linux_dsm_epyc7002/fs/autofs4/root.c

1106 lines
28 KiB
C
Raw Normal View History

/* -*- c -*- --------------------------------------------------------------- *
*
* linux/fs/autofs/root.c
*
* Copyright 1997-1998 Transmeta Corporation -- All Rights Reserved
* Copyright 1999-2000 Jeremy Fitzhardinge <jeremy@goop.org>
* Copyright 2001-2006 Ian Kent <raven@themaw.net>
*
* This file is part of the Linux kernel and is made available under
* the terms of the GNU General Public License, version 2, or at your
* option, any later version, incorporated herein by reference.
*
* ------------------------------------------------------------------------- */
#include <linux/capability.h>
#include <linux/errno.h>
#include <linux/stat.h>
#include <linux/param.h>
#include <linux/time.h>
#include "autofs_i.h"
static int autofs4_dir_symlink(struct inode *,struct dentry *,const char *);
static int autofs4_dir_unlink(struct inode *,struct dentry *);
static int autofs4_dir_rmdir(struct inode *,struct dentry *);
static int autofs4_dir_mkdir(struct inode *,struct dentry *,int);
static int autofs4_root_ioctl(struct inode *, struct file *,unsigned int,unsigned long);
static int autofs4_dir_open(struct inode *inode, struct file *file);
static int autofs4_dir_close(struct inode *inode, struct file *file);
static int autofs4_dir_readdir(struct file * filp, void * dirent, filldir_t filldir);
static int autofs4_root_readdir(struct file * filp, void * dirent, filldir_t filldir);
static struct dentry *autofs4_lookup(struct inode *,struct dentry *, struct nameidata *);
static void *autofs4_follow_link(struct dentry *, struct nameidata *);
const struct file_operations autofs4_root_operations = {
.open = dcache_dir_open,
.release = dcache_dir_close,
.read = generic_read_dir,
.readdir = autofs4_root_readdir,
.ioctl = autofs4_root_ioctl,
};
const struct file_operations autofs4_dir_operations = {
.open = autofs4_dir_open,
.release = autofs4_dir_close,
.read = generic_read_dir,
.readdir = autofs4_dir_readdir,
};
const struct inode_operations autofs4_indirect_root_inode_operations = {
.lookup = autofs4_lookup,
.unlink = autofs4_dir_unlink,
.symlink = autofs4_dir_symlink,
.mkdir = autofs4_dir_mkdir,
.rmdir = autofs4_dir_rmdir,
};
const struct inode_operations autofs4_direct_root_inode_operations = {
.lookup = autofs4_lookup,
.unlink = autofs4_dir_unlink,
.mkdir = autofs4_dir_mkdir,
.rmdir = autofs4_dir_rmdir,
.follow_link = autofs4_follow_link,
};
const struct inode_operations autofs4_dir_inode_operations = {
.lookup = autofs4_lookup,
.unlink = autofs4_dir_unlink,
.symlink = autofs4_dir_symlink,
.mkdir = autofs4_dir_mkdir,
.rmdir = autofs4_dir_rmdir,
};
static int autofs4_root_readdir(struct file *file, void *dirent,
filldir_t filldir)
{
struct autofs_sb_info *sbi = autofs4_sbi(file->f_path.dentry->d_sb);
int oz_mode = autofs4_oz_mode(sbi);
DPRINTK("called, filp->f_pos = %lld", file->f_pos);
/*
* Don't set reghost flag if:
* 1) f_pos is larger than zero -- we've already been here.
* 2) we haven't even enabled reghosting in the 1st place.
* 3) this is the daemon doing a readdir
*/
if (oz_mode && file->f_pos == 0 && sbi->reghost_enabled)
sbi->needs_reghost = 1;
DPRINTK("needs_reghost = %d", sbi->needs_reghost);
return dcache_readdir(file, dirent, filldir);
}
static int autofs4_dir_open(struct inode *inode, struct file *file)
{
struct dentry *dentry = file->f_path.dentry;
struct vfsmount *mnt = file->f_path.mnt;
struct autofs_sb_info *sbi = autofs4_sbi(dentry->d_sb);
struct dentry *cursor;
int status;
status = dcache_dir_open(inode, file);
if (status)
goto out;
cursor = file->private_data;
cursor->d_fsdata = NULL;
DPRINTK("file=%p dentry=%p %.*s",
file, dentry, dentry->d_name.len, dentry->d_name.name);
if (autofs4_oz_mode(sbi))
goto out;
if (autofs4_ispending(dentry)) {
DPRINTK("dentry busy");
dcache_dir_close(inode, file);
status = -EBUSY;
goto out;
}
status = -ENOENT;
if (!d_mountpoint(dentry) && dentry->d_op && dentry->d_op->d_revalidate) {
struct nameidata nd;
int empty, ret;
/* In case there are stale directory dentrys from a failed mount */
spin_lock(&dcache_lock);
empty = list_empty(&dentry->d_subdirs);
spin_unlock(&dcache_lock);
if (!empty)
d_invalidate(dentry);
nd.flags = LOOKUP_DIRECTORY;
ret = (dentry->d_op->d_revalidate)(dentry, &nd);
if (ret <= 0) {
if (ret < 0)
status = ret;
dcache_dir_close(inode, file);
goto out;
}
}
if (d_mountpoint(dentry)) {
struct file *fp = NULL;
struct path fp_path = { .dentry = dentry, .mnt = mnt };
path_get(&fp_path);
if (!autofs4_follow_mount(&fp_path.mnt, &fp_path.dentry)) {
path_put(&fp_path);
dcache_dir_close(inode, file);
goto out;
}
fp = dentry_open(fp_path.dentry, fp_path.mnt, file->f_flags);
status = PTR_ERR(fp);
if (IS_ERR(fp)) {
dcache_dir_close(inode, file);
goto out;
}
cursor->d_fsdata = fp;
}
return 0;
out:
return status;
}
static int autofs4_dir_close(struct inode *inode, struct file *file)
{
struct dentry *dentry = file->f_path.dentry;
struct autofs_sb_info *sbi = autofs4_sbi(dentry->d_sb);
struct dentry *cursor = file->private_data;
int status = 0;
DPRINTK("file=%p dentry=%p %.*s",
file, dentry, dentry->d_name.len, dentry->d_name.name);
if (autofs4_oz_mode(sbi))
goto out;
if (autofs4_ispending(dentry)) {
DPRINTK("dentry busy");
status = -EBUSY;
goto out;
}
if (d_mountpoint(dentry)) {
struct file *fp = cursor->d_fsdata;
if (!fp) {
status = -ENOENT;
goto out;
}
filp_close(fp, current->files);
}
out:
dcache_dir_close(inode, file);
return status;
}
static int autofs4_dir_readdir(struct file *file, void *dirent, filldir_t filldir)
{
struct dentry *dentry = file->f_path.dentry;
struct autofs_sb_info *sbi = autofs4_sbi(dentry->d_sb);
struct dentry *cursor = file->private_data;
int status;
DPRINTK("file=%p dentry=%p %.*s",
file, dentry, dentry->d_name.len, dentry->d_name.name);
if (autofs4_oz_mode(sbi))
goto out;
if (autofs4_ispending(dentry)) {
DPRINTK("dentry busy");
return -EBUSY;
}
if (d_mountpoint(dentry)) {
struct file *fp = cursor->d_fsdata;
if (!fp)
return -ENOENT;
if (!fp->f_op || !fp->f_op->readdir)
goto out;
status = vfs_readdir(fp, filldir, dirent);
file->f_pos = fp->f_pos;
if (status)
autofs4_copy_atime(file, fp);
return status;
}
out:
return dcache_readdir(file, dirent, filldir);
}
static int try_to_fill_dentry(struct dentry *dentry, int flags)
{
struct autofs_sb_info *sbi = autofs4_sbi(dentry->d_sb);
struct autofs_info *ino = autofs4_dentry_ino(dentry);
int status;
/* Block on any pending expiry here; invalidate the dentry
when expiration is done to trigger mount request with a new
dentry */
if (ino && (ino->flags & AUTOFS_INF_EXPIRING)) {
DPRINTK("waiting for expire %p name=%.*s",
dentry, dentry->d_name.len, dentry->d_name.name);
status = autofs4_wait(sbi, dentry, NFY_NONE);
DPRINTK("expire done status=%d", status);
/*
* If the directory still exists the mount request must
* continue otherwise it can't be followed at the right
* time during the walk.
