Commit | Line | Data |
---|---|---|
fc513a33 DK |
1 | |
2 | Ext4 Filesystem | |
3 | =============== | |
4 | ||
22359f57 DC |
5 | Ext4 is an an advanced level of the ext3 filesystem which incorporates |
6 | scalability and reliability enhancements for supporting large filesystems | |
7 | (64 bit) in keeping with increasing disk capacities and state-of-the-art | |
8 | feature requirements. | |
fc513a33 | 9 | |
22359f57 DC |
10 | Mailing list: linux-ext4@vger.kernel.org |
11 | Web site: http://ext4.wiki.kernel.org | |
fc513a33 DK |
12 | |
13 | ||
14 | 1. Quick usage instructions: | |
15 | =========================== | |
16 | ||
22359f57 DC |
17 | Note: More extensive information for getting started with ext4 can be |
18 | found at the ext4 wiki site at the URL: | |
19 | http://ext4.wiki.kernel.org/index.php/Ext4_Howto | |
20 | ||
93e3270c | 21 | - Compile and install the latest version of e2fsprogs (as of this |
22359f57 | 22 | writing version 1.41.3) from: |
93e3270c JS |
23 | |
24 | http://sourceforge.net/project/showfiles.php?group_id=2406 | |
25 | ||
26 | or | |
27 | ||
fc513a33 DK |
28 | ftp://ftp.kernel.org/pub/linux/kernel/people/tytso/e2fsprogs/ |
29 | ||
93e3270c JS |
30 | or grab the latest git repository from: |
31 | ||
32 | git://git.kernel.org/pub/scm/fs/ext2/e2fsprogs.git | |
33 | ||
4537398d TT |
34 | - Note that it is highly important to install the mke2fs.conf file |
35 | that comes with the e2fsprogs 1.41.x sources in /etc/mke2fs.conf. If | |
36 | you have edited the /etc/mke2fs.conf file installed on your system, | |
37 | you will need to merge your changes with the version from e2fsprogs | |
38 | 1.41.x. | |
39 | ||
03010a33 | 40 | - Create a new filesystem using the ext4 filesystem type: |
93e3270c | 41 | |
03010a33 | 42 | # mke2fs -t ext4 /dev/hda1 |
93e3270c | 43 | |
22359f57 | 44 | Or to configure an existing ext3 filesystem to support extents: |
fc513a33 | 45 | |
22359f57 | 46 | # tune2fs -O extents /dev/hda1 |
fc513a33 | 47 | |
93e3270c JS |
48 | If the filesystem was created with 128 byte inodes, it can be |
49 | converted to use 256 byte for greater efficiency via: | |
fc513a33 | 50 | |
93e3270c | 51 | # tune2fs -I 256 /dev/hda1 |
fc513a33 | 52 | |
03010a33 | 53 | (Note: we currently do not have tools to convert an ext4 |
93e3270c JS |
54 | filesystem back to ext3; so please do not do try this on production |
55 | filesystems.) | |
fc513a33 | 56 | |
93e3270c JS |
57 | - Mounting: |
58 | ||
03010a33 | 59 | # mount -t ext4 /dev/hda1 /wherever |
fc513a33 | 60 | |
8e1a4857 TT |
61 | - When comparing performance with other filesystems, it's always |
62 | important to try multiple workloads; very often a subtle change in a | |
63 | workload parameter can completely change the ranking of which | |
64 | filesystems do well compared to others. When comparing versus ext3, | |
65 | note that ext4 enables write barriers by default, while ext3 does | |
66 | not enable write barriers by default. So it is useful to use | |
67 | explicitly specify whether barriers are enabled or not when via the | |
68 | '-o barriers=[0|1]' mount option for both ext3 and ext4 filesystems | |
69 | for a fair comparison. When tuning ext3 for best benchmark numbers, | |
70 | it is often worthwhile to try changing the data journaling mode; '-o | |
71 | data=writeback,nobh' can be faster for some workloads. (Note | |
72 | however that running mounted with data=writeback can potentially | |
73 | leave stale data exposed in recently written files in case of an | |
74 | unclean shutdown, which could be a security exposure in some | |
75 | situations.) Configuring the filesystem with a large journal can | |
76 | also be helpful for metadata-intensive workloads. | |
fc513a33 DK |
77 | |
78 | 2. Features | |
79 | =========== | |
80 | ||
81 | 2.1 Currently available | |
82 | ||
93e3270c | 83 | * ability to use filesystems > 16TB (e2fsprogs support not available yet) |
fc513a33 DK |
