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> 305ad4e53bSAl Viro #include <linux/fs_struct.h> 311da177e4SLinus Torvalds #include <asm/uaccess.h> 321da177e4SLinus Torvalds #include <asm/unistd.h> 3307b20889SRam Pai #include "pnode.h" 34948730b0SAdrian Bunk #include "internal.h" 351da177e4SLinus Torvalds 3613f14b4dSEric Dumazet #define HASH_SHIFT ilog2(PAGE_SIZE / sizeof(struct list_head)) 3713f14b4dSEric Dumazet #define HASH_SIZE (1UL << HASH_SHIFT) 3813f14b4dSEric Dumazet 391da177e4SLinus Torvalds /* spinlock for vfsmount related operations, inplace of dcache_lock */ 401da177e4SLinus Torvalds __cacheline_aligned_in_smp DEFINE_SPINLOCK(vfsmount_lock); 411da177e4SLinus Torvalds 425addc5ddSAl Viro static int event; 4373cd49ecSMiklos Szeredi static DEFINE_IDA(mnt_id_ida); 44719f5d7fSMiklos Szeredi static DEFINE_IDA(mnt_group_ida); 45*f21f6220SAl Viro static int mnt_id_start = 0; 46*f21f6220SAl Viro static int mnt_group_start = 1; 475addc5ddSAl Viro 48fa3536ccSEric Dumazet static struct list_head *mount_hashtable __read_mostly; 49e18b890bSChristoph Lameter static struct kmem_cache *mnt_cache __read_mostly; 50390c6843SRam Pai static struct rw_semaphore namespace_sem; 511da177e4SLinus Torvalds 52f87fd4c2SMiklos Szeredi /* /sys/fs */ 5300d26666SGreg Kroah-Hartman struct kobject *fs_kobj; 5400d26666SGreg Kroah-Hartman EXPORT_SYMBOL_GPL(fs_kobj); 55f87fd4c2SMiklos Szeredi 561da177e4SLinus Torvalds static inline unsigned long hash(struct vfsmount *mnt, struct dentry *dentry) 571da177e4SLinus Torvalds { 581da177e4SLinus Torvalds unsigned long tmp = ((unsigned long)mnt / L1_CACHE_BYTES); 591da177e4SLinus Torvalds tmp += ((unsigned long)dentry / L1_CACHE_BYTES); 6013f14b4dSEric Dumazet tmp = tmp + (tmp >> HASH_SHIFT); 6113f14b4dSEric Dumazet return tmp & (HASH_SIZE - 1); 621da177e4SLinus Torvalds } 631da177e4SLinus Torvalds 643d733633SDave Hansen #define MNT_WRITER_UNDERFLOW_LIMIT -(1<<16) 653d733633SDave Hansen 6673cd49ecSMiklos Szeredi /* allocation is serialized by namespace_sem */ 6773cd49ecSMiklos Szeredi static int mnt_alloc_id(struct vfsmount *mnt) 6873cd49ecSMiklos Szeredi { 6973cd49ecSMiklos Szeredi int res; 7073cd49ecSMiklos Szeredi 7173cd49ecSMiklos Szeredi retry: 7273cd49ecSMiklos Szeredi ida_pre_get(&mnt_id_ida, GFP_KERNEL); 7373cd49ecSMiklos Szeredi spin_lock(&vfsmount_lock); 74*f21f6220SAl Viro res = ida_get_new_above(&mnt_id_ida, mnt_id_start, &mnt->mnt_id); 75*f21f6220SAl Viro if (!res) 76*f21f6220SAl Viro mnt_id_start = mnt->mnt_id + 1; 7773cd49ecSMiklos Szeredi spin_unlock(&vfsmount_lock); 7873cd49ecSMiklos Szeredi if (res == -EAGAIN) 7973cd49ecSMiklos Szeredi goto retry; 8073cd49ecSMiklos Szeredi 8173cd49ecSMiklos Szeredi return res; 8273cd49ecSMiklos Szeredi } 8373cd49ecSMiklos Szeredi 8473cd49ecSMiklos Szeredi static void mnt_free_id(struct vfsmount *mnt) 8573cd49ecSMiklos Szeredi { 86*f21f6220SAl Viro int id = mnt->mnt_id; 8773cd49ecSMiklos Szeredi