Commit | Line | Data |
---|---|---|
fc513a33 DK |
1 | |
2 | Ext4 Filesystem | |
3 | =============== | |
4 | ||
5 | This is a development version of the ext4 filesystem, an advanced level | |
6 | of the ext3 filesystem which incorporates scalability and reliability | |
7 | enhancements for supporting large filesystems (64 bit) in keeping with | |
8 | increasing disk capacities and state-of-the-art feature requirements. | |
9 | ||
10 | Mailing list: linux-ext4@vger.kernel.org | |
11 | ||
12 | ||
13 | 1. Quick usage instructions: | |
14 | =========================== | |
15 | ||
93e3270c JS |
16 | - Compile and install the latest version of e2fsprogs (as of this |
17 | writing version 1.41) from: | |
18 | ||
19 | http://sourceforge.net/project/showfiles.php?group_id=2406 | |
20 | ||
21 | or | |
22 | ||
fc513a33 DK |
23 | ftp://ftp.kernel.org/pub/linux/kernel/people/tytso/e2fsprogs/ |
24 | ||
93e3270c JS |
25 | or grab the latest git repository from: |
26 | ||
27 | git://git.kernel.org/pub/scm/fs/ext2/e2fsprogs.git | |
28 | ||
29 | - Create a new filesystem using the ext4dev filesystem type: | |
30 | ||
31 | # mke2fs -t ext4dev /dev/hda1 | |
32 | ||
33 | Or configure an existing ext3 filesystem to support extents and set | |
34 | the test_fs flag to indicate that it's ok for an in-development | |
35 | filesystem to touch this filesystem: | |
fc513a33 | 36 | |
93e3270c | 37 | # tune2fs -O extents -E test_fs /dev/hda1 |
fc513a33 | 38 | |
93e3270c JS |
39 | If the filesystem was created with 128 byte inodes, it can be |
40 | converted to use 256 byte for greater efficiency via: | |
fc513a33 | 41 | |
93e3270c | 42 | # tune2fs -I 256 /dev/hda1 |
fc513a33 | 43 | |
93e3270c JS |
44 | (Note: we currently do not have tools to convert an ext4dev |
45 | filesystem back to ext3; so please do not do try this on production | |
46 | filesystems.) | |
fc513a33 | 47 | |
93e3270c JS |
48 | - Mounting: |
49 | ||
50 | # mount -t ext4dev /dev/hda1 /wherever | |
fc513a33 DK |
51 | |
52 | - When comparing performance with other filesystems, remember that | |
93e3270c JS |
53 | ext3/4 by default offers higher data integrity guarantees than most. |
54 | So when comparing with a metadata-only journalling filesystem, such | |
55 | as ext3, use `mount -o data=writeback'. And you might as well use | |
56 | `mount -o nobh' too along with it. Making the journal larger than | |
57 | the mke2fs default often helps performance with metadata-intensive | |
58 | workloads. | |
fc513a33 DK |
59 | |
60 | 2. Features | |
61 | =========== | |
62 | ||
63 | 2.1 Currently available | |
64 | ||
93e3270c | 65 | * ability to use filesystems > 16TB (e2fsprogs support not available yet) |
fc513a33 DK |
66 | * extent format reduces metadata overhead (RAM, IO for access, transactions) |
67 | * extent format more robust in face of on-disk corruption due to magics, | |
68 | * internal redunancy in tree | |
93e3270c JS |
69 | * improved file allocation (multi-block alloc, delayed alloc) |
70 | * fix 32000 subdirectory limit | |
71 | * nsec timestamps for mtime, atime, ctime, create time | |
72 | * inode version field on disk (NFSv4, Lustre) | |
73 | * reduced e2fsck time via uninit_bg feature | |
74 | * journal checksumming for robustness, performance | |
75 | * persistent file preallocation (e.g for streaming media, databases) | |
76 | * ability to pack bitmaps and inode tables into larger virtual groups via the | |
77 | flex_bg feature | |
78 | * large file support | |
79 | * Inode allocation using large virtual block groups via flex_bg | |
fc513a33 DK |
80 | |
81 | 2.2 Candidate features for future inclusion | |
82 | ||
93e3270c JS |
83 | * Online defrag (patches available but not well tested) |
84 | * reduced mke2fs time via lazy itable initialization in conjuction with | |
85 | the uninit_bg feature (capability to do this is available in e2fsprogs | |
86 | but a kernel thread to do lazy zeroing of unused inode table blocks | |
87 | after filesystem is first mounted is required for safety) | |
fc513a33 | 88 | |
93e3270c JS |
89 | There are several others under discussion, whether they all make it in is |
90 | partly a function of how much time everyone has to work on them. Features like | |
91 | metadata checksumming have been discussed and planned for a bit but no patches | |
92 | exist yet so I'm not sure they're in the near-term roadmap. | |
