1<?xml version="1.0" encoding="UTF-8"?> 2<!DOCTYPE book PUBLIC "-//OASIS//DTD DocBook XML V4.1.2//EN" 3 "http://www.oasis-open.org/docbook/xml/4.1.2/docbookx.dtd" []> 4 5<book id="Linux-filesystems-API"> 6 <bookinfo> 7 <title>Linux Filesystems API</title> 8 9 <legalnotice> 10 <para> 11 This documentation is free software; you can redistribute 12 it and/or modify it under the terms of the GNU General Public 13 License as published by the Free Software Foundation; either 14 version 2 of the License, or (at your option) any later 15 version. 16 </para> 17 18 <para> 19 This program is distributed in the hope that it will be 20 useful, but WITHOUT ANY WARRANTY; without even the implied 21 warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. 22 See the GNU General Public License for more details. 23 </para> 24 25 <para> 26 You should have received a copy of the GNU General Public 27 License along with this program; if not, write to the Free 28 Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, 29 MA 02111-1307 USA 30 </para> 31 32 <para> 33 For more details see the file COPYING in the source 34 distribution of Linux. 35 </para> 36 </legalnotice> 37 </bookinfo> 38 39<toc></toc> 40 41 <chapter id="vfs"> 42 <title>The Linux VFS</title> 43 <sect1 id="the_filesystem_types"><title>The Filesystem types</title> 44!Iinclude/linux/fs.h 45 </sect1> 46 <sect1 id="the_directory_cache"><title>The Directory Cache</title> 47!Efs/dcache.c 48!Iinclude/linux/dcache.h 49 </sect1> 50 <sect1 id="inode_handling"><title>Inode Handling</title> 51!Efs/inode.c 52!Efs/bad_inode.c 53 </sect1> 54 <sect1 id="registration_and_superblocks"><title>Registration and Superblocks</title> 55!Efs/super.c 56 </sect1> 57 <sect1 id="file_locks"><title>File Locks</title> 58!Efs/locks.c 59!Ifs/locks.c 60 </sect1> 61 <sect1 id="other_functions"><title>Other Functions</title> 62!Efs/mpage.c 63!Efs/namei.c 64!Efs/buffer.c 65!Efs/bio.c 66!Efs/seq_file.c 67!Efs/filesystems.c 68!Efs/fs-writeback.c 69!Efs/block_dev.c 70 </sect1> 71 </chapter> 72 73 <chapter id="proc"> 74 <title>The proc filesystem</title> 75 76 <sect1 id="sysctl_interface"><title>sysctl interface</title> 77!Ekernel/sysctl.c 78 </sect1> 79 80 <sect1 id="proc_filesystem_interface"><title>proc filesystem interface</title> 81!Ifs/proc/base.c 82 </sect1> 83 </chapter> 84 85 <chapter id="sysfs"> 86 <title>The Filesystem for Exporting Kernel Objects</title> 87!Efs/sysfs/file.c 88!Efs/sysfs/symlink.c 89!Efs/sysfs/bin.c 90 </chapter> 91 92 <chapter id="debugfs"> 93 <title>The debugfs filesystem</title> 94 95 <sect1 id="debugfs_interface"><title>debugfs interface</title> 96!Efs/debugfs/inode.c 97!Efs/debugfs/file.c 98 </sect1> 99 </chapter> 100 101 <chapter id="LinuxJDBAPI"> 102 <chapterinfo> 103 <title>The Linux Journalling API</title> 104 105 <authorgroup> 106 <author> 107 <firstname>Roger</firstname> 108 <surname>Gammans</surname> 109 <affiliation> 110 <address> 111 <email>rgammans@computer-surgery.co.uk</email> 112 </address> 113 </affiliation> 114 </author> 115 </authorgroup> 116 117 <authorgroup> 118 <author> 119 <firstname>Stephen</firstname> 120 <surname>Tweedie</surname> 121 <affiliation> 122 <address> 123 <email>sct@redhat.com</email> 124 </address> 125 </affiliation> 126 </author> 127 </authorgroup> 128 129 <copyright> 130 <year>2002</year> 131 <holder>Roger Gammans</holder> 132 </copyright> 133 </chapterinfo> 134 135 <title>The Linux Journalling API</title> 136 137 <sect1 id="journaling_overview"> 138 <title>Overview</title> 139 <sect2 id="journaling_details"> 140 <title>Details</title> 141<para> 142The journalling layer is easy to use. You need to 143first of all create a journal_t data structure. There are 144two calls to do this dependent on how you decide to allocate the physical 145media on which the journal resides. The journal_init_inode() call 146is for journals stored in filesystem inodes, or the journal_init_dev() 147call can be use for journal stored on a raw device (in a continuous range 148of blocks). A journal_t is a typedef for a struct pointer, so when 149you are finally finished make sure you call journal_destroy() on it 150to free up any used kernel memory. 151</para> 152 153<para> 154Once you have got your journal_t object you need to 'mount' or load the journal 155file, unless of course you haven't initialised it yet - in which case you 156need to call journal_create(). 157</para> 158 159<para> 160Most of the time however your journal file will already have been created, but 161before you load it you must call journal_wipe() to empty the journal file. 162Hang on, you say , what if the filesystem wasn't cleanly umount()'d . Well, it is the 163job of the client file system to detect this and skip the call to journal_wipe(). 