1 |
On Tue, Apr 28, 2009 at 6:39 PM, Duncan <1i5t5.duncan@×××.net> wrote: |
2 |
> Nikos Chantziaras <realnc@×××××.de> posted gt7r28$cql$1@×××××××××.org, |
3 |
> excerpted below, on Wed, 29 Apr 2009 00:05:33 +0300: |
4 |
> |
5 |
>> I have "defaults,noatime" and yet mount reports: |
6 |
>> |
7 |
>> /dev/root on / type ext4 (rw,noatime,barrier=1,data=ordered) |
8 |
>> |
9 |
>> Where does "rw,barrier=1,data=ordered" come from if not from "defaults"? |
10 |
> |
11 |
> Those (but for rw which is the normal default) are defaults of either |
12 |
> the kernel or the filesystem itself, as set at mkfs time or with tune2fs. |
13 |
> |
14 |
> There's really quite a hierarchy of options and fall-thru defaults: |
15 |
> |
16 |
> Anything set on the mount commandline takes priority (well, after |
17 |
> physical properties such as read-only for CDROM and floppies with write- |
18 |
> protect set), |
19 |
> |
20 |
> followed by what's in fstab or the hal config, |
21 |
> |
22 |
> followed by options set for that particular filesystem at mkfs time or |
23 |
> with the filesystem tuner (tun2fs, reiserfstune, whatever), |
24 |
> |
25 |
> followed by kernel defaults specific to that type of fs (ext3 defaulted |
26 |
> to data=ordered but that's set to change to data=writeback in 2.6.30, |
27 |
> after some fixes they've done but it's still rather controversial), |
28 |
> |
29 |
> followed by mount's normal defaults. |
30 |
|
31 |
Seems root mounts get special treatment as well. Compare my / & /opt |
32 |
- both ext4, both created identically & mounted with just noatime: |
33 |
|
34 |
/dev/vg/root on / type ext4 (rw,noatime,barrier=1,data=ordered) |
35 |
/dev/mapper/vg-opt on /opt type ext4 (rw,noatime) |
36 |
|
37 |
Or perhaps its just more verbose since like Duncan said data=ordered |
38 |
is the default & barriers are currently disabled on lvm. |
39 |
|
40 |
Wil |