fc513a33 | 93 | |
93e3270c JS |
94 | The big performance win will come with mballoc, delalloc and flex_bg |
95 | grouping of bitmaps and inode tables. Some test results available here: | |
fc513a33 | 96 | |
93e3270c JS |
97 | - http://www.bullopensource.org/ext4/20080530/ffsb-write-2.6.26-rc2.html |
98 | - http://www.bullopensource.org/ext4/20080530/ffsb-readwrite-2.6.26-rc2.html | |
fc513a33 DK |
99 | |
100 | 3. Options | |
101 | ========== | |
102 | ||
103 | When mounting an ext4 filesystem, the following option are accepted: | |
104 | (*) == default | |
105 | ||
c9de560d | 106 | extents (*) ext4 will use extents to address file data. The |
fc513a33 DK |
107 | file system will no longer be mountable by ext3. |
108 | ||
c9de560d AT |
109 | noextents ext4 will not use extents for newly created files |
110 | ||
818d276c GS |
111 | journal_checksum Enable checksumming of the journal transactions. |
112 | This will allow the recovery code in e2fsck and the | |
113 | kernel to detect corruption in the kernel. It is a | |
114 | compatible change and will be ignored by older kernels. | |
115 | ||
116 | journal_async_commit Commit block can be written to disk without waiting | |
117 | for descriptor blocks. If enabled older kernels cannot | |
118 | mount the device. This will enable 'journal_checksum' | |
119 | internally. | |
120 | ||
fc513a33 DK |
121 | journal=update Update the ext4 file system's journal to the current |
122 | format. | |
123 | ||
124 | journal=inum When a journal already exists, this option is ignored. | |
125 | Otherwise, it specifies the number of the inode which | |
126 | will represent the ext4 file system's journal file. | |
127 | ||
128 | journal_dev=devnum When the external journal device's major/minor numbers | |
129 | have changed, this option allows the user to specify | |
130 | the new journal location. The journal device is | |
131 | identified through its new major/minor numbers encoded | |
132 | in devnum. | |
133 | ||
134 | noload Don't load the journal on mounting. | |
135 | ||
136 | data=journal All data are committed into the journal prior to being | |
137 | written into the main file system. | |
138 | ||
139 | data=ordered (*) All data are forced directly out to the main file | |
140 | system prior to its metadata being committed to the | |
141 | journal. | |
142 | ||
143 | data=writeback Data ordering is not preserved, data may be written | |
144 | into the main file system after its metadata has been | |
145 | committed to the journal. | |
146 | ||
147 | commit=nrsec (*) Ext4 can be told to sync all its data and metadata | |
148 | every 'nrsec' seconds. The default value is 5 seconds. | |
149 | This means that if you lose your power, you will lose | |
150 | as much as the latest 5 seconds of work (your | |
151 | filesystem will not be damaged though, thanks to the | |
152 | journaling). This default value (or any low value) | |
153 | will hurt performance, but it's good for data-safety. | |
154 | Setting it to 0 will have the same effect as leaving | |
155 | it at the default (5 seconds). | |
156 | Setting it to very large values will improve | |
157 | performance. | |
158 | ||
571640ca ES |
159 | barrier=<0|1(*)> This enables/disables the use of write barriers in |
160 | the jbd code. barrier=0 disables, barrier=1 enables. | |
161 | This also requires an IO stack which can support | |
162 | barriers, and if jbd gets an error on a barrier | |
163 | write, it will disable again with a warning. | |
164 | Write barriers enforce proper on-disk ordering | |
165 | of journal commits, making volatile disk write caches | |
166 | safe to use, at some performance penalty. If | |
167 | your disks are battery-backed in one way or another, | |
168 | disabling barriers may safely improve performance. | |
fc513a33 DK |
169 | |
170 | orlov (*) This enables the new Orlov block allocator. It is | |
171 | enabled by default. | |
172 | ||
173 | oldalloc This disables the Orlov block allocator and enables | |
174 | the old block allocator. Orlov should have better | |
175 | performance - we'd like to get some feedback if it's | |
176 | the contrary for you. | |
177 | ||
178 | user_xattr Enables Extended User Attributes. Additionally, you | |
179 | need to have extended attribute support enabled in the | |
