Esteemed overlayfs maintainers,
Can overlayfs be used for filesystem root ? If not, say no and ignore the rest
of the email.
I have been playing with overlayfs on hikey devlopement board running Android.
(kernel tree based of V4.1). The device does not use initrd/ramdisk
for root instead
an actual block device formatted in squashfs filesystem is used.
hikey:/ # cat /proc/mounts
/dev/root / squashfs ro,seclabel,relatime 0 0
Occasionally we want to make modifications to the rootfs contents for
developement
purposes. Now since, squashfs is a read-only filesystem, we are
experimenting with
overlayfs to absorb the writes when required.
So I mounted overlayfs at /,
hikey:/ # cat /proc/mounts
/dev/root / squashfs ro,seclabel,relatime 0 0
overlayfs / overlay
rw,relatime,lowerdir=/,upperdir=/cache/upper,workdir=/cache/working 0
0
Now if I try to create a directory at /, the mkdir operation complains
that it is a readonly
file system.
hikey:/ # mkdir r2
mkdir: 'r2': Read-only file system
I traced down the error (mnt_is_readonly returns true) and the
following is the call stack.
[ 97.975074] CPU: 5 PID: 2340 Comm: mkdir Tainted: G W
4.1.15-00109-gbc41636-dirty #28
[ 97.975078] Hardware name: HiKey Development Board (DT)
[ 97.975081] Call trace:
[ 97.975094] [<ffffffc000089824>] dump_backtrace+0x0/0x124
[ 97.975100] [<ffffffc00008995c>] show_stack+0x14/0x1c
[ 97.975108] [<ffffffc0009bfc4c>] dump_stack+0x7c/0xbc
[ 97.975116] [<ffffffc0001cfd5c>] __mnt_want_write+0xac/0x168
[ 97.975122] [<ffffffc0001cfe40>] mnt_want_write+0x28/0x4c
[ 97.975128] [<ffffffc0001bb518>] filename_create+0x58/0x24c
[ 97.975134] [<ffffffc0001be968>] SyS_mkdirat+0x70/0x1ec
It seems like current->fs->pwd is pointing to the old mount as in the
original squashfs
mount. I printed out the following.
diff --git a/fs/namei.c b/fs/namei.c
index ccd7f98..6b2ceac 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -1942,6 +1942,11 @@ static int path_init(int dfd, const struct
filename *name, unsigned int flags,
do {
seq = read_seqcount_begin(&fs->seq);
nd->path = fs->pwd;
+ if (printk_ratelimit()) {
+ printk("path %s d_iname: %s
subtype:%s\n", s, fs->pwd.dentry->d_iname,
fs->pwd.mnt->mnt_sb->s_type->name);
+ dump_stack();
+ }
+
nd->seq =
__read_seqcount_begin(&nd->path.dentry->d_seq);
} while (read_seqcount_retry(&fs->seq, seq));
} else {
and the following gets printed:
[ 151.820919] path r2 d_iname: / subtype:squashfs
[ 151.825454] CPU: 4 PID: 2629 Comm: mkdir Tainted: G W
4.1.15-00109-gbc41636-dirty #27
[ 151.834439] Hardware name: HiKey Development Board (DT)
[ 151.839689] Call trace:
[ 151.842141] [<ffffffc000089824>] dump_backtrace+0x0/0x124
[ 151.847565] [<ffffffc00008995c>] show_stack+0x14/0x1c
[ 151.852625] [<ffffffc0009bfc4c>] dump_stack+0x7c/0xbc
[ 151.857703] [<ffffffc0001babdc>] path_init+0x3a8/0x46c
[ 151.862845] [<ffffffc0001bacc8>] path_lookupat+0x28/0x748
[ 151.868268] [<ffffffc0001bb414>] filename_lookup+0x2c/0xd8
[ 151.873757] [<ffffffc0001bb4f0>] filename_create+0x30/0x24c
[ 151.879357] [<ffffffc0001be968>] SyS_mkdirat+0x70/0x1ec
Note that the subtype is squashfs instead of overlayfs !!
I am trying to fix this and I am not sure when and what refreshes the
current->fs->pwd ?
Am I looking at this in the right direction ?
Thanks !!
--
To unsubscribe from this list: send the line "unsubscribe linux-unionfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