spin_lock(&vfsmount_lock); 88*f21f6220SAl Viro ida_remove(&mnt_id_ida, id); 89*f21f6220SAl Viro if (mnt_id_start > id) 90*f21f6220SAl Viro mnt_id_start = id; 9173cd49ecSMiklos Szeredi spin_unlock(&vfsmount_lock); 9273cd49ecSMiklos Szeredi } 9373cd49ecSMiklos Szeredi 94719f5d7fSMiklos Szeredi /* 95719f5d7fSMiklos Szeredi * Allocate a new peer group ID 96719f5d7fSMiklos Szeredi * 97719f5d7fSMiklos Szeredi * mnt_group_ida is protected by namespace_sem 98719f5d7fSMiklos Szeredi */ 99719f5d7fSMiklos Szeredi static int mnt_alloc_group_id(struct vfsmount *mnt) 100719f5d7fSMiklos Szeredi { 101*f21f6220SAl Viro int res; 102*f21f6220SAl Viro 103719f5d7fSMiklos Szeredi if (!ida_pre_get(&mnt_group_ida, GFP_KERNEL)) 104719f5d7fSMiklos Szeredi return -ENOMEM; 105719f5d7fSMiklos Szeredi 106*f21f6220SAl Viro res = ida_get_new_above(&mnt_group_ida, 107*f21f6220SAl Viro mnt_group_start, 108*f21f6220SAl Viro &mnt->mnt_group_id); 109*f21f6220SAl Viro if (!res) 110*f21f6220SAl Viro mnt_group_start = mnt->mnt_group_id + 1; 111*f21f6220SAl Viro 112*f21f6220SAl Viro return res; 113719f5d7fSMiklos Szeredi } 114719f5d7fSMiklos Szeredi 115719f5d7fSMiklos Szeredi /* 116719f5d7fSMiklos Szeredi * Release a peer group ID 117719f5d7fSMiklos Szeredi */ 118719f5d7fSMiklos Szeredi void mnt_release_group_id(struct vfsmount *mnt) 119719f5d7fSMiklos Szeredi { 120*f21f6220SAl Viro int id = mnt->mnt_group_id; 121*f21f6220SAl Viro ida_remove(&mnt_group_ida, id); 122*f21f6220SAl Viro if (mnt_group_start > id) 123*f21f6220SAl Viro mnt_group_start = id; 124719f5d7fSMiklos Szeredi mnt->mnt_group_id = 0; 125719f5d7fSMiklos Szeredi } 126719f5d7fSMiklos Szeredi 1271da177e4SLinus Torvalds struct vfsmount *alloc_vfsmnt(const char *name) 1281da177e4SLinus Torvalds { 129c3762229SRobert P. J. Day struct vfsmount *mnt = kmem_cache_zalloc(mnt_cache, GFP_KERNEL); 1301da177e4SLinus Torvalds if (mnt) { 13173cd49ecSMiklos Szeredi int err; 13273cd49ecSMiklos Szeredi 13373cd49ecSMiklos Szeredi err = mnt_alloc_id(mnt); 13488b38782SLi Zefan if (err) 13588b38782SLi Zefan goto out_free_cache; 13688b38782SLi Zefan 13788b38782SLi Zefan if (name) { 13888b38782SLi Zefan mnt->mnt_devname = kstrdup(name, GFP_KERNEL); 13988b38782SLi Zefan if (!mnt->mnt_devname) 14088b38782SLi Zefan goto out_free_id; 14173cd49ecSMiklos Szeredi } 14273cd49ecSMiklos Szeredi 1431da177e4SLinus Torvalds atomic_set(&mnt->mnt_count, 1); 1441da177e4SLinus Torvalds INIT_LIST_HEAD(&mnt->mnt_hash); 1451da177e4SLinus Torvalds INIT_LIST_HEAD(&mnt->mnt_child); 1461da177e4SLinus Torvalds INIT_LIST_HEAD(&mnt->mnt_mounts); 1471da177e4SLinus Torvalds INIT_LIST_HEAD(&mnt->mnt_list); 14855e700b9SMiklos Szeredi INIT_LIST_HEAD(&mnt->mnt_expire); 14903e06e68SRam Pai INIT_LIST_HEAD(&mnt->mnt_share); 150a58b0eb8SRam Pai INIT_LIST_HEAD(&mnt->mnt_slave_list); 151a58b0eb8SRam Pai INIT_LIST_HEAD(&mnt->mnt_slave); 152d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 