*/
status = d_invalidate(dentry);
if (status != -EBUSY)
return -EAGAIN;
}
DPRINTK("dentry=%p %.*s ino=%p",
dentry, dentry->d_name.len, dentry->d_name.name, dentry->d_inode);
/*
* Wait for a pending mount, triggering one if there
* isn't one already
*/
if (dentry->d_inode == NULL) {
DPRINTK("waiting for mount name=%.*s",
dentry->d_name.len, dentry->d_name.name);
status = autofs4_wait(sbi, dentry, NFY_MOUNT);
DPRINTK("mount done status=%d", status);
/* Turn this into a real negative dentry? */
if (status == -ENOENT) {
spin_lock(&dentry->d_lock);
dentry->d_flags &= ~DCACHE_AUTOFS_PENDING;
spin_unlock(&dentry->d_lock);
return status;
} else if (status) {
/* Return a negative dentry, but leave it "pending" */
return status;
}
/* Trigger mount for path component or follow link */
} else if (flags & (LOOKUP_CONTINUE | LOOKUP_DIRECTORY) ||
current->link_count) {
DPRINTK("waiting for mount name=%.*s",
dentry->d_name.len, dentry->d_name.name);
spin_lock(&dentry->d_lock);
dentry->d_flags |= DCACHE_AUTOFS_PENDING;
spin_unlock(&dentry->d_lock);
status = autofs4_wait(sbi, dentry, NFY_MOUNT);
DPRINTK("mount done status=%d", status);
if (status) {
spin_lock(&dentry->d_lock);
dentry->d_flags &= ~DCACHE_AUTOFS_PENDING;
spin_unlock(&dentry->d_lock);
return status;
}
}
/* Initialize expiry counter after successful mount */
if (ino)
ino->last_used = jiffies;
spin_lock(&dentry->d_lock);
dentry->d_flags &= ~DCACHE_AUTOFS_PENDING;
spin_unlock(&dentry->d_lock);
autofs4: fix execution order race in mount request code Jeff Moyer has identified a race in due to an execution order dependency in the autofs4 function root.c:try_to_fill_dentry(). Jeff's description of this race is: "P1 does a lookup of /mount/submount/foo. Since the VFS can't find an entry for "foo" under /mount/submount, it calls into the autofs4 kernel module to allocate a new dentry, D1. The kernel creates a new waitq for this lookup and calls the daemon to perform the mount. The daemon performs a mkdir of the "foo" directory under /mount/submount, which ends up creating a *new* dentry, D2. Then, P2 does a lookup of /mount/submount/foo. The VFS path walking logic finds a dentry in the dcache, D2, and calls the revalidate function with this. In the autofs4 revalidate code, we then trigger a mount, since the dentry is an empty directory that isn't a mountpoint, and so set DCACHE_AUTOFS_PENDING and call into the wait code to trigger the mount. The wait code finds our existing waitq entry (since it is keyed off of the directory name) and adds itself to the list of waiters. After the daemon finishes the mount, it calls back into the kernel to release the waiters. When this happens, P1 is woken up and goes about clearing the DCACHE_AUTOFS_PENDING flag, but it does this in D1! So, given that P1 in our case is a program that will immediately try to access a file under /mount/submount/foo, we end up finding the dentry D2 which still has the pending flag set, and we set out to wait for a mount *again*! So, one way to address this is to re-do the lookup at the end of try_to_fill_dentry, and to clear the pending flag on the hashed dentry. This seems a sane approach to me." And Jeff's patch does this. Signed-off-by: Jeff Moyer <jmoyer@redhat.com> Signed-off-by-by: Ian Kent <raven@themaw.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-05-01 18:35:08 +07:00
return 0;
}
/* For autofs direct mounts the follow link triggers the mount */
static void *autofs4_follow_link(struct dentry *dentry, struct nameidata *nd)
{
struct autofs_sb_info *sbi = autofs4_sbi(dentry->d_sb);
struct autofs_info *ino = autofs4_dentry_ino(dentry);
int oz_mode = autofs4_oz_mode(sbi);
unsigned int lookup_type;
int status;
DPRINTK("dentry=%p %.*s oz_mode=%d nd->flags=%d",
dentry, dentry->d_name.len, dentry->d_name.name, oz_mode,
nd->flags);
/* If it's our master or we shouldn't trigger a mount we're done */
lookup_type = nd->flags & (LOOKUP_CONTINUE | LOOKUP_DIRECTORY);
if (oz_mode || !lookup_type)
goto done;
/* If an expire request is pending wait for it. */
if (ino && (ino->flags & AUTOFS_INF_EXPIRING)) {
DPRINTK("waiting for active request %p name=%.*s",
dentry, dentry->d_name.len, dentry->d_name.name);
status = autofs4_wait(sbi, dentry, NFY_NONE);
DPRINTK("request done status=%d", status);
}
/*
* If the dentry contains directories then it is an
* autofs multi-mount with no root mount offset. So
* don't try to mount it again.
*/
spin_lock(&dcache_lock);
if (!d_mountpoint(dentry) && __simple_empty(dentry)) {
spin_unlock(&dcache_lock);
status = try_to_fill_dentry(dentry, 0);
if (status)
goto out_error;
/*
* The mount succeeded but if there is no root mount
* it must be an autofs multi-mount with no root offset
* so we don't need to follow the mount.
*/
if (d_mountpoint(dentry)) {
if (!autofs4_follow_mount(&nd->path.mnt,
&nd->path.dentry)) {
status = -ENOENT;
goto out_error;
}
}
goto done;
}
spin_unlock(&dcache_lock);
done:
return NULL;
out_error:
path_put(&nd->path);
return ERR_PTR(status);
}
/*
* Revalidate is called on every cache lookup. Some of those
* cache lookups may actually happen while the dentry is not
* yet completely filled in, and revalidate has to delay such
* lookups..
*/
static int autofs4_revalidate(struct dentry *dentry, struct nameidata *nd)
{
struct inode *dir = dentry->d_parent->d_inode;
struct autofs_sb_info *sbi = autofs4_sbi(dir->i_sb);
int oz_mode = autofs4_oz_mode(sbi);
int flags = nd ? nd->flags : 0;
int status = 1;
/* Pending dentry */
if (autofs4_ispending(dentry)) {
/* The daemon never causes a mount to trigger */
if (oz_mode)
return 1;
/*
* A zero status is success otherwise we have a
* negative error code.
*/
status = try_to_fill_dentry(dentry, flags);
if (status == 0)
return 1;
/*
* A status of EAGAIN here means that the dentry has gone
* away while waiting for an expire to complete. If we are
* racing with expire lookup will wait for it so this must
* be a revalidate and we need to send it to lookup.
*/
if (status == -EAGAIN)
return 0;
return status;
}
/* Negative dentry.. invalidate if "old" */
if (dentry->d_inode == NULL)
return 0;
/* Check for a non-mountpoint directory with no contents */
spin_lock(&dcache_lock);
if (S_ISDIR(dentry->d_inode->i_mode) &&
!d_mountpoint(dentry) &&
__simple_empty(dentry)) {
DPRINTK("dentry=%p %.*s, emptydir",
dentry, dentry->d_name.len, dentry->d_name.name);
spin_unlock(&dcache_lock);
/* The daemon never causes a mount to trigger */
if (oz_mode)
return 1;
/*
* A zero status is success otherwise we have a
* negative error code.
