kernel_optimize_test/fs/afs/dir_silly.c
David Howells f76e0829bb afs: Fix speculative status fetches
[ Upstream commit 22650f148126571be1098d34160eb4931fc77241 ]

The generic/464 xfstest causes kAFS to emit occasional warnings of the
form:

        kAFS: vnode modified {100055:8a} 30->31 YFS.StoreData64 (c=6015)

This indicates that the data version received back from the server did not
match the expected value (the DV should be incremented monotonically for
each individual modification op committed to a vnode).

What is happening is that a lookup call is doing a bulk status fetch
speculatively on a bunch of vnodes in a directory besides getting the
status of the vnode it's actually interested in.  This is racing with a
StoreData operation (though it could also occur with, say, a MakeDir op).

On the client, a modification operation locks the vnode, but the bulk
status fetch only locks the parent directory, so no ordering is imposed
there (thereby avoiding an avenue to deadlock).

On the server, the StoreData op handler doesn't lock the vnode until it's
received all the request data, and downgrades the lock after committing the
data until it has finished sending change notifications to other clients -
which allows the status fetch to occur before it has finished.

This means that:

 - a status fetch can access the target vnode either side of the exclusive
   section of the modification

 - the status fetch could start before the modification, yet finish after,
   and vice-versa.

 - the status fetch and the modification RPCs can complete in either order.

 - the status fetch can return either the before or the after DV from the
   modification.

 - the status fetch might regress the locally cached DV.

Some of these are handled by the previous fix[1], but that's not sufficient
because it checks the DV it received against the DV it cached at the start
of the op, but the DV might've been updated in the meantime by a locally
generated modification op.

Fix this by the following means:

 (1) Keep track of when we're performing a modification operation on a
     vnode.  This is done by marking vnode parameters with a 'modification'
     note that causes the AFS_VNODE_MODIFYING flag to be set on the vnode
     for the duration.

 (2) Alter the speculation race detection to ignore speculative status
     fetches if either the vnode is marked as being modified or the data
     version number is not what we expected.

Note that whilst the "vnode modified" warning does get recovered from as it
causes the client to refetch the status at the next opportunity, it will
also invalidate the pagecache, so changes might get lost.

Fixes: a9e5c87ca7 ("afs: Fix speculative status fetch going out of order wrt to modifications")
Reported-by: Marc Dionne <marc.dionne@auristor.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Tested-and-reviewed-by: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
Link: https://lore.kernel.org/r/160605082531.252452.14708077925602709042.stgit@warthog.procyon.org.uk/ [1]
Link: https://lore.kernel.org/linux-fsdevel/161961335926.39335.2552653972195467566.stgit@warthog.procyon.org.uk/ # v1
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2021-05-14 09:50:45 +02:00

