linux_dsm_epyc7002/fs/autofs4/autofs_i.h

274 lines
7.6 KiB
C
Raw Normal View History

/*
* Copyright 1997-1998 Transmeta Corporation - All Rights Reserved
* Copyright 2005-2006 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.
*/
/* Internal header file for autofs */
#include <linux/auto_fs4.h>
#include <linux/auto_dev-ioctl.h>
#include <linux/mutex.h>
#include <linux/spinlock.h>
#include <linux/list.h>
#include <linux/completion.h>
/* This is the range of ioctl() numbers we claim as ours */
#define AUTOFS_IOC_FIRST AUTOFS_IOC_READY
#define AUTOFS_IOC_COUNT 32
#define AUTOFS_DEV_IOCTL_IOC_FIRST (AUTOFS_DEV_IOCTL_VERSION)
#define AUTOFS_DEV_IOCTL_IOC_COUNT \
(AUTOFS_DEV_IOCTL_ISMOUNTPOINT_CMD - AUTOFS_DEV_IOCTL_VERSION_CMD)
#include <linux/kernel.h>
#include <linux/slab.h>
#include <linux/time.h>
#include <linux/string.h>
#include <linux/wait.h>
#include <linux/sched.h>
#include <linux/mount.h>
#include <linux/namei.h>
#include <asm/current.h>
#include <linux/uaccess.h>
#ifdef pr_fmt
#undef pr_fmt
#endif
#define pr_fmt(fmt) KBUILD_MODNAME ":pid:%d:%s: " fmt, current->pid, __func__
/*
* Unified info structure. This is pointed to by both the dentry and
* inode structures. Each file in the filesystem has an instance of this
* structure. It holds a reference to the dentry, so dentries are never
* flushed while the file exists. All name lookups are dealt with at the
* dentry level, although the filesystem can interfere in the validation
* process. Readdir is implemented by traversing the dentry lists.
*/
struct autofs_info {
struct dentry *dentry;
struct inode *inode;
int flags;
struct completion expire_complete;
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 list_head active;
int active_count;
struct list_head expiring;
struct autofs_sb_info *sbi;
unsigned long last_used;
atomic_t count;
kuid_t uid;
kgid_t gid;
};
#define AUTOFS_INF_EXPIRING (1<<0) /* dentry in the process of expiring */
#define AUTOFS_INF_WANT_EXPIRE (1<<1) /* the dentry is being considered
* for expiry, so RCU_walk is
* not permitted. If it progresses to
* actual expiry attempt, the flag is
* not cleared when EXPIRING is set -
* in that case it gets cleared only
* when it comes to clearing EXPIRING.
*/
#define AUTOFS_INF_PENDING (1<<2) /* dentry pending mount */
struct autofs_wait_queue {
wait_queue_head_t queue;
struct autofs_wait_queue *next;
autofs_wqt_t wait_queue_token;
/* We use the following to see what we are waiting for */
struct qstr name;
u32 dev;
u64 ino;
kuid_t uid;
kgid_t gid;
pid_t pid;
pid_t tgid;
/* This is for status reporting upon return */
int status;
unsigned int wait_ctr;
};
#define AUTOFS_SBI_MAGIC 0x6d4a556d
struct autofs_sb_info {
u32 magic;
int pipefd;
struct file *pipe;
autofs4: allow autofs to work outside the initial PID namespace Enable autofs4 to work in a "container". oz_pgrp is converted from pid_t to struct pid and this is stored at mount time based on the "pgrp=" option or if the option is missing then the current pgrp. The "pgrp=" option is interpreted in the PID namespace of the current process. This option is flawed in that it doesn't carry the namespace information, so it should be deprecated. AFAICS the autofs daemon always sends the current pgrp, which is the default anyway. The oz_pgrp is also set from the AUTOFS_DEV_IOCTL_SETPIPEFD_CMD ioctl. This ioctl sets oz_pgrp to the current pgrp. It is not allowed to change the pid namespace. oz_pgrp is used mainly to determine whether the process traversing the autofs mount tree is the autofs daemon itself or not. This function now compares the pid pointers instead of the pid_t values. One other use of oz_pgrp is in autofs4_show_options. There is shows the virtual pid number (i.e. the one that is valid inside the PID namespace of the calling process) For debugging printk convert oz_pgrp to the value in the initial pid namespace. Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com> Signed-off-by: Miklos Szeredi <mszeredi@suse.cz> Acked-by: Serge Hallyn <serge.hallyn@canonical.com> Cc: Eric Biederman <ebiederm@xmission.com> Acked-by: Ian Kent <raven@themaw.net> Cc: Oleg Nesterov <oleg@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-01-24 06:54:57 +07:00
struct pid *oz_pgrp;
int catatonic;
int version;
int sub_version;
int min_proto;
int max_proto;
unsigned long exp_timeout;
unsigned int type;
struct super_block *sb;
struct mutex wq_mutex;
struct mutex pipe_mutex;
spinlock_t fs_lock;
struct autofs_wait_queue *queues; /* Wait queue pointer */
spinlock_t 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
struct list_head active_list;
struct list_head expiring_list;
struct rcu_head rcu;
};
static inline struct autofs_sb_info *autofs4_sbi(struct super_block *sb)
{
return (struct autofs_sb_info *)(sb->s_fs_info);
}
static inline struct autofs_info *autofs4_dentry_ino(struct dentry *dentry)
{
return (struct autofs_info *)(dentry->d_fsdata);
}
/* autofs4_oz_mode(): do we see the man behind the curtain? (The
* processes which do manipulations for us in user space sees the raw
* filesystem without "magic".)