*/
status = try_to_fill_dentry(dentry, flags);
if (status == 0)
return 1;
return status;
}
spin_unlock(&dcache_lock);
return 1;
}
void autofs4_dentry_release(struct dentry *de)
{
struct autofs_info *inf;
DPRINTK("releasing %p", de);
inf = autofs4_dentry_ino(de);
de->d_fsdata = NULL;
if (inf) {
struct autofs_sb_info *sbi = autofs4_sbi(de->d_sb);
if (sbi) {
spin_lock(&sbi->lookup_lock);
autofs4: use look aside list for lookups A while ago a patch to resolve a deadlock during directory creation was merged. This delayed the hashing of lookup dentrys until the ->mkdir() (or ->symlink()) operation completed to ensure we always went through ->lookup() instead of also having processes go through ->revalidate() so our VFS locking remained consistent. Now we are seeing a couple of side affects of that change in situations with heavy mount activity. Two cases have been identified: 1) When a mount request is triggered, due to the delayed hashing, the directory created by user space for the mount point doesn't have the DCACHE_AUTOFS_PENDING flag set. In the case of an autofs multi-mount where a tree of mount point directories are created this can lead to the path walk continuing rather than the dentry being sent to the wait queue to wait for request completion. This is because, if the pending flag isn't set, the criteria for deciding this is a mount in progress fails to hold, namely that the dentry is not a mount point and has no subdirectories. 2) A mount request dentry is initially created negative and unhashed. It remains this way until the ->mkdir() callback completes. Since it is unhashed a fresh dentry is used when the user space mount request creates the mount point directory. This leaves the original dentry negative and unhashed. But revalidate has no way to tell the VFS that the dentry has changed, other than to force another ->lookup() by returning false, which is at best wastefull and at worst not possible. This results in an -ENOENT return from the original path walk when in fact the mount succeeded. To resolve this we need to ensure that the same dentry is used in all calls to ->lookup() during the course of a mount request. This patch achieves that by adding the initial dentry to a look aside list and removes it at ->mkdir() or ->symlink() completion (or when the dentry is released), since these are the only create operations autofs4 supports. Signed-off-by: Ian Kent <raven@themaw.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-24 11:30:12 +07:00
if (!list_empty(&inf->active))
list_del(&inf->active);
if (!list_empty(&inf->expiring))
list_del(&inf->expiring);
spin_unlock(&sbi->lookup_lock);
}
inf->dentry = NULL;
inf->inode = NULL;
autofs4_free_ino(inf);
}
}
/* For dentries of directories in the root dir */
static struct dentry_operations autofs4_root_dentry_operations = {
.d_revalidate = autofs4_revalidate,
.d_release = autofs4_dentry_release,
};
/* For other dentries */
static struct dentry_operations autofs4_dentry_operations = {
.d_revalidate = autofs4_revalidate,
.d_release = autofs4_dentry_release,
};
autofs4: use look aside list for lookups A while ago a patch to resolve a deadlock during directory creation was merged. This delayed the hashing of lookup dentrys until the ->mkdir() (or ->symlink()) operation completed to ensure we always went through ->lookup() instead of also having processes go through ->revalidate() so our VFS locking remained consistent. Now we are seeing a couple of side affects of that change in situations with heavy mount activity. Two cases have been identified: 1) When a mount request is triggered, due to the delayed hashing, the directory created by user space for the mount point doesn't have the DCACHE_AUTOFS_PENDING flag set. In the case of an autofs multi-mount where a tree of mount point directories are created this can lead to the path walk continuing rather than the dentry being sent to the wait queue to wait for request completion. This is because, if the pending flag isn't set, the criteria for deciding this is a mount in progress fails to hold, namely that the dentry is not a mount point and has no subdirectories. 2) A mount request dentry is initially created negative and unhashed. It remains this way until the ->mkdir() callback completes. Since it is unhashed a fresh dentry is used when the user space mount request creates the mount point directory. This leaves the original dentry negative and unhashed. But revalidate has no way to tell the VFS that the dentry has changed, other than to force another ->lookup() by returning false, which is at best wastefull and at worst not possible. This results in an -ENOENT return from the original path walk when in fact the mount succeeded. To resolve this we need to ensure that the same dentry is used in all calls to ->lookup() during the course of a mount request. This patch achieves that by adding the initial dentry to a look aside list and removes it at ->mkdir() or ->symlink() completion (or when the dentry is released), since these are the only create operations autofs4 supports. Signed-off-by: Ian Kent <raven@themaw.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-24 11:30:12 +07:00
static struct dentry *autofs4_lookup_active(struct autofs_sb_info *sbi, struct dentry *parent, struct qstr *name)
{
unsigned int len = name->len;
unsigned int hash = name->hash;
const unsigned char *str = name->name;
struct list_head *p, *head;
spin_lock(&dcache_lock);
spin_lock(&sbi->lookup_lock);
head = &sbi->active_list;
list_for_each(p, head) {
struct autofs_info *ino;
struct dentry *dentry;
struct qstr *qstr;
ino = list_entry(p, struct autofs_info, active);
dentry = ino->dentry;
spin_lock(&dentry->d_lock);
/* Already gone? */
if (atomic_read(&dentry->d_count) == 0)
goto next;
qstr = &dentry->d_name;
if (dentry->d_name.hash != hash)
goto next;
if (dentry->d_parent != parent)
goto next;
if (qstr->len != len)
goto next;
if (memcmp(qstr->name, str, len))
goto next;
if (d_unhashed(dentry)) {
dget(dentry);
spin_unlock(&dentry->d_lock);
spin_unlock(&sbi->lookup_lock);
spin_unlock(&dcache_lock);
return dentry;
}
next:
spin_unlock(&dentry->d_lock);
}
spin_unlock(&sbi->lookup_lock);
spin_unlock(&dcache_lock);
return NULL;
}
static struct dentry *autofs4_lookup_expiring(struct autofs_sb_info *sbi, struct dentry *parent, struct qstr *name)
{
unsigned int len = name->len;
unsigned int hash = name->hash;
const unsigned char *str = name->name;
struct list_head *p, *head;
spin_lock(&dcache_lock);
spin_lock(&sbi->lookup_lock);
head = &sbi->expiring_list;
list_for_each(p, head) {
struct autofs_info *ino;
struct dentry *dentry;
struct qstr *qstr;
ino = list_entry(p, struct autofs_info, expiring);
dentry = ino->dentry;
spin_lock(&dentry->d_lock);
/* Bad luck, we've already been dentry_iput */
if (!dentry->d_inode)
goto next;
qstr = &dentry->d_name;
if (dentry->d_name.hash != hash)
goto next;
if (dentry->d_parent != parent)
goto next;
if (qstr->len != len)
goto next;
if (memcmp(qstr->name, str, len))
goto next;
if (d_unhashed(dentry)) {
dget(dentry);
spin_unlock(&dentry->d_lock);
spin_unlock(&sbi->lookup_lock);
spin_unlock(&dcache_lock);
return dentry;
}
next:
spin_unlock(&dentry->d_lock);
}
spin_unlock(&sbi->lookup_lock);
spin_unlock(&dcache_lock);
return NULL;
}
/* Lookups in the root directory */
static struct dentry *autofs4_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
{
struct autofs_sb_info *sbi;
autofs4: use look aside list for lookups A while ago a patch to resolve a deadlock during directory creation was merged. This delayed the hashing of lookup dentrys until the ->mkdir() (or ->symlink()) operation completed to ensure we always went through ->lookup() instead of also having processes go through ->revalidate() so our VFS locking remained consistent. Now we are seeing a couple of side affects of that change in situations with heavy mount activity. Two cases have been identified: 1) When a mount request is triggered, due to the delayed hashing, the directory created by user space for the mount point doesn't have the DCACHE_AUTOFS_PENDING flag set. In the case of an autofs multi-mount where a tree of mount point directories are created this can lead to the path walk continuing rather than the dentry being sent to the wait queue to wait for request completion. This is because, if the pending flag isn't set, the criteria for deciding this is a mount in progress fails to hold, namely that the dentry is not a mount point and has no subdirectories. 2) A mount request dentry is initially created negative and unhashed. It remains this way until the ->mkdir() callback completes. Since it is unhashed a fresh dentry is used when the user space mount request creates the mount point directory. This leaves the original dentry negative and unhashed. But revalidate has no way to tell the VFS that the dentry has changed, other than to force another ->lookup() by returning false, which is at best wastefull and at worst not possible. This results in an -ENOENT return from the original path walk when in fact the mount succeeded. To resolve this we need to ensure that the same dentry is used in all calls to ->lookup() during the course of a mount request. This patch achieves that by adding the initial dentry to a look aside list and removes it at ->mkdir() or ->symlink() completion (or when the dentry is released), since these are the only create operations autofs4 supports. Signed-off-by: Ian Kent <raven@themaw.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-24 11:30:12 +07:00
struct autofs_info *ino;
struct dentry *expiring, *unhashed;
int oz_mode;
DPRINTK("name = %.*s",
dentry->d_name.len, dentry->d_name.name);
/* File name too long to exist */
if (dentry->d_name.len > NAME_MAX)
return ERR_PTR(-ENAMETOOLONG);
sbi = autofs4_sbi(dir->i_sb);
oz_mode = autofs4_oz_mode(sbi);
DPRINTK("pid = %u, pgrp = %u, catatonic = %d, oz_mode = %d",
current->pid, task_pgrp_nr(current), sbi->catatonic, oz_mode);
expiring = autofs4_lookup_expiring(sbi, dentry->d_parent, &dentry->d_name);
if (expiring) {
/*
* If we are racing with expire the request might not
* be quite complete but the directory has been removed
* so it must have been successful, so just wait for it.
