11da177e4SLinus Torvalds /* 21da177e4SLinus Torvalds * linux/fs/namespace.c 31da177e4SLinus Torvalds * 41da177e4SLinus Torvalds * (C) Copyright Al Viro 2000, 2001 51da177e4SLinus Torvalds * Released under GPL v2. 61da177e4SLinus Torvalds * 71da177e4SLinus Torvalds * Based on code from fs/super.c, copyright Linus Torvalds and others. 81da177e4SLinus Torvalds * Heavily rewritten. 91da177e4SLinus Torvalds */ 101da177e4SLinus Torvalds 111da177e4SLinus Torvalds #include <linux/syscalls.h> 121da177e4SLinus Torvalds #include <linux/slab.h> 131da177e4SLinus Torvalds #include <linux/sched.h> 141da177e4SLinus Torvalds #include <linux/smp_lock.h> 151da177e4SLinus Torvalds #include <linux/init.h> 1615a67dd8SRandy Dunlap #include <linux/kernel.h> 171da177e4SLinus Torvalds #include <linux/acct.h> 1816f7e0feSRandy Dunlap #include <linux/capability.h> 193d733633SDave Hansen #include <linux/cpumask.h> 201da177e4SLinus Torvalds #include <linux/module.h> 21f20a9eadSAndrew Morton #include <linux/sysfs.h> 221da177e4SLinus Torvalds #include <linux/seq_file.h> 236b3286edSKirill Korotaev #include <linux/mnt_namespace.h> 241da177e4SLinus Torvalds #include <linux/namei.h> 251da177e4SLinus Torvalds #include <linux/security.h> 261da177e4SLinus Torvalds #include <linux/mount.h> 2707f3f05cSDavid Howells #include <linux/ramfs.h> 2813f14b4dSEric Dumazet #include <linux/log2.h> 2973cd49ecSMiklos Szeredi #include <linux/idr.h> 301da177e4SLinus Torvalds #include <asm/uaccess.h> 311da177e4SLinus Torvalds #include <asm/unistd.h> 3207b20889SRam Pai #include "pnode.h" 33948730b0SAdrian Bunk #include "internal.h" 341da177e4SLinus Torvalds 3513f14b4dSEric Dumazet #define HASH_SHIFT ilog2(PAGE_SIZE / sizeof(struct list_head)) 3613f14b4dSEric Dumazet #define HASH_SIZE (1UL << HASH_SHIFT) 3713f14b4dSEric Dumazet 381da177e4SLinus Torvalds /* spinlock for vfsmount related operations, inplace of dcache_lock */ 391da177e4SLinus Torvalds __cacheline_aligned_in_smp DEFINE_SPINLOCK(vfsmount_lock); 401da177e4SLinus Torvalds 415addc5ddSAl Viro static int event; 4273cd49ecSMiklos Szeredi static DEFINE_IDA(mnt_id_ida); 43719f5d7fSMiklos Szeredi static DEFINE_IDA(mnt_group_ida); 445addc5ddSAl Viro 45fa3536ccSEric Dumazet static struct list_head *mount_hashtable __read_mostly; 46e18b890bSChristoph Lameter static struct kmem_cache *mnt_cache __read_mostly; 47390c6843SRam Pai static struct rw_semaphore namespace_sem; 481da177e4SLinus Torvalds 49f87fd4c2SMiklos Szeredi /* /sys/fs */ 5000d26666SGreg Kroah-Hartman struct kobject *fs_kobj; 5100d26666SGreg Kroah-Hartman EXPORT_SYMBOL_GPL(fs_kobj); 52f87fd4c2SMiklos Szeredi 531da177e4SLinus Torvalds static inline unsigned long hash(struct vfsmount *mnt, struct dentry *dentry) 541da177e4SLinus Torvalds { 551da177e4SLinus Torvalds unsigned long tmp = ((unsigned long)mnt / L1_CACHE_BYTES); 561da177e4SLinus Torvalds tmp += ((unsigned long)dentry / L1_CACHE_BYTES); 5713f14b4dSEric Dumazet tmp = tmp + (tmp >> HASH_SHIFT); 5813f14b4dSEric Dumazet return tmp & (HASH_SIZE - 1); 591da177e4SLinus Torvalds } 601da177e4SLinus Torvalds 613d733633SDave Hansen #define MNT_WRITER_UNDERFLOW_LIMIT -(1<<16) 623d733633SDave Hansen 6373cd49ecSMiklos Szeredi /* allocation is serialized by namespace_sem */ 6473cd49ecSMiklos Szeredi static int mnt_alloc_id(struct vfsmount *mnt) 6573cd49ecSMiklos Szeredi { 6673cd49ecSMiklos Szeredi int res; 6773cd49ecSMiklos Szeredi 6873cd49ecSMiklos Szeredi retry: 6973cd49ecSMiklos Szeredi ida_pre_get(&mnt_id_ida, GFP_KERNEL); 7073cd49ecSMiklos Szeredi spin_lock(&vfsmount_lock); 7173cd49ecSMiklos Szeredi res = ida_get_new(&mnt_id_ida, &mnt->mnt_id); 7273cd49ecSMiklos Szeredi spin_unlock(&vfsmount_lock); 7373cd49ecSMiklos Szeredi if (res == -EAGAIN) 7473cd49ecSMiklos Szeredi goto retry; 7573cd49ecSMiklos Szeredi 7673cd49ecSMiklos Szeredi return res; 7773cd49ecSMiklos Szeredi } 7873cd49ecSMiklos Szeredi 7973cd49ecSMiklos Szeredi static void mnt_free_id(struct vfsmount *mnt) 8073cd49ecSMiklos Szeredi { 8173cd49ecSMiklos Szeredi spin_lock(&vfsmount_lock); 8273cd49ecSMiklos Szeredi ida_remove(&mnt_id_ida, mnt->mnt_id); 8373cd49ecSMiklos Szeredi spin_unlock(&vfsmount_lock); 8473cd49ecSMiklos Szeredi } 8573cd49ecSMiklos Szeredi 86719f5d7fSMiklos Szeredi /* 87719f5d7fSMiklos Szeredi * Allocate a new peer group ID 88719f5d7fSMiklos Szeredi * 89719f5d7fSMiklos Szeredi * mnt_group_ida is protected by namespace_sem 90719f5d7fSMiklos Szeredi */ 91719f5d7fSMiklos Szeredi static int mnt_alloc_group_id(struct vfsmount *mnt) 92719f5d7fSMiklos Szeredi { 93719f5d7fSMiklos Szeredi if (!ida_pre_get(&mnt_group_ida, GFP_KERNEL)) 94719f5d7fSMiklos Szeredi return -ENOMEM; 95719f5d7fSMiklos Szeredi 96719f5d7fSMiklos Szeredi return ida_get_new_above(&mnt_group_ida, 1, &mnt->mnt_group_id); 97719f5d7fSMiklos Szeredi } 98719f5d7fSMiklos Szeredi 99719f5d7fSMiklos Szeredi /* 100719f5d7fSMiklos Szeredi * Release a peer group ID 101719f5d7fSMiklos Szeredi */ 102719f5d7fSMiklos Szeredi void mnt_release_group_id(struct vfsmount *mnt) 103719f5d7fSMiklos Szeredi { 104719f5d7fSMiklos Szeredi ida_remove(&mnt_group_ida, mnt->mnt_group_id); 105719f5d7fSMiklos Szeredi mnt->mnt_group_id = 0; 106719f5d7fSMiklos Szeredi } 107719f5d7fSMiklos Szeredi 1081da177e4SLinus Torvalds struct vfsmount *alloc_vfsmnt(const char *name) 1091da177e4SLinus Torvalds { 110c3762229SRobert P. J. Day struct vfsmount *mnt = kmem_cache_zalloc(mnt_cache, GFP_KERNEL); 1111da177e4SLinus Torvalds if (mnt) { 11273cd49ecSMiklos Szeredi int err; 11373cd49ecSMiklos Szeredi 11473cd49ecSMiklos Szeredi err = mnt_alloc_id(mnt); 11588b38782SLi Zefan if (err) 11688b38782SLi Zefan goto out_free_cache; 11788b38782SLi Zefan 11888b38782SLi Zefan if (name) { 11988b38782SLi Zefan mnt->mnt_devname = kstrdup(name, GFP_KERNEL); 12088b38782SLi Zefan if (!mnt->mnt_devname) 12188b38782SLi Zefan goto out_free_id; 12273cd49ecSMiklos Szeredi } 12373cd49ecSMiklos Szeredi 1241da177e4SLinus Torvalds atomic_set(&mnt->mnt_count, 1); 1251da177e4SLinus Torvalds INIT_LIST_HEAD(&mnt->mnt_hash); 1261da177e4SLinus Torvalds INIT_LIST_HEAD(&mnt->mnt_child); 1271da177e4SLinus Torvalds INIT_LIST_HEAD(&mnt->mnt_mounts); 1281da177e4SLinus Torvalds INIT_LIST_HEAD(&mnt->mnt_list); 12955e700b9SMiklos Szeredi INIT_LIST_HEAD(&mnt->mnt_expire); 13003e06e68SRam Pai INIT_LIST_HEAD(&mnt->mnt_share); 131a58b0eb8SRam Pai INIT_LIST_HEAD(&mnt->mnt_slave_list); 132a58b0eb8SRam Pai INIT_LIST_HEAD(&mnt->mnt_slave); 1333d733633SDave Hansen atomic_set(&mnt->__mnt_writers, 0); 1341da177e4SLinus Torvalds } 1351da177e4SLinus Torvalds return mnt; 13688b38782SLi Zefan 13788b38782SLi Zefan out_free_id: 13888b38782SLi Zefan mnt_free_id(mnt); 13988b38782SLi Zefan out_free_cache: 14088b38782SLi Zefan kmem_cache_free(mnt_cache, mnt); 14188b38782SLi Zefan return NULL; 1421da177e4SLinus Torvalds } 1431da177e4SLinus Torvalds 1448366025eSDave Hansen /* 1458366025eSDave Hansen * Most r/o checks on a fs are for operations that take 1468366025eSDave Hansen * discrete amounts of time, like a write() or unlink(). 1478366025eSDave Hansen * We must keep track of when those operations start 1488366025eSDave Hansen * (for permission checks) and when they end, so that 1498366025eSDave Hansen * we can determine when writes are able to occur to 1508366025eSDave Hansen * a filesystem. 1518366025eSDave Hansen */ 1523d733633SDave Hansen /* 1533d733633SDave Hansen * __mnt_is_readonly: check whether a mount is read-only 1543d733633SDave Hansen * @mnt: the mount to check for its write status 1553d733633SDave Hansen * 1563d733633SDave Hansen * This shouldn't be used directly ouside of the VFS. 1573d733633SDave Hansen * It does not guarantee that the filesystem will stay 1583d733633SDave Hansen * r/w, just that it is right *now*. This can not and 1593d733633SDave Hansen * should not be used in place of IS_RDONLY(inode). 1603d733633SDave Hansen * mnt_want/drop_write() will _keep_ the filesystem 1613d733633SDave Hansen * r/w. 1623d733633SDave Hansen */ 1633d733633SDave Hansen int __mnt_is_readonly(struct vfsmount *mnt) 1643d733633SDave Hansen { 1652e4b7fcdSDave Hansen if (mnt->mnt_flags & MNT_READONLY) 1662e4b7fcdSDave Hansen return 1; 1672e4b7fcdSDave Hansen if (mnt->mnt_sb->s_flags & MS_RDONLY) 1682e4b7fcdSDave Hansen return 1; 1692e4b7fcdSDave Hansen return 0; 1703d733633SDave Hansen } 1713d733633SDave Hansen EXPORT_SYMBOL_GPL(__mnt_is_readonly); 1723d733633SDave Hansen 1733d733633SDave Hansen struct mnt_writer { 1743d733633SDave Hansen /* 1753d733633SDave Hansen * If holding multiple instances of this lock, they 1763d733633SDave Hansen * must be ordered by cpu number. 1773d733633SDave Hansen */ 1783d733633SDave Hansen spinlock_t lock; 1793d733633SDave Hansen struct lock_class_key lock_class; /* compiles out with !lockdep */ 1803d733633SDave Hansen unsigned long count; 1813d733633SDave Hansen struct vfsmount *mnt; 1823d733633SDave Hansen } ____cacheline_aligned_in_smp; 1833d733633SDave Hansen static DEFINE_PER_CPU(struct mnt_writer, mnt_writers); 1843d733633SDave Hansen 1853d733633SDave Hansen static int __init init_mnt_writers(void) 1863d733633SDave Hansen { 1873d733633SDave Hansen int cpu; 1883d733633SDave Hansen for_each_possible_cpu(cpu) { 1893d733633SDave Hansen struct mnt_writer *writer = &per_cpu(mnt_writers, cpu); 1903d733633SDave Hansen spin_lock_init(&writer->lock); 1913d733633SDave Hansen lockdep_set_class(&writer->lock, &writer->lock_class); 1923d733633SDave Hansen writer->count = 0; 1933d733633SDave Hansen } 1943d733633SDave Hansen return 0; 1953d733633SDave Hansen } 1963d733633SDave Hansen fs_initcall(init_mnt_writers); 1973d733633SDave Hansen 1983d733633SDave Hansen static void unlock_mnt_writers(void) 1993d733633SDave Hansen { 2003d733633SDave Hansen int cpu; 2013d733633SDave Hansen struct mnt_writer *cpu_writer; 2023d733633SDave Hansen 2033d733633SDave Hansen for_each_possible_cpu(cpu) { 2043d733633SDave Hansen cpu_writer = &per_cpu(mnt_writers, cpu); 2053d733633SDave Hansen spin_unlock(&cpu_writer->lock); 2063d733633SDave Hansen } 2073d733633SDave Hansen } 2083d733633SDave Hansen 2093d733633SDave Hansen static inline void __clear_mnt_count(struct mnt_writer *cpu_writer) 2103d733633SDave Hansen { 2113d733633SDave Hansen if (!cpu_writer->mnt) 2123d733633SDave Hansen return; 2133d733633SDave Hansen /* 2143d733633SDave Hansen * This is in case anyone ever leaves an invalid, 2153d733633SDave Hansen * old ->mnt and a count of 0. 2163d733633SDave Hansen */ 2173d733633SDave Hansen if (!cpu_writer->count) 2183d733633SDave Hansen return; 2193d733633SDave Hansen atomic_add(cpu_writer->count, &cpu_writer->mnt->__mnt_writers); 2203d733633SDave Hansen cpu_writer->count = 0; 2213d733633SDave Hansen } 2223d733633SDave Hansen /* 2233d733633SDave Hansen * must hold cpu_writer->lock 2243d733633SDave Hansen */ 2253d733633SDave Hansen static inline void use_cpu_writer_for_mount(struct mnt_writer *cpu_writer, 2263d733633SDave Hansen struct vfsmount *mnt) 2273d733633SDave Hansen { 2283d733633SDave Hansen if (cpu_writer->mnt == mnt) 2293d733633SDave Hansen return; 2303d733633SDave Hansen __clear_mnt_count(cpu_writer); 2313d733633SDave Hansen cpu_writer->mnt = mnt; 2323d733633SDave Hansen } 2333d733633SDave Hansen 2343d733633SDave Hansen /* 2353d733633SDave Hansen * Most r/o checks on a fs are for operations that take 2363d733633SDave Hansen * discrete amounts of time, like a write() or unlink(). 2373d733633SDave Hansen * We must keep track of when those operations start 2383d733633SDave Hansen * (for permission checks) and when they end, so that 2393d733633SDave Hansen * we can determine when writes are able to occur to 2403d733633SDave Hansen * a filesystem. 2413d733633SDave Hansen */ 2428366025eSDave Hansen /** 2438366025eSDave Hansen * mnt_want_write - get write access to a mount 2448366025eSDave Hansen * @mnt: the mount on which to take a write 2458366025eSDave Hansen * 2468366025eSDave Hansen * This tells the low-level filesystem that a write is 2478366025eSDave Hansen * about to be performed to it, and makes sure that 2488366025eSDave Hansen * writes are allowed before returning success. When 2498366025eSDave Hansen * the write operation is finished, mnt_drop_write() 2508366025eSDave Hansen * must be called. This is effectively a refcount. 2518366025eSDave Hansen */ 2528366025eSDave Hansen int mnt_want_write(struct vfsmount *mnt) 2538366025eSDave Hansen { 2543d733633SDave Hansen int ret = 0; 2553d733633SDave Hansen struct mnt_writer *cpu_writer; 2563d733633SDave Hansen 2573d733633SDave Hansen cpu_writer = &get_cpu_var(mnt_writers); 2583d733633SDave Hansen spin_lock(&cpu_writer->lock); 2593d733633SDave Hansen if (__mnt_is_readonly(mnt)) { 2603d733633SDave Hansen ret = -EROFS; 2613d733633SDave Hansen goto out; 2623d733633SDave Hansen } 2633d733633SDave Hansen use_cpu_writer_for_mount(cpu_writer, mnt); 2643d733633SDave Hansen cpu_writer->count++; 2653d733633SDave Hansen out: 2663d733633SDave Hansen spin_unlock(&cpu_writer->lock); 2673d733633SDave Hansen put_cpu_var(mnt_writers); 2683d733633SDave Hansen return ret; 2698366025eSDave Hansen } 2708366025eSDave Hansen EXPORT_SYMBOL_GPL(mnt_want_write); 2718366025eSDave Hansen 2723d733633SDave Hansen static void lock_mnt_writers(void) 2733d733633SDave Hansen { 2743d733633SDave Hansen int cpu; 2753d733633SDave Hansen struct mnt_writer *cpu_writer; 2763d733633SDave Hansen 2773d733633SDave Hansen for_each_possible_cpu(cpu) { 2783d733633SDave Hansen cpu_writer = &per_cpu(mnt_writers, cpu); 2793d733633SDave Hansen spin_lock(&cpu_writer->lock); 2803d733633SDave Hansen __clear_mnt_count(cpu_writer); 2813d733633SDave Hansen cpu_writer->mnt = NULL; 2823d733633SDave Hansen } 2833d733633SDave Hansen } 2843d733633SDave Hansen 2853d733633SDave Hansen /* 2863d733633SDave Hansen * These per-cpu write counts are not guaranteed to have 2873d733633SDave Hansen * matched increments and decrements on any given cpu. 2883d733633SDave Hansen * A file open()ed for write on one cpu and close()d on 2893d733633SDave Hansen * another cpu will imbalance this count. Make sure it 2903d733633SDave Hansen * does not get too far out of whack. 