linux_dsm_epyc7002/fs/nfs/unlink.c

564 lines
14 KiB
C
Raw Normal View History

/*
* linux/fs/nfs/unlink.c
*
* nfs sillydelete handling
*
*/
#include <linux/slab.h>
#include <linux/string.h>
#include <linux/dcache.h>
#include <linux/sunrpc/sched.h>
#include <linux/sunrpc/clnt.h>
#include <linux/nfs_fs.h>
#include <linux/sched.h>
#include <linux/wait.h>
#include <linux/namei.h>
#include "internal.h"
#include "nfs4_fs.h"
#include "iostat.h"
#include "delegation.h"
/**
* nfs_free_unlinkdata - release data from a sillydelete operation.
* @data: pointer to unlink structure.
*/
static void
nfs_free_unlinkdata(struct nfs_unlinkdata *data)
{
iput(data->dir);
put_rpccred(data->cred);
kfree(data->args.name.name);
kfree(data);
}
#define NAME_ALLOC_LEN(len) ((len+16) & ~15)
/**
* nfs_copy_dname - copy dentry name to data structure
* @dentry: pointer to dentry
* @data: nfs_unlinkdata
*/
static int nfs_copy_dname(struct dentry *dentry, struct nfs_unlinkdata *data)
{
char *str;
int len = dentry->d_name.len;
str = kmemdup(dentry->d_name.name, NAME_ALLOC_LEN(len), GFP_KERNEL);
if (!str)
return -ENOMEM;
data->args.name.len = len;
data->args.name.name = str;
return 0;
}
static void nfs_free_dname(struct nfs_unlinkdata *data)
{
kfree(data->args.name.name);
data->args.name.name = NULL;
data->args.name.len = 0;
}
static void nfs_dec_sillycount(struct inode *dir)
{
struct nfs_inode *nfsi = NFS_I(dir);
if (atomic_dec_return(&nfsi->silly_count) == 1)
wake_up(&nfsi->waitqueue);
}
/**
* nfs_async_unlink_done - Sillydelete post-processing
* @task: rpc_task of the sillydelete
*
* Do the directory attribute update.
*/
static void nfs_async_unlink_done(struct rpc_task *task, void *calldata)
{
struct nfs_unlinkdata *data = calldata;
struct inode *dir = data->dir;
if (!NFS_PROTO(dir)->unlink_done(task, dir))
rpc_restart_call_prepare(task);
}
/**
* nfs_async_unlink_release - Release the sillydelete data.
* @task: rpc_task of the sillydelete
*
* We need to call nfs_put_unlinkdata as a 'tk_release' task since the
* rpc_task would be freed too.
*/
static void nfs_async_unlink_release(void *calldata)
{
struct nfs_unlinkdata *data = calldata;
struct super_block *sb = data->dir->i_sb;
nfs_dec_sillycount(data->dir);
nfs_free_unlinkdata(data);
nfs_sb_deactive(sb);
}
static void nfs_unlink_prepare(struct rpc_task *task, void *calldata)
{
struct nfs_unlinkdata *data = calldata;
NFS_PROTO(data->dir)->unlink_rpc_prepare(task, data);
}
static const struct rpc_call_ops nfs_unlink_ops = {
.rpc_call_done = nfs_async_unlink_done,
.rpc_release = nfs_async_unlink_release,
.rpc_call_prepare = nfs_unlink_prepare,
};
static int nfs_do_call_unlink(struct dentry *parent, struct inode *dir, struct nfs_unlinkdata *data)
{
struct rpc_message msg = {
.rpc_argp = &data->args,
.rpc_resp = &data->res,
.rpc_cred = data->cred,
};
struct rpc_task_setup task_setup_data = {
.rpc_message = &msg,
.callback_ops = &nfs_unlink_ops,
.callback_data = data,
.workqueue = nfsiod_workqueue,
.flags = RPC_TASK_ASYNC,
};
struct rpc_task *task;
struct dentry *alias;
alias = d_lookup(parent, &data->args.name);
if (alias != NULL) {
int ret;
void *devname_garbage = NULL;
/*
* Hey, we raced with lookup... See if we need to transfer
* the sillyrename information to the aliased dentry.
