Bug 55495
Summary: | Can't (re)mount ext3 with data=writeback or journal | ||
---|---|---|---|
Product: | [Retired] Red Hat Linux | Reporter: | fred-m |
Component: | kernel | Assignee: | Stephen Tweedie <sct> |
Status: | CLOSED WONTFIX | QA Contact: | Brian Brock <bbrock> |
Severity: | medium | Docs Contact: | |
Priority: | medium | ||
Version: | 7.2 | CC: | dark.lite, jn, sct |
Target Milestone: | --- | Keywords: | FutureFeature |
Target Release: | --- | ||
Hardware: | i386 | ||
OS: | Linux | ||
Whiteboard: | |||
Fixed In Version: | Doc Type: | Enhancement | |
Doc Text: | Story Points: | --- | |
Clone Of: | Environment: | ||
Last Closed: | 2003-07-08 13:21:16 UTC | Type: | --- |
Regression: | --- | Mount Type: | --- |
Documentation: | --- | CRM: | |
Verified Versions: | Category: | --- | |
oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
Cloudforms Team: | --- | Target Upstream Version: | |
Embargoed: |
Description
fred-m
2001-11-01 10:26:05 UTC
Reproduced on my 7.2 machine. [taw@trixie taw]$ uname -a Linux trixie.meridian.redhat.com 2.4.9-13smp #1 SMP Tue Oct 30 19:57:16 EST 2001 i686 unknown I simply added data=journal to my /etc/fstab (for the / partition) and got simular results. And therefore boot process fails. Right --- changing the data mode after mount is not a supported function. All sorts of things can go wrong if you change journal semantics live while the filesystem is running. Having said that, there _is_ locking inside ext3 which would help to achieve this: we'd have to suspend the entire filesystem, sync everything to disk, flush the cache, flush the journal, and _then_ change the mode, but it should be possible. I'll re-categorise this as a RFE (request for enhancement), but it's not an ext3 bug. First of all, thanks!!! This explains everything. IMHO, two things need to be done before we touch ext3 itself: - Improving the documentation ("mount" manual, Red Hat web site, etc.), to say that "changing the data mode after mount is not (yet) a supported function". - Include a selection of the mode in the Red Hat installer (on "expert" mode *only*), in case ext3 is not enchanced until the next release. In response to the comment by sct on 12/17: in the general case, all the stuff you mentioned (sync & flush) would need to take place. But in the specific case of remounting / during boot-up, it doesn't really apply because / was mounted read-only! Even the default mode of "data=ordered" isn't necessary because nothing will be written to the disk. Clever observation, I must say, and it's too bad I can't take credit for it myself: I found in on the vger.kernel.org mail list. There is a secondary problem in that after the remount fails, the other file systems are mounted normally but "mount" will not show them as mounted (mount-2.11g). For the reader who finds themself unable to boot because of a non-default mount option for their root file system, yet somehow able to read this bug report, you can get out of your predicament this way: 1. Reboot linux in "single" mode. 2. Remount / with: mount -t ext3 -o defaults,remount,rw /dev/hda5 / (change /dev/hda5 as appropriate). If you don't spell out the options, some versions of mount try to get them from /etc/fstab. 3. Edit /etc/fstab to remove "data=journal" for / 4. Reboot. I have all of my file systems as ext3, and / is the only one I have this trouble with. /boot is OK as data=journal. |