these decisions. 95</p><p> 96Another wriggle to watch out for is your on-disk block allocation strategy. 97why? Because, if you undo a delete, you need to ensure you haven't reused any 98of the freed blocks in a later transaction. One simple way of doing this 99is make sure any blocks you allocate only have checkpointed transactions 100listed against them. Ext3 does this in ext3_test_allocatable(). 101</p><p> 102Lock is also providing through journal_{un,}lock_updates(), 103ext3 uses this when it wants a window with a clean and stable fs for a moment. 104eg. 105</p><pre class="programlisting"> 106 107 journal_lock_updates() //stop new stuff happening.. 108 journal_flush() // checkpoint everything. 109 ..do stuff on stable fs 110 journal_unlock_updates() // carry on with filesystem use. 111</pre><p> 112The opportunities for abuse and DOS attacks with this should be obvious, 113if you allow unprivileged userspace to trigger codepaths containing these 114calls. 115</p><p> 116A new feature of jbd since 2.5.25 is commit callbacks with the new 117journal_callback_set() function you can now ask the journalling layer 118to call you back when the transaction is finally committed to disk, so that 119you can do some of your own management. The key to this is the journal_callback 120struct, this maintains the internal callback information but you can 121extend it like this:- 122</p><pre class="programlisting"> 123 struct myfs_callback_s { 124 //Data structure element required by jbd.. 125 struct journal_callback for_jbd; 126 // Stuff for myfs allocated together. 127 myfs_inode* i_commited; 128 129 } 130</pre><p> 131this would be useful if you needed to know when data was committed to a 132particular inode. 133</p></div><div class="sect2"><div class="titlepage"><div><div><h3 class="title"><a name="jbd_summary"></a>Summary</h3></div></div></div><p> 134Using the journal is a matter of wrapping the different context changes, 135being each mount, each modification (transaction) and each changed buffer 136to tell the journalling layer about them. 137</p><p> 138Here is a some pseudo code to give you an idea of how it works, as 139an example. 140</p><pre class="programlisting"> 141 journal_t* my_jnrl = journal_create(); 142 journal_init_{dev,inode}(jnrl,...) 143 if (clean) journal_wipe(); 144 journal_load(); 145 146 foreach(transaction) { /*transactions must be 147 completed before 148 a syscall returns to 149 userspace*/ 150 151 handle_t * xct=journal_start(my_jnrl); 152 foreach(bh) { 153 journal_get_{create,write,undo}_access(xact,bh); 154 if ( myfs_modify(bh) ) { /* returns true 155 if makes changes */ 156 journal_dirty_{meta,}data(xact,bh); 157 } else { 158 journal_forget(bh); 159 } 160 } 161 journal_stop(xct); 162 } 163 journal_destroy(my_jrnl); 164</pre></div></div></div><div class="navfooter"><hr><table width="100%" summary="Navigation footer"><tr><td width="40%" align="left"><a accesskey="p" href="API-debugfs-create-devm-seqfile.html">Prev</a> </td><td width="20%" align="center"> </td><td width="40%" align="right"> <a accesskey="n" href="data_types.html">Next</a></td></tr><tr><td width="40%" align="left" valign="top"><span class="phrase">debugfs_create_devm_seqfile</span> </td><td width="20%" align="center"><a accesskey="h" href="index.html">Home</a></td><td width="40%" align="right" valign="top"> Data Types</td></tr></table></div></body></html> 165