*/
nfs_free_dname(data);
ret = nfs_copy_dname(alias, data);
spin_lock(&alias->d_lock);
if (ret == 0 && alias->d_inode != NULL &&
!(alias->d_flags & DCACHE_NFSFS_RENAMED)) {
devname_garbage = alias->d_fsdata;
alias->d_fsdata = data;
alias->d_flags |= DCACHE_NFSFS_RENAMED;
ret = 1;
} else
ret = 0;
spin_unlock(&alias->d_lock);
nfs_dec_sillycount(dir);
dput(alias);
/*
* If we'd displaced old cached devname, free it. At that
* point dentry is definitely not a root, so we won't need
* that anymore.
*/
kfree(devname_garbage);
return ret;
}
data->dir = igrab(dir);
if (!data->dir) {
nfs_dec_sillycount(dir);
return 0;
}
nfs_sb_active(dir->i_sb);
data->args.fh = NFS_FH(dir);
nfs_fattr_init(data->res.dir_attr);
NFS_PROTO(dir)->unlink_setup(&msg, dir);
task_setup_data.rpc_client = NFS_CLIENT(dir);
task = rpc_run_task(&task_setup_data);
if (!IS_ERR(task))
SUNRPC: Close a race in __rpc_wait_for_completion_task() Although they run as rpciod background tasks, under normal operation (i.e. no SIGKILL), functions like nfs_sillyrename(), nfs4_proc_unlck() and nfs4_do_close() want to be fully synchronous. This means that when we exit, we want all references to the rpc_task to be gone, and we want any dentry references etc. held by that task to be released. For this reason these functions call __rpc_wait_for_completion_task(), followed by rpc_put_task() in the expectation that the latter will be releasing the last reference to the rpc_task, and thus ensuring that the callback_ops->rpc_release() has been called synchronously. This patch fixes a race which exists due to the fact that rpciod calls rpc_complete_task() (in order to wake up the callers of __rpc_wait_for_completion_task()) and then subsequently calls rpc_put_task() without ensuring that these two steps are done atomically. In order to avoid adding new spin locks, the patch uses the existing waitqueue spin lock to order the rpc_task reference count releases between the waiting process and rpciod. The common case where nobody is waiting for completion is optimised for by checking if the RPC_TASK_ASYNC flag is cleared and/or if the rpc_task reference count is 1: in those cases we drop trying to grab the spin lock, and immediately free up the rpc_task. Those few processes that need to put the rpc_task from inside an asynchronous context and that do not care about ordering are given a new helper: rpc_put_task_async(). Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2011-02-22 02:05:41 +07:00
rpc_put_task_async(task);
return 1;
}
static int nfs_call_unlink(struct dentry *dentry, struct nfs_unlinkdata *data)
{
struct dentry *parent;
struct inode *dir;
int ret = 0;
parent = dget_parent(dentry);
if (parent == NULL)
goto out_free;
dir = parent->d_inode;
/* Non-exclusive lock protects against concurrent lookup() calls */
spin_lock(&dir->i_lock);
if (atomic_inc_not_zero(&NFS_I(dir)->silly_count) == 0) {
/* Deferred delete */
hlist_add_head(&data->list, &NFS_I(dir)->silly_list);
spin_unlock(&dir->i_lock);
ret = 1;
goto out_dput;
}
spin_unlock(&dir->i_lock);
ret = nfs_do_call_unlink(parent, dir, data);
out_dput:
dput(parent);
out_free:
return ret;
}
void nfs_block_sillyrename(struct dentry *dentry)
{
struct nfs_inode *nfsi = NFS_I(dentry->d_inode);
wait_event(nfsi->waitqueue, atomic_cmpxchg(&nfsi->silly_count, 1, 0) == 1);
}
void nfs_unblock_sillyrename(struct dentry *dentry)
{
struct inode *dir = dentry->d_inode;
struct nfs_inode *nfsi = NFS_I(dir);
struct nfs_unlinkdata *data;
atomic_inc(&nfsi->silly_count);
spin_lock(&dir->i_lock);
while (!hlist_empty(&nfsi->silly_list)) {
if (!atomic_inc_not_zero(&nfsi->silly_count))
break;
data = hlist_entry(nfsi->silly_list.first, struct nfs_unlinkdata, list);
hlist_del(&data->list);
spin_unlock(&dir->i_lock);
if (nfs_do_call_unlink(dentry, dir, data) == 0)
nfs_free_unlinkdata(data);
spin_lock(&dir->i_lock);
}
spin_unlock(&dir->i_lock);
}
/**
* nfs_async_unlink - asynchronous unlinking of a file
* @dir: parent directory of dentry
* @dentry: dentry to unlink
*/
static int
nfs_async_unlink(struct inode *dir, struct dentry *dentry)
{
struct nfs_unlinkdata *data;
int status = -ENOMEM;
void *devname_garbage = NULL;
data = kzalloc(sizeof(*data), GFP_KERNEL);
if (data == NULL)
goto out;
data->cred = rpc_lookup_cred();
if (IS_ERR(data->cred)) {
status = PTR_ERR(data->cred);
goto out_free;
}
data->res.dir_attr = &data->dir_attr;
status = -EBUSY;
spin_lock(&dentry->d_lock);
if (dentry->d_flags & DCACHE_NFSFS_RENAMED)
goto out_unlock;
dentry->d_flags |= DCACHE_NFSFS_RENAMED;
devname_garbage = dentry->d_fsdata;
dentry->d_fsdata = data;
spin_unlock(&dentry->d_lock);
/*
* If we'd displaced old cached devname, free it. At that
* point dentry is definitely not a root, so we won't need
* that anymore.