2913d733633SDave Hansen */ 2923d733633SDave Hansen static void handle_write_count_underflow(struct vfsmount *mnt) 2933d733633SDave Hansen { 2943d733633SDave Hansen if (atomic_read(&mnt->__mnt_writers) >= 2953d733633SDave Hansen MNT_WRITER_UNDERFLOW_LIMIT) 2963d733633SDave Hansen return; 2973d733633SDave Hansen /* 2983d733633SDave Hansen * It isn't necessary to hold all of the locks 2993d733633SDave Hansen * at the same time, but doing it this way makes 3003d733633SDave Hansen * us share a lot more code. 3013d733633SDave Hansen */ 3023d733633SDave Hansen lock_mnt_writers(); 3033d733633SDave Hansen /* 3043d733633SDave Hansen * vfsmount_lock is for mnt_flags. 3053d733633SDave Hansen */ 3063d733633SDave Hansen spin_lock(&vfsmount_lock); 3073d733633SDave Hansen /* 3083d733633SDave Hansen * If coalescing the per-cpu writer counts did not 3093d733633SDave Hansen * get us back to a positive writer count, we have 3103d733633SDave Hansen * a bug. 3113d733633SDave Hansen */ 3123d733633SDave Hansen if ((atomic_read(&mnt->__mnt_writers) < 0) && 3133d733633SDave Hansen !(mnt->mnt_flags & MNT_IMBALANCED_WRITE_COUNT)) { 3145c752ad9SArjan van de Ven WARN(1, KERN_DEBUG "leak detected on mount(%p) writers " 3153d733633SDave Hansen "count: %d\n", 3163d733633SDave Hansen mnt, atomic_read(&mnt->__mnt_writers)); 3173d733633SDave Hansen /* use the flag to keep the dmesg spam down */ 3183d733633SDave Hansen mnt->mnt_flags |= MNT_IMBALANCED_WRITE_COUNT; 3193d733633SDave Hansen } 3203d733633SDave Hansen spin_unlock(&vfsmount_lock); 3213d733633SDave Hansen unlock_mnt_writers(); 3223d733633SDave Hansen } 3233d733633SDave Hansen 3248366025eSDave Hansen /** 3258366025eSDave Hansen * mnt_drop_write - give up write access to a mount 3268366025eSDave Hansen * @mnt: the mount on which to give up write access 3278366025eSDave Hansen * 3288366025eSDave Hansen * Tells the low-level filesystem that we are done 3298366025eSDave Hansen * performing writes to it. Must be matched with 3308366025eSDave Hansen * mnt_want_write() call above. 3318366025eSDave Hansen */ 3328366025eSDave Hansen void mnt_drop_write(struct vfsmount *mnt) 3338366025eSDave Hansen { 3343d733633SDave Hansen int must_check_underflow = 0; 3353d733633SDave Hansen struct mnt_writer *cpu_writer; 3363d733633SDave Hansen 3373d733633SDave Hansen cpu_writer = &get_cpu_var(mnt_writers); 3383d733633SDave Hansen spin_lock(&cpu_writer->lock); 3393d733633SDave Hansen 3403d733633SDave Hansen use_cpu_writer_for_mount(cpu_writer, mnt); 3413d733633SDave Hansen if (cpu_writer->count > 0) { 3423d733633SDave Hansen cpu_writer->count--; 3433d733633SDave Hansen } else { 3443d733633SDave Hansen must_check_underflow = 1; 3453d733633SDave Hansen atomic_dec(&mnt->__mnt_writers); 3463d733633SDave Hansen } 3473d733633SDave Hansen 3483d733633SDave Hansen spin_unlock(&cpu_writer->lock); 3493d733633SDave Hansen /* 3503d733633SDave Hansen * Logically, we could call this each time, 3513d733633SDave Hansen * but the __mnt_writers cacheline tends to 3523d733633SDave Hansen * be cold, and makes this expensive. 3533d733633SDave Hansen */ 3543d733633SDave Hansen if (must_check_underflow) 3553d733633SDave Hansen handle_write_count_underflow(mnt); 3563d733633SDave Hansen /* 3573d733633SDave Hansen * This could be done right after the spinlock 3583d733633SDave Hansen * is taken because the spinlock keeps us on 3593d733633SDave Hansen * the cpu, and disables preemption. However, 3603d733633SDave Hansen * putting it here bounds the amount that 3613d733633SDave Hansen * __mnt_writers can underflow. Without it, 3623d733633SDave Hansen * we could theoretically wrap __mnt_writers. 3633d733633SDave Hansen */ 3643d733633SDave Hansen put_cpu_var(mnt_writers); 3658366025eSDave Hansen } 3668366025eSDave Hansen EXPORT_SYMBOL_GPL(mnt_drop_write); 3678366025eSDave Hansen 3682e4b7fcdSDave Hansen static int mnt_make_readonly(struct vfsmount *mnt) 3698366025eSDave Hansen { 3703d733633SDave Hansen int ret = 0; 3713d733633SDave Hansen 3723d733633SDave Hansen lock_mnt_writers(); 3733d733633SDave Hansen /* 3743d733633SDave Hansen * With all the locks held, this value is stable 3753d733633SDave Hansen */ 3763d733633SDave Hansen if (atomic_read(&mnt->__mnt_writers) > 0) { 3773d733633SDave Hansen ret = -EBUSY; 3783d733633SDave Hansen goto out; 3798366025eSDave Hansen } 3803d733633SDave Hansen /* 3812e4b7fcdSDave Hansen * nobody can do a successful mnt_want_write() with all 3822e4b7fcdSDave Hansen * of the counts in MNT_DENIED_WRITE and the locks held. 3833d733633SDave Hansen */ 3842e4b7fcdSDave Hansen spin_lock(&vfsmount_lock); 3852e4b7fcdSDave Hansen if (!ret) 3862e4b7fcdSDave Hansen mnt->mnt_flags |= MNT_READONLY; 3872e4b7fcdSDave Hansen spin_unlock(&vfsmount_lock); 3883d733633SDave Hansen out: 3893d733633SDave Hansen unlock_mnt_writers(); 3903d733633SDave Hansen return ret; 3913d733633SDave Hansen } 3928366025eSDave Hansen 3932e4b7fcdSDave Hansen static void __mnt_unmake_readonly(struct vfsmount *mnt) 3942e4b7fcdSDave Hansen { 3952e4b7fcdSDave Hansen spin_lock(&vfsmount_lock); 3962e4b7fcdSDave Hansen mnt->mnt_flags &= ~MNT_READONLY; 3972e4b7fcdSDave Hansen spin_unlock(&vfsmount_lock); 3982e4b7fcdSDave Hansen } 3992e4b7fcdSDave Hansen 400454e2398SDavid Howells int simple_set_mnt(struct vfsmount *mnt, struct super_block *sb) 401454e2398SDavid Howells { 402454e2398SDavid Howells mnt->mnt_sb = sb; 403454e2398SDavid Howells mnt->mnt_root = dget(sb->s_root); 404454e2398SDavid Howells return 0; 405454e2398SDavid Howells } 406454e2398SDavid Howells 407454e2398SDavid Howells EXPORT_SYMBOL(simple_set_mnt); 408454e2398SDavid Howells 4091da177e4SLinus Torvalds void free_vfsmnt(struct vfsmount *mnt) 4101da177e4SLinus Torvalds { 4111da177e4SLinus Torvalds kfree(mnt->mnt_devname); 41273cd49ecSMiklos Szeredi mnt_free_id(mnt); 4131da177e4SLinus Torvalds kmem_cache_free(mnt_cache, mnt); 4141da177e4SLinus Torvalds } 4151da177e4SLinus Torvalds 4161da177e4SLinus Torvalds /* 417a05964f3SRam Pai * find the first or last mount at @dentry on vfsmount @mnt depending on 418a05964f3SRam Pai * @dir. If @dir is set return the first mount else return the last mount. 4191da177e4SLinus Torvalds */ 420a05964f3SRam Pai struct vfsmount *__lookup_mnt(struct vfsmount *mnt, struct dentry *dentry, 421a05964f3SRam Pai int dir) 4221da177e4SLinus Torvalds { 4231da177e4SLinus Torvalds struct list_head *head = mount_hashtable + hash(mnt, dentry); 4241da177e4SLinus Torvalds struct list_head *tmp = head; 4251da177e4SLinus Torvalds struct vfsmount *p, *found = NULL; 4261da177e4SLinus Torvalds 4271da177e4SLinus Torvalds for (;;) { 428a05964f3SRam Pai tmp = dir ? tmp->next : tmp->prev; 4291da177e4SLinus Torvalds p = NULL; 4301da177e4SLinus Torvalds if (tmp == head) 4311da177e4SLinus Torvalds break; 4321da177e4SLinus Torvalds p = list_entry(tmp, struct vfsmount, mnt_hash); 4331da177e4SLinus Torvalds if (p->mnt_parent == mnt && p->mnt_mountpoint == dentry) { 434a05964f3SRam Pai found = p; 4351da177e4SLinus Torvalds break; 4361da177e4SLinus Torvalds } 4371da177e4SLinus Torvalds } 4381da177e4SLinus Torvalds return found; 4391da177e4SLinus Torvalds } 4401da177e4SLinus Torvalds 441a05964f3SRam Pai /* 442a05964f3SRam Pai * lookup_mnt increments the ref count before returning 443a05964f3SRam Pai * the vfsmount struct. 444a05964f3SRam Pai */ 445a05964f3SRam Pai struct vfsmount *lookup_mnt(struct vfsmount *mnt, struct dentry *dentry) 446a05964f3SRam Pai { 447a05964f3SRam Pai struct vfsmount *child_mnt; 448a05964f3SRam Pai spin_lock(&vfsmount_lock); 449a05964f3SRam Pai if ((child_mnt = __lookup_mnt(mnt, dentry, 1))) 450a05964f3SRam Pai mntget(child_mnt); 451a05964f3SRam Pai spin_unlock(&vfsmount_lock); 452a05964f3SRam Pai return child_mnt; 453a05964f3SRam Pai } 454a05964f3SRam Pai 4551da177e4SLinus Torvalds static inline int check_mnt(struct vfsmount *mnt) 4561da177e4SLinus Torvalds { 4576b3286edSKirill Korotaev return mnt->mnt_ns == current->nsproxy->mnt_ns; 4581da177e4SLinus Torvalds } 4591da177e4SLinus Torvalds 4606b3286edSKirill Korotaev static void touch_mnt_namespace(struct mnt_namespace *ns) 4615addc5ddSAl Viro { 4625addc5ddSAl Viro if (ns) { 4635addc5ddSAl Viro ns->event = ++event; 4645addc5ddSAl Viro wake_up_interruptible(&ns->poll); 4655addc5ddSAl Viro } 4665addc5ddSAl Viro } 4675addc5ddSAl Viro 4686b3286edSKirill Korotaev static void __touch_mnt_namespace(struct mnt_namespace *ns) 4695addc5ddSAl Viro { 4705addc5ddSAl Viro if (ns && ns->event != event) { 4715addc5ddSAl Viro ns->event = event; 4725addc5ddSAl Viro wake_up_interruptible(&ns->poll); 4735addc5ddSAl Viro } 4745addc5ddSAl Viro } 4755addc5ddSAl Viro 4761a390689SAl Viro static void detach_mnt(struct vfsmount *mnt, struct path *old_path) 4771da177e4SLinus Torvalds { 4781a390689SAl Viro old_path->dentry = mnt->mnt_mountpoint; 4791a390689SAl Viro old_path->mnt = mnt->mnt_parent; 4801da177e4SLinus Torvalds mnt->mnt_parent = mnt; 4811da177e4SLinus Torvalds mnt->mnt_mountpoint = mnt->mnt_root; 4821da177e4SLinus Torvalds list_del_init(&mnt->mnt_child); 4831da177e4SLinus Torvalds list_del_init(&mnt->mnt_hash); 4841a390689SAl Viro old_path->dentry->d_mounted--; 4851da177e4SLinus Torvalds } 4861da177e4SLinus Torvalds 487b90fa9aeSRam Pai void mnt_set_mountpoint(struct vfsmount *mnt, struct dentry *dentry, 488b90fa9aeSRam Pai struct vfsmount *child_mnt) 489b90fa9aeSRam Pai { 490b90fa9aeSRam Pai child_mnt->mnt_parent = mntget(mnt); 491b90fa9aeSRam Pai child_mnt->mnt_mountpoint = dget(dentry); 492b90fa9aeSRam Pai dentry->d_mounted++; 493b90fa9aeSRam Pai } 494b90fa9aeSRam Pai 4951a390689SAl Viro static void attach_mnt(struct vfsmount *mnt, struct path *path) 4961da177e4SLinus Torvalds { 4971a390689SAl Viro mnt_set_mountpoint(path->mnt, path->dentry, mnt); 498b90fa9aeSRam Pai list_add_tail(&mnt->mnt_hash, mount_hashtable + 4991a390689SAl Viro hash(path->mnt, path->dentry)); 5001a390689SAl Viro list_add_tail(&mnt->mnt_child, &path->mnt->mnt_mounts); 501b90fa9aeSRam Pai } 502b90fa9aeSRam Pai 503b90fa9aeSRam Pai /* 504b90fa9aeSRam Pai * the caller must hold vfsmount_lock 505b90fa9aeSRam Pai */ 506b90fa9aeSRam Pai static void commit_tree(struct vfsmount *mnt) 507b90fa9aeSRam Pai { 508b90fa9aeSRam Pai struct vfsmount *parent = mnt->mnt_parent; 509b90fa9aeSRam Pai struct vfsmount *m; 510b90fa9aeSRam Pai LIST_HEAD(head); 5116b3286edSKirill Korotaev struct mnt_namespace *n = parent->mnt_ns; 512b90fa9aeSRam Pai 513b90fa9aeSRam Pai BUG_ON(parent == mnt); 514b90fa9aeSRam Pai 515b90fa9aeSRam Pai list_add_tail(&head, &mnt->mnt_list); 516b90fa9aeSRam Pai list_for_each_entry(m, &head, mnt_list) 5176b3286edSKirill Korotaev m->mnt_ns = n; 518b90fa9aeSRam Pai list_splice(&head, n->list.prev); 519b90fa9aeSRam Pai 520b90fa9aeSRam Pai list_add_tail(&mnt->mnt_hash, mount_hashtable + 521b90fa9aeSRam Pai hash(parent, mnt->mnt_mountpoint)); 522b90fa9aeSRam Pai list_add_tail(&mnt->mnt_child, &parent->mnt_mounts); 5236b3286edSKirill Korotaev touch_mnt_namespace(n); 5241da177e4SLinus Torvalds } 5251da177e4SLinus Torvalds 5261da177e4SLinus Torvalds static struct vfsmount *next_mnt(struct vfsmount *p, struct vfsmount *root) 5271da177e4SLinus Torvalds { 5281da177e4SLinus Torvalds struct list_head *next = p->mnt_mounts.next; 5291da177e4SLinus Torvalds if (next == &p->mnt_mounts) { 5301da177e4SLinus Torvalds while (1) { 5311da177e4SLinus Torvalds if (p == root) 5321da177e4SLinus Torvalds return NULL; 5331da177e4SLinus Torvalds next = p->mnt_child.next; 5341da177e4SLinus Torvalds if (next != &p->mnt_parent->mnt_mounts) 5351da177e4SLinus Torvalds break; 5361da177e4SLinus Torvalds p = p->mnt_parent; 5371da177e4SLinus Torvalds } 5381da177e4SLinus Torvalds } 5391da177e4SLinus Torvalds return list_entry(next, struct vfsmount, mnt_child); 5401da177e4SLinus Torvalds } 5411da177e4SLinus Torvalds 5429676f0c6SRam Pai static struct vfsmount *skip_mnt_tree(struct vfsmount *p) 5439676f0c6SRam Pai { 5449676f0c6SRam Pai struct list_head *prev = p->mnt_mounts.prev; 5459676f0c6SRam Pai while (prev != &p->mnt_mounts) { 5469676f0c6SRam Pai p = list_entry(prev, struct vfsmount, mnt_child); 5479676f0c6SRam Pai prev = p->mnt_mounts.prev; 5489676f0c6SRam Pai } 5499676f0c6SRam Pai return p; 5509676f0c6SRam Pai } 5519676f0c6SRam Pai 55236341f64SRam Pai static struct vfsmount *clone_mnt(struct vfsmount *old, struct dentry *root, 55336341f64SRam Pai int flag) 5541da177e4SLinus Torvalds { 5551da177e4SLinus Torvalds struct super_block *sb = old->mnt_sb; 5561da177e4SLinus Torvalds struct vfsmount *mnt = alloc_vfsmnt(old->mnt_devname); 5571da177e4SLinus Torvalds 5581da177e4SLinus Torvalds if (mnt) { 559719f5d7fSMiklos Szeredi if (flag & (CL_SLAVE | CL_PRIVATE)) 560719f5d7fSMiklos Szeredi mnt->mnt_group_id = 0; /* not a peer of original */ 561719f5d7fSMiklos Szeredi else 562719f5d7fSMiklos Szeredi mnt->mnt_group_id = old->mnt_group_id; 563719f5d7fSMiklos Szeredi 564719f5d7fSMiklos Szeredi if ((flag & CL_MAKE_SHARED) && !mnt->mnt_group_id) { 565719f5d7fSMiklos Szeredi int err = mnt_alloc_group_id(mnt); 566719f5d7fSMiklos Szeredi if (err) 567719f5d7fSMiklos Szeredi goto out_free; 568719f5d7fSMiklos Szeredi } 569719f5d7fSMiklos Szeredi 5701da177e4SLinus Torvalds mnt->mnt_flags = old->mnt_flags; 5711da177e4SLinus Torvalds atomic_inc(&sb->s_active); 5721da177e4SLinus Torvalds mnt->mnt_sb = sb; 5731da177e4SLinus Torvalds mnt->mnt_root = dget(root); 5741da177e4SLinus Torvalds mnt->mnt_mountpoint = mnt->mnt_root; 5751da177e4SLinus Torvalds mnt->mnt_parent = mnt; 576b90fa9aeSRam Pai 5775afe0022SRam Pai if (flag & CL_SLAVE) { 5785afe0022SRam Pai list_add(&mnt->mnt_slave, &old->mnt_slave_list); 5795afe0022SRam Pai mnt->mnt_master = old; 5805afe0022SRam Pai CLEAR_MNT_SHARED(mnt); 5818aec0809SAl Viro } else if (!