84 | * extent format reduces metadata overhead (RAM, IO for access, transactions) |
85 | * extent format more robust in face of on-disk corruption due to magics, | |
8e1a4857 | 86 | * internal redundancy in tree |
49f1487b | 87 | * improved file allocation (multi-block alloc) |
722bde68 | 88 | * lift 32000 subdirectory limit imposed by i_links_count[1] |
93e3270c JS |
89 | * nsec timestamps for mtime, atime, ctime, create time |
90 | * inode version field on disk (NFSv4, Lustre) | |
91 | * reduced e2fsck time via uninit_bg feature | |
92 | * journal checksumming for robustness, performance | |
93 | * persistent file preallocation (e.g for streaming media, databases) | |
94 | * ability to pack bitmaps and inode tables into larger virtual groups via the | |
95 | flex_bg feature | |
96 | * large file support | |
97 | * Inode allocation using large virtual block groups via flex_bg | |
49f1487b MC |
98 | * delayed allocation |
99 | * large block (up to pagesize) support | |
100 | * efficent new ordered mode in JBD2 and ext4(avoid using buffer head to force | |
101 | the ordering) | |
fc513a33 | 102 | |
722bde68 TT |
103 | [1] Filesystems with a block size of 1k may see a limit imposed by the |
104 | directory hash tree having a maximum depth of two. | |
105 | ||
fc513a33 DK |
106 | 2.2 Candidate features for future inclusion |
107 | ||
93e3270c JS |
108 | * Online defrag (patches available but not well tested) |
109 | * reduced mke2fs time via lazy itable initialization in conjuction with | |
110 | the uninit_bg feature (capability to do this is available in e2fsprogs | |
111 | but a kernel thread to do lazy zeroing of unused inode table blocks | |
112 | after filesystem is first mounted is required for safety) | |
fc513a33 | 113 | |
93e3270c JS |
114 | There are several others under discussion, whether they all make it in is |
115 | partly a function of how much time everyone has to work on them. Features like | |
116 | metadata checksumming have been discussed and planned for a bit but no patches | |
117 | exist yet so I'm not sure they're in the near-term roadmap. | |
fc513a33 | 118 | |
93e3270c JS |
119 | The big performance win will come with mballoc, delalloc and flex_bg |
120 | grouping of bitmaps and inode tables. Some test results available here: | |
fc513a33 | 121 | |
22359f57 DC |
122 | - http://www.bullopensource.org/ext4/20080818-ffsb/ffsb-write-2.6.27-rc1.html |
123 | - http://www.bullopensource.org/ext4/20080818-ffsb/ffsb-readwrite-2.6.27-rc1.html | |
fc513a33 DK |
124 | |
125 | 3. Options | |
126 | ========== | |
127 | ||
128 | When mounting an ext4 filesystem, the following option are accepted: | |
129 | (*) == default | |
130 | ||
8e1a4857 TT |
131 | ro Mount filesystem read only. Note that ext4 will |
132 | replay the journal (and thus write to the | |
133 | partition) even when mounted "read only". The | |
134 | mount options "ro,noload" can be used to prevent | |
135 | writes to the filesystem. | |
136 | ||
818d276c GS |
137 | journal_async_commit Commit block can be written to disk without waiting |
138 | for descriptor blocks. If enabled older kernels cannot | |
d0646f7b | 139 | mount the device. |
818d276c | 140 | |
fc513a33 DK |
141 | journal=update Update the ext4 file system's journal to the current |
142 | format. | |
143 | ||
fc513a33 DK |
144 | journal_dev=devnum When the external journal device's major/minor numbers |
145 | have changed, this option allows the user to specify | |
146 | the new journal location. The journal device is | |
147 | identified through its new major/minor numbers encoded | |
148 | in devnum. | |
149 | ||
8e1a4857 TT |
150 | noload Don't load the journal on mounting. Note that |
151 | if the filesystem was not unmounted cleanly, | |
152 | skipping the journal replay will lead to the | |
153 | filesystem containing inconsistencies that can | |
154 | lead to any number of problems. | |
fc513a33 DK |
155 | |