153d3ef3d73Snpiggin@suse.de mnt->mnt_writers = alloc_percpu(int); 154d3ef3d73Snpiggin@suse.de if (!mnt->mnt_writers) 155d3ef3d73Snpiggin@suse.de goto out_free_devname; 156d3ef3d73Snpiggin@suse.de #else 157d3ef3d73Snpiggin@suse.de mnt->mnt_writers = 0; 158d3ef3d73Snpiggin@suse.de #endif 1591da177e4SLinus Torvalds } 1601da177e4SLinus Torvalds return mnt; 16188b38782SLi Zefan 162d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 163d3ef3d73Snpiggin@suse.de out_free_devname: 164d3ef3d73Snpiggin@suse.de kfree(mnt->mnt_devname); 165d3ef3d73Snpiggin@suse.de #endif 16688b38782SLi Zefan out_free_id: 16788b38782SLi Zefan mnt_free_id(mnt); 16888b38782SLi Zefan out_free_cache: 16988b38782SLi Zefan kmem_cache_free(mnt_cache, mnt); 17088b38782SLi Zefan return NULL; 1711da177e4SLinus Torvalds } 1721da177e4SLinus Torvalds 1738366025eSDave Hansen /* 1748366025eSDave Hansen * Most r/o checks on a fs are for operations that take 1758366025eSDave Hansen * discrete amounts of time, like a write() or unlink(). 1768366025eSDave Hansen * We must keep track of when those operations start 1778366025eSDave Hansen * (for permission checks) and when they end, so that 1788366025eSDave Hansen * we can determine when writes are able to occur to 1798366025eSDave Hansen * a filesystem. 1808366025eSDave Hansen */ 1813d733633SDave Hansen /* 1823d733633SDave Hansen * __mnt_is_readonly: check whether a mount is read-only 1833d733633SDave Hansen * @mnt: the mount to check for its write status 1843d733633SDave Hansen * 1853d733633SDave Hansen * This shouldn't be used directly ouside of the VFS. 1863d733633SDave Hansen * It does not guarantee that the filesystem will stay 1873d733633SDave Hansen * r/w, just that it is right *now*. This can not and 1883d733633SDave Hansen * should not be used in place of IS_RDONLY(inode). 1893d733633SDave Hansen * mnt_want/drop_write() will _keep_ the filesystem 1903d733633SDave Hansen * r/w. 1913d733633SDave Hansen */ 1923d733633SDave Hansen int __mnt_is_readonly(struct vfsmount *mnt) 1933d733633SDave Hansen { 1942e4b7fcdSDave Hansen if (mnt->mnt_flags & MNT_READONLY) 1952e4b7fcdSDave Hansen return 1; 1962e4b7fcdSDave Hansen if (mnt->mnt_sb->s_flags & MS_RDONLY) 1972e4b7fcdSDave Hansen return 1; 1982e4b7fcdSDave Hansen return 0; 1993d733633SDave Hansen } 2003d733633SDave Hansen EXPORT_SYMBOL_GPL(__mnt_is_readonly); 2013d733633SDave Hansen 202d3ef3d73Snpiggin@suse.de static inline void inc_mnt_writers(struct vfsmount *mnt) 2033d733633SDave Hansen { 204d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 205d3ef3d73Snpiggin@suse.de (*per_cpu_ptr(mnt->mnt_writers, smp_processor_id()))++; 206d3ef3d73Snpiggin@suse.de #else 207d3ef3d73Snpiggin@suse.de mnt->mnt_writers++; 208d3ef3d73Snpiggin@suse.de #endif 2093d733633SDave Hansen } 2103d733633SDave Hansen 211d3ef3d73Snpiggin@suse.de static inline void dec_mnt_writers(struct vfsmount *mnt) 2123d733633SDave Hansen { 