(flag & CL_PRIVATE)) { 582b90fa9aeSRam Pai if ((flag & CL_PROPAGATION) || IS_MNT_SHARED(old)) 583b90fa9aeSRam Pai list_add(&mnt->mnt_share, &old->mnt_share); 5845afe0022SRam Pai if (IS_MNT_SLAVE(old)) 5855afe0022SRam Pai list_add(&mnt->mnt_slave, &old->mnt_slave); 5865afe0022SRam Pai mnt->mnt_master = old->mnt_master; 5875afe0022SRam Pai } 588b90fa9aeSRam Pai if (flag & CL_MAKE_SHARED) 589b90fa9aeSRam Pai set_mnt_shared(mnt); 5901da177e4SLinus Torvalds 5911da177e4SLinus Torvalds /* stick the duplicate mount on the same expiry list 5921da177e4SLinus Torvalds * as the original if that was on one */ 59336341f64SRam Pai if (flag & CL_EXPIRE) { 59455e700b9SMiklos Szeredi if (!list_empty(&old->mnt_expire)) 59555e700b9SMiklos Szeredi list_add(&mnt->mnt_expire, &old->mnt_expire); 5961da177e4SLinus Torvalds } 59736341f64SRam Pai } 5981da177e4SLinus Torvalds return mnt; 599719f5d7fSMiklos Szeredi 600719f5d7fSMiklos Szeredi out_free: 601719f5d7fSMiklos Szeredi free_vfsmnt(mnt); 602719f5d7fSMiklos Szeredi return NULL; 6031da177e4SLinus Torvalds } 6041da177e4SLinus Torvalds 6057b7b1aceSAl Viro static inline void __mntput(struct vfsmount *mnt) 6061da177e4SLinus Torvalds { 6073d733633SDave Hansen int cpu; 6081da177e4SLinus Torvalds struct super_block *sb = mnt->mnt_sb; 6093d733633SDave Hansen /* 6103d733633SDave Hansen * We don't have to hold all of the locks at the 6113d733633SDave Hansen * same time here because we know that we're the 6123d733633SDave Hansen * last reference to mnt and that no new writers 6133d733633SDave Hansen * can come in. 6143d733633SDave Hansen */ 6153d733633SDave Hansen for_each_possible_cpu(cpu) { 6163d733633SDave Hansen struct mnt_writer *cpu_writer = &per_cpu(mnt_writers, cpu); 6173d733633SDave Hansen if (cpu_writer->mnt != mnt) 6183d733633SDave Hansen continue; 6193d733633SDave Hansen spin_lock(&cpu_writer->lock); 6203d733633SDave Hansen atomic_add(cpu_writer->count, &mnt->__mnt_writers); 6213d733633SDave Hansen cpu_writer->count = 0; 6223d733633SDave Hansen /* 6233d733633SDave Hansen * Might as well do this so that no one 6243d733633SDave Hansen * ever sees the pointer and expects 6253d733633SDave Hansen * it to be valid. 6263d733633SDave Hansen */ 6273d733633SDave Hansen cpu_writer->mnt = NULL; 6283d733633SDave Hansen spin_unlock(&cpu_writer->lock); 6293d733633SDave Hansen } 6303d733633SDave Hansen /* 6313d733633SDave Hansen * This probably indicates that somebody messed 6323d733633SDave Hansen * up a mnt_want/drop_write() pair. If this 6333d733633SDave Hansen * happens, the filesystem was probably unable 6343d733633SDave Hansen * to make r/w->r/o transitions. 6353d733633SDave Hansen */ 6363d733633SDave Hansen WARN_ON(atomic_read(&mnt->__mnt_writers)); 6371da177e4SLinus Torvalds dput(mnt->mnt_root); 6381da177e4SLinus Torvalds free_vfsmnt(mnt); 6391da177e4SLinus Torvalds deactivate_super(sb); 6401da177e4SLinus Torvalds } 6411da177e4SLinus Torvalds 6427b7b1aceSAl Viro void mntput_no_expire(struct vfsmount *mnt) 6437b7b1aceSAl Viro { 6447b7b1aceSAl Viro repeat: 6457b7b1aceSAl Viro if (atomic_dec_and_lock(&mnt->mnt_count, &vfsmount_lock)) { 6467b7b1aceSAl Viro if (likely(!mnt->mnt_pinned)) { 6477b7b1aceSAl Viro spin_unlock(&vfsmount_lock); 6487b7b1aceSAl Viro __mntput(mnt); 6497b7b1aceSAl Viro return; 6507b7b1aceSAl Viro } 6517b7b1aceSAl Viro atomic_add(mnt->mnt_pinned + 1, &mnt->mnt_count); 6527b7b1aceSAl Viro mnt->mnt_pinned = 0; 6537b7b1aceSAl Viro spin_unlock(&vfsmount_lock); 6547b7b1aceSAl Viro acct_auto_close_mnt(mnt); 6557b7b1aceSAl Viro security_sb_umount_close(mnt); 6567b7b1aceSAl Viro goto repeat; 6577b7b1aceSAl Viro } 6587b7b1aceSAl Viro } 6597b7b1aceSAl Viro 6607b7b1aceSAl Viro EXPORT_SYMBOL(mntput_no_expire); 6617b7b1aceSAl Viro 6627b7b1aceSAl Viro void mnt_pin(struct vfsmount *mnt) 6637b7b1aceSAl Viro { 6647b7b1aceSAl Viro spin_lock(&vfsmount_lock); 6657b7b1aceSAl Viro mnt->mnt_pinned++; 6667b7b1aceSAl Viro spin_unlock(&vfsmount_lock); 6677b7b1aceSAl Viro } 6687b7b1aceSAl Viro 6697b7b1aceSAl Viro EXPORT_SYMBOL(mnt_pin); 6707b7b1aceSAl Viro 6717b7b1aceSAl Viro void mnt_unpin(struct vfsmount *mnt) 6727b7b1aceSAl Viro { 6737b7b1aceSAl Viro spin_lock(&vfsmount_lock); 6747b7b1aceSAl Viro if (mnt->mnt_pinned) { 6757b7b1aceSAl Viro atomic_inc(&mnt->mnt_count); 6767b7b1aceSAl Viro mnt->mnt_pinned--; 6777b7b1aceSAl Viro } 6787b7b1aceSAl Viro spin_unlock(&vfsmount_lock); 6797b7b1aceSAl Viro } 6807b7b1aceSAl Viro 6817b7b1aceSAl Viro EXPORT_SYMBOL(mnt_unpin); 6821da177e4SLinus Torvalds 683b3b304a2SMiklos Szeredi static inline void mangle(struct seq_file *m, const char *s) 684b3b304a2SMiklos Szeredi { 685b3b304a2SMiklos Szeredi seq_escape(m, s, " \t\n\\"); 686b3b304a2SMiklos Szeredi } 687b3b304a2SMiklos Szeredi 688b3b304a2SMiklos Szeredi /* 689b3b304a2SMiklos Szeredi * Simple .show_options callback for filesystems which don't want to 690b3b304a2SMiklos Szeredi * implement more complex mount option showing. 691b3b304a2SMiklos Szeredi * 692b3b304a2SMiklos Szeredi * See also save_mount_options(). 693b3b304a2SMiklos Szeredi */ 694b3b304a2SMiklos Szeredi int generic_show_options(struct seq_file *m, struct vfsmount *mnt) 695b3b304a2SMiklos Szeredi { 696b3b304a2SMiklos Szeredi const char *options = mnt->mnt_sb->s_options; 697b3b304a2SMiklos Szeredi 698b3b304a2SMiklos Szeredi if (options != NULL && options[0]) { 699b3b304a2SMiklos Szeredi seq_putc(m, ','); 700b3b304a2SMiklos Szeredi mangle(m, options); 701b3b304a2SMiklos Szeredi } 702b3b304a2SMiklos Szeredi 703b3b304a2SMiklos Szeredi return 0; 704b3b304a2SMiklos Szeredi } 705b3b304a2SMiklos Szeredi EXPORT_SYMBOL(generic_show_options); 706b3b304a2SMiklos Szeredi 707b3b304a2SMiklos Szeredi /* 708b3b304a2SMiklos Szeredi * If filesystem uses generic_show_options(), this function should be 709b3b304a2SMiklos Szeredi * called from the fill_super() callback. 710b3b304a2SMiklos Szeredi * 711b3b304a2SMiklos Szeredi * The .remount_fs callback usually needs to be handled in a special 712b3b304a2SMiklos Szeredi * way, to make sure, that previous options are not overwritten if the 713b3b304a2SMiklos Szeredi * remount fails. 714b3b304a2SMiklos Szeredi * 715b3b304a2SMiklos Szeredi * Also note, that if the filesystem's .remount_fs function doesn't 716b3b304a2SMiklos Szeredi * reset all options to their default value, but changes only newly 717b3b304a2SMiklos Szeredi * given options, then the displayed options will not reflect reality 718b3b304a2SMiklos Szeredi * any more. 719b3b304a2SMiklos Szeredi */ 720b3b304a2SMiklos Szeredi void save_mount_options(struct super_block *sb, char *options) 721b3b304a2SMiklos Szeredi { 722b3b304a2SMiklos Szeredi kfree(sb->s_options); 723b3b304a2SMiklos Szeredi sb->s_options = kstrdup(options, GFP_KERNEL); 724b3b304a2SMiklos Szeredi } 725b3b304a2SMiklos Szeredi EXPORT_SYMBOL(save_mount_options); 726b3b304a2SMiklos Szeredi 727a1a2c409SMiklos Szeredi #ifdef CONFIG_PROC_FS 7281da177e4SLinus Torvalds /* iterator */ 7291da177e4SLinus Torvalds static void *m_start(struct seq_file *m, loff_t *pos) 7301da177e4SLinus Torvalds { 731a1a2c409SMiklos Szeredi struct proc_mounts *p = m->private; 7321da177e4SLinus Torvalds 733390c6843SRam Pai down_read(&namespace_sem); 734a1a2c409SMiklos Szeredi return seq_list_start(&p->ns->list, *pos); 7351da177e4SLinus Torvalds } 7361da177e4SLinus Torvalds 7371da177e4SLinus Torvalds static void *m_next(struct seq_file *m, void *v, loff_t *pos) 7381da177e4SLinus Torvalds { 739a1a2c409SMiklos Szeredi struct proc_mounts *p = m->private; 740b0765fb8SPavel Emelianov 741a1a2c409SMiklos Szeredi return seq_list_next(v, &p->ns->list, pos); 7421da177e4SLinus Torvalds } 7431da177e4SLinus Torvalds 7441da177e4SLinus Torvalds static void m_stop(struct seq_file *m, void *v) 7451da177e4SLinus Torvalds { 746390c6843SRam Pai up_read(&namespace_sem); 7471da177e4SLinus Torvalds } 7481da177e4SLinus Torvalds 7492d4d4864SRam Pai struct proc_fs_info { 7501da177e4SLinus Torvalds int flag; 7512d4d4864SRam Pai const char *str; 7522d4d4864SRam Pai }; 7532d4d4864SRam Pai 7542069f457SEric Paris static int show_sb_opts(struct seq_file *m, struct super_block *sb) 7552d4d4864SRam Pai { 7562d4d4864SRam Pai static const struct proc_fs_info fs_info[] = { 7571da177e4SLinus Torvalds { MS_SYNCHRONOUS, ",sync" }, 7581da177e4SLinus Torvalds { MS_DIRSYNC, ",dirsync" }, 7591da177e4SLinus Torvalds { MS_MANDLOCK, ",mand" }, 7601da177e4SLinus Torvalds { 0, NULL } 7611da177e4SLinus Torvalds }; 7622d4d4864SRam Pai const struct proc_fs_info *fs_infop; 7632d4d4864SRam Pai 7642d4d4864SRam Pai for (fs_infop = fs_info; fs_infop->flag; fs_infop++) { 7652d4d4864SRam Pai if (sb->s_flags & fs_infop->flag) 7662d4d4864SRam Pai seq_puts(m, fs_infop->str); 7672d4d4864SRam Pai } 7682069f457SEric Paris 7692069f457SEric Paris return security_sb_show_options(m, sb); 7702d4d4864SRam Pai } 7712d4d4864SRam Pai 7722d4d4864SRam Pai static void show_mnt_opts(struct seq_file *m, struct vfsmount *mnt) 7732d4d4864SRam Pai { 7742d4d4864SRam Pai static const struct proc_fs_info mnt_info[] = { 7751da177e4SLinus Torvalds { MNT_NOSUID, ",nosuid" }, 7761da177e4SLinus Torvalds { MNT_NODEV, ",nodev" }, 7771da177e4SLinus Torvalds { MNT_NOEXEC, ",noexec" }, 778fc33a7bbSChristoph Hellwig { MNT_NOATIME, ",noatime" }, 779fc33a7bbSChristoph Hellwig { MNT_NODIRATIME, ",nodiratime" }, 78047ae32d6SValerie Henson { MNT_RELATIME, ",relatime" }, 7811da177e4SLinus Torvalds { 0, NULL } 7821da177e4SLinus Torvalds }; 7832d4d4864SRam Pai const struct proc_fs_info *fs_infop; 7842d4d4864SRam Pai 7852d4d4864SRam Pai for (fs_infop = mnt_info; fs_infop->flag; fs_infop++) { 7862d4d4864SRam Pai if (mnt->mnt_flags & fs_infop->flag) 7872d4d4864SRam Pai seq_puts(m, fs_infop->str); 7882d4d4864SRam Pai } 7892d4d4864SRam Pai } 7902d4d4864SRam Pai 7912d4d4864SRam Pai static void show_type(struct seq_file *m, struct super_block *sb) 7922d4d4864SRam Pai { 7932d4d4864SRam Pai mangle(m, sb->s_type->name); 7942d4d4864SRam Pai if (sb->s_subtype && sb->s_subtype[0]) { 7952d4d4864SRam Pai seq_putc(m, '.'); 7962d4d4864SRam Pai mangle(m, sb->s_subtype); 7972d4d4864SRam Pai } 7982d4d4864SRam Pai } 7992d4d4864SRam Pai 8002d4d4864SRam Pai static int show_vfsmnt(struct seq_file *m, void *v) 8012d4d4864SRam Pai { 8022d4d4864SRam Pai struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); 8032d4d4864SRam Pai int err = 0; 804c32c2f63SJan Blunck struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; 8051da177e4SLinus Torvalds 8061da177e4SLinus Torvalds mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none"); 8071da177e4SLinus Torvalds seq_putc(m, ' '); 808c32c2f63SJan Blunck seq_path(m, &mnt_path, " \t\n\\"); 8091da177e4SLinus Torvalds seq_putc(m, ' '); 8102d4d4864SRam Pai show_type(m, mnt->mnt_sb); 8112e4b7fcdSDave Hansen seq_puts(m, __mnt_is_readonly(mnt) ? " ro" : " rw"); 8122069f457SEric Paris err = show_sb_opts(m, mnt->mnt_sb); 8132069f457SEric Paris if (err) 8142069f457SEric Paris goto out; 8152d4d4864SRam Pai show_mnt_opts(m, mnt); 8161da177e4SLinus Torvalds if (mnt->mnt_sb->s_op->show_options) 8171da177e4SLinus Torvalds err = mnt->mnt_sb->s_op->show_options(m, mnt); 8181da177e4SLinus Torvalds seq_puts(m, " 0 0\n"); 8192069f457SEric Paris out: 8201da177e4SLinus Torvalds return err; 8211da177e4SLinus Torvalds } 8221da177e4SLinus Torvalds 823a1a2c409SMiklos Szeredi const struct seq_operations mounts_op = { 8241da177e4SLinus Torvalds .start = m_start, 8251da177e4SLinus Torvalds .next = m_next, 8261da177e4SLinus Torvalds .stop = m_stop, 8271da177e4SLinus Torvalds .show = show_vfsmnt 8281da177e4SLinus Torvalds }; 8291da177e4SLinus Torvalds 8302d4d4864SRam Pai static int show_mountinfo(struct seq_file *m, void *v) 8312d4d4864SRam Pai { 8322d4d4864SRam Pai struct proc_mounts *p = m->private; 8332d4d4864SRam Pai struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); 8342d4d4864SRam Pai struct super_block *sb = mnt->mnt_sb; 8352d4d4864SRam Pai struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; 8362d4d4864SRam Pai struct path root = p->root; 8372d4d4864SRam Pai int err = 0; 8382d4d4864SRam Pai 8392d4d4864SRam Pai seq_printf(m, "%i %i %u:%u ", mnt->mnt_id, mnt->mnt_parent->mnt_id, 8402d4d4864SRam Pai MAJOR(sb->s_dev), MINOR(sb->s_dev)); 8412d4d4864SRam Pai seq_dentry(m, mnt->mnt_root, " \t\n\\"); 8422d4d4864SRam Pai seq_putc(m, ' '); 8432d4d4864SRam Pai seq_path_root(m, &mnt_path, &root, " \t\n\\"); 8442d4d4864SRam Pai if (root.mnt != p->root.mnt || root.dentry != p->root.dentry) { 8452d4d4864SRam Pai /* 8462d4d4864SRam Pai * Mountpoint is outside root, discard that one. Ugly, 8472d4d4864SRam Pai * but less so than trying to do that in iterator in a 8482d4d4864SRam Pai * race-free way (due to renames). 8492d4d4864SRam Pai */ 8502d4d4864SRam Pai return SEQ_SKIP; 8512d4d4864SRam Pai } 8522d4d4864SRam Pai seq_puts(m, mnt->mnt_flags & MNT_READONLY ? " ro" : " rw"); 8532d4d4864SRam Pai show_mnt_opts(m, mnt); 8542d4d4864SRam Pai 8552d4d4864SRam Pai /* Tagged fields ("foo:X" or "bar") */ 8562d4d4864SRam Pai if (IS_MNT_SHARED(mnt)) 8572d4d4864SRam Pai seq_printf(m, " shared:%i", mnt->mnt_group_id); 85897e7e0f7SMiklos Szeredi if (IS_MNT_SLAVE(mnt)) { 85997e7e0f7SMiklos Szeredi int master = mnt->mnt_master->mnt_group_id; 86097e7e0f7SMiklos Szeredi int dom = get_dominating_id(mnt, &p->root); 86197e7e0f7SMiklos Szeredi seq_printf(m, " master:%i", master); 86297e7e0f7SMiklos Szeredi if (dom && dom != master) 86397e7e0f7SMiklos Szeredi seq_printf(m, " propagate_from:%i", dom); 86497e7e0f7SMiklos Szeredi } 8652d4d4864SRam Pai if (IS_MNT_UNBINDABLE(mnt)) 8662d4d4864SRam Pai seq_puts(m, " unbindable"); 8672d4d4864SRam Pai 8682d4d4864SRam Pai /* Filesystem specific data */ 8692d4d4864SRam Pai seq_puts(m, " - "); 8702d4d4864SRam Pai show_type(m, sb); 8712d4d4864SRam Pai seq_putc(m, ' '); 8722d4d4864SRam Pai mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none"); 8732d4d4864SRam Pai seq_puts(m, sb->s_flags & MS_RDONLY ? " ro" : " rw"); 8742069f457SEric Paris err = show_sb_opts(m, sb); 8752069f457SEric Paris if (err) 8762069f457SEric Paris goto out; 8772d4d4864SRam Pai if (sb->s_op->show_options) 8782d4d4864SRam Pai err = sb->s_op->show_options(m, mnt); 8792d4d4864SRam Pai seq_putc(m, '\n'); 8802069f457SEric Paris out: 8812d4d4864SRam Pai return err; 8822d4d4864SRam Pai } 8832d4d4864SRam Pai 8842d4d4864SRam Pai const struct seq_operations mountinfo_op = { 8852d4d4864SRam Pai .start = m_start, 8862d4d4864SRam Pai .next = m_next, 8872d4d4864SRam Pai .stop = m_stop, 8882d4d4864SRam Pai .show = show_mountinfo, 8892d4d4864SRam Pai }; 8902d4d4864SRam Pai 891b4629fe2SChuck Lever static int show_vfsstat(struct seq_file *m, void *v) 892b4629fe2SChuck Lever { 893b0765fb8SPavel Emelianov struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); 894c32c2f63SJan Blunck struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; 895b4629fe2SChuck Lever int err = 0; 896b4629fe2SChuck Lever 897b4629fe2SChuck Lever /* device */ 898b4629fe2SChuck Lever if (mnt->mnt_devname) { 899b4629fe2SChuck Lever seq_puts(m, "device "); 900b4629fe2SChuck Lever mangle(m, mnt->mnt_devname); 901b4629fe2SChuck Lever } else 902b4629fe2SChuck Lever seq_puts(m, "no device"); 903b4629fe2SChuck Lever 904b4629fe2SChuck Lever /* mount point */ 905b4629fe2SChuck Lever seq_puts(m, " mounted on "); 906c32c2f63SJan Blunck seq_path(m, &mnt_path, " \t\n\\"); 