*/
autofs4: use look aside list for lookups A while ago a patch to resolve a deadlock during directory creation was merged. This delayed the hashing of lookup dentrys until the ->mkdir() (or ->symlink()) operation completed to ensure we always went through ->lookup() instead of also having processes go through ->revalidate() so our VFS locking remained consistent. Now we are seeing a couple of side affects of that change in situations with heavy mount activity. Two cases have been identified: 1) When a mount request is triggered, due to the delayed hashing, the directory created by user space for the mount point doesn't have the DCACHE_AUTOFS_PENDING flag set. In the case of an autofs multi-mount where a tree of mount point directories are created this can lead to the path walk continuing rather than the dentry being sent to the wait queue to wait for request completion. This is because, if the pending flag isn't set, the criteria for deciding this is a mount in progress fails to hold, namely that the dentry is not a mount point and has no subdirectories. 2) A mount request dentry is initially created negative and unhashed. It remains this way until the ->mkdir() callback completes. Since it is unhashed a fresh dentry is used when the user space mount request creates the mount point directory. This leaves the original dentry negative and unhashed. But revalidate has no way to tell the VFS that the dentry has changed, other than to force another ->lookup() by returning false, which is at best wastefull and at worst not possible. This results in an -ENOENT return from the original path walk when in fact the mount succeeded. To resolve this we need to ensure that the same dentry is used in all calls to ->lookup() during the course of a mount request. This patch achieves that by adding the initial dentry to a look aside list and removes it at ->mkdir() or ->symlink() completion (or when the dentry is released), since these are the only create operations autofs4 supports. Signed-off-by: Ian Kent <raven@themaw.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-24 11:30:12 +07:00
ino = autofs4_dentry_ino(expiring);
while (ino && (ino->flags & AUTOFS_INF_EXPIRING)) {
DPRINTK("wait for incomplete expire %p name=%.*s",
expiring, expiring->d_name.len,
expiring->d_name.name);
autofs4_wait(sbi, expiring, NFY_NONE);
DPRINTK("request completed");
}
spin_lock(&sbi->lookup_lock);
if (!list_empty(&ino->expiring))
list_del_init(&ino->expiring);
spin_unlock(&sbi->lookup_lock);
dput(expiring);
}
autofs4: use look aside list for lookups A while ago a patch to resolve a deadlock during directory creation was merged. This delayed the hashing of lookup dentrys until the ->mkdir() (or ->symlink()) operation completed to ensure we always went through ->lookup() instead of also having processes go through ->revalidate() so our VFS locking remained consistent. Now we are seeing a couple of side affects of that change in situations with heavy mount activity. Two cases have been identified: 1) When a mount request is triggered, due to the delayed hashing, the directory created by user space for the mount point doesn't have the DCACHE_AUTOFS_PENDING flag set. In the case of an autofs multi-mount where a tree of mount point directories are created this can lead to the path walk continuing rather than the dentry being sent to the wait queue to wait for request completion. This is because, if the pending flag isn't set, the criteria for deciding this is a mount in progress fails to hold, namely that the dentry is not a mount point and has no subdirectories. 2) A mount request dentry is initially created negative and unhashed. It remains this way until the ->mkdir() callback completes. Since it is unhashed a fresh dentry is used when the user space mount request creates the mount point directory. This leaves the original dentry negative and unhashed. But revalidate has no way to tell the VFS that the dentry has changed, other than to force another ->lookup() by returning false, which is at best wastefull and at worst not possible. This results in an -ENOENT return from the original path walk when in fact the mount succeeded. To resolve this we need to ensure that the same dentry is used in all calls to ->lookup() during the course of a mount request. This patch achieves that by adding the initial dentry to a look aside list and removes it at ->mkdir() or ->symlink() completion (or when the dentry is released), since these are the only create operations autofs4 supports. Signed-off-by: Ian Kent <raven@themaw.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-24 11:30:12 +07:00
unhashed = autofs4_lookup_active(sbi, dentry->d_parent, &dentry->d_name);
if (unhashed)
dentry = unhashed;
else {
/*
* Mark the dentry incomplete but don't hash it. We do this
* to serialize our inode creation operations (symlink and
* mkdir) which prevents deadlock during the callback to
* the daemon. Subsequent user space lookups for the same
* dentry are placed on the wait queue while the daemon
* itself is allowed passage unresticted so the create
* operation itself can then hash the dentry. Finally,
* we check for the hashed dentry and return the newly
* hashed dentry.
*/
dentry->d_op = &autofs4_root_dentry_operations;
/*
* And we need to ensure that the same dentry is used for
* all following lookup calls until it is hashed so that
* the dentry flags are persistent throughout the request.
*/
ino = autofs4_init_ino(NULL, sbi, 0555);
if (!ino)
return ERR_PTR(-ENOMEM);
dentry->d_fsdata = ino;
ino->dentry = dentry;
spin_lock(&sbi->lookup_lock);
list_add(&ino->active, &sbi->active_list);
spin_unlock(&sbi->lookup_lock);
autofs4: use look aside list for lookups A while ago a patch to resolve a deadlock during directory creation was merged. This delayed the hashing of lookup dentrys until the ->mkdir() (or ->symlink()) operation completed to ensure we always went through ->lookup() instead of also having processes go through ->revalidate() so our VFS locking remained consistent. Now we are seeing a couple of side affects of that change in situations with heavy mount activity. Two cases have been identified: 1) When a mount request is triggered, due to the delayed hashing, the directory created by user space for the mount point doesn't have the DCACHE_AUTOFS_PENDING flag set. In the case of an autofs multi-mount where a tree of mount point directories are created this can lead to the path walk continuing rather than the dentry being sent to the wait queue to wait for request completion. This is because, if the pending flag isn't set, the criteria for deciding this is a mount in progress fails to hold, namely that the dentry is not a mount point and has no subdirectories. 2) A mount request dentry is initially created negative and unhashed. It remains this way until the ->mkdir() callback completes. Since it is unhashed a fresh dentry is used when the user space mount request creates the mount point directory. This leaves the original dentry negative and unhashed. But revalidate has no way to tell the VFS that the dentry has changed, other than to force another ->lookup() by returning false, which is at best wastefull and at worst not possible. This results in an -ENOENT return from the original path walk when in fact the mount succeeded. To resolve this we need to ensure that the same dentry is used in all calls to ->lookup() during the course of a mount request. This patch achieves that by adding the initial dentry to a look aside list and removes it at ->mkdir() or ->symlink() completion (or when the dentry is released), since these are the only create operations autofs4 supports. Signed-off-by: Ian Kent <raven@themaw.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-24 11:30:12 +07:00
d_instantiate(dentry, NULL);
}
if (!oz_mode) {
spin_lock(&dentry->d_lock);
dentry->d_flags |= DCACHE_AUTOFS_PENDING;
spin_unlock(&dentry->d_lock);
if (dentry->d_op && dentry->d_op->d_revalidate) {
mutex_unlock(&dir->i_mutex);
(dentry->d_op->d_revalidate)(dentry, nd);
mutex_lock(&dir->i_mutex);
}
}
/*
* If we are still pending, check if we had to handle
* a signal. If so we can force a restart..