213d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 214d3ef3d73Snpiggin@suse.de (*per_cpu_ptr(mnt->mnt_writers, smp_processor_id()))--; 215d3ef3d73Snpiggin@suse.de #else 216d3ef3d73Snpiggin@suse.de mnt->mnt_writers--; 217d3ef3d73Snpiggin@suse.de #endif 218d3ef3d73Snpiggin@suse.de } 219d3ef3d73Snpiggin@suse.de 220d3ef3d73Snpiggin@suse.de static unsigned int count_mnt_writers(struct vfsmount *mnt) 221d3ef3d73Snpiggin@suse.de { 222d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 223d3ef3d73Snpiggin@suse.de unsigned int count = 0; 2243d733633SDave Hansen int cpu; 2253d733633SDave Hansen 2263d733633SDave Hansen for_each_possible_cpu(cpu) { 227d3ef3d73Snpiggin@suse.de count += *per_cpu_ptr(mnt->mnt_writers, cpu); 2283d733633SDave Hansen } 2293d733633SDave Hansen 230d3ef3d73Snpiggin@suse.de return count; 231d3ef3d73Snpiggin@suse.de #else 232d3ef3d73Snpiggin@suse.de return mnt->mnt_writers; 233d3ef3d73Snpiggin@suse.de #endif 2343d733633SDave Hansen } 2353d733633SDave Hansen 2363d733633SDave Hansen /* 2373d733633SDave Hansen * Most r/o checks on a fs are for operations that take 2383d733633SDave Hansen * discrete amounts of time, like a write() or unlink(). 2393d733633SDave Hansen * We must keep track of when those operations start 2403d733633SDave Hansen * (for permission checks) and when they end, so that 2413d733633SDave Hansen * we can determine when writes are able to occur to 2423d733633SDave Hansen * a filesystem. 2433d733633SDave Hansen */ 2448366025eSDave Hansen /** 2458366025eSDave Hansen * mnt_want_write - get write access to a mount 2468366025eSDave Hansen * @mnt: the mount on which to take a write 2478366025eSDave Hansen * 2488366025eSDave Hansen * This tells the low-level filesystem that a write is 2498366025eSDave Hansen * about to be performed to it, and makes sure that 2508366025eSDave Hansen * writes are allowed before returning success. When 2518366025eSDave Hansen * the write operation is finished, mnt_drop_write() 2528366025eSDave Hansen * must be called. This is effectively a refcount. 2538366025eSDave Hansen */ 2548366025eSDave Hansen int mnt_want_write(struct vfsmount *mnt) 2558366025eSDave Hansen { 2563d733633SDave Hansen int ret = 0; 2573d733633SDave Hansen 258d3ef3d73Snpiggin@suse.de preempt_disable(); 259d3ef3d73Snpiggin@suse.de inc_mnt_writers(mnt); 260d3ef3d73Snpiggin@suse.de /* 261d3ef3d73Snpiggin@suse.de * The store to inc_mnt_writers must be visible before we pass 262d3ef3d73Snpiggin@suse.de * MNT_WRITE_HOLD loop below, so that the slowpath can see our 263d3ef3d73Snpiggin@suse.de * incremented count after it has set MNT_WRITE_HOLD. 264d3ef3d73Snpiggin@suse.de */ 265d3ef3d73Snpiggin@suse.de smp_mb(); 266d3ef3d73Snpiggin@suse.de while (mnt->mnt_flags & MNT_WRITE_HOLD) 267d3ef3d73Snpiggin@suse.de cpu_relax(); 268d3ef3d73Snpiggin@suse.de /* 269d3ef3d73Snpiggin@suse.de * After the slowpath clears MNT_WRITE_HOLD, mnt_is_readonly will 270d3ef3d73Snpiggin@suse.de * be set to match its requirements. So we must not load that until 271d3ef3d73Snpiggin@suse.de * MNT_WRITE_HOLD is cleared. 