907b4629fe2SChuck Lever seq_putc(m, ' '); 908b4629fe2SChuck Lever 909b4629fe2SChuck Lever /* file system type */ 910b4629fe2SChuck Lever seq_puts(m, "with fstype "); 9112d4d4864SRam Pai show_type(m, mnt->mnt_sb); 912b4629fe2SChuck Lever 913b4629fe2SChuck Lever /* optional statistics */ 914b4629fe2SChuck Lever if (mnt->mnt_sb->s_op->show_stats) { 915b4629fe2SChuck Lever seq_putc(m, ' '); 916b4629fe2SChuck Lever err = mnt->mnt_sb->s_op->show_stats(m, mnt); 917b4629fe2SChuck Lever } 918b4629fe2SChuck Lever 919b4629fe2SChuck Lever seq_putc(m, '\n'); 920b4629fe2SChuck Lever return err; 921b4629fe2SChuck Lever } 922b4629fe2SChuck Lever 923a1a2c409SMiklos Szeredi const struct seq_operations mountstats_op = { 924b4629fe2SChuck Lever .start = m_start, 925b4629fe2SChuck Lever .next = m_next, 926b4629fe2SChuck Lever .stop = m_stop, 927b4629fe2SChuck Lever .show = show_vfsstat, 928b4629fe2SChuck Lever }; 929a1a2c409SMiklos Szeredi #endif /* CONFIG_PROC_FS */ 930b4629fe2SChuck Lever 9311da177e4SLinus Torvalds /** 9321da177e4SLinus Torvalds * may_umount_tree - check if a mount tree is busy 9331da177e4SLinus Torvalds * @mnt: root of mount tree 9341da177e4SLinus Torvalds * 9351da177e4SLinus Torvalds * This is called to check if a tree of mounts has any 9361da177e4SLinus Torvalds * open files, pwds, chroots or sub mounts that are 9371da177e4SLinus Torvalds * busy. 9381da177e4SLinus Torvalds */ 9391da177e4SLinus Torvalds int may_umount_tree(struct vfsmount *mnt) 9401da177e4SLinus Torvalds { 94136341f64SRam Pai int actual_refs = 0; 94236341f64SRam Pai int minimum_refs = 0; 94336341f64SRam Pai struct vfsmount *p; 9441da177e4SLinus Torvalds 9451da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 94636341f64SRam Pai for (p = mnt; p; p = next_mnt(p, mnt)) { 9471da177e4SLinus Torvalds actual_refs += atomic_read(&p->mnt_count); 9481da177e4SLinus Torvalds minimum_refs += 2; 9491da177e4SLinus Torvalds } 9501da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 9511da177e4SLinus Torvalds 9521da177e4SLinus Torvalds if (actual_refs > minimum_refs) 9531da177e4SLinus Torvalds return 0; 954e3474a8eSIan Kent 955e3474a8eSIan Kent return 1; 9561da177e4SLinus Torvalds } 9571da177e4SLinus Torvalds 9581da177e4SLinus Torvalds EXPORT_SYMBOL(may_umount_tree); 9591da177e4SLinus Torvalds 9601da177e4SLinus Torvalds /** 9611da177e4SLinus Torvalds * may_umount - check if a mount point is busy 9621da177e4SLinus Torvalds * @mnt: root of mount 9631da177e4SLinus Torvalds * 9641da177e4SLinus Torvalds * This is called to check if a mount point has any 9651da177e4SLinus Torvalds * open files, pwds, chroots or sub mounts. If the 9661da177e4SLinus Torvalds * mount has sub mounts this will return busy 9671da177e4SLinus Torvalds * regardless of whether the sub mounts are busy. 9681da177e4SLinus Torvalds * 9691da177e4SLinus Torvalds * Doesn't take quota and stuff into account. IOW, in some cases it will 9701da177e4SLinus Torvalds * give false negatives. The main reason why it's here is that we need 9711da177e4SLinus Torvalds * a non-destructive way to look for easily umountable filesystems. 9721da177e4SLinus Torvalds */ 9731da177e4SLinus Torvalds int may_umount(struct vfsmount *mnt) 9741da177e4SLinus Torvalds { 975e3474a8eSIan Kent int ret = 1; 976a05964f3SRam Pai spin_lock(&vfsmount_lock); 977a05964f3SRam Pai if (propagate_mount_busy(mnt, 2)) 978e3474a8eSIan Kent ret = 0; 979a05964f3SRam Pai spin_unlock(&vfsmount_lock); 980a05964f3SRam Pai return ret; 9811da177e4SLinus Torvalds } 9821da177e4SLinus Torvalds 9831da177e4SLinus Torvalds EXPORT_SYMBOL(may_umount); 9841da177e4SLinus Torvalds 985b90fa9aeSRam Pai void release_mounts(struct list_head *head) 9861da177e4SLinus Torvalds { 98770fbcdf4SRam Pai struct vfsmount *mnt; 98870fbcdf4SRam Pai while (!list_empty(head)) { 989b5e61818SPavel Emelianov mnt = list_first_entry(head, struct vfsmount, mnt_hash); 99070fbcdf4SRam Pai list_del_init(&mnt->mnt_hash); 99170fbcdf4SRam Pai if (mnt->mnt_parent != mnt) { 99270fbcdf4SRam Pai struct dentry *dentry; 99370fbcdf4SRam Pai struct vfsmount *m; 99470fbcdf4SRam Pai spin_lock(&vfsmount_lock); 99570fbcdf4SRam Pai dentry = mnt->mnt_mountpoint; 99670fbcdf4SRam Pai m = mnt->mnt_parent; 99770fbcdf4SRam Pai mnt->mnt_mountpoint = mnt->mnt_root; 99870fbcdf4SRam Pai mnt->mnt_parent = mnt; 9997c4b93d8SAl Viro m->mnt_ghosts--; 10001da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 100170fbcdf4SRam Pai dput(dentry); 100270fbcdf4SRam Pai mntput(m); 10031da177e4SLinus Torvalds } 10041da177e4SLinus Torvalds mntput(mnt); 100570fbcdf4SRam Pai } 100670fbcdf4SRam Pai } 100770fbcdf4SRam Pai 1008a05964f3SRam Pai void umount_tree(struct vfsmount *mnt, int propagate, struct list_head *kill) 100970fbcdf4SRam Pai { 101070fbcdf4SRam Pai struct vfsmount *p; 101170fbcdf4SRam Pai 10121bfba4e8SAkinobu Mita for (p = mnt; p; p = next_mnt(p, mnt)) 10131bfba4e8SAkinobu Mita list_move(&p->mnt_hash, kill); 101470fbcdf4SRam Pai 1015a05964f3SRam Pai if (propagate) 1016a05964f3SRam Pai propagate_umount(kill); 1017a05964f3SRam Pai 101870fbcdf4SRam Pai list_for_each_entry(p, kill, mnt_hash) { 101970fbcdf4SRam Pai list_del_init(&p->mnt_expire); 102070fbcdf4SRam Pai list_del_init(&p->mnt_list); 10216b3286edSKirill Korotaev __touch_mnt_namespace(p->mnt_ns); 10226b3286edSKirill Korotaev p->mnt_ns = NULL; 102370fbcdf4SRam Pai list_del_init(&p->mnt_child); 10247c4b93d8SAl Viro if (p->mnt_parent != p) { 10257c4b93d8SAl Viro p->mnt_parent->mnt_ghosts++; 1026f30ac319SAl Viro p->mnt_mountpoint->d_mounted--; 10277c4b93d8SAl Viro } 1028a05964f3SRam Pai change_mnt_propagation(p, MS_PRIVATE); 10291da177e4SLinus Torvalds } 10301da177e4SLinus Torvalds } 10311da177e4SLinus Torvalds 1032c35038beSAl Viro static void shrink_submounts(struct vfsmount *mnt, struct list_head *umounts); 1033c35038beSAl Viro 10341da177e4SLinus Torvalds static int do_umount(struct vfsmount *mnt, int flags) 10351da177e4SLinus Torvalds { 10361da177e4SLinus Torvalds struct super_block *sb = mnt->mnt_sb; 10371da177e4SLinus Torvalds int retval; 103870fbcdf4SRam Pai LIST_HEAD(umount_list); 10391da177e4SLinus Torvalds 10401da177e4SLinus Torvalds retval = security_sb_umount(mnt, flags); 10411da177e4SLinus Torvalds if (retval) 10421da177e4SLinus Torvalds return retval; 10431da177e4SLinus Torvalds 10441da177e4SLinus Torvalds /* 10451da177e4SLinus Torvalds * Allow userspace to request a mountpoint be expired rather than 10461da177e4SLinus Torvalds * unmounting unconditionally. Unmount only happens if: 10471da177e4SLinus Torvalds * (1) the mark is already set (the mark is cleared by mntput()) 10481da177e4SLinus Torvalds * (2) the usage count == 1 [parent vfsmount] + 1 [sys_umount] 10491da177e4SLinus Torvalds */ 10501da177e4SLinus Torvalds if (flags & MNT_EXPIRE) { 10516ac08c39SJan Blunck if (mnt == current->fs->root.mnt || 10521da177e4SLinus Torvalds flags & (MNT_FORCE | MNT_DETACH)) 10531da177e4SLinus Torvalds return -EINVAL; 10541da177e4SLinus Torvalds 10551da177e4SLinus Torvalds if (atomic_read(&mnt->mnt_count) != 2) 10561da177e4SLinus Torvalds return -EBUSY; 10571da177e4SLinus Torvalds 10581da177e4SLinus Torvalds if (!xchg(&mnt->mnt_expiry_mark, 1)) 10591da177e4SLinus Torvalds return -EAGAIN; 10601da177e4SLinus Torvalds } 10611da177e4SLinus Torvalds 10621da177e4SLinus Torvalds /* 10631da177e4SLinus Torvalds * If we may have to abort operations to get out of this 10641da177e4SLinus Torvalds * mount, and they will themselves hold resources we must 10651da177e4SLinus Torvalds * allow the fs to do things. In the Unix tradition of 10661da177e4SLinus Torvalds * 'Gee thats tricky lets do it in userspace' the umount_begin 10671da177e4SLinus Torvalds * might fail to complete on the first run through as other tasks 10681da177e4SLinus Torvalds * must return, and the like. Thats for the mount program to worry 10691da177e4SLinus Torvalds * about for the moment. 10701da177e4SLinus Torvalds */ 10711da177e4SLinus Torvalds 107242faad99SAl Viro if (flags & MNT_FORCE && sb->s_op->umount_begin) { 10731da177e4SLinus Torvalds lock_kernel(); 107442faad99SAl Viro sb->s_op->umount_begin(sb); 10751da177e4SLinus Torvalds unlock_kernel(); 107642faad99SAl Viro } 10771da177e4SLinus Torvalds 10781da177e4SLinus Torvalds /* 10791da177e4SLinus Torvalds * No sense to grab the lock for this test, but test itself looks 10801da177e4SLinus Torvalds * somewhat bogus. Suggestions for better replacement? 10811da177e4SLinus Torvalds * Ho-hum... In principle, we might treat that as umount + switch 10821da177e4SLinus Torvalds * to rootfs. GC would eventually take care of the old vfsmount. 10831da177e4SLinus Torvalds * Actually it makes sense, especially if rootfs would contain a 10841da177e4SLinus Torvalds * /reboot - static binary that would close all descriptors and 10851da177e4SLinus Torvalds * call reboot(9). Then init(8) could umount root and exec /reboot. 10861da177e4SLinus Torvalds */ 10876ac08c39SJan Blunck if (mnt == current->fs->root.mnt && !(flags & MNT_DETACH)) { 10881da177e4SLinus Torvalds /* 10891da177e4SLinus Torvalds * Special case for "unmounting" root ... 10901da177e4SLinus Torvalds * we just try to remount it readonly. 10911da177e4SLinus Torvalds */ 10921da177e4SLinus Torvalds down_write(&sb->s_umount); 10931da177e4SLinus Torvalds if (!(sb->s_flags & MS_RDONLY)) { 10941da177e4SLinus Torvalds lock_kernel(); 10951da177e4SLinus Torvalds retval = do_remount_sb(sb, MS_RDONLY, NULL, 0); 10961da177e4SLinus Torvalds unlock_kernel(); 10971da177e4SLinus Torvalds } 10981da177e4SLinus Torvalds up_write(&sb->s_umount); 10991da177e4SLinus Torvalds return retval; 11001da177e4SLinus Torvalds } 11011da177e4SLinus Torvalds 1102390c6843SRam Pai down_write(&namespace_sem); 11031da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 11045addc5ddSAl Viro event++; 11051da177e4SLinus Torvalds 1106c35038beSAl Viro if (!(flags & MNT_DETACH)) 1107c35038beSAl Viro shrink_submounts(mnt, &umount_list); 1108c35038beSAl Viro 11091da177e4SLinus Torvalds retval = -EBUSY; 1110a05964f3SRam Pai if (flags & MNT_DETACH || !propagate_mount_busy(mnt, 2)) { 11111da177e4SLinus Torvalds if (!list_empty(&mnt->mnt_list)) 1112a05964f3SRam Pai umount_tree(mnt, 1, &umount_list); 11131da177e4SLinus Torvalds retval = 0; 11141da177e4SLinus Torvalds } 11151da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 11161da177e4SLinus Torvalds if (retval) 11171da177e4SLinus Torvalds security_sb_umount_busy(mnt); 1118390c6843SRam Pai up_write(&namespace_sem); 111970fbcdf4SRam Pai release_mounts(&umount_list); 11201da177e4SLinus Torvalds return retval; 11211da177e4SLinus Torvalds } 11221da177e4SLinus Torvalds 11231da177e4SLinus Torvalds /* 11241da177e4SLinus Torvalds * Now umount can handle mount points as well as block devices. 11251da177e4SLinus Torvalds * This is important for filesystems which use unnamed block devices. 11261da177e4SLinus Torvalds * 11271da177e4SLinus Torvalds * We now support a flag for forced unmount like the other 'big iron' 11281da177e4SLinus Torvalds * unixes. Our API is identical to OSF/1 to avoid making a mess of AMD 11291da177e4SLinus Torvalds */ 11301da177e4SLinus Torvalds 11311da177e4SLinus Torvalds asmlinkage long sys_umount(char __user * name, int flags) 11321da177e4SLinus Torvalds { 11332d8f3038SAl Viro struct path path; 11341da177e4SLinus Torvalds int retval; 11351da177e4SLinus Torvalds 11362d8f3038SAl Viro retval = user_path(name, &path); 11371da177e4SLinus Torvalds if (retval) 11381da177e4SLinus Torvalds goto out; 11391da177e4SLinus Torvalds retval = -EINVAL; 11402d8f3038SAl Viro if (path.dentry != path.mnt->mnt_root) 11411da177e4SLinus Torvalds goto dput_and_out; 11422d8f3038SAl Viro if (!check_mnt(path.mnt)) 11431da177e4SLinus Torvalds goto dput_and_out; 11441da177e4SLinus Torvalds 11451da177e4SLinus Torvalds retval = -EPERM; 11461da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 11471da177e4SLinus Torvalds goto dput_and_out; 11481da177e4SLinus Torvalds 11492d8f3038SAl Viro retval = do_umount(path.mnt, flags); 11501da177e4SLinus Torvalds dput_and_out: 1151429731b1SJan Blunck /* we mustn't call path_put() as that would clear mnt_expiry_mark */ 11522d8f3038SAl Viro dput(path.dentry); 11532d8f3038SAl Viro mntput_no_expire(path.mnt); 11541da177e4SLinus Torvalds out: 11551da177e4SLinus Torvalds return retval; 11561da177e4SLinus Torvalds } 11571da177e4SLinus Torvalds 11581da177e4SLinus Torvalds #ifdef __ARCH_WANT_SYS_OLDUMOUNT 11591da177e4SLinus Torvalds 11601da177e4SLinus Torvalds /* 11611da177e4SLinus Torvalds * The 2.0 compatible umount. No flags. 