*/
if (dentry->d_flags & DCACHE_AUTOFS_PENDING) {
/* See if we were interrupted */
if (signal_pending(current)) {
sigset_t *sigset = &current->pending.signal;
if (sigismember (sigset, SIGKILL) ||
sigismember (sigset, SIGQUIT) ||
sigismember (sigset, SIGINT)) {
autofs4: use look aside list for lookups A while ago a patch to resolve a deadlock during directory creation was merged. This delayed the hashing of lookup dentrys until the ->mkdir() (or ->symlink()) operation completed to ensure we always went through ->lookup() instead of also having processes go through ->revalidate() so our VFS locking remained consistent. Now we are seeing a couple of side affects of that change in situations with heavy mount activity. Two cases have been identified: 1) When a mount request is triggered, due to the delayed hashing, the directory created by user space for the mount point doesn't have the DCACHE_AUTOFS_PENDING flag set. In the case of an autofs multi-mount where a tree of mount point directories are created this can lead to the path walk continuing rather than the dentry being sent to the wait queue to wait for request completion. This is because, if the pending flag isn't set, the criteria for deciding this is a mount in progress fails to hold, namely that the dentry is not a mount point and has no subdirectories. 2) A mount request dentry is initially created negative and unhashed. It remains this way until the ->mkdir() callback completes. Since it is unhashed a fresh dentry is used when the user space mount request creates the mount point directory. This leaves the original dentry negative and unhashed. But revalidate has no way to tell the VFS that the dentry has changed, other than to force another ->lookup() by returning false, which is at best wastefull and at worst not possible. This results in an -ENOENT return from the original path walk when in fact the mount succeeded. To resolve this we need to ensure that the same dentry is used in all calls to ->lookup() during the course of a mount request. This patch achieves that by adding the initial dentry to a look aside list and removes it at ->mkdir() or ->symlink() completion (or when the dentry is released), since these are the only create operations autofs4 supports. Signed-off-by: Ian Kent <raven@themaw.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-24 11:30:12 +07:00
if (unhashed)
dput(unhashed);
return ERR_PTR(-ERESTARTNOINTR);
}
}
autofs4: use look aside list for lookups A while ago a patch to resolve a deadlock during directory creation was merged. This delayed the hashing of lookup dentrys until the ->mkdir() (or ->symlink()) operation completed to ensure we always went through ->lookup() instead of also having processes go through ->revalidate() so our VFS locking remained consistent. Now we are seeing a couple of side affects of that change in situations with heavy mount activity. Two cases have been identified: 1) When a mount request is triggered, due to the delayed hashing, the directory created by user space for the mount point doesn't have the DCACHE_AUTOFS_PENDING flag set. In the case of an autofs multi-mount where a tree of mount point directories are created this can lead to the path walk continuing rather than the dentry being sent to the wait queue to wait for request completion. This is because, if the pending flag isn't set, the criteria for deciding this is a mount in progress fails to hold, namely that the dentry is not a mount point and has no subdirectories. 2) A mount request dentry is initially created negative and unhashed. It remains this way until the ->mkdir() callback completes. Since it is unhashed a fresh dentry is used when the user space mount request creates the mount point directory. This leaves the original dentry negative and unhashed. But revalidate has no way to tell the VFS that the dentry has changed, other than to force another ->lookup() by returning false, which is at best wastefull and at worst not possible. This results in an -ENOENT return from the original path walk when in fact the mount succeeded. To resolve this we need to ensure that the same dentry is used in all calls to ->lookup() during the course of a mount request. This patch achieves that by adding the initial dentry to a look aside list and removes it at ->mkdir() or ->symlink() completion (or when the dentry is released), since these are the only create operations autofs4 supports. Signed-off-by: Ian Kent <raven@themaw.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-24 11:30:12 +07:00
if (!oz_mode) {
spin_lock(&dentry->d_lock);
dentry->d_flags &= ~DCACHE_AUTOFS_PENDING;
spin_unlock(&dentry->d_lock);
}
}
/*
* If this dentry is unhashed, then we shouldn't honour this
* lookup. Returning ENOENT here doesn't do the right thing
* for all system calls, but it should be OK for the operations
* we permit from an autofs.
*/
if (!oz_mode && d_unhashed(dentry)) {
/*
* A user space application can (and has done in the past)
* remove and re-create this directory during the callback.
* This can leave us with an unhashed dentry, but a
* successful mount! So we need to perform another
* cached lookup in case the dentry now exists.
*/
struct dentry *parent = dentry->d_parent;
struct dentry *new = d_lookup(parent, &dentry->d_name);
if (new != NULL)
dentry = new;
else
dentry = ERR_PTR(-ENOENT);
autofs4: use look aside list for lookups A while ago a patch to resolve a deadlock during directory creation was merged. This delayed the hashing of lookup dentrys until the ->mkdir() (or ->symlink()) operation completed to ensure we always went through ->lookup() instead of also having processes go through ->revalidate() so our VFS locking remained consistent. Now we are seeing a couple of side affects of that change in situations with heavy mount activity. Two cases have been identified: 1) When a mount request is triggered, due to the delayed hashing, the directory created by user space for the mount point doesn't have the DCACHE_AUTOFS_PENDING flag set. In the case of an autofs multi-mount where a tree of mount point directories are created this can lead to the path walk continuing rather than the dentry being sent to the wait queue to wait for request completion. This is because, if the pending flag isn't set, the criteria for deciding this is a mount in progress fails to hold, namely that the dentry is not a mount point and has no subdirectories. 2) A mount request dentry is initially created negative and unhashed. It remains this way until the ->mkdir() callback completes. Since it is unhashed a fresh dentry is used when the user space mount request creates the mount point directory. This leaves the original dentry negative and unhashed. But revalidate has no way to tell the VFS that the dentry has changed, other than to force another ->lookup() by returning false, which is at best wastefull and at worst not possible. This results in an -ENOENT return from the original path walk when in fact the mount succeeded. To resolve this we need to ensure that the same dentry is used in all calls to ->lookup() during the course of a mount request. This patch achieves that by adding the initial dentry to a look aside list and removes it at ->mkdir() or ->symlink() completion (or when the dentry is released), since these are the only create operations autofs4 supports. Signed-off-by: Ian Kent <raven@themaw.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-24 11:30:12 +07:00
if (unhashed)
dput(unhashed);
return dentry;
}
autofs4: use look aside list for lookups A while ago a patch to resolve a deadlock during directory creation was merged. This delayed the hashing of lookup dentrys until the ->mkdir() (or ->symlink()) operation completed to ensure we always went through ->lookup() instead of also having processes go through ->revalidate() so our VFS locking remained consistent. Now we are seeing a couple of side affects of that change in situations with heavy mount activity. Two cases have been identified: 1) When a mount request is triggered, due to the delayed hashing, the directory created by user space for the mount point doesn't have the DCACHE_AUTOFS_PENDING flag set. In the case of an autofs multi-mount where a tree of mount point directories are created this can lead to the path walk continuing rather than the dentry being sent to the wait queue to wait for request completion. This is because, if the pending flag isn't set, the criteria for deciding this is a mount in progress fails to hold, namely that the dentry is not a mount point and has no subdirectories. 2) A mount request dentry is initially created negative and unhashed. It remains this way until the ->mkdir() callback completes. Since it is unhashed a fresh dentry is used when the user space mount request creates the mount point directory. This leaves the original dentry negative and unhashed. But revalidate has no way to tell the VFS that the dentry has changed, other than to force another ->lookup() by returning false, which is at best wastefull and at worst not possible. This results in an -ENOENT return from the original path walk when in fact the mount succeeded. To resolve this we need to ensure that the same dentry is used in all calls to ->lookup() during the course of a mount request. This patch achieves that by adding the initial dentry to a look aside list and removes it at ->mkdir() or ->symlink() completion (or when the dentry is released), since these are the only create operations autofs4 supports. Signed-off-by: Ian Kent <raven@themaw.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-24 11:30:12 +07:00
if (unhashed)
return unhashed;
return NULL;
}
static int autofs4_dir_symlink(struct inode *dir,
struct dentry *dentry,
const char *symname)
{
struct autofs_sb_info *sbi = autofs4_sbi(dir->i_sb);
struct autofs_info *ino = autofs4_dentry_ino(dentry);
struct autofs_info *p_ino;
struct inode *inode;
char *cp;
DPRINTK("%s <- %.*s", symname,
dentry->d_name.len, dentry->d_name.name);
if (!autofs4_oz_mode(sbi))
return -EACCES;
ino = autofs4_init_ino(ino, sbi, S_IFLNK | 0555);