156 | data=journal All data are committed into the journal prior to being | |
157 | written into the main file system. | |
158 | ||
159 | data=ordered (*) All data are forced directly out to the main file | |
160 | system prior to its metadata being committed to the | |
161 | journal. | |
162 | ||
163 | data=writeback Data ordering is not preserved, data may be written | |
164 | into the main file system after its metadata has been | |
165 | committed to the journal. | |
166 | ||
167 | commit=nrsec (*) Ext4 can be told to sync all its data and metadata | |
168 | every 'nrsec' seconds. The default value is 5 seconds. | |
169 | This means that if you lose your power, you will lose | |
170 | as much as the latest 5 seconds of work (your | |
171 | filesystem will not be damaged though, thanks to the | |
172 | journaling). This default value (or any low value) | |
173 | will hurt performance, but it's good for data-safety. | |
174 | Setting it to 0 will have the same effect as leaving | |
175 | it at the default (5 seconds). | |
176 | Setting it to very large values will improve | |
177 | performance. | |
178 | ||
571640ca | 179 | barrier=<0|1(*)> This enables/disables the use of write barriers in |
06705bff TT |
180 | barrier(*) the jbd code. barrier=0 disables, barrier=1 enables. |
181 | nobarrier This also requires an IO stack which can support | |
571640ca ES |
182 | barriers, and if jbd gets an error on a barrier |
183 | write, it will disable again with a warning. | |
184 | Write barriers enforce proper on-disk ordering | |
185 | of journal commits, making volatile disk write caches | |
186 | safe to use, at some performance penalty. If | |
187 | your disks are battery-backed in one way or another, | |
188 | disabling barriers may safely improve performance. | |
06705bff TT |
189 | The mount options "barrier" and "nobarrier" can |
190 | also be used to enable or disable barriers, for | |
191 | consistency with other ext4 mount options. | |
fc513a33 | 192 | |
240799cd TT |
193 | inode_readahead=n This tuning parameter controls the maximum |
194 | number of inode table blocks that ext4's inode | |
195 | table readahead algorithm will pre-read into | |
196 | the buffer cache. The default value is 32 blocks. | |
197 | ||
fc513a33 DK |
198 | orlov (*) This enables the new Orlov block allocator. It is |
199 | enabled by default. | |
200 | ||
201 | oldalloc This disables the Orlov block allocator and enables | |
202 | the old block allocator. Orlov should have better | |
203 | performance - we'd like to get some feedback if it's | |
204 | the contrary for you. | |
205 | ||
206 | user_xattr Enables Extended User Attributes. Additionally, you | |
207 | need to have extended attribute support enabled in the | |
208 | kernel configuration (CONFIG_EXT4_FS_XATTR). See the | |
209 | attr(5) manual page and http://acl.bestbits.at/ to | |
210 | learn more about extended attributes. | |
211 | ||
212 | nouser_xattr Disables Extended User Attributes. | |
213 | ||
214 | acl Enables POSIX Access Control Lists support. | |
215 | Additionally, you need to have ACL support enabled in | |
216 | the kernel configuration (CONFIG_EXT4_FS_POSIX_ACL). | |
217 | See the acl(5) manual page and http://acl.bestbits.at/ | |
218 | for more information. | |
219 | ||
220 | noacl This option disables POSIX Access Control List | |
221 | support. | |
222 | ||
223 | reservation | |
224 | ||
225 | noreservation | |
226 | ||
227 | bsddf (*) Make 'df' act like BSD. | |
228 | minixdf Make 'df' act like Minix. | |
229 | ||
fc513a33 DK |
230 | debug Extra debugging information is sent to syslog. |
231 | ||
8a8a2050 TT |
232 | abort Simulate the effects of calling ext4_abort() for |
233 | debugging purposes. This is normally used while | |
234 | remounting a filesystem which is already mounted. | |
235 | ||
8e1a4857 | 236 | errors=remount-ro Remount the filesystem read-only on an error. |
fc513a33 DK |
237 | errors=continue Keep going on a filesystem error. |