11621da177e4SLinus Torvalds */ 11631da177e4SLinus Torvalds asmlinkage long sys_oldumount(char __user * name) 11641da177e4SLinus Torvalds { 11651da177e4SLinus Torvalds return sys_umount(name, 0); 11661da177e4SLinus Torvalds } 11671da177e4SLinus Torvalds 11681da177e4SLinus Torvalds #endif 11691da177e4SLinus Torvalds 11702d92ab3cSAl Viro static int mount_is_safe(struct path *path) 11711da177e4SLinus Torvalds { 11721da177e4SLinus Torvalds if (capable(CAP_SYS_ADMIN)) 11731da177e4SLinus Torvalds return 0; 11741da177e4SLinus Torvalds return -EPERM; 11751da177e4SLinus Torvalds #ifdef notyet 11762d92ab3cSAl Viro if (S_ISLNK(path->dentry->d_inode->i_mode)) 11771da177e4SLinus Torvalds return -EPERM; 11782d92ab3cSAl Viro if (path->dentry->d_inode->i_mode & S_ISVTX) { 1179*da9592edSDavid Howells if (current_uid() != path->dentry->d_inode->i_uid) 11801da177e4SLinus Torvalds return -EPERM; 11811da177e4SLinus Torvalds } 11822d92ab3cSAl Viro if (inode_permission(path->dentry->d_inode, MAY_WRITE)) 11831da177e4SLinus Torvalds return -EPERM; 11841da177e4SLinus Torvalds return 0; 11851da177e4SLinus Torvalds #endif 11861da177e4SLinus Torvalds } 11871da177e4SLinus Torvalds 1188b90fa9aeSRam Pai struct vfsmount *copy_tree(struct vfsmount *mnt, struct dentry *dentry, 118936341f64SRam Pai int flag) 11901da177e4SLinus Torvalds { 11911da177e4SLinus Torvalds struct vfsmount *res, *p, *q, *r, *s; 11921a390689SAl Viro struct path path; 11931da177e4SLinus Torvalds 11949676f0c6SRam Pai if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(mnt)) 11959676f0c6SRam Pai return NULL; 11969676f0c6SRam Pai 119736341f64SRam Pai res = q = clone_mnt(mnt, dentry, flag); 11981da177e4SLinus Torvalds if (!q) 11991da177e4SLinus Torvalds goto Enomem; 12001da177e4SLinus Torvalds q->mnt_mountpoint = mnt->mnt_mountpoint; 12011da177e4SLinus Torvalds 12021da177e4SLinus Torvalds p = mnt; 1203fdadd65fSDomen Puncer list_for_each_entry(r, &mnt->mnt_mounts, mnt_child) { 12047ec02ef1SJan Blunck if (!is_subdir(r->mnt_mountpoint, dentry)) 12051da177e4SLinus Torvalds continue; 12061da177e4SLinus Torvalds 12071da177e4SLinus Torvalds for (s = r; s; s = next_mnt(s, r)) { 12089676f0c6SRam Pai if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(s)) { 12099676f0c6SRam Pai s = skip_mnt_tree(s); 12109676f0c6SRam Pai continue; 12119676f0c6SRam Pai } 12121da177e4SLinus Torvalds while (p != s->mnt_parent) { 12131da177e4SLinus Torvalds p = p->mnt_parent; 12141da177e4SLinus Torvalds q = q->mnt_parent; 12151da177e4SLinus Torvalds } 12161da177e4SLinus Torvalds p = s; 12171a390689SAl Viro path.mnt = q; 12181a390689SAl Viro path.dentry = p->mnt_mountpoint; 121936341f64SRam Pai q = clone_mnt(p, p->mnt_root, flag); 12201da177e4SLinus Torvalds if (!q) 12211da177e4SLinus Torvalds goto Enomem; 12221da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 12231da177e4SLinus Torvalds list_add_tail(&q->mnt_list, &res->mnt_list); 12241a390689SAl Viro attach_mnt(q, &path); 12251da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 12261da177e4SLinus Torvalds } 12271da177e4SLinus Torvalds } 12281da177e4SLinus Torvalds return res; 12291da177e4SLinus Torvalds Enomem: 12301da177e4SLinus Torvalds if (res) { 123170fbcdf4SRam Pai LIST_HEAD(umount_list); 12321da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 1233a05964f3SRam Pai umount_tree(res, 0, &umount_list); 12341da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 123570fbcdf4SRam Pai release_mounts(&umount_list); 12361da177e4SLinus Torvalds } 12371da177e4SLinus Torvalds return NULL; 12381da177e4SLinus Torvalds } 12391da177e4SLinus Torvalds 12408aec0809SAl Viro struct vfsmount *collect_mounts(struct vfsmount *mnt, struct dentry *dentry) 12418aec0809SAl Viro { 12428aec0809SAl Viro struct vfsmount *tree; 12431a60a280SAl Viro down_write(&namespace_sem); 12448aec0809SAl Viro tree = copy_tree(mnt, dentry, CL_COPY_ALL | CL_PRIVATE); 12451a60a280SAl Viro up_write(&namespace_sem); 12468aec0809SAl Viro return tree; 12478aec0809SAl Viro } 12488aec0809SAl Viro 12498aec0809SAl Viro void drop_collected_mounts(struct vfsmount *mnt) 12508aec0809SAl Viro { 12518aec0809SAl Viro LIST_HEAD(umount_list); 12521a60a280SAl Viro down_write(&namespace_sem); 12538aec0809SAl Viro spin_lock(&vfsmount_lock); 12548aec0809SAl Viro umount_tree(mnt, 0, &umount_list); 12558aec0809SAl Viro spin_unlock(&vfsmount_lock); 12561a60a280SAl Viro up_write(&namespace_sem); 12578aec0809SAl Viro release_mounts(&umount_list); 12588aec0809SAl Viro } 12598aec0809SAl Viro 1260719f5d7fSMiklos Szeredi static void cleanup_group_ids(struct vfsmount *mnt, struct vfsmount *end) 1261719f5d7fSMiklos Szeredi { 1262719f5d7fSMiklos Szeredi struct vfsmount *p; 1263719f5d7fSMiklos Szeredi 1264719f5d7fSMiklos Szeredi for (p = mnt; p != end; p = next_mnt(p, mnt)) { 1265719f5d7fSMiklos Szeredi if (p->mnt_group_id && !IS_MNT_SHARED(p)) 1266719f5d7fSMiklos Szeredi mnt_release_group_id(p); 1267719f5d7fSMiklos Szeredi } 1268719f5d7fSMiklos Szeredi } 1269719f5d7fSMiklos Szeredi 1270719f5d7fSMiklos Szeredi static int invent_group_ids(struct vfsmount *mnt, bool recurse) 1271719f5d7fSMiklos Szeredi { 1272719f5d7fSMiklos Szeredi struct vfsmount *p; 1273719f5d7fSMiklos Szeredi 1274719f5d7fSMiklos Szeredi for (p = mnt; p; p = recurse ? next_mnt(p, mnt) : NULL) { 1275719f5d7fSMiklos Szeredi if (!p->mnt_group_id && !IS_MNT_SHARED(p)) { 1276719f5d7fSMiklos Szeredi int err = mnt_alloc_group_id(p); 1277719f5d7fSMiklos Szeredi if (err) { 1278719f5d7fSMiklos Szeredi cleanup_group_ids(mnt, p); 1279719f5d7fSMiklos Szeredi return err; 1280719f5d7fSMiklos Szeredi } 1281719f5d7fSMiklos Szeredi } 1282719f5d7fSMiklos Szeredi } 1283719f5d7fSMiklos Szeredi 1284719f5d7fSMiklos Szeredi return 0; 1285719f5d7fSMiklos Szeredi } 1286719f5d7fSMiklos Szeredi 1287b90fa9aeSRam Pai /* 1288b90fa9aeSRam Pai * @source_mnt : mount tree to be attached 1289b90fa9aeSRam Pai * @nd : place the mount tree @source_mnt is attached 129021444403SRam Pai * @parent_nd : if non-null, detach the source_mnt from its parent and 129121444403SRam Pai * store the parent mount and mountpoint dentry. 129221444403SRam Pai * (done when source_mnt is moved) 1293b90fa9aeSRam Pai * 1294b90fa9aeSRam Pai * NOTE: in the table below explains the semantics when a source mount 1295b90fa9aeSRam Pai * of a given type is attached to a destination mount of a given type. 12969676f0c6SRam Pai * --------------------------------------------------------------------------- 1297b90fa9aeSRam Pai * | BIND MOUNT OPERATION | 12989676f0c6SRam Pai * |************************************************************************** 12999676f0c6SRam Pai * | source-->| shared | private | slave | unbindable | 13009676f0c6SRam Pai * | dest | | | | | 13019676f0c6SRam Pai * | | | | | | | 13029676f0c6SRam Pai * | v | | | | | 13039676f0c6SRam Pai * |************************************************************************** 13049676f0c6SRam Pai * | shared | shared (++) | shared (+) | shared(+++)| invalid | 13055afe0022SRam Pai * | | | | | | 13069676f0c6SRam Pai * |non-shared| shared (+) | private | slave (*) | invalid | 13079676f0c6SRam Pai * *************************************************************************** 1308b90fa9aeSRam Pai * A bind operation clones the source mount and mounts the clone on the 1309b90fa9aeSRam Pai * destination mount. 1310b90fa9aeSRam Pai * 1311b90fa9aeSRam Pai * (++) the cloned mount is propagated to all the mounts in the propagation 1312b90fa9aeSRam Pai * tree of the destination mount and the cloned mount is added to 1313b90fa9aeSRam Pai * the peer group of the source mount. 1314b90fa9aeSRam Pai * (+) the cloned mount is created under the destination mount and is marked 1315b90fa9aeSRam Pai * as shared. The cloned mount is added to the peer group of the source 1316b90fa9aeSRam Pai * mount. 13175afe0022SRam Pai * (+++) the mount is propagated to all the mounts in the propagation tree 13185afe0022SRam Pai * of the destination mount and the cloned mount is made slave 13195afe0022SRam Pai * of the same master as that of the source mount. The cloned mount 13205afe0022SRam Pai * is marked as 'shared and slave'. 13215afe0022SRam Pai * (*) the cloned mount is made a slave of the same master as that of the 13225afe0022SRam Pai * source mount. 13235afe0022SRam Pai * 13249676f0c6SRam Pai * --------------------------------------------------------------------------- 132521444403SRam Pai * | MOVE MOUNT OPERATION | 13269676f0c6SRam Pai * |************************************************************************** 13279676f0c6SRam Pai * | source-->| shared | private | slave | unbindable | 13289676f0c6SRam Pai * | dest | | | | | 13299676f0c6SRam Pai * | | | | | | | 13309676f0c6SRam Pai * | v | | | | | 13319676f0c6SRam Pai * |************************************************************************** 13329676f0c6SRam Pai * | shared | shared (+) | shared (+) | shared(+++) | invalid | 13335afe0022SRam Pai * | | | | | | 13349676f0c6SRam Pai * |non-shared| shared (+*) | private | slave (*) | unbindable | 13359676f0c6SRam Pai * *************************************************************************** 13365afe0022SRam Pai * 13375afe0022SRam Pai * (+) the mount is moved to the destination. And is then propagated to 13385afe0022SRam Pai * all the mounts in the propagation tree of the destination mount. 133921444403SRam Pai * (+*) the mount is moved to the destination. 13405afe0022SRam Pai * (+++) the mount is moved to the destination and is then propagated to 13415afe0022SRam Pai * all the mounts belonging to the destination mount's propagation tree. 13425afe0022SRam Pai * the mount is marked as 'shared and slave'. 13435afe0022SRam Pai * (*) the mount continues to be a slave at the new location. 1344b90fa9aeSRam Pai * 1345b90fa9aeSRam Pai * if the source mount is a tree, the operations explained above is 1346b90fa9aeSRam Pai * applied to each mount in the tree. 1347b90fa9aeSRam Pai * Must be called without spinlocks held, since this function can sleep 1348b90fa9aeSRam Pai * in allocations. 1349b90fa9aeSRam Pai */ 1350b90fa9aeSRam Pai static int attach_recursive_mnt(struct vfsmount *source_mnt, 13511a390689SAl Viro struct path *path, struct path *parent_path) 1352b90fa9aeSRam Pai { 1353b90fa9aeSRam Pai LIST_HEAD(tree_list); 13541a390689SAl Viro struct vfsmount *dest_mnt = path->mnt; 13551a390689SAl Viro struct dentry *dest_dentry = path->dentry; 1356b90fa9aeSRam Pai struct vfsmount *child, *p; 1357719f5d7fSMiklos Szeredi int err; 1358b90fa9aeSRam Pai 1359719f5d7fSMiklos Szeredi if (IS_MNT_SHARED(dest_mnt)) { 1360719f5d7fSMiklos Szeredi err = invent_group_ids(source_mnt, true); 1361719f5d7fSMiklos Szeredi if (err) 1362719f5d7fSMiklos Szeredi goto out; 1363719f5d7fSMiklos Szeredi } 1364719f5d7fSMiklos Szeredi err = propagate_mnt(dest_mnt, dest_dentry, source_mnt, &tree_list); 1365719f5d7fSMiklos Szeredi if (err) 1366719f5d7fSMiklos Szeredi goto out_cleanup_ids; 1367b90fa9aeSRam Pai 1368b90fa9aeSRam Pai if (IS_MNT_SHARED(dest_mnt)) { 1369b90fa9aeSRam Pai for (p = source_mnt; p; p = next_mnt(p, source_mnt)) 1370b90fa9aeSRam Pai set_mnt_shared(p); 1371b90fa9aeSRam Pai } 1372b90fa9aeSRam Pai 1373b90fa9aeSRam Pai spin_lock(&vfsmount_lock); 13741a390689SAl Viro if (parent_path) { 13751a390689SAl Viro detach_mnt(source_mnt, parent_path); 13761a390689SAl Viro attach_mnt(source_mnt, path); 13776b3286edSKirill Korotaev touch_mnt_namespace(current->nsproxy->mnt_ns); 137821444403SRam Pai } else { 1379b90fa9aeSRam Pai mnt_set_mountpoint(dest_mnt, dest_dentry, source_mnt); 1380b90fa9aeSRam Pai commit_tree(source_mnt); 138121444403SRam Pai } 1382b90fa9aeSRam Pai 1383b90fa9aeSRam Pai list_for_each_entry_safe(child, p, &tree_list, mnt_hash) { 1384b90fa9aeSRam Pai list_del_init(&child->mnt_hash); 1385b90fa9aeSRam Pai commit_tree(child); 1386b90fa9aeSRam Pai } 1387b90fa9aeSRam Pai spin_unlock(&vfsmount_lock); 1388b90fa9aeSRam Pai return 0; 1389719f5d7fSMiklos Szeredi 1390719f5d7fSMiklos Szeredi out_cleanup_ids: 1391719f5d7fSMiklos Szeredi if (IS_MNT_SHARED(dest_mnt)) 1392719f5d7fSMiklos Szeredi cleanup_group_ids(source_mnt, NULL); 1393719f5d7fSMiklos Szeredi out: 1394719f5d7fSMiklos Szeredi return err; 1395b90fa9aeSRam Pai } 1396b90fa9aeSRam Pai 13978c3ee42eSAl Viro static int graft_tree(struct vfsmount *mnt, struct path *path) 13981da177e4SLinus Torvalds { 13991da177e4SLinus Torvalds int err; 14001da177e4SLinus Torvalds if (mnt->mnt_sb->s_flags & MS_NOUSER) 14011da177e4SLinus Torvalds return -EINVAL; 14021da177e4SLinus Torvalds 14038c3ee42eSAl Viro if (S_ISDIR(path->dentry->d_inode->i_mode) != 14041da177e4SLinus Torvalds S_ISDIR(mnt->mnt_root->d_inode->i_mode)) 14051da177e4SLinus Torvalds return -ENOTDIR; 14061da177e4SLinus Torvalds 14071da177e4SLinus Torvalds err = -ENOENT; 14088c3ee42eSAl Viro mutex_lock(&path->dentry->d_inode->i_mutex); 14098c3ee42eSAl Viro if (IS_DEADDIR(path->dentry->d_inode)) 14101da177e4SLinus Torvalds goto out_unlock; 14111da177e4SLinus Torvalds 14128c3ee42eSAl Viro err = security_sb_check_sb(mnt, path); 14131da177e4SLinus Torvalds if (err) 14141da177e4SLinus Torvalds goto out_unlock; 14151da177e4SLinus Torvalds 14161da177e4SLinus Torvalds err = -ENOENT; 14178c3ee42eSAl Viro if (IS_ROOT(path->dentry) || !d_unhashed(path->dentry)) 14188c3ee42eSAl Viro err = attach_recursive_mnt(mnt, path, NULL); 14191da177e4SLinus Torvalds out_unlock: 14208c3ee42eSAl Viro mutex_unlock(&path->dentry->d_inode->i_mutex); 14211da177e4SLinus Torvalds if (!err) 14228c3ee42eSAl Viro security_sb_post_addmount(mnt, path); 14231da177e4SLinus Torvalds return err; 14241da177e4SLinus Torvalds } 14251da177e4SLinus Torvalds 14261da177e4SLinus Torvalds /* 142707b20889SRam Pai * recursively change the type of the mountpoint. 142807b20889SRam Pai */ 14290a0d8a46SAl Viro static int do_change_type(struct path *path, int flag) 143007b20889SRam Pai { 14312d92ab3cSAl Viro struct vfsmount *m, *mnt = path->mnt; 143207b20889SRam Pai int recurse = flag & MS_REC; 143307b20889SRam Pai int type = flag & ~MS_REC; 1434719f5d7fSMiklos Szeredi int err = 0; 143507b20889SRam Pai 1436ee6f9582SMiklos Szeredi if (!capable(CAP_SYS_ADMIN)) 1437ee6f9582SMiklos Szeredi return -EPERM; 1438ee6f9582SMiklos Szeredi 14392d92ab3cSAl Viro if (path->dentry != path->mnt->mnt_root) 144007b20889SRam Pai return -EINVAL; 144107b20889SRam Pai 144207b20889SRam Pai down_write(&namespace_sem); 1443719f5d7fSMiklos Szeredi if (type == MS_SHARED) { 1444719f5d7fSMiklos Szeredi err = invent_group_ids(mnt, recurse); 1445719f5d7fSMiklos Szeredi if (err) 1446719f5d7fSMiklos Szeredi goto out_unlock; 1447719f5d7fSMiklos Szeredi } 1448719f5d7fSMiklos Szeredi 144907b20889SRam Pai spin_lock(&vfsmount_lock); 145007b20889SRam Pai for (m = mnt; m; m = (recurse ? next_mnt(m, mnt) : NULL)) 145107b20889SRam Pai change_mnt_propagation(m, type); 145207b20889SRam Pai spin_unlock(&vfsmount_lock); 1453719f5d7fSMiklos Szeredi 1454719f5d7fSMiklos Szeredi out_unlock: 145507b20889SRam Pai up_write(&namespace_sem); 1456719f5d7fSMiklos Szeredi return err; 145707b20889SRam Pai } 145807b20889SRam Pai 145907b20889SRam Pai /* 14601da177e4SLinus Torvalds * do loopback mount. 14611da177e4SLinus Torvalds */ 14620a0d8a46SAl Viro static int do_loopback(struct path *path, char *old_name, 14632dafe1c4SEric Sandeen int recurse) 14641da177e4SLinus Torvalds { 14652d92ab3cSAl Viro struct path old_path; 14661da177e4SLinus Torvalds struct vfsmount *mnt = NULL; 14672d92ab3cSAl Viro int err = mount_is_safe(path); 14681da177e4SLinus Torvalds if (err) 14691da177e4SLinus Torvalds return err; 14701da177e4SLinus Torvalds if (!old_name || !