Hi all,
While doing some quick tests with unionfs here, I noticed that with the
following setup:
[EMAIL PROTECTED] unionfs-test]# ll */oi
-rw-r--r-- 1 root root 4 2008-02-19 11:34 mount/oi
-rw-r--r-- 1 root root 3 2008-02-19 11:30 ro/oi
-rw-r--r-- 1 root root 4 2008-02-19 11:34 rw/oi
[EMAIL PROTECTED] unionfs-test]# pwd
/tmp/unionfs-test
[EMAIL PROTECTED] unionfs-test]# grep unionfs-test /proc/mounts
none /tmp/unionfs-test/mount unionfs
rw,dirs=/tmp/unionfs-test/rw=rw:/tmp/unionfs-test/ro=ro 0 0
unionctl --query didn't work if the file was present on both branches, like
the 'oi' file.
[EMAIL PROTECTED] mount]# unionctl oi --query
Segmentation fault
Then I noticed that unionctl 0.2.1 did a step wrongly. There goes a patch,
which explains it better than any word I guess :)
--- unionfs_utils-0.2.1/ctl.c.old 2008-02-19 13:56:43.000000000 -0300
+++ unionfs_utils-0.2.1/ctl.c 2008-02-19 13:56:43.000000000 -0300
@@ -87,7 +87,7 @@ int unionfs_query(const char *file_path,
if (FD_ISSET(i, &branchlist)) {
*ufs_branches = realloc(*ufs_branches,
sizeof(struct unionfs_branch)*(ret+1));
- (*ufs_branches)[ret].path =
malloc(strlen(branches[ret]+1));
+ (*ufs_branches)[ret].path =
malloc(strlen(branches[i])+1);
strcpy((*ufs_branches)[ret].path, branches[i]);
(*ufs_branches)[ret].perms = branchperms[i];
ret++;
2 bugs on that line:
a) branches[] got indexed by two different variables, differently
controlled. malloc used 'ret' index while strcpy used 'i' index.
b) the '+1' for counting the null-byte should be outside strlen(pointer),
otherwise we will have a 2 bytes offset. This may be worst on chrooted
systems, which would make it malloc(0) bytes.
What do you think?
Please, Cc me as currently I'm not subscribed to the list.
Thanks,
Marcelo.
_______________________________________________
unionfs mailing list: http://unionfs.filesystems.org/
unionfs@mail.fsl.cs.sunysb.edu
http://www.fsl.cs.sunysb.edu/mailman/listinfo/unionfs