272d3ef3d73Snpiggin@suse.de */ 273d3ef3d73Snpiggin@suse.de smp_rmb(); 2743d733633SDave Hansen if (__mnt_is_readonly(mnt)) { 275d3ef3d73Snpiggin@suse.de dec_mnt_writers(mnt); 2763d733633SDave Hansen ret = -EROFS; 2773d733633SDave Hansen goto out; 2783d733633SDave Hansen } 2793d733633SDave Hansen out: 280d3ef3d73Snpiggin@suse.de preempt_enable(); 2813d733633SDave Hansen return ret; 2828366025eSDave Hansen } 2838366025eSDave Hansen EXPORT_SYMBOL_GPL(mnt_want_write); 2848366025eSDave Hansen 2858366025eSDave Hansen /** 28696029c4eSnpiggin@suse.de * mnt_clone_write - get write access to a mount 28796029c4eSnpiggin@suse.de * @mnt: the mount on which to take a write 28896029c4eSnpiggin@suse.de * 28996029c4eSnpiggin@suse.de * This is effectively like mnt_want_write, except 29096029c4eSnpiggin@suse.de * it must only be used to take an extra write reference 29196029c4eSnpiggin@suse.de * on a mountpoint that we already know has a write reference 29296029c4eSnpiggin@suse.de * on it. This allows some optimisation. 29396029c4eSnpiggin@suse.de * 29496029c4eSnpiggin@suse.de * After finished, mnt_drop_write must be called as usual to 29596029c4eSnpiggin@suse.de * drop the reference. 29696029c4eSnpiggin@suse.de */ 29796029c4eSnpiggin@suse.de int mnt_clone_write(struct vfsmount *mnt) 29896029c4eSnpiggin@suse.de { 29996029c4eSnpiggin@suse.de /* superblock may be r/o */ 30096029c4eSnpiggin@suse.de if (__mnt_is_readonly(mnt)) 30196029c4eSnpiggin@suse.de return -EROFS; 30296029c4eSnpiggin@suse.de preempt_disable(); 30396029c4eSnpiggin@suse.de inc_mnt_writers(mnt); 30496029c4eSnpiggin@suse.de preempt_enable(); 30596029c4eSnpiggin@suse.de return 0; 30696029c4eSnpiggin@suse.de } 30796029c4eSnpiggin@suse.de EXPORT_SYMBOL_GPL(mnt_clone_write); 30896029c4eSnpiggin@suse.de 30996029c4eSnpiggin@suse.de /** 31096029c4eSnpiggin@suse.de * mnt_want_write_file - get write access to a file's mount 31196029c4eSnpiggin@suse.de * @file: the file who's mount on which to take a write 31296029c4eSnpiggin@suse.de * 31396029c4eSnpiggin@suse.de * This is like mnt_want_write, but it takes a file and can 31496029c4eSnpiggin@suse.de * do some optimisations if the file is open for write already 31596029c4eSnpiggin@suse.de */ 31696029c4eSnpiggin@suse.de int mnt_want_write_file(struct file *file) 31796029c4eSnpiggin@suse.de { 31896029c4eSnpiggin@suse.de if (!(file->f_mode & FMODE_WRITE)) 31996029c4eSnpiggin@suse.de return mnt_want_write(file->f_path.mnt); 32096029c4eSnpiggin@suse.de else 32196029c4eSnpiggin@suse.de return mnt_clone_write(file->f_path.mnt); 32296029c4eSnpiggin@suse.de } 32396029c4eSnpiggin@suse.de EXPORT_SYMBOL_GPL(mnt_want_write_file); 32496029c4eSnpiggin@suse.de 32596029c4eSnpiggin@suse.de /** 3268366025eSDave Hansen * mnt_drop_write - give up write access to a mount 3278366025eSDave Hansen * @mnt: the mount on which to give up write access 3288366025eSDave Hansen * 3298366025eSDave Hansen * Tells the low-level filesystem that we are done 3308366025eSDave Hansen * performing writes to it. Must be matched with 3318366025eSDave Hansen * mnt_want_write() call above. 