*/
static inline int autofs4_oz_mode(struct autofs_sb_info *sbi)
{
autofs4: allow autofs to work outside the initial PID namespace Enable autofs4 to work in a "container". oz_pgrp is converted from pid_t to struct pid and this is stored at mount time based on the "pgrp=" option or if the option is missing then the current pgrp. The "pgrp=" option is interpreted in the PID namespace of the current process. This option is flawed in that it doesn't carry the namespace information, so it should be deprecated. AFAICS the autofs daemon always sends the current pgrp, which is the default anyway. The oz_pgrp is also set from the AUTOFS_DEV_IOCTL_SETPIPEFD_CMD ioctl. This ioctl sets oz_pgrp to the current pgrp. It is not allowed to change the pid namespace. oz_pgrp is used mainly to determine whether the process traversing the autofs mount tree is the autofs daemon itself or not. This function now compares the pid pointers instead of the pid_t values. One other use of oz_pgrp is in autofs4_show_options. There is shows the virtual pid number (i.e. the one that is valid inside the PID namespace of the calling process) For debugging printk convert oz_pgrp to the value in the initial pid namespace. Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com> Signed-off-by: Miklos Szeredi <mszeredi@suse.cz> Acked-by: Serge Hallyn <serge.hallyn@canonical.com> Cc: Eric Biederman <ebiederm@xmission.com> Acked-by: Ian Kent <raven@themaw.net> Cc: Oleg Nesterov <oleg@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-01-24 06:54:57 +07:00
return sbi->catatonic || task_pgrp(current) == sbi->oz_pgrp;
}
struct inode *autofs4_get_inode(struct super_block *, umode_t);
void autofs4_free_ino(struct autofs_info *);
/* Expiration */
int is_autofs4_dentry(struct dentry *);
int autofs4_expire_wait(const struct path *path, int rcu_walk);
int autofs4_expire_run(struct super_block *, struct vfsmount *,
struct autofs_sb_info *,
struct autofs_packet_expire __user *);
int autofs4_do_expire_multi(struct super_block *sb, struct vfsmount *mnt,
struct autofs_sb_info *sbi, int when);
int autofs4_expire_multi(struct super_block *, struct vfsmount *,
struct autofs_sb_info *, int __user *);
struct dentry *autofs4_expire_direct(struct super_block *sb,
struct vfsmount *mnt,
struct autofs_sb_info *sbi, int how);
struct dentry *autofs4_expire_indirect(struct super_block *sb,
struct vfsmount *mnt,
struct autofs_sb_info *sbi, int how);
/* Device node initialization */
int autofs_dev_ioctl_init(void);
void autofs_dev_ioctl_exit(void);
/* Operations structures */
extern const struct inode_operations autofs4_symlink_inode_operations;
extern const struct inode_operations autofs4_dir_inode_operations;
extern const struct file_operations autofs4_dir_operations;
extern const struct file_operations autofs4_root_operations;
extern const struct dentry_operations autofs4_dentry_operations;
/* VFS automount flags management functions */
static inline void __managed_dentry_set_managed(struct dentry *dentry)
{
dentry->d_flags |= (DCACHE_NEED_AUTOMOUNT|DCACHE_MANAGE_TRANSIT);
}
static inline void managed_dentry_set_managed(struct dentry *dentry)
{
spin_lock(&dentry->d_lock);
__managed_dentry_set_managed(dentry);
spin_unlock(&dentry->d_lock);
}
static inline void __managed_dentry_clear_managed(struct dentry *dentry)
{
dentry->d_flags &= ~(DCACHE_NEED_AUTOMOUNT|DCACHE_MANAGE_TRANSIT);
}
static inline void managed_dentry_clear_managed(struct dentry *dentry)
{
spin_lock(&dentry->d_lock);
__managed_dentry_clear_managed(dentry);
spin_unlock(&dentry->d_lock);
}
/* Initializing function */
int autofs4_fill_super(struct super_block *, void *, int);
struct autofs_info *autofs4_new_ino(struct autofs_sb_info *);
void autofs4_clean_ino(struct autofs_info *);
autofs: make the autofsv5 packet file descriptor use a packetized pipe The autofs packet size has had a very unfortunate size problem on x86: because the alignment of 'u64' differs in 32-bit and 64-bit modes, and because the packet data was not 8-byte aligned, the size of the autofsv5 packet structure differed between 32-bit and 64-bit modes despite looking otherwise identical (300 vs 304 bytes respectively). We first fixed that up by making the 64-bit compat mode know about this problem in commit a32744d4abae ("autofs: work around unhappy compat problem on x86-64"), and that made a 32-bit 'systemd' work happily on a 64-bit kernel because everything then worked the same way as on a 32-bit kernel. But it turned out that 'automount' had actually