*/
kfree(devname_garbage);
return 0;
out_unlock:
spin_unlock(&dentry->d_lock);
put_rpccred(data->cred);
out_free:
kfree(data);
out:
return status;
}
/**
* nfs_complete_unlink - Initialize completion of the sillydelete
* @dentry: dentry to delete
* @inode: inode
*
* Since we're most likely to be called by dentry_iput(), we
* only use the dentry to find the sillydelete. We then copy the name
* into the qstr.
*/
void
nfs_complete_unlink(struct dentry *dentry, struct inode *inode)
{
struct nfs_unlinkdata *data = NULL;
spin_lock(&dentry->d_lock);
if (dentry->d_flags & DCACHE_NFSFS_RENAMED) {
dentry->d_flags &= ~DCACHE_NFSFS_RENAMED;
data = dentry->d_fsdata;
dentry->d_fsdata = NULL;
}
spin_unlock(&dentry->d_lock);
if (data != NULL && (NFS_STALE(inode) || !nfs_call_unlink(dentry, data)))
nfs_free_unlinkdata(data);
}
/* Cancel a queued async unlink. Called when a sillyrename run fails. */
static void
nfs_cancel_async_unlink(struct dentry *dentry)
{
spin_lock(&dentry->d_lock);
if (dentry->d_flags & DCACHE_NFSFS_RENAMED) {
struct nfs_unlinkdata *data = dentry->d_fsdata;
dentry->d_flags &= ~DCACHE_NFSFS_RENAMED;
dentry->d_fsdata = NULL;
spin_unlock(&dentry->d_lock);
nfs_free_unlinkdata(data);
return;
}
spin_unlock(&dentry->d_lock);
}
/**
* nfs_async_rename_done - Sillyrename post-processing
* @task: rpc_task of the sillyrename
* @calldata: nfs_renamedata for the sillyrename
*
* Do the directory attribute updates and the d_move
*/
static void nfs_async_rename_done(struct rpc_task *task, void *calldata)
{
struct nfs_renamedata *data = calldata;
struct inode *old_dir = data->old_dir;
struct inode *new_dir = data->new_dir;
struct dentry *old_dentry = data->old_dentry;
if (!NFS_PROTO(old_dir)->rename_done(task, old_dir, new_dir)) {
rpc_restart_call_prepare(task);
return;
}
NFS: Don't allow NFS silly-renamed files to be deleted, no signal Commit 73ca100 broke the code that prevents the client from deleting a silly renamed dentry. This affected "delete on last close" semantics as after that commit, nothing prevented removal of silly-renamed files. As a result, a process holding a file open could easily get an ESTALE on the file in a directory where some other process issued 'rm -rf some_dir_containing_the_file' twice. Before the commit, any attempt at unlinking silly renamed files would fail inside may_delete() with -EBUSY because of the DCACHE_NFSFS_RENAMED flag. The following testcase demonstrates the problem: tail -f /nfsmnt/dir/file & rm -rf /nfsmnt/dir rm -rf /nfsmnt/dir # second removal does not fail, 'tail' process receives ESTALE The problem with the above commit is that it unhashes the old and new dentries from the lookup path, even in the normal case when a signal is not encountered and it would have been safe to call d_move. Unfortunately the old dentry has the special DCACHE_NFSFS_RENAMED flag set on it. Unhashing has the side-effect that future lookups call d_alloc(), allocating a new dentry without the special flag for any silly-renamed files. As a result, subsequent calls to unlink silly renamed files do not fail but allow the removal to go through. This will result in ESTALE errors for any other process doing operations on the file. To fix this, go back to using d_move on success. For the signal case, it's unclear what we may safely do beyond d_drop. Reported-by: Dave Wysochanski <dwysocha@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Acked-by: Jeff Layton <jlayton@redhat.com> Cc: stable@vger.kernel.org
2013-02-23 00:53:43 +07:00
if (task->tk_status != 0)
nfs_cancel_async_unlink(old_dentry);
}
/**
* nfs_async_rename_release - Release the sillyrename data.
