Skip to content

Commit

Permalink
afs: Fix speculative status fetches
Browse files Browse the repository at this point in the history
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: a9e5c87 ("afs: Fix speculative status fetch going out of order wrt to modifications")
Reported-by: Marc Dionne <[email protected]>
Signed-off-by: David Howells <[email protected]>
Tested-and-reviewed-by: Marc Dionne <[email protected]>
cc: [email protected]
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 <[email protected]>
  • Loading branch information
dhowells authored and torvalds committed May 1, 2021
1 parent 7af81cd commit 22650f1
Show file tree
Hide file tree
Showing 6 changed files with 23 additions and 2 deletions.
7 changes: 7 additions & 0 deletions fs/afs/dir.c
Original file line number Diff line number Diff line change
Expand Up @@ -1419,6 +1419,7 @@ static int afs_mkdir(struct user_namespace *mnt_userns, struct inode *dir,

afs_op_set_vnode(op, 0, dvnode);
op->file[0].dv_delta = 1;
op->file[0].modification = true;
op->file[0].update_ctime = true;
op->dentry = dentry;
op->create.mode = S_IFDIR | mode;
Expand Down Expand Up @@ -1500,6 +1501,7 @@ static int afs_rmdir(struct inode *dir, struct dentry *dentry)

afs_op_set_vnode(op, 0, dvnode);
op->file[0].dv_delta = 1;
op->file[0].modification = true;
op->file[0].update_ctime = true;

op->dentry = dentry;
Expand Down Expand Up @@ -1636,6 +1638,7 @@ static int afs_unlink(struct inode *dir, struct dentry *dentry)

afs_op_set_vnode(op, 0, dvnode);
op->file[0].dv_delta = 1;
op->file[0].modification = true;
op->file[0].update_ctime = true;

/* Try to make sure we have a callback promise on the victim. */
Expand Down Expand Up @@ -1718,6 +1721,7 @@ static int afs_create(struct user_namespace *mnt_userns, struct inode *dir,

afs_op_set_vnode(op, 0, dvnode);
op->file[0].dv_delta = 1;
op->file[0].modification = true;
op->file[0].update_ctime = true;

op->dentry = dentry;
Expand Down Expand Up @@ -1792,6 +1796,7 @@ static int afs_link(struct dentry *from, struct inode *dir,
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].update_ctime = true;

Expand Down Expand Up @@ -1987,6 +1992,8 @@ static int afs_rename(struct user_namespace *mnt_userns, struct inode *old_dir,
afs_op_set_vnode(op, 1, new_dvnode); /* May be same as orig_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;

Expand Down
3 changes: 3 additions & 0 deletions fs/afs/dir_silly.c
Original file line number Diff line number Diff line change
Expand Up @@ -73,6 +73,8 @@ static int afs_do_silly_rename(struct afs_vnode *dvnode, struct afs_vnode *vnode
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;

Expand Down Expand Up @@ -201,6 +203,7 @@ static int afs_do_silly_unlink(struct afs_vnode *dvnode, struct afs_vnode *vnode
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;
Expand Down
6 changes: 6 additions & 0 deletions fs/afs/fs_operation.c
Original file line number Diff line number Diff line change
Expand Up @@ -118,6 +118,8 @@ static void afs_prepare_vnode(struct afs_operation *op, struct afs_vnode_param *
vp->cb_break_before = afs_calc_vnode_cb_break(vnode);
if (vnode->lock_state != AFS_VNODE_LOCK_NONE)
op->flags |= AFS_OPERATION_CUR_ONLY;
if (vp->modification)
set_bit(AFS_VNODE_MODIFYING, &vnode->flags);
}

if (vp->fid.vnode)
Expand Down Expand Up @@ -225,6 +227,10 @@ int afs_put_operation(struct afs_operation *op)

if (op->ops && op->ops->put)
op->ops->put(op);
if (op->file[0].modification)
clear_bit(AFS_VNODE_MODIFYING, &op->file[0].vnode->flags);
if (op->file[1].modification && op->file[1].vnode != op->file[0].vnode)
clear_bit(AFS_VNODE_MODIFYING, &op->file[1].vnode->flags);
if (op->file[0].put_vnode)
iput(&op->file[0].vnode->vfs_inode);
if (op->file[1].put_vnode)
Expand Down
6 changes: 4 additions & 2 deletions fs/afs/inode.c
Original file line number Diff line number Diff line change
Expand Up @@ -294,8 +294,9 @@ void afs_vnode_commit_status(struct afs_operation *op, struct afs_vnode_param *v
op->flags &= ~AFS_OPERATION_DIR_CONFLICT;
}
} else if (vp->scb.have_status) {
if (vp->dv_before + vp->dv_delta != vp->scb.status.data_version &&
vp->speculative)
if (vp->speculative &&
(test_bit(AFS_VNODE_MODIFYING, &vnode->flags) ||
vp->dv_before != vnode->status.data_version))
/* Ignore the result of a speculative bulk status fetch
* if it splits around a modification op, thereby
* appearing to regress the data version.
Expand Down Expand Up @@ -911,6 +912,7 @@ int afs_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
}
op->ctime = attr->ia_ctime;
op->file[0].update_ctime = 1;
op->file[0].modification = true;

op->ops = &afs_setattr_operation;
ret = afs_do_sync_operation(op);
Expand Down
2 changes: 2 additions & 0 deletions fs/afs/internal.h
Original file line number Diff line number Diff line change
Expand Up @@ -645,6 +645,7 @@ struct afs_vnode {
#define AFS_VNODE_PSEUDODIR 7 /* set if Vnode is a pseudo directory */
#define AFS_VNODE_NEW_CONTENT 8 /* Set if file has new content (create/trunc-0) */
#define AFS_VNODE_SILLY_DELETED 9 /* Set if file has been silly-deleted */
#define AFS_VNODE_MODIFYING 10 /* Set if we're performing a modification op */

struct list_head wb_keys; /* List of keys available for writeback */
struct list_head pending_locks; /* locks waiting to be granted */
Expand Down Expand Up @@ -762,6 +763,7 @@ struct afs_vnode_param {
bool set_size:1; /* Must update i_size */
bool op_unlinked:1; /* True if file was unlinked by op */
bool speculative:1; /* T if speculative status fetch (no vnode lock) */
bool modification:1; /* Set if the content gets modified */
};

/*
Expand Down
1 change: 1 addition & 0 deletions fs/afs/write.c
Original file line number Diff line number Diff line change
Expand Up @@ -377,6 +377,7 @@ static int afs_store_data(struct afs_vnode *vnode, struct iov_iter *iter, loff_t

afs_op_set_vnode(op, 0, vnode);
op->file[0].dv_delta = 1;
op->file[0].modification = true;
op->store.write_iter = iter;
op->store.pos = pos;
op->store.size = size;
Expand Down

0 comments on commit 22650f1

Please sign in to comment.