164</para> 165 166<para> 167In either case the next call should be to journal_load() which prepares the 168journal file for use. Note that journal_wipe(..,0) calls journal_skip_recovery() 169for you if it detects any outstanding transactions in the journal and similarly 170journal_load() will call journal_recover() if necessary. 171I would advise reading fs/ext3/super.c for examples on this stage. 172[RGG: Why is the journal_wipe() call necessary - doesn't this needlessly 173complicate the API. Or isn't a good idea for the journal layer to hide 174dirty mounts from the client fs] 175</para> 176 177<para> 178Now you can go ahead and start modifying the underlying 179filesystem. Almost. 180</para> 181 182<para> 183 184You still need to actually journal your filesystem changes, this 185is done by wrapping them into transactions. Additionally you 186also need to wrap the modification of each of the buffers 187with calls to the journal layer, so it knows what the modifications 188you are actually making are. To do this use journal_start() which 189returns a transaction handle. 190</para> 191 192<para> 193journal_start() 194and its counterpart journal_stop(), which indicates the end of a transaction 195are nestable calls, so you can reenter a transaction if necessary, 196but remember you must call journal_stop() the same number of times as 197journal_start() before the transaction is completed (or more accurately 198leaves the update phase). Ext3/VFS makes use of this feature to simplify 199quota support. 200</para> 201 202<para> 203Inside each transaction you need to wrap the modifications to the 204individual buffers (blocks). Before you start to modify a buffer you 205need to call journal_get_{create,write,undo}_access() as appropriate, 206this allows the journalling layer to copy the unmodified data if it 207needs to. After all the buffer may be part of a previously uncommitted 208transaction. 209At this point you are at last ready to modify a buffer, and once 210you are have done so you need to call journal_dirty_{meta,}data(). 211Or if you've asked for access to a buffer you now know is now longer 212required to be pushed back on the device you can call journal_forget() 213in much the same way as you might have used bforget() in the past. 214</para> 215 216<para> 217A journal_flush() may be called at any time to commit and checkpoint 218all your transactions. 219</para> 220 221<para> 222Then at umount time , in your put_super() (2.4) or write_super() (2.5) 223you can then call journal_destroy() to clean up your in-core journal object. 224</para> 225 226<para> 227Unfortunately there a couple of ways the journal layer can cause a deadlock. 228The first thing to note is that each task can only have 229a single outstanding transaction at any one time, remember nothing 230commits until the outermost journal_stop(). This means 231you must complete the transaction at the end of each file/inode/address 232etc. operation you perform, so that the journalling system isn't re-entered 233on another journal. Since transactions can't be nested/batched 234across differing journals, and another filesystem other than 235yours (say ext3) may be modified in a later syscall. 236</para> 237 238<para> 239The second case to bear in mind is that journal_start() can 240block if there isn't enough space in the journal for your transaction 241(based on the passed nblocks param) - when it blocks it merely(!) needs to 242wait for transactions to complete and be committed from other tasks, 243so essentially we are waiting for journal_stop(). So to avoid 244deadlocks you must treat journal_start/stop() as if they 245were semaphores and include them in your semaphore ordering rules to prevent 246deadlocks. Note that journal_extend() has similar blocking behaviour to 247journal_start() so you can deadlock here just as easily as on journal_start(). 248</para> 249 250<para> 251Try to reserve the right number of blocks the first time. ;-). This will 252be the maximum number of blocks you are going to touch in this transaction. 253I advise having a look at at least ext3_jbd.h to see the basis on which 254ext3 uses to make these decisions. 255</para> 256 257<para> 258Another wriggle to watch out for is your on-disk block allocation strategy. 259why? Because, if you undo a delete, you need to ensure you haven't reused any 260of the freed blocks in a later transaction. One simple way of doing this 261is make sure any blocks you allocate only have checkpointed transactions 262listed against them. Ext3 does this in ext3_test_allocatable(). 