autofs4: use look aside list for lookups A while ago a patch to resolve a deadlock during directory creation was merged. This delayed the hashing of lookup dentrys until the ->mkdir() (or ->symlink()) operation completed to ensure we always went through ->lookup() instead of also having processes go through ->revalidate() so our VFS locking remained consistent. Now we are seeing a couple of side affects of that change in situations with heavy mount activity. Two cases have been identified: 1) When a mount request is triggered, due to the delayed hashing, the directory created by user space for the mount point doesn't have the DCACHE_AUTOFS_PENDING flag set. In the case of an autofs multi-mount where a tree of mount point directories are created this can lead to the path walk continuing rather than the dentry being sent to the wait queue to wait for request completion. This is because, if the pending flag isn't set, the criteria for deciding this is a mount in progress fails to hold, namely that the dentry is not a mount point and has no subdirectories. 2) A mount request dentry is initially created negative and unhashed. It remains this way until the ->mkdir() callback completes. Since it is unhashed a fresh dentry is used when the user space mount request creates the mount point directory. This leaves the original dentry negative and unhashed. But revalidate has no way to tell the VFS that the dentry has changed, other than to force another ->lookup() by returning false, which is at best wastefull and at worst not possible. This results in an -ENOENT return from the original path walk when in fact the mount succeeded. To resolve this we need to ensure that the same dentry is used in all calls to ->lookup() during the course of a mount request. This patch achieves that by adding the initial dentry to a look aside list and removes it at ->mkdir() or ->symlink() completion (or when the dentry is released), since these are the only create operations autofs4 supports. Signed-off-by: Ian Kent <raven@themaw.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-24 11:30:12 +07:00
if (!ino)
return -ENOMEM;
autofs4: use look aside list for lookups A while ago a patch to resolve a deadlock during directory creation was merged. This delayed the hashing of lookup dentrys until the ->mkdir() (or ->symlink()) operation completed to ensure we always went through ->lookup() instead of also having processes go through ->revalidate() so our VFS locking remained consistent. Now we are seeing a couple of side affects of that change in situations with heavy mount activity. Two cases have been identified: 1) When a mount request is triggered, due to the delayed hashing, the directory created by user space for the mount point doesn't have the DCACHE_AUTOFS_PENDING flag set. In the case of an autofs multi-mount where a tree of mount point directories are created this can lead to the path walk continuing rather than the dentry being sent to the wait queue to wait for request completion. This is because, if the pending flag isn't set, the criteria for deciding this is a mount in progress fails to hold, namely that the dentry is not a mount point and has no subdirectories. 2) A mount request dentry is initially created negative and unhashed. It remains this way until the ->mkdir() callback completes. Since it is unhashed a fresh dentry is used when the user space mount request creates the mount point directory. This leaves the original dentry negative and unhashed. But revalidate has no way to tell the VFS that the dentry has changed, other than to force another ->lookup() by returning false, which is at best wastefull and at worst not possible. This results in an -ENOENT return from the original path walk when in fact the mount succeeded. To resolve this we need to ensure that the same dentry is used in all calls to ->lookup() during the course of a mount request. This patch achieves that by adding the initial dentry to a look aside list and removes it at ->mkdir() or ->symlink() completion (or when the dentry is released), since these are the only create operations autofs4 supports. Signed-off-by: Ian Kent <raven@themaw.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-24 11:30:12 +07:00
spin_lock(&sbi->lookup_lock);
if (!list_empty(&ino->active))
list_del_init(&ino->active);
spin_unlock(&sbi->lookup_lock);
ino->size = strlen(symname);
autofs4: use look aside list for lookups A while ago a patch to resolve a deadlock during directory creation was merged. This delayed the hashing of lookup dentrys until the ->mkdir() (or ->symlink()) operation completed to ensure we always went through ->lookup() instead of also having processes go through ->revalidate() so our VFS locking remained consistent. Now we are seeing a couple of side affects of that change in situations with heavy mount activity. Two cases have been identified: 1) When a mount request is triggered, due to the delayed hashing, the directory created by user space for the mount point doesn't have the DCACHE_AUTOFS_PENDING flag set. In the case of an autofs multi-mount where a tree of mount point directories are created this can lead to the path walk continuing rather than the dentry being sent to the wait queue to wait for request completion. This is because, if the pending flag isn't set, the criteria for deciding this is a mount in progress fails to hold, namely that the dentry is not a mount point and has no subdirectories. 2) A mount request dentry is initially created negative and unhashed. It remains this way until the ->mkdir() callback completes. Since it is unhashed a fresh dentry is used when the user space mount request creates the mount point directory. This leaves the original dentry negative and unhashed. But revalidate has no way to tell the VFS that the dentry has changed, other than to force another ->lookup() by returning false, which is at best wastefull and at worst not possible. This results in an -ENOENT return from the original path walk when in fact the mount succeeded. To resolve this we need to ensure that the same dentry is used in all calls to ->lookup() during the course of a mount request. This patch achieves that by adding the initial dentry to a look aside list and removes it at ->mkdir() or ->symlink() completion (or when the dentry is released), since these are the only create operations autofs4 supports. Signed-off-by: Ian Kent <raven@themaw.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-24 11:30:12 +07:00
cp = kmalloc(ino->size + 1, GFP_KERNEL);
if (!cp) {
if (!dentry->d_fsdata)
kfree(ino);
return -ENOMEM;
}
strcpy(cp, symname);
inode = autofs4_get_inode(dir->i_sb, ino);
autofs4: use look aside list for lookups A while ago a patch to resolve a deadlock during directory creation was merged. This delayed the hashing of lookup dentrys until the ->mkdir() (or ->symlink()) operation completed to ensure we always went through ->lookup() instead of also having processes go through ->revalidate() so our VFS locking remained consistent. Now we are seeing a couple of side affects of that change in situations with heavy mount activity. Two cases have been identified: 1) When a mount request is triggered, due to the delayed hashing, the directory created by user space for the mount point doesn't have the DCACHE_AUTOFS_PENDING flag set. In the case of an autofs multi-mount where a tree of mount point directories are created this can lead to the path walk continuing rather than the dentry being sent to the wait queue to wait for request completion. This is because, if the pending flag isn't set, the criteria for deciding this is a mount in progress fails to hold, namely that the dentry is not a mount point and has no subdirectories. 2) A mount request dentry is initially created negative and unhashed. It remains this way until the ->mkdir() callback completes. Since it is unhashed a fresh dentry is used when the user space mount request creates the mount point directory. This leaves the original dentry negative and unhashed. But revalidate has no way to tell the VFS that the dentry has changed, other than to force another ->lookup() by returning false, which is at best wastefull and at worst not possible. This results in an -ENOENT return from the original path walk when in fact the mount succeeded. To resolve this we need to ensure that the same dentry is used in all calls to ->lookup() during the course of a mount request. This patch achieves that by adding the initial dentry to a look aside list and removes it at ->mkdir() or ->symlink() completion (or when the dentry is released), since these are the only create operations autofs4 supports. Signed-off-by: Ian Kent <raven@themaw.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-24 11:30:12 +07:00
if (!inode) {
kfree(cp);
if (!dentry->d_fsdata)
kfree(ino);
return -ENOMEM;
}
d_add(dentry, inode);
if (dir == dir->i_sb->s_root->d_inode)
dentry->d_op = &autofs4_root_dentry_operations;
else
dentry->d_op = &autofs4_dentry_operations;
dentry->d_fsdata = ino;
ino->dentry = dget(dentry);
atomic_inc(&ino->count);
p_ino = autofs4_dentry_ino(dentry->d_parent);
if (p_ino && dentry->d_parent != dentry)
atomic_inc(&p_ino->count);
ino->inode = inode;
autofs4: use look aside list for lookups A while ago a patch to resolve a deadlock during directory creation was merged. This delayed the hashing of lookup dentrys until the ->mkdir() (or ->symlink()) operation completed to ensure we always went through ->lookup() instead of also having processes go through ->revalidate() so our VFS locking remained consistent. Now we are seeing a couple of side affects of that change in situations with heavy mount activity. Two cases have been identified: 1) When a mount request is triggered, due to the delayed hashing, the directory created by user space for the mount point doesn't have the DCACHE_AUTOFS_PENDING flag set. In the case of an autofs multi-mount where a tree of mount point directories are created this can lead to the path walk continuing rather than the dentry being sent to the wait queue to wait for request completion. This is because, if the pending flag isn't set, the criteria for deciding this is a mount in progress fails to hold, namely that the dentry is not a mount point and has no subdirectories. 2) A mount request dentry is initially created negative and unhashed. It remains this way until the ->mkdir() callback completes. Since it is unhashed a fresh dentry is used when the user space mount request creates the mount point directory. This leaves the original dentry negative and unhashed. But revalidate has no way to tell the VFS that the dentry has changed, other than to force another ->lookup() by returning false, which is at best wastefull and at worst not possible. This results in an -ENOENT return from the original path walk when in fact the mount succeeded. To resolve this we need to ensure that the same dentry is used in all calls to ->lookup() during the course of a mount request. This patch achieves that by adding the initial dentry to a look aside list and removes it at ->mkdir() or ->symlink() completion (or when the dentry is released), since these are the only create operations autofs4 supports. Signed-off-by: Ian Kent <raven@themaw.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-24 11:30:12 +07:00
ino->u.symlink = cp;
dir->i_mtime = CURRENT_TIME;
return 0;
}
/*
* NOTE!
