On Sat, May 26, 2018 at 12:53:26AM +0100, Al Viro wrote: > Continuation of the series - more simple cases: > openpromfs: switch to d_splice_alias() > orangefs_lookup: simplify > omfs_lookup(): report IO errors, use d_splice_alias() > hfs: use d_splice_alias() > hfs: don't allow mounting over .../rsrc > hfsplus: switch to d_splice_alias() > ncp_lookup(): use d_splice_alias() > 9p: unify paths in v9fs_vfs_lookup() > cifs_lookup(): cifs_get_inode_...() never returns 0 with *inode left NULL > cifs_lookup(): switch to d_splice_alias() ... and a lot more convoluted ones - procfs side. There we have a bloody misleading games around pid_revalidate() and friends; we do *NOT* need those for any kind of race protection anymore (and hadn't for years). However, these calls cannot be simply dropped - some of their side effects (namely, setting ownership and permissions) are needed. Separating those into common helpers used by revalidate and by lookups allows to untangle that mess. I would really like an ACK/NAK from Alexey on that part...
