Re: [PATCH v7 07/12] kernfs: hold kernfs_iattr_rwsem when initializing a new inode

Op 16-10-2026 10:50 EDT schreef Greg Kroah-Hartman [off-list ref]:

 
On Fri, Oct 16, 2026 at 04:38:01PM +0200, Jori Koolstra wrote:

quoted

kernfs_refresh_inode() copies the uid, gid, timestamps and nlink of a
kernfs_node into its VFS inode. Since commit 9caf69614225 ("kernfs:
Introduce separate rwsem to protect inode attributes.") these are
protected by kernfs_iattr_rwsem.

kernfs_init_inode() calls kernfs_refresh_inode(), but was not converted.
Its callers in turn hold kernfs_rwsem (kernfs_iop_lookup(),
kernfs_fill_super()) or no lock at all (__kernfs_fh_to_dentry()),
neither of which excludes a concurrent
kernfs_setattr()/kernfs_iop_setattr().

Take kernfs_iattr_rwsem for reading around the refresh in
kernfs_init_inode(). This nests inside kernfs_rwsem, which matches
kernfs_link_sibling() and kernfs_unlink_sibling().

Also, kernfs_get_inode() then does not need to hold kernfs_rwsem, so
drop it from kernfs_fill_super().

Oh, nice, this fixes the issue of kernfs not keeping the inode like we
talked about at LPC?  Can't argue with this tiny patch, nice work!

No, that part is included in the patch following this one. This is a bit of
preliminary work to fix the locking around kernfs_refresh_inode(). I think the
memory requirement for the allocated inodes is not a blocker (but you judge, since
you told me that's the primary concern here). The inode allocation only happens
for userspace mkdir(2) calls afaict, and then you probably want to do a lookuo
anyway.

I see there's some work by a Meta team on kernfs. Since I'm in that code now
I'll try to review that soon.

Enjoy the weekend!
Jori.

Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>