*
* Normal filesystems would do a "d_delete()" to tell the VFS dcache
* that the file no longer exists. However, doing that means that the
* VFS layer can turn the dentry into a negative dentry. We don't want
* this, because the unlink is probably the result of an expire.
* We simply d_drop it and add it to a expiring list in the super block,
* which allows the dentry lookup to check for an incomplete expire.
*
* If a process is blocked on the dentry waiting for the expire to finish,
* it will invalidate the dentry and try to mount with a new one.
*
* Also see autofs4_dir_rmdir()..
*/
static int autofs4_dir_unlink(struct inode *dir, struct dentry *dentry)
{
struct autofs_sb_info *sbi = autofs4_sbi(dir->i_sb);
struct autofs_info *ino = autofs4_dentry_ino(dentry);
struct autofs_info *p_ino;
/* This allows root to remove symlinks */
if (!autofs4_oz_mode(sbi) && !capable(CAP_SYS_ADMIN))
return -EACCES;
if (atomic_dec_and_test(&ino->count)) {
p_ino = autofs4_dentry_ino(dentry->d_parent);
if (p_ino && dentry->d_parent != dentry)
atomic_dec(&p_ino->count);
}
dput(ino->dentry);
dentry->d_inode->i_size = 0;
clear_nlink(dentry->d_inode);
dir->i_mtime = CURRENT_TIME;
spin_lock(&dcache_lock);
spin_lock(&sbi->lookup_lock);
autofs4: use look aside list for lookups A while ago a patch to resolve a deadlock during directory creation was merged. This delayed the hashing of lookup dentrys until the ->mkdir() (or ->symlink()) operation completed to ensure we always went through ->lookup() instead of also having processes go through ->revalidate() so our VFS locking remained consistent. Now we are seeing a couple of side affects of that change in situations with heavy mount activity. Two cases have been identified: 1) When a mount request is triggered, due to the delayed hashing, the directory created by user space for the mount point doesn't have the DCACHE_AUTOFS_PENDING flag set. In the case of an autofs multi-mount where a tree of mount point directories are created this can lead to the path walk continuing rather than the dentry being sent to the wait queue to wait for request completion. This is because, if the pending flag isn't set, the criteria for deciding this is a mount in progress fails to hold, namely that the dentry is not a mount point and has no subdirectories. 2) A mount request dentry is initially created negative and unhashed. It remains this way until the ->mkdir() callback completes. Since it is unhashed a fresh dentry is used when the user space mount request creates the mount point directory. This leaves the original dentry negative and unhashed. But revalidate has no way to tell the VFS that the dentry has changed, other than to force another ->lookup() by returning false, which is at best wastefull and at worst not possible. This results in an -ENOENT return from the original path walk when in fact the mount succeeded. To resolve this we need to ensure that the same dentry is used in all calls to ->lookup() during the course of a mount request. This patch achieves that by adding the initial dentry to a look aside list and removes it at ->mkdir() or ->symlink() completion (or when the dentry is released), since these are the only create operations autofs4 supports. Signed-off-by: Ian Kent <raven@themaw.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-24 11:30:12 +07:00
if (list_empty(&ino->expiring))
list_add(&ino->expiring, &sbi->expiring_list);
spin_unlock(&sbi->lookup_lock);
spin_lock(&dentry->d_lock);
__d_drop(dentry);
spin_unlock(&dentry->d_lock);
spin_unlock(&dcache_lock);
return 0;
}
static int autofs4_dir_rmdir(struct inode *dir, struct dentry *dentry)
{
struct autofs_sb_info *sbi = autofs4_sbi(dir->i_sb);
struct autofs_info *ino = autofs4_dentry_ino(dentry);
struct autofs_info *p_ino;
DPRINTK("dentry %p, removing %.*s",
dentry, dentry->d_name.len, dentry->d_name.name);
if (!autofs4_oz_mode(sbi))
return -EACCES;
spin_lock(&dcache_lock);
if (!list_empty(&dentry->d_subdirs)) {
spin_unlock(&dcache_lock);
return -ENOTEMPTY;
}
spin_lock(&sbi->lookup_lock);
autofs4: use look aside list for lookups A while ago a patch to resolve a deadlock during directory creation was merged. This delayed the hashing of lookup dentrys until the ->mkdir() (or ->symlink()) operation completed to ensure we always went through ->lookup() instead of also having processes go through ->revalidate() so our VFS locking remained consistent. Now we are seeing a couple of side affects of that change in situations with heavy mount activity. Two cases have been identified: 1) When a mount request is triggered, due to the delayed hashing, the directory created by user space for the mount point doesn't have the DCACHE_AUTOFS_PENDING flag set. In the case of an autofs multi-mount where a tree of mount point directories are created this can lead to the path walk continuing rather than the dentry being sent to the wait queue to wait for request completion. This is because, if the pending flag isn't set, the criteria for deciding this is a mount in progress fails to hold, namely that the dentry is not a mount point and has no subdirectories. 2) A mount request dentry is initially created negative and unhashed. It remains this way until the ->mkdir() callback completes. Since it is unhashed a fresh dentry is used when the user space mount request creates the mount point directory. This leaves the original dentry negative and unhashed. But revalidate has no way to tell the VFS that the dentry has changed, other than to force another ->lookup() by returning false, which is at best wastefull and at worst not possible. This results in an -ENOENT return from the original path walk when in fact the mount succeeded. To resolve this we need to ensure that the same dentry is used in all calls to ->lookup() during the course of a mount request. This patch achieves that by adding the initial dentry to a look aside list and removes it at ->mkdir() or ->symlink() completion (or when the dentry is released), since these are the only create operations autofs4 supports. Signed-off-by: Ian Kent <raven@themaw.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-24 11:30:12 +07:00
if (list_empty(&ino->expiring))
list_add(&ino->expiring, &sbi->expiring_list);
spin_unlock(&sbi->lookup_lock);
spin_lock(&dentry->d_lock);
__d_drop(dentry);
spin_unlock(&dentry->d_lock);
spin_unlock(&dcache_lock);
if (atomic_dec_and_test(&ino->count)) {
p_ino = autofs4_dentry_ino(dentry->d_parent);
if (p_ino && dentry->d_parent != dentry)
atomic_dec(&p_ino->count);
}
dput(ino->dentry);
dentry->d_inode->i_size = 0;
clear_nlink(dentry->d_inode);
if (dir->i_nlink)
drop_nlink(dir);
return 0;
}
static int autofs4_dir_mkdir(struct inode *dir, struct dentry *dentry, int mode)
{
struct autofs_sb_info *sbi = autofs4_sbi(dir->i_sb);
struct autofs_info *ino = autofs4_dentry_ino(dentry);
struct autofs_info *p_ino;
struct inode *inode;
if (!autofs4_oz_mode(sbi))
return -EACCES;
DPRINTK("dentry %p, creating %.*s",
dentry, dentry->d_name.len, dentry->d_name.name);
ino = autofs4_init_ino(ino, sbi, S_IFDIR | 0555);