283 lines
7.4 KiB
C

// SPDX-License-Identifier: GPL-2.0-or-later
/* AFS silly rename handling
*
* Copyright (C) 2019 Red Hat, Inc. All Rights Reserved.
* Written by David Howells (dhowells@redhat.com)
* - Derived from NFS's sillyrename.
*/
#include <linux/kernel.h>
#include <linux/fs.h>
#include <linux/namei.h>
#include <linux/fsnotify.h>
#include "internal.h"
static void afs_silly_rename_success(struct afs_operation *op)
{
_enter("op=%08x", op->debug_id);
afs_check_dir_conflict(op, &op->file[0]);
afs_vnode_commit_status(op, &op->file[0]);
}
static void afs_silly_rename_edit_dir(struct afs_operation *op)
{
struct afs_vnode_param *dvp = &op->file[0];
struct afs_vnode *dvnode = dvp->vnode;
struct afs_vnode *vnode = AFS_FS_I(d_inode(op->dentry));
struct dentry *old = op->dentry;
struct dentry *new = op->dentry_2;
spin_lock(&old->d_lock);
old->d_flags |= DCACHE_NFSFS_RENAMED;
spin_unlock(&old->d_lock);
if (dvnode->silly_key != op->key) {
key_put(dvnode->silly_key);
dvnode->silly_key = key_get(op->key);
}
down_write(&dvnode->validate_lock);
if (test_bit(AFS_VNODE_DIR_VALID, &dvnode->flags) &&
dvnode->status.data_version == dvp->dv_before + dvp->dv_delta) {
afs_edit_dir_remove(dvnode, &old->d_name,
afs_edit_dir_for_silly_0);
afs_edit_dir_add(dvnode, &new->d_name,
&vnode->fid, afs_edit_dir_for_silly_1);
}
up_write(&dvnode->validate_lock);
}
static const struct afs_operation_ops afs_silly_rename_operation = {
.issue_afs_rpc = afs_fs_rename,
.issue_yfs_rpc = yfs_fs_rename,
.success = afs_silly_rename_success,
.edit_dir = afs_silly_rename_edit_dir,
};
/*
* Actually perform the silly rename step.
*/
static int afs_do_silly_rename(struct afs_vnode *dvnode, struct afs_vnode *vnode,
struct dentry *old, struct dentry *new,
struct key *key)
{
struct afs_operation *op;
_enter("%pd,%pd", old, new);
op = afs_alloc_operation(key, dvnode->volume);
if (IS_ERR(op))
return PTR_ERR(op);
afs_op_set_vnode(op, 0, dvnode);
afs_op_set_vnode(op, 1, dvnode);
op->file[0].dv_delta = 1;
op->file[1].dv_delta = 1;
op->file[0].modification = true;
op->file[1].modification = true;
op->file[0].update_ctime = true;
op->file[1].update_ctime = true;
op->dentry = old;
op->dentry_2 = new;
op->ops = &afs_silly_rename_operation;
trace_afs_silly_rename(vnode, false);
return afs_do_sync_operation(op);
}
/**
* afs_sillyrename - Perform a silly-rename of a dentry
*
* AFS is stateless and the server doesn't know when the client is holding a
* file open. To prevent application problems when a file is unlinked while
* it's still open, the client performs a "silly-rename". That is, it renames
* the file to a hidden file in the same directory, and only performs the
* unlink once the last reference to it is put.
*
* The final cleanup is done during dentry_iput.
*/
int afs_sillyrename(struct afs_vnode *dvnode, struct afs_vnode *vnode,
struct dentry *dentry, struct key *key)
{
static unsigned int sillycounter;
struct dentry *sdentry = NULL;
unsigned char silly[16];
int ret = -EBUSY;
_enter("");
/* We don't allow a dentry to be silly-renamed twice. */
if (dentry->d_flags & DCACHE_NFSFS_RENAMED)
return -EBUSY;
sdentry = NULL;
do {
int slen;
dput(sdentry);
sillycounter++;
/* Create a silly name. Note that the ".__afs" prefix is
* understood by the salvager and must not be changed.
*/
slen = scnprintf(silly, sizeof(silly), ".__afs%04X", sillycounter);
sdentry = lookup_one_len(silly, dentry->d_parent, slen);
/* N.B. Better to return EBUSY here ... it could be dangerous
* to delete the file while it's in use.
*/
if (IS_ERR(sdentry))
goto out;
} while (!d_is_negative(sdentry));
ihold(&vnode->vfs_inode);
ret = afs_do_silly_rename(dvnode, vnode, dentry, sdentry, key);
