Bug 3533

Summary:	HFSPLUS does not take care of uid/gid options
Product:	File System	Reporter:	Carlos Perell (carlos)
Component:	HFS/HFSPLUS	Assignee:	Roman Zippel (zippel)
Status:	REJECTED DOCUMENTED
Severity:	normal	CC:	bunk, protasnb, spkane00
Priority:	P2
Hardware:	i386
OS:	Linux
Kernel Version:	2.6.8.1	Subsystem:
Regression:	---	Bisected commit-id:
Attachments:	Patch to fix the bug Second try, this time following Apple specification fix broken logic for mount uid=xxx,gid=xxx

Description Carlos Perell 2004-10-09 06:08:43 UTC

Distribution: Ubuntu
Hardware Environment: Powerbook G4 550Mhz
Software Environment: 
Problem Description: When you mount a hfsplus filesystems using the options
uid/gid, those options are ignored and the filesystem does not use them.

Steps to reproduce:
1.- plug a device with a hfsplus filesystem
2.- mount it using the options "uid=1000,gid=1000" (or any other number you want)
3.- execute a ls -l in the root directory where it's mounted and you will see
that it's not following those uid and gid numbers.


I did a patch and also tested it and works perfectly here with my iPod.

Comment 1 Carlos Perell 2004-10-09 06:10:58 UTC

Created attachment 3790 [details]
Patch to fix the bug

Comment 2 Carlos Perell 2004-10-09 08:58:01 UTC

I'm working on a new patch, this one is not correct.

The new one will follow the Apple specification at:

http://developer.apple.com/technotes/tn/tn1150.html#HFSPlusPermissions

Comment 3 Carlos Perell 2004-10-10 10:40:10 UTC

Created attachment 3798 [details]
Second try, this time following Apple specification

Here you have a proposal patch to fix it.

It does not have activated the dynamic owner change if you use uid=99 because I

don't know a way to invalidate the inode cache and thus, "ls" always shows the
uid of the files as the first user that accessed to it.

Comment 4 Adrian Bunk 2006-02-04 14:08:09 UTC

Roman, can you comment on this patch?

Comment 5 Sean P. Kane 2007-06-22 08:45:35 UTC

Created attachment 11849 [details]
fix broken logic for mount uid=xxx,gid=xxx

Patched by: Wyatt Banks <wyatt[at]banksresearch.com>

and originally posted here: http://lkml.org/lkml/2007/6/20/364

Comment 6 Roman Zippel 2007-06-24 09:44:50 UTC

This is not really a bug, this option is not intended to override the uid/gid, it's only used as default for uninitialized values.
The uid remapping for uid 99 should depend on a different kernel option and it only the uid should be remapped, there is no gid remapping.

Comment 7 Natalie Protasevich 2007-09-22 20:49:38 UTC

Any update on this problem please.
Thanks.

Comment 8 Natalie Protasevich 2008-03-04 00:57:45 UTC

I think the bug was sufficiently explained in #6, so I suppose we can close it as it "works as intended".