Open files, which aren't present in the snapshot, which is being
roll-backed to, need to disappear from the visible VFS image of
the dataset.

Kernel provides __d_drop function to locklessly drop invalid entry
from the dcache, but inode can be referenced by multiple dentries.

The introduced spl_d_drop_aliases function walks and invalidates
all aliases of an inode.

Signed-off-by: Pavel Snajdr <snajpa@snajpa.net>

This is in preparation for RENAME_EXCHANGE and RENAME_WHITEOUT support
for ZoL, but the changes here allow for far nicer fallbacks than the
previous implementation (the source and target are re-linked in case of
the final link failing).

In addition, a small cleanup was done for the "target exists but is a
different type" codepath so that it's more understandable.

Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
Signed-off-by: Pavel Snajdr <snajpa@snajpa.net>

Implement support for Linux's RENAME_* flags (for renameat2). Aside from
being quite useful for userspace (providing race-free ways to exchange
paths and implement mv --no-clobber), they are used by overlayfs and are
thus required in order to use overlayfs-on-ZFS.

In order for us to represent the new renameat2(2) flags in the ZIL, we
need to create a new transaction type (to be backwards-compatible).
Since RENAME_EXCHANGE and RENAME_WHITEOUT are mutually exclusive they
deserve separate types. We just re-use the logic of
zfs_{log,replay}_rename() with the only change being the transaction
types and the associate vfsflags passed to zfs_rename().

RENAME_NOREPLACE doesn't need an entry because if the renameat2(2) fails
because of RENAME_NOREPLACE there won't be a ZIL entry for the operation
(and if it succeeds then it should also succeed on-replay).

Unfortunately, more work is required in order use overlayfs-on-ZFS
(namely we have to remove our .d_revalidate hook, since overlayfs
refuses to use a filesystem with d_revalidate as an upperdir).

Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
Signed-off-by: Pavel Snajdr <snajpa@snajpa.net>

Since mv(1) doesn't expose the RENAME_* flags we need to have our own
variation in the tests/ tree. The tests are fairly obvious functional
tests, though in the future (once we solve the d_revalidate issue) we
might also add a full-stack overlayfs integration test.

Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
Signed-off-by: Pavel Snajdr <snajpa@snajpa.net>

Removing new txtypes in favor of compound ZIL operations, see comment in
module/zfs/zfs_log.c.

Other notable changes:

- unlock after the inodes are updated
- pass whiteout znode pointer to zfs_log_rename_whiteout
- don't wrap code directly in ASSERT*(), it turn to noop on non-debug
  builds
- update configure time tests for rename2 to support kernels
  from 3.5 to 4.8

Fixes openzfs#2256
Fixes openzfs#8648
Fixes openzfs#8774

Signed-off-by: Pavel Snajdr <snajpa@snajpa.net>

Signed-off-by: Pavel Snajdr <snajpa@snajpa.net>

EL7 has .rename2 op upper in the inode_operations_wrapper,
so let's add a wrapping ifdef to dir inode_operations
and ifdef the ops assignment too.

Configure test lives in kernel-rename.m4 for now, as rename2/EL7 case
is the only reason why we need it - for now, until support for
tmpfile is needed. then it would me more logical to move this check
to a separate file.

Signed-off-by: Pavel Snajdr <snajpa@snajpa.net>