*old_name) 14711da177e4SLinus Torvalds return -EINVAL; 14722d92ab3cSAl Viro err = kern_path(old_name, LOOKUP_FOLLOW, &old_path); 14731da177e4SLinus Torvalds if (err) 14741da177e4SLinus Torvalds return err; 14751da177e4SLinus Torvalds 1476390c6843SRam Pai down_write(&namespace_sem); 14771da177e4SLinus Torvalds err = -EINVAL; 14782d92ab3cSAl Viro if (IS_MNT_UNBINDABLE(old_path.mnt)) 14799676f0c6SRam Pai goto out; 14809676f0c6SRam Pai 14812d92ab3cSAl Viro if (!check_mnt(path->mnt) || !check_mnt(old_path.mnt)) 1482ccd48bc7SAl Viro goto out; 1483ccd48bc7SAl Viro 14841da177e4SLinus Torvalds err = -ENOMEM; 14851da177e4SLinus Torvalds if (recurse) 14862d92ab3cSAl Viro mnt = copy_tree(old_path.mnt, old_path.dentry, 0); 14871da177e4SLinus Torvalds else 14882d92ab3cSAl Viro mnt = clone_mnt(old_path.mnt, old_path.dentry, 0); 14891da177e4SLinus Torvalds 1490ccd48bc7SAl Viro if (!mnt) 1491ccd48bc7SAl Viro goto out; 1492ccd48bc7SAl Viro 14932d92ab3cSAl Viro err = graft_tree(mnt, path); 14941da177e4SLinus Torvalds if (err) { 149570fbcdf4SRam Pai LIST_HEAD(umount_list); 14961da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 1497a05964f3SRam Pai umount_tree(mnt, 0, &umount_list); 14981da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 149970fbcdf4SRam Pai release_mounts(&umount_list); 15005b83d2c5SRam Pai } 15011da177e4SLinus Torvalds 1502ccd48bc7SAl Viro out: 1503390c6843SRam Pai up_write(&namespace_sem); 15042d92ab3cSAl Viro path_put(&old_path); 15051da177e4SLinus Torvalds return err; 15061da177e4SLinus Torvalds } 15071da177e4SLinus Torvalds 15082e4b7fcdSDave Hansen static int change_mount_flags(struct vfsmount *mnt, int ms_flags) 15092e4b7fcdSDave Hansen { 15102e4b7fcdSDave Hansen int error = 0; 15112e4b7fcdSDave Hansen int readonly_request = 0; 15122e4b7fcdSDave Hansen 15132e4b7fcdSDave Hansen if (ms_flags & MS_RDONLY) 15142e4b7fcdSDave Hansen readonly_request = 1; 15152e4b7fcdSDave Hansen if (readonly_request == __mnt_is_readonly(mnt)) 15162e4b7fcdSDave Hansen return 0; 15172e4b7fcdSDave Hansen 15182e4b7fcdSDave Hansen if (readonly_request) 15192e4b7fcdSDave Hansen error = mnt_make_readonly(mnt); 15202e4b7fcdSDave Hansen else 15212e4b7fcdSDave Hansen __mnt_unmake_readonly(mnt); 15222e4b7fcdSDave Hansen return error; 15232e4b7fcdSDave Hansen } 15242e4b7fcdSDave Hansen 15251da177e4SLinus Torvalds /* 15261da177e4SLinus Torvalds * change filesystem flags. dir should be a physical root of filesystem. 15271da177e4SLinus Torvalds * If you've mounted a non-root directory somewhere and want to do remount 15281da177e4SLinus Torvalds * on it - tough luck. 15291da177e4SLinus Torvalds */ 15300a0d8a46SAl Viro static int do_remount(struct path *path, int flags, int mnt_flags, 15311da177e4SLinus Torvalds void *data) 15321da177e4SLinus Torvalds { 15331da177e4SLinus Torvalds int err; 15342d92ab3cSAl Viro struct super_block *sb = path->mnt->mnt_sb; 15351da177e4SLinus Torvalds 15361da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 15371da177e4SLinus Torvalds return -EPERM; 15381da177e4SLinus Torvalds 15392d92ab3cSAl Viro if (!check_mnt(path->mnt)) 15401da177e4SLinus Torvalds return -EINVAL; 15411da177e4SLinus Torvalds 15422d92ab3cSAl Viro if (path->dentry != path->mnt->mnt_root) 15431da177e4SLinus Torvalds return -EINVAL; 15441da177e4SLinus Torvalds 15451da177e4SLinus Torvalds down_write(&sb->s_umount); 15462e4b7fcdSDave Hansen if (flags & MS_BIND) 15472d92ab3cSAl Viro err = change_mount_flags(path->mnt, flags); 15482e4b7fcdSDave Hansen else 15491da177e4SLinus Torvalds err = do_remount_sb(sb, flags, data, 0); 15501da177e4SLinus Torvalds if (!err) 15512d92ab3cSAl Viro path->mnt->mnt_flags = mnt_flags; 15521da177e4SLinus Torvalds up_write(&sb->s_umount); 15530e55a7ccSDan Williams if (!err) { 15542d92ab3cSAl Viro security_sb_post_remount(path->mnt, flags, data); 15550e55a7ccSDan Williams 15560e55a7ccSDan Williams spin_lock(&vfsmount_lock); 15570e55a7ccSDan Williams touch_mnt_namespace(path->mnt->mnt_ns); 15580e55a7ccSDan Williams spin_unlock(&vfsmount_lock); 15590e55a7ccSDan Williams } 15601da177e4SLinus Torvalds return err; 15611da177e4SLinus Torvalds } 15621da177e4SLinus Torvalds 15639676f0c6SRam Pai static inline int tree_contains_unbindable(struct vfsmount *mnt) 15649676f0c6SRam Pai { 15659676f0c6SRam Pai struct vfsmount *p; 15669676f0c6SRam Pai for (p = mnt; p; p = next_mnt(p, mnt)) { 15679676f0c6SRam Pai if (IS_MNT_UNBINDABLE(p)) 15689676f0c6SRam Pai return 1; 15699676f0c6SRam Pai } 15709676f0c6SRam Pai return 0; 15719676f0c6SRam Pai } 15729676f0c6SRam Pai 15730a0d8a46SAl Viro static int do_move_mount(struct path *path, char *old_name) 15741da177e4SLinus Torvalds { 15752d92ab3cSAl Viro struct path old_path, parent_path; 15761da177e4SLinus Torvalds struct vfsmount *p; 15771da177e4SLinus Torvalds int err = 0; 15781da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 15791da177e4SLinus Torvalds return -EPERM; 15801da177e4SLinus Torvalds if (!old_name || !*old_name) 15811da177e4SLinus Torvalds return -EINVAL; 15822d92ab3cSAl Viro err = kern_path(old_name, LOOKUP_FOLLOW, &old_path); 15831da177e4SLinus Torvalds if (err) 15841da177e4SLinus Torvalds return err; 15851da177e4SLinus Torvalds 1586390c6843SRam Pai down_write(&namespace_sem); 15872d92ab3cSAl Viro while (d_mountpoint(path->dentry) && 15882d92ab3cSAl Viro follow_down(&path->mnt, &path->dentry)) 15891da177e4SLinus Torvalds ; 15901da177e4SLinus Torvalds err = -EINVAL; 15912d92ab3cSAl Viro if (!check_mnt(path->mnt) || !check_mnt(old_path.mnt)) 15921da177e4SLinus Torvalds goto out; 15931da177e4SLinus Torvalds 15941da177e4SLinus Torvalds err = -ENOENT; 15952d92ab3cSAl Viro mutex_lock(&path->dentry->d_inode->i_mutex); 15962d92ab3cSAl Viro if (IS_DEADDIR(path->dentry->d_inode)) 15971da177e4SLinus Torvalds goto out1; 15981da177e4SLinus Torvalds 15992d92ab3cSAl Viro if (!IS_ROOT(path->dentry) && d_unhashed(path->dentry)) 160021444403SRam Pai goto out1; 16011da177e4SLinus Torvalds 16021da177e4SLinus Torvalds err = -EINVAL; 16032d92ab3cSAl Viro if (old_path.dentry != old_path.mnt->mnt_root) 160421444403SRam Pai goto out1; 16051da177e4SLinus Torvalds 16062d92ab3cSAl Viro if (old_path.mnt == old_path.mnt->mnt_parent) 160721444403SRam Pai goto out1; 16081da177e4SLinus Torvalds 16092d92ab3cSAl Viro if (S_ISDIR(path->dentry->d_inode->i_mode) != 16102d92ab3cSAl Viro S_ISDIR(old_path.dentry->d_inode->i_mode)) 161121444403SRam Pai goto out1; 161221444403SRam Pai /* 161321444403SRam Pai * Don't move a mount residing in a shared parent. 161421444403SRam Pai */ 16152d92ab3cSAl Viro if (old_path.mnt->mnt_parent && 16162d92ab3cSAl Viro IS_MNT_SHARED(old_path.mnt->mnt_parent)) 161721444403SRam Pai goto out1; 16189676f0c6SRam Pai /* 16199676f0c6SRam Pai * Don't move a mount tree containing unbindable mounts to a destination 16209676f0c6SRam Pai * mount which is shared. 16219676f0c6SRam Pai */ 16222d92ab3cSAl Viro if (IS_MNT_SHARED(path->mnt) && 16232d92ab3cSAl Viro tree_contains_unbindable(old_path.mnt)) 16249676f0c6SRam Pai goto out1; 16251da177e4SLinus Torvalds err = -ELOOP; 16262d92ab3cSAl Viro for (p = path->mnt; p->mnt_parent != p; p = p->mnt_parent) 16272d92ab3cSAl Viro if (p == old_path.mnt) 162821444403SRam Pai goto out1; 16291da177e4SLinus Torvalds 16302d92ab3cSAl Viro err = attach_recursive_mnt(old_path.mnt, path, &parent_path); 16314ac91378SJan Blunck if (err) 163221444403SRam Pai goto out1; 16331da177e4SLinus Torvalds 16341da177e4SLinus Torvalds /* if the mount is moved, it should no longer be expire 16351da177e4SLinus Torvalds * automatically */ 16362d92ab3cSAl Viro list_del_init(&old_path.mnt->mnt_expire); 16371da177e4SLinus Torvalds out1: 16382d92ab3cSAl Viro mutex_unlock(&path->dentry->d_inode->i_mutex); 16391da177e4SLinus Torvalds out: 1640390c6843SRam Pai up_write(&namespace_sem); 16411da177e4SLinus Torvalds if (!err) 16421a390689SAl Viro path_put(&parent_path); 16432d92ab3cSAl Viro path_put(&old_path); 16441da177e4SLinus Torvalds return err; 16451da177e4SLinus Torvalds } 16461da177e4SLinus Torvalds 16471da177e4SLinus Torvalds /* 16481da177e4SLinus Torvalds * create a new mount for userspace and request it to be added into the 16491da177e4SLinus Torvalds * namespace's tree 16501da177e4SLinus Torvalds */ 16510a0d8a46SAl Viro static int do_new_mount(struct path *path, char *type, int flags, 16521da177e4SLinus Torvalds int mnt_flags, char *name, void *data) 16531da177e4SLinus Torvalds { 16541da177e4SLinus Torvalds struct vfsmount *mnt; 16551da177e4SLinus Torvalds 16561da177e4SLinus Torvalds if (!type || !memchr(type, 0, PAGE_SIZE)) 16571da177e4SLinus Torvalds return -EINVAL; 16581da177e4SLinus Torvalds 16591da177e4SLinus Torvalds /* we need capabilities... */ 16601da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 16611da177e4SLinus Torvalds return -EPERM; 16621da177e4SLinus Torvalds 16631da177e4SLinus Torvalds mnt = do_kern_mount(type, flags, name, data); 16641da177e4SLinus Torvalds if (IS_ERR(mnt)) 16651da177e4SLinus Torvalds return PTR_ERR(mnt); 16661da177e4SLinus Torvalds 16672d92ab3cSAl Viro return do_add_mount(mnt, path, mnt_flags, NULL); 16681da177e4SLinus Torvalds } 16691da177e4SLinus Torvalds 16701da177e4SLinus Torvalds /* 16711da177e4SLinus Torvalds * add a mount into a namespace's mount tree 16721da177e4SLinus Torvalds * - provide the option of adding the new mount to an expiration list 16731da177e4SLinus Torvalds */ 16748d66bf54SAl Viro int do_add_mount(struct vfsmount *newmnt, struct path *path, 16751da177e4SLinus Torvalds int mnt_flags, struct list_head *fslist) 16761da177e4SLinus Torvalds { 16771da177e4SLinus Torvalds int err; 16781da177e4SLinus Torvalds 1679390c6843SRam Pai down_write(&namespace_sem); 16801da177e4SLinus Torvalds /* Something was mounted here while we slept */ 16818d66bf54SAl Viro while (d_mountpoint(path->dentry) && 16828d66bf54SAl Viro follow_down(&path->mnt, &path->dentry)) 16831da177e4SLinus Torvalds ; 16841da177e4SLinus Torvalds err = -EINVAL; 16858d66bf54SAl Viro if (!check_mnt(path->mnt)) 16861da177e4SLinus Torvalds goto unlock; 16871da177e4SLinus Torvalds 16881da177e4SLinus Torvalds /* Refuse the same filesystem on the same mount point */ 16891da177e4SLinus Torvalds err = -EBUSY; 16908d66bf54SAl Viro if (path->mnt->mnt_sb == newmnt->mnt_sb && 16918d66bf54SAl Viro path->mnt->mnt_root == path->dentry) 16921da177e4SLinus Torvalds goto unlock; 16931da177e4SLinus Torvalds 16941da177e4SLinus Torvalds err = -EINVAL; 16951da177e4SLinus Torvalds if (S_ISLNK(newmnt->mnt_root->d_inode->i_mode)) 16961da177e4SLinus Torvalds goto unlock; 16971da177e4SLinus Torvalds 16981da177e4SLinus Torvalds newmnt->mnt_flags = mnt_flags; 16998d66bf54SAl Viro if ((err = graft_tree(newmnt, path))) 17005b83d2c5SRam Pai goto unlock; 17011da177e4SLinus Torvalds 17026758f953SAl Viro if (fslist) /* add to the specified expiration list */ 170355e700b9SMiklos Szeredi list_add_tail(&newmnt->mnt_expire, fslist); 17046758f953SAl Viro 1705390c6843SRam Pai up_write(&namespace_sem); 17065b83d2c5SRam Pai return 0; 17071da177e4SLinus Torvalds 17081da177e4SLinus Torvalds unlock: 1709390c6843SRam Pai up_write(&namespace_sem); 17101da177e4SLinus Torvalds mntput(newmnt); 17111da177e4SLinus Torvalds return err; 17121da177e4SLinus Torvalds } 17131da177e4SLinus Torvalds 17141da177e4SLinus Torvalds EXPORT_SYMBOL_GPL(do_add_mount); 17151da177e4SLinus Torvalds 17165528f911STrond Myklebust /* 17171da177e4SLinus Torvalds * process a list of expirable mountpoints with the intent of discarding any 17181da177e4SLinus Torvalds * mountpoints that aren't in use and haven't been touched since last we came 17191da177e4SLinus Torvalds * here 17201da177e4SLinus Torvalds */ 17211da177e4SLinus Torvalds void mark_mounts_for_expiry(struct list_head *mounts) 17221da177e4SLinus Torvalds { 17231da177e4SLinus Torvalds struct vfsmount *mnt, *next; 17241da177e4SLinus Torvalds LIST_HEAD(graveyard); 1725bcc5c7d2SAl Viro LIST_HEAD(umounts); 17261da177e4SLinus Torvalds 17271da177e4SLinus Torvalds if (list_empty(mounts)) 17281da177e4SLinus Torvalds return; 17291da177e4SLinus Torvalds 1730bcc5c7d2SAl Viro down_write(&namespace_sem); 17311da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 17321da177e4SLinus Torvalds 17331da177e4SLinus Torvalds /* extract from the expiration list every vfsmount that matches the 17341da177e4SLinus Torvalds * following criteria: 17351da177e4SLinus Torvalds * - only referenced by its parent vfsmount 17361da177e4SLinus Torvalds * - still marked for expiry (marked on the last call here; marks are 17371da177e4SLinus Torvalds * cleared by mntput()) 17381da177e4SLinus Torvalds */ 173955e700b9SMiklos Szeredi list_for_each_entry_safe(mnt, next, mounts, mnt_expire) { 17401da177e4SLinus Torvalds if (!xchg(&mnt->mnt_expiry_mark, 1) || 1741bcc5c7d2SAl Viro propagate_mount_busy(mnt, 1)) 17421da177e4SLinus Torvalds continue; 174355e700b9SMiklos Szeredi list_move(&mnt->mnt_expire, &graveyard); 17441da177e4SLinus Torvalds } 1745bcc5c7d2SAl Viro while (!list_empty(&graveyard)) { 1746bcc5c7d2SAl Viro mnt = list_first_entry(&graveyard, struct vfsmount, mnt_expire); 1747bcc5c7d2SAl Viro touch_mnt_namespace(mnt->mnt_ns); 1748bcc5c7d2SAl Viro umount_tree(mnt, 1, &umounts); 1749bcc5c7d2SAl Viro } 17501da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 1751bcc5c7d2SAl Viro up_write(&namespace_sem); 1752bcc5c7d2SAl Viro 1753bcc5c7d2SAl Viro release_mounts(&umounts); 17541da177e4SLinus Torvalds } 17551da177e4SLinus Torvalds 17561da177e4SLinus Torvalds EXPORT_SYMBOL_GPL(mark_mounts_for_expiry); 17571da177e4SLinus Torvalds 17581da177e4SLinus Torvalds /* 17595528f911STrond Myklebust * Ripoff of 'select_parent()' 17605528f911STrond Myklebust * 17615528f911STrond Myklebust * search the list of submounts for a given mountpoint, and move any 17625528f911STrond Myklebust * shrinkable submounts to the 'graveyard' list. 17635528f911STrond Myklebust */ 17645528f911STrond Myklebust static int select_submounts(struct vfsmount *parent, struct list_head *graveyard) 17655528f911STrond Myklebust { 17665528f911STrond Myklebust struct vfsmount *this_parent = parent; 17675528f911STrond Myklebust struct list_head *next; 17685528f911STrond Myklebust int found = 0; 17695528f911STrond Myklebust 17705528f911STrond Myklebust repeat: 17715528f911STrond Myklebust next = this_parent->mnt_mounts.next; 17725528f911STrond Myklebust resume: 17735528f911STrond Myklebust while (next != &this_parent->mnt_mounts) { 17745528f911STrond Myklebust struct list_head *tmp = next; 17755528f911STrond Myklebust struct vfsmount *mnt = list_entry(tmp, struct vfsmount, mnt_child); 17765528f911STrond Myklebust 17775528f911STrond Myklebust next = tmp->next; 17785528f911STrond Myklebust if (!(mnt->mnt_flags & MNT_SHRINKABLE)) 17795528f911STrond Myklebust continue; 17805528f911STrond Myklebust /* 17815528f911STrond Myklebust * Descend a level if the d_mounts list is non-empty. 