Skip to contentSkip to conversation

restoring hard linked files from zisofs/iso9660 w. RR

DISCUSSION Frank van Maarseveen 3 participants 6 messages Jan 25, 2002 – Jan 28, 2002

lore
Share & tools
lore
#0 lore raw

This doesn't seem to work due to inode number differences. Is this a fundamental problem or can it be solved somehow, e.g. by an attribute which refers to a sort of "original" inode number?

or by a more advanced inode number synthesis in fs/isofs?

Signature · 1 line
Frank
#1 lore raw
Quoted 8 lines · Frank van Maarseveen

This doesn't seem to work due to inode number differences. Is this a fundamental problem or can it be solved somehow, e.g. by an attribute which refers to a sort of "original" inode number?

or by a more advanced inode number synthesis in fs/isofs?

Don't know about isofs, but hardlinks do not use an inode. It's just another entry in a directory pointing to a "used" inode (there is another entry refering to the same inode)

fast symlinks do occupy an inode, symlinks an inode + 1 diskblock.

#2 lore raw
In reply to Frank van Maarseveen
Followup to:  <[email protected]>
By author:    Frank van Maarseveen <[email protected]>
In newsgroup: linux.dev.kernel
Quoted 6 lines

This doesn't seem to work due to inode number differences. Is this a fundamental problem or can it be solved somehow, e.g. by an attribute which refers to a sort of "original" inode number?

WHAT doesn't work?

Quoted 1 line

or by a more advanced inode number synthesis in fs/isofs?

There is, I belive, an inode number RR attribute. Last I checked I was happily using hard links with RockRidge...

-hpa

Signature · 3 lines
<[email protected]> at work, <[email protected]> in private!
"Unix gives you enough rope to shoot yourself in the foot."
http://www.zytor.com/~hpa/puzzle.txt	<[email protected]>
#3 lore raw
In reply to Frank van Maarseveen
Quoted 4 lines · H. Peter Anvin

WHAT doesn't work?
There is, I belive, an inode number RR attribute. Last I checked I was happily using hard links with RockRidge...

Try restoring a few hard linked files from such a CD. The links will break because the inodes of hard linked objects on CD do not have an identical inode number anymore.

Signature · 1 line
Frank
#4 lore raw
In reply to Frank van Maarseveen
Quoted 10 lines · Frank van Maarseveen
Quoted 3 lines · H. Peter Anvin

WHAT doesn't work?
There is, I belive, an inode number RR attribute. Last I checked I was happily using hard links with RockRidge...

Try restoring a few hard linked files from such a CD. The links will break because the inodes of hard linked objects on CD do not have an identical inode number anymore.

What do you mean with "restore" in this context... (dump/restore clearly don't apply.)

Please give a "I used the following mkisofs options to make the CD and then I see the following when I execute this command after mounting the CD image with the following mount options..."

-hpa

#5 lore raw
In reply to Frank van Maarseveen

Okay... I have successfully reproduced this bug. It's definitely a Linux kernel bug -- mkisofs correctly puts file numbers (inode numbers) in the PX Rock Ridge entries.

Unfortunately it is a pretty deep decision in the Linux isofs that the inode number is the offset of the directory entry; isofs pretty much needs to be rewritten to be dentry-centric rather than inode-centric on order to resolve this. Note that dentry-centric operation is the sane thing for something like isofs, so this wouldn't be a bad idea.

Unfortunately, that doesn't solve all problems: there is no structure to the inode numbers provided in the PX entries, and there is the potential for collision when used on a disc which is only partially RockRidge (which is legal.) This has to be taken into account, and all of these things have to be done without reading the whole disc first...

-hpa