known and worked around this problem in user space, so fixing the kernel to do the proper 32-bit compatibility handling actually *broke* 32-bit automount on a 64-bit kernel, because it knew that the packet sizes were wrong and expected those incorrect sizes. As a result, we ended up reverting that compatibility mode fix, and thus breaking systemd again, in commit fcbf94b9dedd. With both automount and systemd doing a single read() system call, and verifying that they get *exactly* the size they expect but using different sizes, it seemed that fixing one of them inevitably seemed to break the other. At one point, a patch I seriously considered applying from Michael Tokarev did a "strcmp()" to see if it was automount that was doing the operation. Ugly, ugly. However, a prettier solution exists now thanks to the packetized pipe mode. By marking the communication pipe as being packetized (by simply setting the O_DIRECT flag), we can always just write the bigger packet size, and if user-space does a smaller read, it will just get that partial end result and the extra alignment padding will simply be thrown away. This makes both automount and systemd happy, since they now get the size they asked for, and the kernel side of autofs simply no longer needs to care - it could pad out the packet arbitrarily. Of course, if there is some *other* user of autofs (please, please, please tell me it ain't so - and we haven't heard of any) that tries to read the packets with multiple writes, that other user will now be broken - the whole point of the packetized mode is that one system call gets exactly one packet, and you cannot read a packet in pieces. Tested-by: Michael Tokarev <mjt@tls.msk.ru> Cc: Alan Cox <alan@lxorguk.ukuu.org.uk> Cc: David Miller <davem@davemloft.net> Cc: Ian Kent <raven@themaw.net> Cc: Thomas Meyer <thomas@m3y3r.de> Cc: stable@kernel.org Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-04-30 03:30:08 +07:00
static inline int autofs_prepare_pipe(struct file *pipe)
{
if (!(pipe->f_mode & FMODE_CAN_WRITE))
autofs: make the autofsv5 packet file descriptor use a packetized pipe The autofs packet size has had a very unfortunate size problem on x86: because the alignment of 'u64' differs in 32-bit and 64-bit modes, and because the packet data was not 8-byte aligned, the size of the autofsv5 packet structure differed between 32-bit and 64-bit modes despite looking otherwise identical (300 vs 304 bytes respectively). We first fixed that up by making the 64-bit compat mode know about this problem in commit a32744d4abae ("autofs: work around unhappy compat problem on x86-64"), and that made a 32-bit 'systemd' work happily on a 64-bit kernel because everything then worked the same way as on a 32-bit kernel. But it turned out that 'automount' had actually known and worked around this problem in user space, so fixing the kernel to do the proper 32-bit compatibility handling actually *broke* 32-bit automount on a 64-bit kernel, because it knew that the packet sizes were wrong and expected those incorrect sizes. As a result, we ended up reverting that compatibility mode fix, and thus breaking systemd again, in commit fcbf94b9dedd. With both automount and systemd doing a single read() system call, and verifying that they get *exactly* the size they expect but using different sizes, it seemed that fixing one of them inevitably seemed to break the other. At one point, a patch I seriously considered applying from Michael Tokarev did a "strcmp()" to see if it was automount that was doing the operation. Ugly, ugly. However, a prettier solution exists now thanks to the packetized pipe mode. By marking the communication pipe as being packetized (by simply setting the O_DIRECT flag), we can always just write the bigger packet size, and if user-space does a smaller read, it will just get that partial end result and the extra alignment padding will simply be thrown away. This makes both automount and systemd happy, since they now get the size they asked for, and the kernel side of autofs simply no longer needs to care - it could pad out the packet arbitrarily. Of course, if there is some *other* user of autofs (please, please, please tell me it ain't so - and we haven't heard of any) that tries to read the packets with multiple writes, that other user will now be broken - the whole point of the packetized mode is that one system call gets exactly one packet, and you cannot read a packet in pieces. Tested-by: Michael Tokarev <mjt@tls.msk.ru> Cc: Alan Cox <alan@lxorguk.ukuu.org.uk> Cc: David Miller <davem@davemloft.net> Cc: Ian Kent <raven@themaw.net> Cc: Thomas Meyer <thomas@m3y3r.de> Cc: stable@kernel.org Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-04-30 03:30:08 +07:00
