On 5/19/15 10:46 AM, Reindl Harald wrote:
>
> Am 19.05.2015 um 17:25 schrieb Eric Sandeen:
>> On 5/19/15 4:14 AM, Reindl Harald wrote:
>>> i switched from "noatime,nodiratime" to
>>> "relatime,nodiratime,lazytime" which seems to lead in unexpected
>>> mtime changes for a lot of files leading in rkhunter alerts and
>>> likely rsync problems
>>
>> The cargo-cult tuning, it burns! ;)
>>
>> noatime implies nodiratime, no need to specify both:
>
> and "relatime" implies it too?
Nope, relatime is relatime. ;)
The no[dir]atime options disable atime updates completely; relatime
acts as documented in mount(8), opportunistically updating atime
when it's more efficient to do so, and skipping it other times.
>> touch_atime()
>> {
>> ...
>> if (mnt->mnt_flags & MNT_NOATIME)
>> return;
>> if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode))
>> return;
>
> different story
I simply mean that if you have set noatime, you never even get to the
nodiratime test; hence noatime,nodiratime is redundant.
>> and relatime is default, so no need to re-specify:
>>
>> /* Default to relatime unless overriden */
>> if (!(flags & MS_NOATIME))
>> mnt_flags |= MNT_RELATIME;
>>
>> Anyway ... you didn't say what kernel you were running under, or what
>> filesystem.
>
> surely you just stripped that part from your reply
>
> May 15 13:35:51 Installed: kernel-core-4.0.3-201.fc21.x86_64
> May 15 13:35:54 Installed: kernel-modules-4.0.3-201.fc21.x86_64
> May 15 13:35:55 Updated: kernel-headers-4.0.3-201.fc21.x86_64
You were installing that kernel, but I didn't know what kernel was
running when you installed those packages. :)
>> If it's ext4, there have been some flaws there; recently:
>
> yes, "lazytime" implies ext4
Oh, I guess so, I forgot this was another ext4-only special hack,
sorry.
>>> commit 8f4d855839179f410fa910a26eb81d646d628f26
>>> Author: Theodore Ts'o <tytso at mit.edu>
>>> Date: Thu May 14 18:19:01 2015 -0400
>>>
>>> ext4: fix lazytime optimization
>>>
>>> We had a fencepost error in the lazytime optimization which means that
>>> timestamp would get written to the wrong inode.
>>>
>>> Cc: stable at vger.kernel.org
>>> Signed-off-by: Theodore Ts'o <tytso at mit.edu>
>>
>> which looks like it's probably the culprit. This was broken since:
>>
>>> commit a26f49926da938f47561f386be56a83dd37a496d
>>> Author: Theodore Ts'o <tytso at mit.edu>
>>> Date: Mon Feb 2 00:37:02 2015 -0500
>>>
>>> ext4: add optimization for the lazytime mount option
>
> wow - and nobody noticed all the months that mtimes are randomly
> changing all over the system while it takes exactly one day to get
> rkhunter mails and rsync listing a ton of untouched files
Apparently not. :)
-Eric