17825528f911STrond Myklebust */ 17835528f911STrond Myklebust if (!list_empty(&mnt->mnt_mounts)) { 17845528f911STrond Myklebust this_parent = mnt; 17855528f911STrond Myklebust goto repeat; 17865528f911STrond Myklebust } 17875528f911STrond Myklebust 17885528f911STrond Myklebust if (!propagate_mount_busy(mnt, 1)) { 17895528f911STrond Myklebust list_move_tail(&mnt->mnt_expire, graveyard); 17905528f911STrond Myklebust found++; 17915528f911STrond Myklebust } 17925528f911STrond Myklebust } 17935528f911STrond Myklebust /* 17945528f911STrond Myklebust * All done at this level ... ascend and resume the search 17955528f911STrond Myklebust */ 17965528f911STrond Myklebust if (this_parent != parent) { 17975528f911STrond Myklebust next = this_parent->mnt_child.next; 17985528f911STrond Myklebust this_parent = this_parent->mnt_parent; 17995528f911STrond Myklebust goto resume; 18005528f911STrond Myklebust } 18015528f911STrond Myklebust return found; 18025528f911STrond Myklebust } 18035528f911STrond Myklebust 18045528f911STrond Myklebust /* 18055528f911STrond Myklebust * process a list of expirable mountpoints with the intent of discarding any 18065528f911STrond Myklebust * submounts of a specific parent mountpoint 18075528f911STrond Myklebust */ 1808c35038beSAl Viro static void shrink_submounts(struct vfsmount *mnt, struct list_head *umounts) 18095528f911STrond Myklebust { 18105528f911STrond Myklebust LIST_HEAD(graveyard); 1811c35038beSAl Viro struct vfsmount *m; 18125528f911STrond Myklebust 18135528f911STrond Myklebust /* extract submounts of 'mountpoint' from the expiration list */ 1814c35038beSAl Viro while (select_submounts(mnt, &graveyard)) { 1815bcc5c7d2SAl Viro while (!list_empty(&graveyard)) { 1816c35038beSAl Viro m = list_first_entry(&graveyard, struct vfsmount, 1817bcc5c7d2SAl Viro mnt_expire); 1818bcc5c7d2SAl Viro touch_mnt_namespace(mnt->mnt_ns); 1819c35038beSAl Viro umount_tree(mnt, 1, umounts); 1820bcc5c7d2SAl Viro } 1821bcc5c7d2SAl Viro } 18225528f911STrond Myklebust } 18235528f911STrond Myklebust 18245528f911STrond Myklebust /* 18251da177e4SLinus Torvalds * Some copy_from_user() implementations do not return the exact number of 18261da177e4SLinus Torvalds * bytes remaining to copy on a fault. But copy_mount_options() requires that. 18271da177e4SLinus Torvalds * Note that this function differs from copy_from_user() in that it will oops 18281da177e4SLinus Torvalds * on bad values of `to', rather than returning a short copy. 18291da177e4SLinus Torvalds */ 1830b58fed8bSRam Pai static long exact_copy_from_user(void *to, const void __user * from, 1831b58fed8bSRam Pai unsigned long n) 18321da177e4SLinus Torvalds { 18331da177e4SLinus Torvalds char *t = to; 18341da177e4SLinus Torvalds const char __user *f = from; 18351da177e4SLinus Torvalds char c; 18361da177e4SLinus Torvalds 18371da177e4SLinus Torvalds if (!access_ok(VERIFY_READ, from, n)) 18381da177e4SLinus Torvalds return n; 18391da177e4SLinus Torvalds 18401da177e4SLinus Torvalds while (n) { 18411da177e4SLinus Torvalds if (__get_user(c, f)) { 18421da177e4SLinus Torvalds memset(t, 0, n); 18431da177e4SLinus Torvalds break; 18441da177e4SLinus Torvalds } 18451da177e4SLinus Torvalds *t++ = c; 18461da177e4SLinus Torvalds f++; 18471da177e4SLinus Torvalds n--; 18481da177e4SLinus Torvalds } 18491da177e4SLinus Torvalds return n; 18501da177e4SLinus Torvalds } 18511da177e4SLinus Torvalds 18521da177e4SLinus Torvalds int copy_mount_options(const void __user * data, unsigned long *where) 18531da177e4SLinus Torvalds { 18541da177e4SLinus Torvalds int i; 18551da177e4SLinus Torvalds unsigned long page; 18561da177e4SLinus Torvalds unsigned long size; 18571da177e4SLinus Torvalds 18581da177e4SLinus Torvalds *where = 0; 18591da177e4SLinus Torvalds if (!data) 18601da177e4SLinus Torvalds return 0; 18611da177e4SLinus Torvalds 18621da177e4SLinus Torvalds if (!(page = __get_free_page(GFP_KERNEL))) 18631da177e4SLinus Torvalds return -ENOMEM; 18641da177e4SLinus Torvalds 18651da177e4SLinus Torvalds /* We only care that *some* data at the address the user 18661da177e4SLinus Torvalds * gave us is valid. Just in case, we'll zero 18671da177e4SLinus Torvalds * the remainder of the page. 18681da177e4SLinus Torvalds */ 18691da177e4SLinus Torvalds /* copy_from_user cannot cross TASK_SIZE ! */ 18701da177e4SLinus Torvalds size = TASK_SIZE - (unsigned long)data; 18711da177e4SLinus Torvalds if (size > PAGE_SIZE) 18721da177e4SLinus Torvalds size = PAGE_SIZE; 18731da177e4SLinus Torvalds 18741da177e4SLinus Torvalds i = size - exact_copy_from_user((void *)page, data, size); 18751da177e4SLinus Torvalds if (!i) { 18761da177e4SLinus Torvalds free_page(page); 18771da177e4SLinus Torvalds return -EFAULT; 18781da177e4SLinus Torvalds } 18791da177e4SLinus Torvalds if (i != PAGE_SIZE) 18801da177e4SLinus Torvalds memset((char *)page + i, 0, PAGE_SIZE - i); 18811da177e4SLinus Torvalds *where = page; 18821da177e4SLinus Torvalds return 0; 18831da177e4SLinus Torvalds } 18841da177e4SLinus Torvalds 18851da177e4SLinus Torvalds /* 18861da177e4SLinus Torvalds * Flags is a 32-bit value that allows up to 31 non-fs dependent flags to 18871da177e4SLinus Torvalds * be given to the mount() call (ie: read-only, no-dev, no-suid etc). 18881da177e4SLinus Torvalds * 18891da177e4SLinus Torvalds * data is a (void *) that can point to any structure up to 18901da177e4SLinus Torvalds * PAGE_SIZE-1 bytes, which can contain arbitrary fs-dependent 18911da177e4SLinus Torvalds * information (or be NULL). 18921da177e4SLinus Torvalds * 18931da177e4SLinus Torvalds * Pre-0.97 versions of mount() didn't have a flags word. 18941da177e4SLinus Torvalds * When the flags word was introduced its top half was required 18951da177e4SLinus Torvalds * to have the magic value 0xC0ED, and this remained so until 2.4.0-test9. 18961da177e4SLinus Torvalds * Therefore, if this magic number is present, it carries no information 18971da177e4SLinus Torvalds * and must be discarded. 18981da177e4SLinus Torvalds */ 18991da177e4SLinus Torvalds long do_mount(char *dev_name, char *dir_name, char *type_page, 19001da177e4SLinus Torvalds unsigned long flags, void *data_page) 19011da177e4SLinus Torvalds { 19022d92ab3cSAl Viro struct path path; 19031da177e4SLinus Torvalds int retval = 0; 19041da177e4SLinus Torvalds int mnt_flags = 0; 19051da177e4SLinus Torvalds 19061da177e4SLinus Torvalds /* Discard magic */ 19071da177e4SLinus Torvalds if ((flags & MS_MGC_MSK) == MS_MGC_VAL) 19081da177e4SLinus Torvalds flags &= ~MS_MGC_MSK; 19091da177e4SLinus Torvalds 19101da177e4SLinus Torvalds /* Basic sanity checks */ 19111da177e4SLinus Torvalds 19121da177e4SLinus Torvalds if (!dir_name || !*dir_name || !memchr(dir_name, 0, PAGE_SIZE)) 19131da177e4SLinus Torvalds return -EINVAL; 19141da177e4SLinus Torvalds if (dev_name && !memchr(dev_name, 0, PAGE_SIZE)) 19151da177e4SLinus Torvalds return -EINVAL; 19161da177e4SLinus Torvalds 19171da177e4SLinus Torvalds if (data_page) 19181da177e4SLinus Torvalds ((char *)data_page)[PAGE_SIZE - 1] = 0; 19191da177e4SLinus Torvalds 19201da177e4SLinus Torvalds /* Separate the per-mountpoint flags */ 19211da177e4SLinus Torvalds if (flags & MS_NOSUID) 19221da177e4SLinus Torvalds mnt_flags |= MNT_NOSUID; 19231da177e4SLinus Torvalds if (flags & MS_NODEV) 19241da177e4SLinus Torvalds mnt_flags |= MNT_NODEV; 19251da177e4SLinus Torvalds if (flags & MS_NOEXEC) 19261da177e4SLinus Torvalds mnt_flags |= MNT_NOEXEC; 1927fc33a7bbSChristoph Hellwig if (flags & MS_NOATIME) 1928fc33a7bbSChristoph Hellwig mnt_flags |= MNT_NOATIME; 1929fc33a7bbSChristoph Hellwig if (flags & MS_NODIRATIME) 1930fc33a7bbSChristoph Hellwig mnt_flags |= MNT_NODIRATIME; 193147ae32d6SValerie Henson if (flags & MS_RELATIME) 193247ae32d6SValerie Henson mnt_flags |= MNT_RELATIME; 19332e4b7fcdSDave Hansen if (flags & MS_RDONLY) 19342e4b7fcdSDave Hansen mnt_flags |= MNT_READONLY; 1935fc33a7bbSChristoph Hellwig 1936fc33a7bbSChristoph Hellwig flags &= ~(MS_NOSUID | MS_NOEXEC | MS_NODEV | MS_ACTIVE | 19378bf9725cSPavel Emelyanov MS_NOATIME | MS_NODIRATIME | MS_RELATIME| MS_KERNMOUNT); 19381da177e4SLinus Torvalds 19391da177e4SLinus Torvalds /* ... and get the mountpoint */ 19402d92ab3cSAl Viro retval = kern_path(dir_name, LOOKUP_FOLLOW, &path); 19411da177e4SLinus Torvalds if (retval) 19421da177e4SLinus Torvalds return retval; 19431da177e4SLinus Torvalds 19442d92ab3cSAl Viro retval = security_sb_mount(dev_name, &path, 1945b5266eb4SAl Viro type_page, flags, data_page); 19461da177e4SLinus Torvalds if (retval) 19471da177e4SLinus Torvalds goto dput_out; 19481da177e4SLinus Torvalds 19491da177e4SLinus Torvalds if (flags & MS_REMOUNT) 19502d92ab3cSAl Viro retval = do_remount(&path, flags & ~MS_REMOUNT, mnt_flags, 19511da177e4SLinus Torvalds data_page); 19521da177e4SLinus Torvalds else if (flags & MS_BIND) 19532d92ab3cSAl Viro retval = do_loopback(&path, dev_name, flags & MS_REC); 19549676f0c6SRam Pai else if (flags & (MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE)) 19552d92ab3cSAl Viro retval = do_change_type(&path, flags); 19561da177e4SLinus Torvalds else if (flags & MS_MOVE) 19572d92ab3cSAl Viro retval = do_move_mount(&path, dev_name); 19581da177e4SLinus Torvalds else 19592d92ab3cSAl Viro retval = do_new_mount(&path, type_page, flags, mnt_flags, 19601da177e4SLinus Torvalds dev_name, data_page); 19611da177e4SLinus Torvalds dput_out: 19622d92ab3cSAl Viro path_put(&path); 19631da177e4SLinus Torvalds return retval; 19641da177e4SLinus Torvalds } 19651da177e4SLinus Torvalds 1966741a2951SJANAK DESAI /* 1967741a2951SJANAK DESAI * Allocate a new namespace structure and populate it with contents 1968741a2951SJANAK DESAI * copied from the namespace of the passed in task structure. 1969741a2951SJANAK DESAI */ 1970e3222c4eSBadari Pulavarty static struct mnt_namespace *dup_mnt_ns(struct mnt_namespace *mnt_ns, 19716b3286edSKirill Korotaev struct fs_struct *fs) 19721da177e4SLinus Torvalds { 19736b3286edSKirill Korotaev struct mnt_namespace *new_ns; 19747f2da1e7SAl Viro struct vfsmount *rootmnt = NULL, *pwdmnt = NULL; 19751da177e4SLinus Torvalds struct vfsmount *p, *q; 19761da177e4SLinus Torvalds 19776b3286edSKirill Korotaev new_ns = kmalloc(sizeof(struct mnt_namespace), GFP_KERNEL); 19781da177e4SLinus Torvalds if (!new_ns) 1979467e9f4bSCedric Le Goater return ERR_PTR(-ENOMEM); 19801da177e4SLinus Torvalds 19811da177e4SLinus Torvalds atomic_set(&new_ns->count, 1); 19821da177e4SLinus Torvalds INIT_LIST_HEAD(&new_ns->list); 19835addc5ddSAl Viro init_waitqueue_head(&new_ns->poll); 19845addc5ddSAl Viro new_ns->event = 0; 19851da177e4SLinus Torvalds 1986390c6843SRam Pai down_write(&namespace_sem); 19871da177e4SLinus Torvalds /* First pass: copy the tree topology */ 19886b3286edSKirill Korotaev new_ns->root = copy_tree(mnt_ns->root, mnt_ns->root->mnt_root, 19899676f0c6SRam Pai CL_COPY_ALL | CL_EXPIRE); 19901da177e4SLinus Torvalds if (!new_ns->root) { 1991390c6843SRam Pai up_write(&namespace_sem); 19921da177e4SLinus Torvalds kfree(new_ns); 1993467e9f4bSCedric Le Goater return ERR_PTR(-ENOMEM);; 19941da177e4SLinus Torvalds } 19951da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 19961da177e4SLinus Torvalds list_add_tail(&new_ns->list, &new_ns->root->mnt_list); 19971da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 19981da177e4SLinus Torvalds 19991da177e4SLinus Torvalds /* 20001da177e4SLinus Torvalds * Second pass: switch the tsk->fs->* elements and mark new vfsmounts 20011da177e4SLinus Torvalds * as belonging to new namespace. We have already acquired a private 20021da177e4SLinus Torvalds * fs_struct, so tsk->fs->lock is not needed. 20031da177e4SLinus Torvalds */ 20046b3286edSKirill Korotaev p = mnt_ns->root; 20051da177e4SLinus Torvalds q = new_ns->root; 20061da177e4SLinus Torvalds while (p) { 20076b3286edSKirill Korotaev q->mnt_ns = new_ns; 20081da177e4SLinus Torvalds if (fs) { 20096ac08c39SJan Blunck if (p == fs->root.mnt) { 20101da177e4SLinus Torvalds rootmnt = p; 20116ac08c39SJan Blunck fs->root.mnt = mntget(q); 20121da177e4SLinus Torvalds } 20136ac08c39SJan Blunck if (p == fs->pwd.mnt) { 20141da177e4SLinus Torvalds pwdmnt = p; 20156ac08c39SJan Blunck fs->pwd.mnt = mntget(q); 20161da177e4SLinus Torvalds } 20171da177e4SLinus Torvalds } 20186b3286edSKirill Korotaev p = next_mnt(p, mnt_ns->root); 20191da177e4SLinus Torvalds q = next_mnt(q, new_ns->root); 20201da177e4SLinus Torvalds } 2021390c6843SRam Pai up_write(&namespace_sem); 20221da177e4SLinus Torvalds 20231da177e4SLinus Torvalds if (rootmnt) 20241da177e4SLinus Torvalds mntput(rootmnt); 20251da177e4SLinus Torvalds if (pwdmnt) 20261da177e4SLinus Torvalds mntput(pwdmnt); 20271da177e4SLinus Torvalds 2028741a2951SJANAK DESAI return new_ns; 2029741a2951SJANAK DESAI } 2030741a2951SJANAK DESAI 2031213dd266SEric W. Biederman struct mnt_namespace *copy_mnt_ns(unsigned long flags, struct mnt_namespace *ns, 2032e3222c4eSBadari Pulavarty struct fs_struct *new_fs) 2033741a2951SJANAK DESAI { 20346b3286edSKirill Korotaev struct mnt_namespace *new_ns; 2035741a2951SJANAK DESAI 2036e3222c4eSBadari Pulavarty BUG_ON(!ns); 20376b3286edSKirill Korotaev get_mnt_ns(ns); 2038741a2951SJANAK DESAI 2039741a2951SJANAK DESAI if (!(flags & CLONE_NEWNS)) 2040e3222c4eSBadari Pulavarty return ns; 2041741a2951SJANAK DESAI 2042e3222c4eSBadari Pulavarty new_ns = dup_mnt_ns(ns, new_fs); 2043741a2951SJANAK DESAI 20446b3286edSKirill Korotaev put_mnt_ns(ns); 2045e3222c4eSBadari Pulavarty return new_ns; 20461da177e4SLinus Torvalds } 20471da177e4SLinus Torvalds 20481da177e4SLinus Torvalds asmlinkage long sys_mount(char __user * dev_name, char __user * dir_name, 20491da177e4SLinus Torvalds char __user * type, unsigned long flags, 20501da177e4SLinus Torvalds void __user * data) 20511da177e4SLinus Torvalds { 20521da177e4SLinus Torvalds int retval; 20531da177e4SLinus Torvalds unsigned long data_page; 20541da177e4SLinus Torvalds unsigned long type_page; 20551da177e4SLinus Torvalds unsigned long dev_page; 20561da177e4SLinus Torvalds char *dir_page; 20571da177e4SLinus Torvalds 20581da177e4SLinus Torvalds retval = copy_mount_options(type, &type_page); 20591da177e4SLinus Torvalds if (retval < 0) 20601da177e4SLinus Torvalds return retval; 20611da177e4SLinus Torvalds 20621da177e4SLinus Torvalds dir_page = getname(dir_name); 20631da177e4SLinus Torvalds retval = PTR_ERR(dir_page); 20641da177e4SLinus Torvalds if (IS_ERR(dir_page)) 20651da177e4SLinus Torvalds goto out1; 20661da177e4SLinus Torvalds 20671da177e4SLinus Torvalds retval = copy_mount_options(dev_name, &dev_page); 20681da177e4SLinus Torvalds if (retval < 0) 20691da177e4SLinus Torvalds goto out2; 20701da177e4SLinus Torvalds 20711da177e4SLinus Torvalds retval = copy_mount_options(data, &data_page); 20721da177e4SLinus Torvalds if (retval < 0) 20731da177e4SLinus Torvalds goto out3; 20741da177e4SLinus Torvalds 20751da177e4SLinus Torvalds lock_kernel(); 20761da177e4SLinus Torvalds retval = do_mount((char *)dev_page, dir_page, (char *)type_page, 20771da177e4SLinus Torvalds flags, (void *)data_page); 20781da177e4SLinus Torvalds unlock_kernel(); 20791da177e4SLinus Torvalds free_page(data_page); 20801da177e4SLinus Torvalds 20811da177e4SLinus Torvalds out3: 20821da177e4SLinus Torvalds free_page(dev_page); 20831da177e4SLinus Torvalds out2: 20841da177e4SLinus Torvalds putname(dir_page); 20851da177e4SLinus Torvalds out1: 20861da177e4SLinus Torvalds free_page(type_page); 20871da177e4SLinus Torvalds return retval; 20881da177e4SLinus Torvalds } 20891da177e4SLinus Torvalds 20901da177e4SLinus Torvalds /* 20911da177e4SLinus Torvalds * Replace the fs->{rootmnt,root} with {mnt,dentry}. Put the old values. 