238 | errors=panic Panic and halt the machine if an error occurs. | |
8e1a4857 TT |
239 | (These mount options override the errors behavior |
240 | specified in the superblock, which can be configured | |
241 | using tune2fs) | |
fc513a33 | 242 | |
5bf5683a HK |
243 | data_err=ignore(*) Just print an error message if an error occurs |
244 | in a file data buffer in ordered mode. | |
245 | data_err=abort Abort the journal if an error occurs in a file | |
246 | data buffer in ordered mode. | |
247 | ||
fc513a33 DK |
248 | grpid Give objects the same group ID as their creator. |
249 | bsdgroups | |
250 | ||
251 | nogrpid (*) New objects have the group ID of their creator. | |
252 | sysvgroups | |
253 | ||
254 | resgid=n The group ID which may use the reserved blocks. | |
255 | ||
256 | resuid=n The user ID which may use the reserved blocks. | |
257 | ||
258 | sb=n Use alternate superblock at this location. | |
259 | ||
1358870d JK |
260 | quota These options are ignored by the filesystem. They |
261 | noquota are used only by quota tools to recognize volumes | |
262 | grpquota where quota should be turned on. See documentation | |
263 | usrquota in the quota-tools package for more details | |
264 | (http://sourceforge.net/projects/linuxquota). | |
265 | ||
266 | jqfmt=<quota type> These options tell filesystem details about quota | |
267 | usrjquota=<file> so that quota information can be properly updated | |
268 | grpjquota=<file> during journal replay. They replace the above | |
269 | quota options. See documentation in the quota-tools | |
270 | package for more details | |
271 | (http://sourceforge.net/projects/linuxquota). | |
fc513a33 DK |
272 | |
273 | bh (*) ext4 associates buffer heads to data pages to | |
274 | nobh (a) cache disk block mapping information | |
275 | (b) link pages into transaction to provide | |
276 | ordering guarantees. | |
277 | "bh" option forces use of buffer heads. | |
278 | "nobh" option tries to avoid associating buffer | |
279 | heads (supported only for "writeback" mode). | |
280 | ||
c9de560d AT |
281 | stripe=n Number of filesystem blocks that mballoc will try |
282 | to use for allocation size and alignment. For RAID5/6 | |
283 | systems this should be the number of data | |
284 | disks * RAID chunk size in file system blocks. | |
83653888 JK |
285 | |
286 | delalloc (*) Defer block allocation until just before ext4 | |
287 | writes out the block(s) in question. This | |
288 | allows ext4 to better allocation decisions | |
289 | more efficiently. | |
290 | nodelalloc Disable delayed allocation. Blocks are allocated | |
291 | when the data is copied from userspace to the | |
292 | page cache, either via the write(2) system call | |
293 | or when an mmap'ed page which was previously | |
294 | unallocated is written for the first time. | |
240799cd | 295 | |
30773840 TT |
296 | max_batch_time=usec Maximum amount of time ext4 should wait for |
297 | additional filesystem operations to be batch | |
298 | together with a synchronous write operation. | |
299 | Since a synchronous write operation is going to | |
300 | force a commit and then a wait for the I/O | |
301 | complete, it doesn't cost much, and can be a | |
302 | huge throughput win, we wait for a small amount | |
303 | of time to see if any other transactions can | |
304 | piggyback on the synchronous write. The | |
305 | algorithm used is designed to automatically tune | |
306 | for the speed of the disk, by measuring the | |
307 | amount of time (on average) that it takes to | |
308 | finish committing a transaction. Call this time | |
309 | the "commit time". If the time that the | |
19f59460 | 310 | transaction has been running is less than the |
30773840 TT |
311 | commit time, ext4 will try sleeping for the |
312 | commit time to see if other operations will join | |
313 | the transaction. The commit time is capped by | |
314 | the max_batch_time, which defaults to 15000us | |