263</para> 264 265<para> 266Lock is also providing through journal_{un,}lock_updates(), 267ext3 uses this when it wants a window with a clean and stable fs for a moment. 268eg. 269</para> 270 271<programlisting> 272 273 journal_lock_updates() //stop new stuff happening.. 274 journal_flush() // checkpoint everything. 275 ..do stuff on stable fs 276 journal_unlock_updates() // carry on with filesystem use. 277</programlisting> 278 279<para> 280The opportunities for abuse and DOS attacks with this should be obvious, 281if you allow unprivileged userspace to trigger codepaths containing these 282calls. 283</para> 284 285<para> 286A new feature of jbd since 2.5.25 is commit callbacks with the new 287journal_callback_set() function you can now ask the journalling layer 288to call you back when the transaction is finally committed to disk, so that 289you can do some of your own management. The key to this is the journal_callback 290struct, this maintains the internal callback information but you can 291extend it like this:- 292</para> 293<programlisting> 294 struct myfs_callback_s { 295 //Data structure element required by jbd.. 296 struct journal_callback for_jbd; 297 // Stuff for myfs allocated together. 298 myfs_inode* i_commited; 299 300 } 301</programlisting> 302 303<para> 304this would be useful if you needed to know when data was committed to a 305particular inode. 306</para> 307 308 </sect2> 309 310 <sect2 id="jbd_summary"> 311 <title>Summary</title> 312<para> 313Using the journal is a matter of wrapping the different context changes, 314being each mount, each modification (transaction) and each changed buffer 315to tell the journalling layer about them. 316</para> 317 318<para> 319Here is a some pseudo code to give you an idea of how it works, as 320an example. 321</para> 322 323<programlisting> 324 journal_t* my_jnrl = journal_create(); 325 journal_init_{dev,inode}(jnrl,...) 326 if (clean) journal_wipe(); 327 journal_load(); 328 329 foreach(transaction) { /*transactions must be 330 completed before 331 a syscall returns to 332 userspace*/ 333 334 handle_t * xct=journal_start(my_jnrl); 335 foreach(bh) { 336 journal_get_{create,write,undo}_access(xact,bh); 337 if ( myfs_modify(bh) ) { /* returns true 338 if makes changes */ 339 journal_dirty_{meta,}data(xact,bh); 340 } else { 341 journal_forget(bh); 342 } 343 } 344 journal_stop(xct); 345 } 346 journal_destroy(my_jrnl); 347</programlisting> 348 </sect2> 349 350 </sect1> 351 352 <sect1 id="data_types"> 353 <title>Data Types</title> 354 <para> 355 The journalling layer uses typedefs to 'hide' the concrete definitions 356 of the structures used. As a client of the JBD layer you can 357 just rely on the using the pointer as a magic cookie of some sort. 358 359 Obviously the hiding is not enforced as this is 'C'. 360 </para> 361 <sect2 id="structures"><title>Structures</title> 362!Iinclude/linux/jbd.h 363 </sect2> 364 </sect1> 365 366 <sect1 id="functions"> 367 <title>Functions</title> 368 <para> 369 The functions here are split into two groups those that 370 affect a journal as a whole, and those which are used to 371 manage transactions 372 </para> 373 <sect2 id="journal_level"><title>Journal Level</title> 374!Efs/jbd/journal.c 375!Ifs/jbd/recovery.c 376 </sect2> 377 <sect2 id="transaction_level"><title>Transasction Level</title> 378!Efs/jbd/transaction.c 379 </sect2> 380 </sect1> 381 <sect1 id="see_also"> 382 <title>See also</title> 383 <para> 384 <citation> 385 <ulink url="ftp://ftp.uk.linux.org/pub/linux/sct/fs/jfs/journal-design.ps.gz"> 386 Journaling the Linux ext2fs Filesystem, LinuxExpo 98, Stephen Tweedie 387 </ulink> 388 </citation> 389 </para> 390 <para> 391 <citation> 392 <ulink url="http://olstrans.sourceforge.net/release/OLS2000-ext3/OLS2000-ext3.html"> 393 Ext3 Journalling FileSystem, OLS 2000, Dr. Stephen Tweedie 394 </ulink> 395 </citation> 396 </para> 397 </sect1> 398 399 </chapter> 400 401 <chapter id="splice"> 402 <title>splice API</title> 403 <para> 404 splice is a method for moving blocks of data around inside the 405 kernel, without continually transferring them between the kernel 406 and user space. 407 </para> 408!Ffs/splice.c 409 </chapter> 410 411 <chapter id="pipes"> 412 <title>pipes API</title> 413 <para> 414 Pipe interfaces are all for in-kernel (builtin image) use. 415 They are not exported for use by modules. 416 </para> 417!Iinclude/linux/pipe_fs_i.h 418!Ffs/pipe.c 419 </chapter> 420 421</book> 422