3328366025eSDave Hansen */ 3338366025eSDave Hansen void mnt_drop_write(struct vfsmount *mnt) 3348366025eSDave Hansen { 335d3ef3d73Snpiggin@suse.de preempt_disable(); 336d3ef3d73Snpiggin@suse.de dec_mnt_writers(mnt); 337d3ef3d73Snpiggin@suse.de preempt_enable(); 3388366025eSDave Hansen } 3398366025eSDave Hansen EXPORT_SYMBOL_GPL(mnt_drop_write); 3408366025eSDave Hansen 3412e4b7fcdSDave Hansen static int mnt_make_readonly(struct vfsmount *mnt) 3428366025eSDave Hansen { 3433d733633SDave Hansen int ret = 0; 3443d733633SDave Hansen 3452e4b7fcdSDave Hansen spin_lock(&vfsmount_lock); 346d3ef3d73Snpiggin@suse.de mnt->mnt_flags |= MNT_WRITE_HOLD; 347d3ef3d73Snpiggin@suse.de /* 348d3ef3d73Snpiggin@suse.de * After storing MNT_WRITE_HOLD, we'll read the counters. This store 349d3ef3d73Snpiggin@suse.de * should be visible before we do. 350d3ef3d73Snpiggin@suse.de */ 351d3ef3d73Snpiggin@suse.de smp_mb(); 352d3ef3d73Snpiggin@suse.de 353d3ef3d73Snpiggin@suse.de /* 354d3ef3d73Snpiggin@suse.de * With writers on hold, if this value is zero, then there are 355d3ef3d73Snpiggin@suse.de * definitely no active writers (although held writers may subsequently 356d3ef3d73Snpiggin@suse.de * increment the count, they'll have to wait, and decrement it after 357d3ef3d73Snpiggin@suse.de * seeing MNT_READONLY). 358d3ef3d73Snpiggin@suse.de * 359d3ef3d73Snpiggin@suse.de * It is OK to have counter incremented on one CPU and decremented on 360d3ef3d73Snpiggin@suse.de * another: the sum will add up correctly. The danger would be when we 361d3ef3d73Snpiggin@suse.de * sum up each counter, if we read a counter before it is incremented, 362d3ef3d73Snpiggin@suse.de * but then read another CPU's count which it has been subsequently 363d3ef3d73Snpiggin@suse.de * decremented from -- we would see more decrements than we should. 364d3ef3d73Snpiggin@suse.de * MNT_WRITE_HOLD protects against this scenario, because 365d3ef3d73Snpiggin@suse.de * mnt_want_write first increments count, then smp_mb, then spins on 366d3ef3d73Snpiggin@suse.de * MNT_WRITE_HOLD, so it can't be decremented by another CPU while 367d3ef3d73Snpiggin@suse.de * we're counting up here. 368d3ef3d73Snpiggin@suse.de */ 369d3ef3d73Snpiggin@suse.de if (count_mnt_writers(mnt) > 0) 370d3ef3d73Snpiggin@suse.de ret = -EBUSY; 371d3ef3d73Snpiggin@suse.de else 3722e4b7fcdSDave Hansen mnt->mnt_flags |= MNT_READONLY; 373d3ef3d73Snpiggin@suse.de /* 374d3ef3d73Snpiggin@suse.de * MNT_READONLY must become visible before ~MNT_WRITE_HOLD, so writers 375d3ef3d73Snpiggin@suse.de * that become unheld will see MNT_READONLY. 