315 | (15ms). This optimization can be turned off | |
316 | entirely by setting max_batch_time to 0. | |
317 | ||
318 | min_batch_time=usec This parameter sets the commit time (as | |
319 | described above) to be at least min_batch_time. | |
320 | It defaults to zero microseconds. Increasing | |
321 | this parameter may improve the throughput of | |
322 | multi-threaded, synchronous workloads on very | |
323 | fast disks, at the cost of increasing latency. | |
324 | ||
b3881f74 TT |
325 | journal_ioprio=prio The I/O priority (from 0 to 7, where 0 is the |
326 | highest priorty) which should be used for I/O | |
327 | operations submitted by kjournald2 during a | |
328 | commit operation. This defaults to 3, which is | |
329 | a slightly higher priority than the default I/O | |
330 | priority. | |
331 | ||
06705bff TT |
332 | auto_da_alloc(*) Many broken applications don't use fsync() when |
333 | noauto_da_alloc replacing existing files via patterns such as | |
334 | fd = open("foo.new")/write(fd,..)/close(fd)/ | |
335 | rename("foo.new", "foo"), or worse yet, | |
336 | fd = open("foo", O_TRUNC)/write(fd,..)/close(fd). | |
337 | If auto_da_alloc is enabled, ext4 will detect | |
338 | the replace-via-rename and replace-via-truncate | |
339 | patterns and force that any delayed allocation | |
340 | blocks are allocated such that at the next | |
341 | journal commit, in the default data=ordered | |
342 | mode, the data blocks of the new file are forced | |
343 | to disk before the rename() operation is | |
19f59460 | 344 | committed. This provides roughly the same level |
06705bff TT |
345 | of guarantees as ext3, and avoids the |
346 | "zero-length" problem that can happen when a | |
347 | system crashes before the delayed allocation | |
348 | blocks are forced to disk. | |
349 | ||
fc513a33 | 350 | Data Mode |
93e3270c | 351 | ========= |
fc513a33 DK |
352 | There are 3 different data modes: |
353 | ||
354 | * writeback mode | |
355 | In data=writeback mode, ext4 does not journal data at all. This mode provides | |
356 | a similar level of journaling as that of XFS, JFS, and ReiserFS in its default | |
357 | mode - metadata journaling. A crash+recovery can cause incorrect data to | |
358 | appear in files which were written shortly before the crash. This mode will | |
359 | typically provide the best ext4 performance. | |
360 | ||
361 | * ordered mode | |
362 | In data=ordered mode, ext4 only officially journals metadata, but it logically | |
49f1487b MC |
363 | groups metadata information related to data changes with the data blocks into a |
364 | single unit called a transaction. When it's time to write the new metadata | |
365 | out to disk, the associated data blocks are written first. In general, | |
366 | this mode performs slightly slower than writeback but significantly faster than journal mode. | |
fc513a33 DK |
367 | |
368 | * journal mode | |
369 | data=journal mode provides full data and metadata journaling. All new data is | |
370 | written to the journal first, and then to its final location. | |
371 | In the event of a crash, the journal can be replayed, bringing both data and | |
372 | metadata into a consistent state. This mode is the slowest except when data | |
373 | needs to be read from and written to disk at the same time where it | |
19f59460 | 374 | outperforms all others modes. Currently ext4 does not have delayed |
49f1487b | 375 | allocation support if this data journalling mode is selected. |
fc513a33 DK |
376 | |
377 | References | |
378 | ========== | |
379 | ||
380 | kernel source: <file:fs/ext4/> | |
381 | <file:fs/jbd2/> | |
382 | ||
383 | programs: http://e2fsprogs.sourceforge.net/ | |
fc513a33 DK |
384 | |
385 | useful links: http://fedoraproject.org/wiki/ext3-devel | |
386 | http://www.bullopensource.org/ext4/ | |
93e3270c JS |
387 | http://ext4.wiki.kernel.org/index.php/Main_Page |
388 | http://fedoraproject.org/wiki/Features/Ext4 |