switch (ret) {
case 0:
/* The rename succeeded. */
set_bit(AFS_VNODE_SILLY_DELETED, &vnode->flags);
d_move(dentry, sdentry);
break;
case -ERESTARTSYS:
/* The result of the rename is unknown. Play it safe by forcing
* a new lookup.
*/
d_drop(dentry);
d_drop(sdentry);
}
iput(&vnode->vfs_inode);
dput(sdentry);
out:
_leave(" = %d", ret);
return ret;
}
static void afs_silly_unlink_success(struct afs_operation *op)
{
_enter("op=%08x", op->debug_id);
afs_check_dir_conflict(op, &op->file[0]);
afs_vnode_commit_status(op, &op->file[0]);
afs_vnode_commit_status(op, &op->file[1]);
afs_update_dentry_version(op, &op->file[0], op->dentry);
}
static void afs_silly_unlink_edit_dir(struct afs_operation *op)
{
struct afs_vnode_param *dvp = &op->file[0];
struct afs_vnode *dvnode = dvp->vnode;
_enter("op=%08x", op->debug_id);
down_write(&dvnode->validate_lock);
if (test_bit(AFS_VNODE_DIR_VALID, &dvnode->flags) &&
dvnode->status.data_version == dvp->dv_before + dvp->dv_delta)
afs_edit_dir_remove(dvnode, &op->dentry->d_name,
afs_edit_dir_for_unlink);
up_write(&dvnode->validate_lock);
}
static const struct afs_operation_ops afs_silly_unlink_operation = {
.issue_afs_rpc = afs_fs_remove_file,
.issue_yfs_rpc = yfs_fs_remove_file,
.success = afs_silly_unlink_success,
.aborted = afs_check_for_remote_deletion,
.edit_dir = afs_silly_unlink_edit_dir,
};
/*
* Tell the server to remove a sillyrename file.
*/
static int afs_do_silly_unlink(struct afs_vnode *dvnode, struct afs_vnode *vnode,
struct dentry *dentry, struct key *key)
{
struct afs_operation *op;
_enter("");
op = afs_alloc_operation(NULL, dvnode->volume);
if (IS_ERR(op))
return PTR_ERR(op);
afs_op_set_vnode(op, 0, dvnode);
afs_op_set_vnode(op, 1, vnode);
op->file[0].dv_delta = 1;
op->file[0].modification = true;
op->file[0].update_ctime = true;
op->file[1].op_unlinked = true;
op->file[1].update_ctime = true;
op->dentry = dentry;
op->ops = &afs_silly_unlink_operation;
trace_afs_silly_rename(vnode, true);
afs_begin_vnode_operation(op);
afs_wait_for_operation(op);
/* If there was a conflict with a third party, check the status of the
* unlinked vnode.
*/
if (op->error == 0 && (op->flags & AFS_OPERATION_DIR_CONFLICT)) {
op->file[1].update_ctime = false;
op->fetch_status.which = 1;
op->ops = &afs_fetch_status_operation;
afs_begin_vnode_operation(op);
afs_wait_for_operation(op);
}
return afs_put_operation(op);
}
/*
* Remove sillyrename file on iput.
*/
int afs_silly_iput(struct dentry *dentry, struct inode *inode)
{
struct afs_vnode *dvnode = AFS_FS_I(d_inode(dentry->d_parent));
struct afs_vnode *vnode = AFS_FS_I(inode);
struct dentry *alias;
int ret;
DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq);
_enter("%p{%pd},%llx", dentry, dentry, vnode->fid.vnode);
down_read(&dvnode->rmdir_lock);
alias = d_alloc_parallel(dentry->d_parent, &dentry->d_name, &wq);
if (IS_ERR(alias)) {
up_read(&dvnode->rmdir_lock);
return 0;
}
if (!d_in_lookup(alias)) {
/* We raced with lookup... See if we need to transfer the
* sillyrename information to the aliased dentry.
*/
ret = 0;
spin_lock(&alias->d_lock);
if (d_really_is_positive(alias) &&
!(alias->d_flags & DCACHE_NFSFS_RENAMED)) {
alias->d_flags |= DCACHE_NFSFS_RENAMED;
ret = 1;
}
spin_unlock(&alias->d_lock);
up_read(&dvnode->rmdir_lock);
dput(alias);
return ret;
}
/* Stop lock-release from complaining. */
spin_lock(&vnode->lock);
vnode->lock_state = AFS_VNODE_LOCK_DELETED;
trace_afs_flock_ev(vnode, NULL, afs_flock_silly_delete, 0);
spin_unlock(&vnode->lock);
afs_do_silly_unlink(dvnode, vnode, dentry, dvnode->silly_key);
up_read(&dvnode->rmdir_lock);
d_lookup_done(alias);
dput(alias);
return 1;
}