autofs4: use look aside list for lookups A while ago a patch to resolve a deadlock during directory creation was merged. This delayed the hashing of lookup dentrys until the ->mkdir() (or ->symlink()) operation completed to ensure we always went through ->lookup() instead of also having processes go through ->revalidate() so our VFS locking remained consistent. Now we are seeing a couple of side affects of that change in situations with heavy mount activity. Two cases have been identified: 1) When a mount request is triggered, due to the delayed hashing, the directory created by user space for the mount point doesn't have the DCACHE_AUTOFS_PENDING flag set. In the case of an autofs multi-mount where a tree of mount point directories are created this can lead to the path walk continuing rather than the dentry being sent to the wait queue to wait for request completion. This is because, if the pending flag isn't set, the criteria for deciding this is a mount in progress fails to hold, namely that the dentry is not a mount point and has no subdirectories. 2) A mount request dentry is initially created negative and unhashed. It remains this way until the ->mkdir() callback completes. Since it is unhashed a fresh dentry is used when the user space mount request creates the mount point directory. This leaves the original dentry negative and unhashed. But revalidate has no way to tell the VFS that the dentry has changed, other than to force another ->lookup() by returning false, which is at best wastefull and at worst not possible. This results in an -ENOENT return from the original path walk when in fact the mount succeeded. To resolve this we need to ensure that the same dentry is used in all calls to ->lookup() during the course of a mount request. This patch achieves that by adding the initial dentry to a look aside list and removes it at ->mkdir() or ->symlink() completion (or when the dentry is released), since these are the only create operations autofs4 supports. Signed-off-by: Ian Kent <raven@themaw.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-24 11:30:12 +07:00
if (!ino)
return -ENOMEM;
spin_lock(&sbi->lookup_lock);
if (!list_empty(&ino->active))
list_del_init(&ino->active);
spin_unlock(&sbi->lookup_lock);
inode = autofs4_get_inode(dir->i_sb, ino);
autofs4: use look aside list for lookups A while ago a patch to resolve a deadlock during directory creation was merged. This delayed the hashing of lookup dentrys until the ->mkdir() (or ->symlink()) operation completed to ensure we always went through ->lookup() instead of also having processes go through ->revalidate() so our VFS locking remained consistent. Now we are seeing a couple of side affects of that change in situations with heavy mount activity. Two cases have been identified: 1) When a mount request is triggered, due to the delayed hashing, the directory created by user space for the mount point doesn't have the DCACHE_AUTOFS_PENDING flag set. In the case of an autofs multi-mount where a tree of mount point directories are created this can lead to the path walk continuing rather than the dentry being sent to the wait queue to wait for request completion. This is because, if the pending flag isn't set, the criteria for deciding this is a mount in progress fails to hold, namely that the dentry is not a mount point and has no subdirectories. 2) A mount request dentry is initially created negative and unhashed. It remains this way until the ->mkdir() callback completes. Since it is unhashed a fresh dentry is used when the user space mount request creates the mount point directory. This leaves the original dentry negative and unhashed. But revalidate has no way to tell the VFS that the dentry has changed, other than to force another ->lookup() by returning false, which is at best wastefull and at worst not possible. This results in an -ENOENT return from the original path walk when in fact the mount succeeded. To resolve this we need to ensure that the same dentry is used in all calls to ->lookup() during the course of a mount request. This patch achieves that by adding the initial dentry to a look aside list and removes it at ->mkdir() or ->symlink() completion (or when the dentry is released), since these are the only create operations autofs4 supports. Signed-off-by: Ian Kent <raven@themaw.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-24 11:30:12 +07:00
if (!inode) {
if (!dentry->d_fsdata)
kfree(ino);
return -ENOMEM;
}
d_add(dentry, inode);
if (dir == dir->i_sb->s_root->d_inode)
dentry->d_op = &autofs4_root_dentry_operations;
else
dentry->d_op = &autofs4_dentry_operations;
dentry->d_fsdata = ino;
ino->dentry = dget(dentry);
atomic_inc(&ino->count);
p_ino = autofs4_dentry_ino(dentry->d_parent);
if (p_ino && dentry->d_parent != dentry)
atomic_inc(&p_ino->count);
ino->inode = inode;
inc_nlink(dir);
dir->i_mtime = CURRENT_TIME;
return 0;
}
/* Get/set timeout ioctl() operation */
static inline int autofs4_get_set_timeout(struct autofs_sb_info *sbi,
unsigned long __user *p)
{
int rv;
unsigned long ntimeout;
if ((rv = get_user(ntimeout, p)) ||
(rv = put_user(sbi->exp_timeout/HZ, p)))
return rv;
if (ntimeout > ULONG_MAX/HZ)
sbi->exp_timeout = 0;
else
sbi->exp_timeout = ntimeout * HZ;
return 0;
}
/* Return protocol version */
static inline int autofs4_get_protover(struct autofs_sb_info *sbi, int __user *p)
{
return put_user(sbi->version, p);
}
/* Return protocol sub version */
static inline int autofs4_get_protosubver(struct autofs_sb_info *sbi, int __user *p)
{
return put_user(sbi->sub_version, p);
}
/*
* Tells the daemon whether we need to reghost or not. Also, clears
* the reghost_needed flag.
*/
static inline int autofs4_ask_reghost(struct autofs_sb_info *sbi, int __user *p)
{
int status;
DPRINTK("returning %d", sbi->needs_reghost);
status = put_user(sbi->needs_reghost, p);
if (status)
return status;
sbi->needs_reghost = 0;
return 0;
}
/*
* Enable / Disable reghosting ioctl() operation
*/
static inline int autofs4_toggle_reghost(struct autofs_sb_info *sbi, int __user *p)
{
int status;
int val;
status = get_user(val, p);
DPRINTK("reghost = %d", val);
if (status)
return status;
/* turn on/off reghosting, with the val */
sbi->reghost_enabled = val;
return 0;
}
/*
* Tells the daemon whether it can umount the autofs mount.
*/
static inline int autofs4_ask_umount(struct vfsmount *mnt, int __user *p)
{
int status = 0;
if (may_umount(mnt))
status = 1;
DPRINTK("returning %d", status);
status = put_user(status, p);
return status;
}
/* Identify autofs4_dentries - this is so we can tell if there's
an extra dentry refcount or not. We only hold a refcount on the
dentry if its non-negative (ie, d_inode != NULL)
*/
int is_autofs4_dentry(struct dentry *dentry)
{
return dentry && dentry->d_inode &&
(dentry->d_op == &autofs4_root_dentry_operations ||
dentry->d_op == &autofs4_dentry_operations) &&
dentry->d_fsdata != NULL;
}
/*
* ioctl()'s on the root directory is the chief method for the daemon to
* generate kernel reactions
*/
static int autofs4_root_ioctl(struct inode *inode, struct file *filp,
unsigned int cmd, unsigned long arg)
{
struct autofs_sb_info *sbi = autofs4_sbi(inode->i_sb);
void __user *p = (void __user *)arg;
DPRINTK("cmd = 0x%08x, arg = 0x%08lx, sbi = %p, pgrp = %u",
cmd,arg,sbi,task_pgrp_nr(current));
if (_IOC_TYPE(cmd) != _IOC_TYPE(AUTOFS_IOC_FIRST) ||
_IOC_NR(cmd) - _IOC_NR(AUTOFS_IOC_FIRST) >= AUTOFS_IOC_COUNT)
return -ENOTTY;
if (!autofs4_oz_mode(sbi) && !capable(CAP_SYS_ADMIN))
return -EPERM;
switch(cmd) {
case AUTOFS_IOC_READY: /* Wait queue: go ahead and retry */
return autofs4_wait_release(sbi,(autofs_wqt_t)arg,0);
case AUTOFS_IOC_FAIL: /* Wait queue: fail with ENOENT */
return autofs4_wait_release(sbi,(autofs_wqt_t)arg,-ENOENT);
case AUTOFS_IOC_CATATONIC: /* Enter catatonic mode (daemon shutdown) */
autofs4_catatonic_mode(sbi);
return 0;
case AUTOFS_IOC_PROTOVER: /* Get protocol version */
return autofs4_get_protover(sbi, p);
case AUTOFS_IOC_PROTOSUBVER: /* Get protocol sub version */
return autofs4_get_protosubver(sbi, p);
case AUTOFS_IOC_SETTIMEOUT:
return autofs4_get_set_timeout(sbi, p);
case AUTOFS_IOC_TOGGLEREGHOST:
return autofs4_toggle_reghost(sbi, p);
case AUTOFS_IOC_ASKREGHOST:
return autofs4_ask_reghost(sbi, p);
case AUTOFS_IOC_ASKUMOUNT:
return autofs4_ask_umount(filp->f_path.mnt, p);
/* return a single thing to expire */
case AUTOFS_IOC_EXPIRE:
return autofs4_expire_run(inode->i_sb,filp->f_path.mnt,sbi, p);
/* same as above, but can send multiple expires through pipe */
case AUTOFS_IOC_EXPIRE_MULTI:
return autofs4_expire_multi(inode->i_sb,filp->f_path.mnt,sbi, p);
default:
return -ENOSYS;
}
}