return -EINVAL;
if (!S_ISFIFO(file_inode(pipe)->i_mode))
autofs: make the autofsv5 packet file descriptor use a packetized pipe The autofs packet size has had a very unfortunate size problem on x86: because the alignment of 'u64' differs in 32-bit and 64-bit modes, and because the packet data was not 8-byte aligned, the size of the autofsv5 packet structure differed between 32-bit and 64-bit modes despite looking otherwise identical (300 vs 304 bytes respectively). We first fixed that up by making the 64-bit compat mode know about this problem in commit a32744d4abae ("autofs: work around unhappy compat problem on x86-64"), and that made a 32-bit 'systemd' work happily on a 64-bit kernel because everything then worked the same way as on a 32-bit kernel. But it turned out that 'automount' had actually known and worked around this problem in user space, so fixing the kernel to do the proper 32-bit compatibility handling actually *broke* 32-bit automount on a 64-bit kernel, because it knew that the packet sizes were wrong and expected those incorrect sizes. As a result, we ended up reverting that compatibility mode fix, and thus breaking systemd again, in commit fcbf94b9dedd. With both automount and systemd doing a single read() system call, and verifying that they get *exactly* the size they expect but using different sizes, it seemed that fixing one of them inevitably seemed to break the other. At one point, a patch I seriously considered applying from Michael Tokarev did a "strcmp()" to see if it was automount that was doing the operation. Ugly, ugly. However, a prettier solution exists now thanks to the packetized pipe mode. By marking the communication pipe as being packetized (by simply setting the O_DIRECT flag), we can always just write the bigger packet size, and if user-space does a smaller read, it will just get that partial end result and the extra alignment padding will simply be thrown away. This makes both automount and systemd happy, since they now get the size they asked for, and the kernel side of autofs simply no longer needs to care - it could pad out the packet arbitrarily. Of course, if there is some *other* user of autofs (please, please, please tell me it ain't so - and we haven't heard of any) that tries to read the packets with multiple writes, that other user will now be broken - the whole point of the packetized mode is that one system call gets exactly one packet, and you cannot read a packet in pieces. Tested-by: Michael Tokarev <mjt@tls.msk.ru> Cc: Alan Cox <alan@lxorguk.ukuu.org.uk> Cc: David Miller <davem@davemloft.net> Cc: Ian Kent <raven@themaw.net> Cc: Thomas Meyer <thomas@m3y3r.de> Cc: stable@kernel.org Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-04-30 03:30:08 +07:00
return -EINVAL;
/* We want a packet pipe */
pipe->f_flags |= O_DIRECT;
return 0;
}
/* Queue management functions */
int autofs4_wait(struct autofs_sb_info *,
const struct path *, enum autofs_notify);
int autofs4_wait_release(struct autofs_sb_info *, autofs_wqt_t, int);
void autofs4_catatonic_mode(struct autofs_sb_info *);
static inline u32 autofs4_get_dev(struct autofs_sb_info *sbi)
{
return new_encode_dev(sbi->sb->s_dev);
}
static inline u64 autofs4_get_ino(struct autofs_sb_info *sbi)
{
return d_inode(sbi->sb->s_root)->i_ino;
}
static inline void __autofs4_add_expiring(struct dentry *dentry)
{
struct autofs_sb_info *sbi = autofs4_sbi(dentry->d_sb);
struct autofs_info *ino = autofs4_dentry_ino(dentry);
if (ino) {
if (list_empty(&ino->expiring))
list_add(&ino->expiring, &sbi->expiring_list);
}
}
static inline void autofs4_add_expiring(struct dentry *dentry)
{
struct autofs_sb_info *sbi = autofs4_sbi(dentry->d_sb);
struct autofs_info *ino = autofs4_dentry_ino(dentry);
if (ino) {
spin_lock(&sbi->lookup_lock);
if (list_empty(&ino->expiring))
list_add(&ino->expiring, &sbi->expiring_list);
spin_unlock(&sbi->lookup_lock);
}
}
static inline void autofs4_del_expiring(struct dentry *dentry)
{
struct autofs_sb_info *sbi = autofs4_sbi(dentry->d_sb);
struct autofs_info *ino = autofs4_dentry_ino(dentry);
if (ino) {
spin_lock(&sbi->lookup_lock);
if (!list_empty(&ino->expiring))
list_del_init(&ino->expiring);
spin_unlock(&sbi->lookup_lock);
}
}
void autofs4_kill_sb(struct super_block *);