180 | kernel configuration (CONFIG_EXT4_FS_XATTR). See the | |
181 | attr(5) manual page and http://acl.bestbits.at/ to | |
182 | learn more about extended attributes. | |
183 | ||
184 | nouser_xattr Disables Extended User Attributes. | |
185 | ||
186 | acl Enables POSIX Access Control Lists support. | |
187 | Additionally, you need to have ACL support enabled in | |
188 | the kernel configuration (CONFIG_EXT4_FS_POSIX_ACL). | |
189 | See the acl(5) manual page and http://acl.bestbits.at/ | |
190 | for more information. | |
191 | ||
192 | noacl This option disables POSIX Access Control List | |
193 | support. | |
194 | ||
195 | reservation | |
196 | ||
197 | noreservation | |
198 | ||
199 | bsddf (*) Make 'df' act like BSD. | |
200 | minixdf Make 'df' act like Minix. | |
201 | ||
202 | check=none Don't do extra checking of bitmaps on mount. | |
203 | nocheck | |
204 | ||
205 | debug Extra debugging information is sent to syslog. | |
206 | ||
207 | errors=remount-ro(*) Remount the filesystem read-only on an error. | |
208 | errors=continue Keep going on a filesystem error. | |
209 | errors=panic Panic and halt the machine if an error occurs. | |
210 | ||
211 | grpid Give objects the same group ID as their creator. | |
212 | bsdgroups | |
213 | ||
214 | nogrpid (*) New objects have the group ID of their creator. | |
215 | sysvgroups | |
216 | ||
217 | resgid=n The group ID which may use the reserved blocks. | |
218 | ||
219 | resuid=n The user ID which may use the reserved blocks. | |
220 | ||
221 | sb=n Use alternate superblock at this location. | |
222 | ||
223 | quota | |
224 | noquota | |
225 | grpquota | |
226 | usrquota | |
227 | ||
228 | bh (*) ext4 associates buffer heads to data pages to | |
229 | nobh (a) cache disk block mapping information | |
230 | (b) link pages into transaction to provide | |
231 | ordering guarantees. | |
232 | "bh" option forces use of buffer heads. | |
233 | "nobh" option tries to avoid associating buffer | |
234 | heads (supported only for "writeback" mode). | |
235 | ||
c9de560d AT |
236 | mballoc (*) Use the multiple block allocator for block allocation |
237 | nomballoc disabled multiple block allocator for block allocation. | |
238 | stripe=n Number of filesystem blocks that mballoc will try | |
239 | to use for allocation size and alignment. For RAID5/6 | |
240 | systems this should be the number of data | |
241 | disks * RAID chunk size in file system blocks. | |
fc513a33 DK |
242 | |
243 | Data Mode | |
93e3270c | 244 | ========= |
fc513a33 DK |
245 | There are 3 different data modes: |
246 | ||
247 | * writeback mode | |
248 | In data=writeback mode, ext4 does not journal data at all. This mode provides | |
249 | a similar level of journaling as that of XFS, JFS, and ReiserFS in its default | |
250 | mode - metadata journaling. A crash+recovery can cause incorrect data to | |
251 | appear in files which were written shortly before the crash. This mode will | |
252 | typically provide the best ext4 performance. | |
253 | ||
254 | * ordered mode | |
255 | In data=ordered mode, ext4 only officially journals metadata, but it logically | |
256 | groups metadata and data blocks into a single unit called a transaction. When | |
257 | it's time to write the new metadata out to disk, the associated data blocks | |
258 | are written first. In general, this mode performs slightly slower than | |
259 | writeback but significantly faster than journal mode. | |
260 | ||
261 | * journal mode | |
262 | data=journal mode provides full data and metadata journaling. All new data is | |
263 | written to the journal first, and then to its final location. | |
264 | In the event of a crash, the journal can be replayed, bringing both data and | |
265 | metadata into a consistent state. This mode is the slowest except when data | |
266 | needs to be read from and written to disk at the same time where it | |
267 | outperforms all others modes. | |
268 | ||
269 | References | |
270 | ========== | |
271 | ||
272 | kernel source: <file:fs/ext4/> | |
273 | <file:fs/jbd2/> | |
274 | ||
275 | programs: http://e2fsprogs.sourceforge.net/ | |
fc513a33 DK |
276 | |
277 | useful links: http://fedoraproject.org/wiki/ext3-devel | |
278 | http://www.bullopensource.org/ext4/ | |
93e3270c JS |
279 | http://ext4.wiki.kernel.org/index.php/Main_Page |
280 | http://fedoraproject.org/wiki/Features/Ext4 |