* @calldata: the struct nfs_renamedata to be released
*/
static void nfs_async_rename_release(void *calldata)
{
struct nfs_renamedata *data = calldata;
struct super_block *sb = data->old_dir->i_sb;
if (data->old_dentry->d_inode)
nfs_mark_for_revalidate(data->old_dentry->d_inode);
dput(data->old_dentry);
dput(data->new_dentry);
iput(data->old_dir);
iput(data->new_dir);
nfs_sb_deactive(sb);
put_rpccred(data->cred);
kfree(data);
}
static void nfs_rename_prepare(struct rpc_task *task, void *calldata)
{
struct nfs_renamedata *data = calldata;
NFS_PROTO(data->old_dir)->rename_rpc_prepare(task, data);
}
static const struct rpc_call_ops nfs_rename_ops = {
.rpc_call_done = nfs_async_rename_done,
.rpc_release = nfs_async_rename_release,
.rpc_call_prepare = nfs_rename_prepare,
};
/**
* nfs_async_rename - perform an asynchronous rename operation
* @old_dir: directory that currently holds the dentry to be renamed
* @new_dir: target directory for the rename
* @old_dentry: original dentry to be renamed
* @new_dentry: dentry to which the old_dentry should be renamed
*
* It's expected that valid references to the dentries and inodes are held
*/
static struct rpc_task *
nfs_async_rename(struct inode *old_dir, struct inode *new_dir,
struct dentry *old_dentry, struct dentry *new_dentry)
{
struct nfs_renamedata *data;
struct rpc_message msg = { };
struct rpc_task_setup task_setup_data = {
.rpc_message = &msg,
.callback_ops = &nfs_rename_ops,
.workqueue = nfsiod_workqueue,
.rpc_client = NFS_CLIENT(old_dir),
.flags = RPC_TASK_ASYNC,
};
data = kzalloc(sizeof(*data), GFP_KERNEL);
if (data == NULL)
return ERR_PTR(-ENOMEM);
task_setup_data.callback_data = data;
data->cred = rpc_lookup_cred();
if (IS_ERR(data->cred)) {
struct rpc_task *task = ERR_CAST(data->cred);
kfree(data);
return task;
}
msg.rpc_argp = &data->args;
msg.rpc_resp = &data->res;
msg.rpc_cred = data->cred;
/* set up nfs_renamedata */
data->old_dir = old_dir;
ihold(old_dir);
data->new_dir = new_dir;
ihold(new_dir);
data->old_dentry = dget(old_dentry);
data->new_dentry = dget(new_dentry);
nfs_fattr_init(&data->old_fattr);
nfs_fattr_init(&data->new_fattr);
/* set up nfs_renameargs */
data->args.old_dir = NFS_FH(old_dir);
data->args.old_name = &old_dentry->d_name;
data->args.new_dir = NFS_FH(new_dir);
data->args.new_name = &new_dentry->d_name;
/* set up nfs_renameres */
data->res.old_fattr = &data->old_fattr;
data->res.new_fattr = &data->new_fattr;
nfs_sb_active(old_dir->i_sb);
NFS_PROTO(data->old_dir)->rename_setup(&msg, old_dir);
return rpc_run_task(&task_setup_data);
}
/**
* nfs_sillyrename - Perform a silly-rename of a dentry
* @dir: inode of directory that contains dentry
* @dentry: dentry to be sillyrenamed
*
* NFSv2/3 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.
