On Wed, May 27, 2020 at 7:02 PM Vivek Goyal <vgoyal@xxxxxxxxxx> wrote:
>
> On Wed, May 27, 2020 at 04:46:37PM +0200, Miklos Szeredi wrote:
> > On Wed, May 27, 2020 at 1:16 PM Amir Goldstein <amir73il@xxxxxxxxx> wrote:
> > >
> > > On Wed, May 27, 2020 at 6:45 AM yangerkun <yangerkun@xxxxxxxxxx> wrote:
> > > >
> > > > Run generic/461 with ext4 upper/lower layer sometimes may trigger the
> > > > bug as below(linux 4.19):
> > > >
> > > > [ 551.001349] overlayfs: failed to get metacopy (-5)
> > > > [ 551.003464] overlayfs: failed to get inode (-5)
> > > > [ 551.004243] overlayfs: cleanup of 'd44/fd51' failed (-5)
> > > > [ 551.004941] overlayfs: failed to get origin (-5)
> > > > [ 551.005199] ------------[ cut here ]------------
> > > > [ 551.006697] WARNING: CPU: 3 PID: 24674 at fs/inode.c:1528 iput+0x33b/0x400
> > > > ...
> > > > [ 551.027219] Call Trace:
> > > > [ 551.027623] ovl_create_object+0x13f/0x170
> > > > [ 551.028268] ovl_create+0x27/0x30
> > > > [ 551.028799] path_openat+0x1a35/0x1ea0
> > > > [ 551.029377] do_filp_open+0xad/0x160
> > > > [ 551.029944] ? vfs_writev+0xe9/0x170
> > > > [ 551.030499] ? page_counter_try_charge+0x77/0x120
> > > > [ 551.031245] ? __alloc_fd+0x160/0x2a0
> > > > [ 551.031832] ? do_sys_open+0x189/0x340
> > > > [ 551.032417] ? get_unused_fd_flags+0x34/0x40
> > > > [ 551.033081] do_sys_open+0x189/0x340
> > > > [ 551.033632] __x64_sys_creat+0x24/0x30
> > > > [ 551.034219] do_syscall_64+0xd5/0x430
> > > > [ 551.034800] entry_SYSCALL_64_after_hwframe+0x44/0xa9
> > > > ...
> > > > [ 556.107515] BUG: Dentry 000000006bc1d73f{i=4129c,n=fd51} still in use (-1) [unmount of ext4 sdb]
> > > > [ 556.108946] ------------[ cut here ]------------
> > > > [ 556.109686] WARNING: CPU: 1 PID: 24682 at fs/dcache.c:1557 umount_check+0x95/0xc0
> > > > [ 556.130343] d_walk+0x10d/0x430
> > > > [ 556.130832] do_one_tree+0x30/0x60
> > > > [ 556.131365] shrink_dcache_for_umount+0x38/0xe0
> > > > [ 556.132063] generic_shutdown_super+0x2e/0x1c0
> > > > [ 556.132747] kill_block_super+0x29/0x80
> > > > [ 556.133338] deactivate_locked_super+0x7a/0x100
> > > > [ 556.134034] deactivate_super+0x9d/0xb0
> > > > [ 556.134627] cleanup_mnt+0x67/0x100
> > > > [ 556.135173] __cleanup_mnt+0x16/0x20
> > > > [ 556.135731] task_work_run+0xdb/0x110
> > > > [ 556.136306] exit_to_usermode_loop+0x197/0x1b0
> > > > [ 556.136991] do_syscall_64+0x3ce/0x430
> > > > [ 556.137571] entry_SYSCALL_64_after_hwframe+0x44/0xa9
> > > > ...
> > > > [ 556.378140] VFS: Busy inodes after unmount of sdb. Self-destruct in 5 seconds. Have a nice day...
> > > >
> > > > After check the code, there may some bug need to fix:
> > > > 1. We need to call iput once ovl_check_metacopy_xattr fail.
> > > > 2. We need to call unlock_new_inode or the above iput(also with iput in
> > > > ovl_create_object) will trigger the a WARN_ON since the I_NEW still
> > > > exists.
> > > > 3. We should move the init for upperdentry to the place below
> > > > ovl_check_metacopy_xattr. Or the dentry reference will decrease to
> > > > -1(error path in ovl_create_upper will inc, ovl_destroy_inode too).
> > > >
> > >
> > > OR we don't check metacopy xattr in ovl_get_inode().
> > >
> > > In ovl_lookup() we already checked metacopy xattr.
> > > No reason to check it again in this subtle context.
> > >
> > > In ovl_lookup() can store value of upper metacopy and after we get
> > > the inode, set the OVL_UPPERDATA inode flag according to
> > > upperdentry && !uppermetacopy.
> > >
> > > That would be consistent with ovl_obtain_alias() which sets the
> > > OVL_UPPERDATA inode flag after getting the inode.
> >
> > I agree that that is a good direction, however for the actual fix I
> > think the following is sufficient (whitespace damaged, only for
> > review).
> >
> > The reason we can skip the metacopy check for the ->newinode != NULL
> > case is that that only happens on object creation, which very
> > obviously won't have metacopy set.
> >
> > Thanks,
> > Miklos
> >
> > diff --git a/fs/overlayfs/inode.c b/fs/overlayfs/inode.c
> > index 3b7ed5d2279c..fd7f1d4adf04 100644
> > --- a/fs/overlayfs/inode.c
> > +++ b/fs/overlayfs/inode.c
> > @@ -889,7 +889,7 @@ struct inode *ovl_get_inode(struct super_block *sb,
> > if (oip->index)
> > ovl_set_flag(OVL_INDEX, inode);
> >
> > - if (upperdentry) {
> > + if (upperdentry && !oip->newinode) {
> > err = ovl_check_metacopy_xattr(upperdentry);
> > if (err < 0)
> > goto out_err;
>
> Hi Miklos and Amir,
>
> How about enahncing above a bit to deal with error. Will this work. Just
> compile tested.
>
Please no. it makes no sense.
Proper fix IMO is to use value checked in ovl_lookup()
instead of rechecking here.
Miklos' fix should works by avoiding hitting this check
in the most likely scenario (that getxattr either fails or succeeds,
but not randomly).
Thanks,
Amir.