20921da177e4SLinus Torvalds * It can block. Requires the big lock held. 20931da177e4SLinus Torvalds */ 2094ac748a09SJan Blunck void set_fs_root(struct fs_struct *fs, struct path *path) 20951da177e4SLinus Torvalds { 20966ac08c39SJan Blunck struct path old_root; 20976ac08c39SJan Blunck 20981da177e4SLinus Torvalds write_lock(&fs->lock); 20991da177e4SLinus Torvalds old_root = fs->root; 2100ac748a09SJan Blunck fs->root = *path; 2101ac748a09SJan Blunck path_get(path); 21021da177e4SLinus Torvalds write_unlock(&fs->lock); 21036ac08c39SJan Blunck if (old_root.dentry) 21046ac08c39SJan Blunck path_put(&old_root); 21051da177e4SLinus Torvalds } 21061da177e4SLinus Torvalds 21071da177e4SLinus Torvalds /* 21081da177e4SLinus Torvalds * Replace the fs->{pwdmnt,pwd} with {mnt,dentry}. Put the old values. 21091da177e4SLinus Torvalds * It can block. Requires the big lock held. 21101da177e4SLinus Torvalds */ 2111ac748a09SJan Blunck void set_fs_pwd(struct fs_struct *fs, struct path *path) 21121da177e4SLinus Torvalds { 21136ac08c39SJan Blunck struct path old_pwd; 21141da177e4SLinus Torvalds 21151da177e4SLinus Torvalds write_lock(&fs->lock); 21161da177e4SLinus Torvalds old_pwd = fs->pwd; 2117ac748a09SJan Blunck fs->pwd = *path; 2118ac748a09SJan Blunck path_get(path); 21191da177e4SLinus Torvalds write_unlock(&fs->lock); 21201da177e4SLinus Torvalds 21216ac08c39SJan Blunck if (old_pwd.dentry) 21226ac08c39SJan Blunck path_put(&old_pwd); 21231da177e4SLinus Torvalds } 21241da177e4SLinus Torvalds 21251a390689SAl Viro static void chroot_fs_refs(struct path *old_root, struct path *new_root) 21261da177e4SLinus Torvalds { 21271da177e4SLinus Torvalds struct task_struct *g, *p; 21281da177e4SLinus Torvalds struct fs_struct *fs; 21291da177e4SLinus Torvalds 21301da177e4SLinus Torvalds read_lock(&tasklist_lock); 21311da177e4SLinus Torvalds do_each_thread(g, p) { 21321da177e4SLinus Torvalds task_lock(p); 21331da177e4SLinus Torvalds fs = p->fs; 21341da177e4SLinus Torvalds if (fs) { 21351da177e4SLinus Torvalds atomic_inc(&fs->count); 21361da177e4SLinus Torvalds task_unlock(p); 21371a390689SAl Viro if (fs->root.dentry == old_root->dentry 21381a390689SAl Viro && fs->root.mnt == old_root->mnt) 21391a390689SAl Viro set_fs_root(fs, new_root); 21401a390689SAl Viro if (fs->pwd.dentry == old_root->dentry 21411a390689SAl Viro && fs->pwd.mnt == old_root->mnt) 21421a390689SAl Viro set_fs_pwd(fs, new_root); 21431da177e4SLinus Torvalds put_fs_struct(fs); 21441da177e4SLinus Torvalds } else 21451da177e4SLinus Torvalds task_unlock(p); 21461da177e4SLinus Torvalds } while_each_thread(g, p); 21471da177e4SLinus Torvalds read_unlock(&tasklist_lock); 21481da177e4SLinus Torvalds } 21491da177e4SLinus Torvalds 21501da177e4SLinus Torvalds /* 21511da177e4SLinus Torvalds * pivot_root Semantics: 21521da177e4SLinus Torvalds * Moves the root file system of the current process to the directory put_old, 21531da177e4SLinus Torvalds * makes new_root as the new root file system of the current process, and sets 21541da177e4SLinus Torvalds * root/cwd of all processes which had them on the current root to new_root. 21551da177e4SLinus Torvalds * 21561da177e4SLinus Torvalds * Restrictions: 21571da177e4SLinus Torvalds * The new_root and put_old must be directories, and must not be on the 21581da177e4SLinus Torvalds * same file system as the current process root. The put_old must be 21591da177e4SLinus Torvalds * underneath new_root, i.e. adding a non-zero number of /.. to the string 21601da177e4SLinus Torvalds * pointed to by put_old must yield the same directory as new_root. No other 21611da177e4SLinus Torvalds * file system may be mounted on put_old. After all, new_root is a mountpoint. 21621da177e4SLinus Torvalds * 21634a0d11faSNeil Brown * Also, the current root cannot be on the 'rootfs' (initial ramfs) filesystem. 21644a0d11faSNeil Brown * See Documentation/filesystems/ramfs-rootfs-initramfs.txt for alternatives 21654a0d11faSNeil Brown * in this situation. 21664a0d11faSNeil Brown * 21671da177e4SLinus Torvalds * Notes: 21681da177e4SLinus Torvalds * - we don't move root/cwd if they are not at the root (reason: if something 21691da177e4SLinus Torvalds * cared enough to change them, it's probably wrong to force them elsewhere) 21701da177e4SLinus Torvalds * - it's okay to pick a root that isn't the root of a file system, e.g. 21711da177e4SLinus Torvalds * /nfs/my_root where /nfs is the mount point. It must be a mountpoint, 21721da177e4SLinus Torvalds * though, so you may need to say mount --bind /nfs/my_root /nfs/my_root 21731da177e4SLinus Torvalds * first. 21741da177e4SLinus Torvalds */ 2175b58fed8bSRam Pai asmlinkage long sys_pivot_root(const char __user * new_root, 2176b58fed8bSRam Pai const char __user * put_old) 21771da177e4SLinus Torvalds { 21781da177e4SLinus Torvalds struct vfsmount *tmp; 21792d8f3038SAl Viro struct path new, old, parent_path, root_parent, root; 21801da177e4SLinus Torvalds int error; 21811da177e4SLinus Torvalds 21821da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 21831da177e4SLinus Torvalds return -EPERM; 21841da177e4SLinus Torvalds 21852d8f3038SAl Viro error = user_path_dir(new_root, &new); 21861da177e4SLinus Torvalds if (error) 21871da177e4SLinus Torvalds goto out0; 21881da177e4SLinus Torvalds error = -EINVAL; 21892d8f3038SAl Viro if (!check_mnt(new.mnt)) 21901da177e4SLinus Torvalds goto out1; 21911da177e4SLinus Torvalds 21922d8f3038SAl Viro error = user_path_dir(put_old, &old); 21931da177e4SLinus Torvalds if (error) 21941da177e4SLinus Torvalds goto out1; 21951da177e4SLinus Torvalds 21962d8f3038SAl Viro error = security_sb_pivotroot(&old, &new); 21971da177e4SLinus Torvalds if (error) { 21982d8f3038SAl Viro path_put(&old); 21991da177e4SLinus Torvalds goto out1; 22001da177e4SLinus Torvalds } 22011da177e4SLinus Torvalds 22021da177e4SLinus Torvalds read_lock(¤t->fs->lock); 22038c3ee42eSAl Viro root = current->fs->root; 22046ac08c39SJan Blunck path_get(¤t->fs->root); 22051da177e4SLinus Torvalds read_unlock(¤t->fs->lock); 2206390c6843SRam Pai down_write(&namespace_sem); 22072d8f3038SAl Viro mutex_lock(&old.dentry->d_inode->i_mutex); 22081da177e4SLinus Torvalds error = -EINVAL; 22092d8f3038SAl Viro if (IS_MNT_SHARED(old.mnt) || 22102d8f3038SAl Viro IS_MNT_SHARED(new.mnt->mnt_parent) || 22118c3ee42eSAl Viro IS_MNT_SHARED(root.mnt->mnt_parent)) 221221444403SRam Pai goto out2; 22138c3ee42eSAl Viro if (!check_mnt(root.mnt)) 22141da177e4SLinus Torvalds goto out2; 22151da177e4SLinus Torvalds error = -ENOENT; 22162d8f3038SAl Viro if (IS_DEADDIR(new.dentry->d_inode)) 22171da177e4SLinus Torvalds goto out2; 22182d8f3038SAl Viro if (d_unhashed(new.dentry) && !IS_ROOT(new.dentry)) 22191da177e4SLinus Torvalds goto out2; 22202d8f3038SAl Viro if (d_unhashed(old.dentry) && !IS_ROOT(old.dentry)) 22211da177e4SLinus Torvalds goto out2; 22221da177e4SLinus Torvalds error = -EBUSY; 22232d8f3038SAl Viro if (new.mnt == root.mnt || 22242d8f3038SAl Viro old.mnt == root.mnt) 22251da177e4SLinus Torvalds goto out2; /* loop, on the same file system */ 22261da177e4SLinus Torvalds error = -EINVAL; 22278c3ee42eSAl Viro if (root.mnt->mnt_root != root.dentry) 22281da177e4SLinus Torvalds goto out2; /* not a mountpoint */ 22298c3ee42eSAl Viro if (root.mnt->mnt_parent == root.mnt) 22300bb6fcc1SMiklos Szeredi goto out2; /* not attached */ 22312d8f3038SAl Viro if (new.mnt->mnt_root != new.dentry) 22321da177e4SLinus Torvalds goto out2; /* not a mountpoint */ 22332d8f3038SAl Viro if (new.mnt->mnt_parent == new.mnt) 22340bb6fcc1SMiklos Szeredi goto out2; /* not attached */ 22354ac91378SJan Blunck /* make sure we can reach put_old from new_root */ 22362d8f3038SAl Viro tmp = old.mnt; 22371da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 22382d8f3038SAl Viro if (tmp != new.mnt) { 22391da177e4SLinus Torvalds for (;;) { 22401da177e4SLinus Torvalds if (tmp->mnt_parent == tmp) 22411da177e4SLinus Torvalds goto out3; /* already mounted on put_old */ 22422d8f3038SAl Viro if (tmp->mnt_parent == new.mnt) 22431da177e4SLinus Torvalds break; 22441da177e4SLinus Torvalds tmp = tmp->mnt_parent; 22451da177e4SLinus Torvalds } 22462d8f3038SAl Viro if (!is_subdir(tmp->mnt_mountpoint, new.dentry)) 22471da177e4SLinus Torvalds goto out3; 22482d8f3038SAl Viro } else if (!is_subdir(old.dentry, new.dentry)) 22491da177e4SLinus Torvalds goto out3; 22502d8f3038SAl Viro detach_mnt(new.mnt, &parent_path); 22518c3ee42eSAl Viro detach_mnt(root.mnt, &root_parent); 22524ac91378SJan Blunck /* mount old root on put_old */ 22532d8f3038SAl Viro attach_mnt(root.mnt, &old); 22544ac91378SJan Blunck /* mount new_root on / */ 22552d8f3038SAl Viro attach_mnt(new.mnt, &root_parent); 22566b3286edSKirill Korotaev touch_mnt_namespace(current->nsproxy->mnt_ns); 22571da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 22582d8f3038SAl Viro chroot_fs_refs(&root, &new); 22592d8f3038SAl Viro security_sb_post_pivotroot(&root, &new); 22601da177e4SLinus Torvalds error = 0; 22611a390689SAl Viro path_put(&root_parent); 22621a390689SAl Viro path_put(&parent_path); 22631da177e4SLinus Torvalds out2: 22642d8f3038SAl Viro mutex_unlock(&old.dentry->d_inode->i_mutex); 2265390c6843SRam Pai up_write(&namespace_sem); 22668c3ee42eSAl Viro path_put(&root); 22672d8f3038SAl Viro path_put(&old); 22681da177e4SLinus Torvalds out1: 22692d8f3038SAl Viro path_put(&new); 22701da177e4SLinus Torvalds out0: 22711da177e4SLinus Torvalds return error; 22721da177e4SLinus Torvalds out3: 22731da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 22741da177e4SLinus Torvalds goto out2; 22751da177e4SLinus Torvalds } 22761da177e4SLinus Torvalds 22771da177e4SLinus Torvalds static void __init init_mount_tree(void) 22781da177e4SLinus Torvalds { 22791da177e4SLinus Torvalds struct vfsmount *mnt; 22806b3286edSKirill Korotaev struct mnt_namespace *ns; 2281ac748a09SJan Blunck struct path root; 22821da177e4SLinus Torvalds 22831da177e4SLinus Torvalds mnt = do_kern_mount("rootfs", 0, "rootfs", NULL); 22841da177e4SLinus Torvalds if (IS_ERR(mnt)) 22851da177e4SLinus Torvalds panic("Can't create rootfs"); 22866b3286edSKirill Korotaev ns = kmalloc(sizeof(*ns), GFP_KERNEL); 22876b3286edSKirill Korotaev if (!ns) 22881da177e4SLinus Torvalds panic("Can't allocate initial namespace"); 22896b3286edSKirill Korotaev atomic_set(&ns->count, 1); 22906b3286edSKirill Korotaev INIT_LIST_HEAD(&ns->list); 22916b3286edSKirill Korotaev init_waitqueue_head(&ns->poll); 22926b3286edSKirill Korotaev ns->event = 0; 22936b3286edSKirill Korotaev list_add(&mnt->mnt_list, &ns->list); 22946b3286edSKirill Korotaev ns->root = mnt; 22956b3286edSKirill Korotaev mnt->mnt_ns = ns; 22961da177e4SLinus Torvalds 22976b3286edSKirill Korotaev init_task.nsproxy->mnt_ns = ns; 22986b3286edSKirill Korotaev get_mnt_ns(ns); 22991da177e4SLinus Torvalds 2300ac748a09SJan Blunck root.mnt = ns->root; 2301ac748a09SJan Blunck root.dentry = ns->root->mnt_root; 2302ac748a09SJan Blunck 2303ac748a09SJan Blunck set_fs_pwd(current->fs, &root); 2304ac748a09SJan Blunck set_fs_root(current->fs, &root); 23051da177e4SLinus Torvalds } 23061da177e4SLinus Torvalds 230774bf17cfSDenis Cheng void __init mnt_init(void) 23081da177e4SLinus Torvalds { 230913f14b4dSEric Dumazet unsigned u; 231015a67dd8SRandy Dunlap int err; 23111da177e4SLinus Torvalds 2312390c6843SRam Pai init_rwsem(&namespace_sem); 2313390c6843SRam Pai 23141da177e4SLinus Torvalds mnt_cache = kmem_cache_create("mnt_cache", sizeof(struct vfsmount), 231520c2df83SPaul Mundt 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC, NULL); 23161da177e4SLinus Torvalds 2317b58fed8bSRam Pai mount_hashtable = (struct list_head *)__get_free_page(GFP_ATOMIC); 23181da177e4SLinus Torvalds 23191da177e4SLinus Torvalds if (!mount_hashtable) 23201da177e4SLinus Torvalds panic("Failed to allocate mount hash table\n"); 23211da177e4SLinus Torvalds 232213f14b4dSEric Dumazet printk("Mount-cache hash table entries: %lu\n", HASH_SIZE); 23231da177e4SLinus Torvalds 232413f14b4dSEric Dumazet for (u = 0; u < HASH_SIZE; u++) 232513f14b4dSEric Dumazet INIT_LIST_HEAD(&mount_hashtable[u]); 23261da177e4SLinus Torvalds 232715a67dd8SRandy Dunlap err = sysfs_init(); 232815a67dd8SRandy Dunlap if (err) 232915a67dd8SRandy Dunlap printk(KERN_WARNING "%s: sysfs_init error: %d\n", 23308e24eea7SHarvey Harrison __func__, err); 233100d26666SGreg Kroah-Hartman fs_kobj = kobject_create_and_add("fs", NULL); 233200d26666SGreg Kroah-Hartman if (!fs_kobj) 23338e24eea7SHarvey Harrison printk(KERN_WARNING "%s: kobj create error\n", __func__); 23341da177e4SLinus Torvalds init_rootfs(); 23351da177e4SLinus Torvalds init_mount_tree(); 23361da177e4SLinus Torvalds } 23371da177e4SLinus Torvalds 23386b3286edSKirill Korotaev void __put_mnt_ns(struct mnt_namespace *ns) 23391da177e4SLinus Torvalds { 23406b3286edSKirill Korotaev struct vfsmount *root = ns->root; 234170fbcdf4SRam Pai LIST_HEAD(umount_list); 23426b3286edSKirill Korotaev ns->root = NULL; 23431ce88cf4SMiklos Szeredi spin_unlock(&vfsmount_lock); 2344390c6843SRam Pai down_write(&namespace_sem); 23451da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 2346a05964f3SRam Pai umount_tree(root, 0, &umount_list); 23471da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 2348390c6843SRam Pai up_write(&namespace_sem); 234970fbcdf4SRam Pai release_mounts(&umount_list); 23506b3286edSKirill Korotaev kfree(ns); 23511da177e4SLinus Torvalds } 2352