*
* (Note: NFSv4 is stateful, and has opens, so in theory an NFSv4 server
* could take responsibility for keeping open files referenced. The server
* would also need to ensure that opened-but-deleted files were kept over
* reboots. However, we may not assume a server does so. (RFC 5661
* does provide an OPEN4_RESULT_PRESERVE_UNLINKED flag that a server can
* use to advertise that it does this; some day we may take advantage of
* it.))
*/
int
nfs_sillyrename(struct inode *dir, struct dentry *dentry)
{
static unsigned int sillycounter;
const int fileidsize = sizeof(NFS_FILEID(dentry->d_inode))*2;
const int countersize = sizeof(sillycounter)*2;
const int slen = sizeof(".nfs")+fileidsize+countersize-1;
char silly[slen+1];
struct dentry *sdentry;
struct rpc_task *task;
int error = -EIO;
dfprintk(VFS, "NFS: silly-rename(%s/%s, ct=%d)\n",
dentry->d_parent->d_name.name, dentry->d_name.name,
dentry->d_count);
nfs_inc_stats(dir, NFSIOS_SILLYRENAME);
/*
* We don't allow a dentry to be silly-renamed twice.
*/
error = -EBUSY;
if (dentry->d_flags & DCACHE_NFSFS_RENAMED)
goto out;
sprintf(silly, ".nfs%*.*Lx",
fileidsize, fileidsize,
(unsigned long long)NFS_FILEID(dentry->d_inode));
/* Return delegation in anticipation of the rename */
NFS_PROTO(dentry->d_inode)->return_delegation(dentry->d_inode);
sdentry = NULL;
do {
char *suffix = silly + slen - countersize;
dput(sdentry);
sillycounter++;
sprintf(suffix, "%*.*x", countersize, countersize, sillycounter);
dfprintk(VFS, "NFS: trying to rename %s to %s\n",
dentry->d_name.name, silly);
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 (sdentry->d_inode != NULL); /* need negative lookup */
/* queue unlink first. Can't do this from rpc_release as it
* has to allocate memory
*/
error = nfs_async_unlink(dir, dentry);
if (error)
goto out_dput;
/* populate unlinkdata with the right dname */
error = nfs_copy_dname(sdentry,
(struct nfs_unlinkdata *)dentry->d_fsdata);
if (error) {
nfs_cancel_async_unlink(dentry);
goto out_dput;
}
/* run the rename task, undo unlink if it fails */
task = nfs_async_rename(dir, dir, dentry, sdentry);
if (IS_ERR(task)) {
error = -EBUSY;
nfs_cancel_async_unlink(dentry);
goto out_dput;
}
/* wait for the RPC task to complete, unless a SIGKILL intervenes */
error = rpc_wait_for_completion_task(task);
if (error == 0)
error = task->tk_status;
NFS: Don't allow NFS silly-renamed files to be deleted, no signal Commit 73ca100 broke the code that prevents the client from deleting a silly renamed dentry. This affected "delete on last close" semantics as after that commit, nothing prevented removal of silly-renamed files. As a result, a process holding a file open could easily get an ESTALE on the file in a directory where some other process issued 'rm -rf some_dir_containing_the_file' twice. Before the commit, any attempt at unlinking silly renamed files would fail inside may_delete() with -EBUSY because of the DCACHE_NFSFS_RENAMED flag. The following testcase demonstrates the problem: tail -f /nfsmnt/dir/file & rm -rf /nfsmnt/dir rm -rf /nfsmnt/dir # second removal does not fail, 'tail' process receives ESTALE The problem with the above commit is that it unhashes the old and new dentries from the lookup path, even in the normal case when a signal is not encountered and it would have been safe to call d_move. Unfortunately the old dentry has the special DCACHE_NFSFS_RENAMED flag set on it. Unhashing has the side-effect that future lookups call d_alloc(), allocating a new dentry without the special flag for any silly-renamed files. As a result, subsequent calls to unlink silly renamed files do not fail but allow the removal to go through. This will result in ESTALE errors for any other process doing operations on the file. To fix this, go back to using d_move on success. For the signal case, it's unclear what we may safely do beyond d_drop. Reported-by: Dave Wysochanski <dwysocha@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Acked-by: Jeff Layton <jlayton@redhat.com> Cc: stable@vger.kernel.org
2013-02-23 00:53:43 +07:00
switch (error) {
case 0:
/* The rename succeeded */
nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
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);
}
rpc_put_task(task);
out_dput:
dput(sdentry);
out:
return error;
}