OESF Portables Forum
Everything Else => Zaurus Distro Support and Discussion => Distros, Development, and Model Specific Forums => Archived Forums => Angstrom & OpenZaurus => Topic started by: obergix on June 27, 2005, 02:33:33 am
-
I’m using rboot to boot an (GPE) image out of an ext2 file stored on my SD.
It works pretty well actually, until there is a problem, for instance if I have to reset...
I'm afraid the / image then becomes corrupt, and I can lose un-synced saved data.
I think that there may be a problem during the pivot-root which mounts the new / without sync, which is not recommended I think for ext2.
The SD is mounted with sync, so any change to the "looped" / image should be written ASAP, but are changed inside the image written ASAP ? If the new / is only mounted ext2 without sync (or appears at such when using mount from inside it), is it good ?
What do others think ?
Would ext3 be preferred ?
I'm also not really sure if at next reboot the uncleanly shutdown filesystem will be checked (fsck'd) before the pivo_root...
Thanks in advance for your comments.
Best regards,
-
I’m using rboot to boot an (GPE) image out of an ext2 file stored on my SD.
It works pretty well actually, until there is a problem, for instance if I have to reset...
I'm afraid the / image then becomes corrupt, and I can lose un-synced saved data.
I think that there may be a problem during the pivot-root which mounts the new / without sync, which is not recommended I think for ext2.
The SD is mounted with sync, so any change to the "looped" / image should be written ASAP,
Correct.
but are changed inside the image written ASAP ? If the new / is only mounted ext2 without sync (or appears at such when using mount from inside it), is it good ?
Nope, if the loop-image is not mounted "sync", data is not written ASAP.
What do others think ?
Running fsck.ext2 on the loop-image prior to mounting is a good idea.
Note: Hacking /sbin/init can be a bitch heh
Would ext3 be preferred ?
Um no, journaling filesystems are a bad idea on flash-based storage devices.
They cause way more writes to the disk as non-journaling filesystems.
I'm also not really sure if at next reboot the uncleanly shutdown filesystem will be checked (fsck'd) before the pivo_root...
The 3.5.3 version of rboot doesn't check the image or the card. It has been some time since I last looked at the script though. Things may have changed.