When using mkisofs 1.12.1 from the ports collection, hard links in the source tree are not preserved in the ISO9660 that mkisofs creates. Joerg Schilling <schilling@fokus.gmd.de>, who is listed on the man page as the maintainer says by email (7 Jul 2000): " . . . send a bugreport to the Freebsd kernel team. Mkisofs is OK, the bug is in the FreeBSD filesystem code." How-To-Repeat: Do this: 1) make a test directory: mkdir /tmp/exampletree 2) make a test file: touch /tmp/exampletree/file1 3) make a link to the test file: ln /tmp/exampletree/file1 /tmp/exampletree/file2 4) take a look and see the hard link: ls -li /tmp/exampletree/ see (notice files have same inode): total 0 14 -rw-r--r-- 2 mackler wheel 0 Jul 8 06:48 file1 14 -rw-r--r-- 2 mackler wheel 0 Jul 8 06:48 file2 5) use mkisofs to make an iso9660 filesystem: mkisofs -o /tmp/iso9660.test /tmp/exampletree/ 6) mount the new iso9660 filesystem: vnconfig -cv vn0 /tmp/iso9660.test mount_cd9660 /dev/vn0c /mnt 7) take a look and see the two files are not linked: ls -li /mnt see (notice files have different inodes): total 0 47172 -r-xr-xr-x 1 root wheel 0 Jul 8 06:48 file1 47214 -r-xr-xr-x 1 root wheel 0 Jul 8 06:48 file2 8) clean up: umount /mnt vnconfig -u vn0 rm /tmp/iso9660.test rm -rf /tmp/exampletree/
Responsible Changed From-To: freebsd-bugs->dirk Dirk, could you chase up on this? I realize that the PR suggests a bug in a part of FreeBSD which is not your area, but since you're the maintainer of the port, you might be a good person to run around asking the right people the right questions. :-)
Responsible Changed From-To: dirk->freebsd-ports-bugs dirk no longer maintains this port
Responsible Changed From-To: freebsd-ports-bugs->marius over to Marius Strobl, who is the port's maintainer and a src committer ;)
State Changed From-To: open->analyzed This was "fixed" in sys/isofs/cd9660/cd9660_vnops.c rev. 1.77, which in turn broke NFS readdir plus support (PR 63446). While this NFS problem is considered second order I somewhat hesitate to just MFC cd9660_vnops.c 1.77 and I'm not yet sure what to do in order to also fix this in RELENG_4.
Responsible Changed From-To: marius->freebsd-bugs@FreeBSD.org Assign back to the pool as I have have no idea what to do about it (the fix in rev. 1.77 of sys/isofs/cd9660/cd9660_vnops.c was backed out in rev. 1.99 again) and this also isn't about cdrtools/mkisofs (which I also no longer maintain). http://www.freebsd.org/cgi/query-pr.cgi?pr=19782 Adding to audit trail manually by bugmeister on 20080125: bugbuster vwerth@ confirms that this is still a problem with 7.0-RC1.
State Changed From-To: analyzed->open unowned PRs must not be in analyzed state
For bugs matching the following criteria: Status: In Progress Changed: (is less than) 2014-06-01 Reset to default assignee and clear in-progress tags. Mail being skipped
mkisofs is not and was not part of base system
It belongs to cdrtools.
Problem still exists.
As already stated in my comments from 2004-05-11 16:10:28 UTC and 2005-08-10 22:28:37 UTC respectively, this in fact isn't a bug in mkisofs. The problem is that cd9660(5) doesn't use unique inode numbers for hard links, which had been plugged with a quick fix in r97094, but that changed has been "temporarily" reverted in r134807 again as it in turn broke NFS readdir plus support. Thus, change the synopsis accordingly and assign back to default.
Tested with 12.2-STABLE/amd64 and problem is still here.