376d3ef3d73Snpiggin@suse.de */ 377d3ef3d73Snpiggin@suse.de smp_wmb(); 378d3ef3d73Snpiggin@suse.de mnt->mnt_flags &= ~MNT_WRITE_HOLD; 3792e4b7fcdSDave Hansen spin_unlock(&vfsmount_lock); 3803d733633SDave Hansen return ret; 3813d733633SDave Hansen } 3828366025eSDave Hansen 3832e4b7fcdSDave Hansen static void __mnt_unmake_readonly(struct vfsmount *mnt) 3842e4b7fcdSDave Hansen { 3852e4b7fcdSDave Hansen spin_lock(&vfsmount_lock); 3862e4b7fcdSDave Hansen mnt->mnt_flags &= ~MNT_READONLY; 3872e4b7fcdSDave Hansen spin_unlock(&vfsmount_lock); 3882e4b7fcdSDave Hansen } 3892e4b7fcdSDave Hansen 390a3ec947cSSukadev Bhattiprolu void simple_set_mnt(struct vfsmount *mnt, struct super_block *sb) 391454e2398SDavid Howells { 392454e2398SDavid Howells mnt->mnt_sb = sb; 393454e2398SDavid Howells mnt->mnt_root = dget(sb->s_root); 394454e2398SDavid Howells } 395454e2398SDavid Howells 396454e2398SDavid Howells EXPORT_SYMBOL(simple_set_mnt); 397454e2398SDavid Howells 3981da177e4SLinus Torvalds void free_vfsmnt(struct vfsmount *mnt) 3991da177e4SLinus Torvalds { 4001da177e4SLinus Torvalds kfree(mnt->mnt_devname); 40173cd49ecSMiklos Szeredi mnt_free_id(mnt); 402d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 403d3ef3d73Snpiggin@suse.de free_percpu(mnt->mnt_writers); 404d3ef3d73Snpiggin@suse.de #endif 4051da177e4SLinus Torvalds kmem_cache_free(mnt_cache, mnt); 4061da177e4SLinus Torvalds } 4071da177e4SLinus Torvalds 4081da177e4SLinus Torvalds /* 409a05964f3SRam Pai * find the first or last mount at @dentry on vfsmount @mnt depending on 410a05964f3SRam Pai * @dir. If @dir is set return the first mount else return the last mount. 4111da177e4SLinus Torvalds */ 412a05964f3SRam Pai struct vfsmount *__lookup_mnt(struct vfsmount *mnt, struct dentry *dentry, 413a05964f3SRam Pai int dir) 4141da177e4SLinus Torvalds { 4151da177e4SLinus Torvalds struct list_head *head = mount_hashtable + hash(mnt, dentry); 4161da177e4SLinus Torvalds struct list_head *tmp = head; 4171da177e4SLinus Torvalds struct vfsmount *p, *found = NULL; 4181da177e4SLinus Torvalds 4191da177e4SLinus Torvalds for (;;) { 420a05964f3SRam Pai tmp = dir ? tmp->next : tmp->prev; 4211da177e4SLinus Torvalds p = NULL; 4221da177e4SLinus Torvalds if (tmp == head) 4231da177e4SLinus Torvalds break; 4241da177e4SLinus Torvalds p = list_entry(tmp, struct vfsmount, mnt_hash); 4251da177e4SLinus Torvalds if (p->mnt_parent == mnt && p->mnt_mountpoint == dentry) { 426a05964f3SRam Pai found = p; 4271da177e4SLinus Torvalds break; 4281da177e4SLinus Torvalds } 4291da177e4SLinus Torvalds } 4301da177e4SLinus Torvalds return found; 4311da177e4SLinus Torvalds } 4321da177e4SLinus Torvalds 433a05964f3SRam Pai /* 434a05964f3SRam Pai * lookup_mnt increments the ref count before returning 435a05964f3SRam Pai * the vfsmount struct. 436a05964f3SRam Pai */ 4371c755af4SAl Viro struct vfsmount *lookup_mnt(struct path *path) 438a05964f3SRam Pai { 439a05964f3SRam Pai struct vfsmount *child_mnt; 440a05964f3SRam Pai spin_lock(&vfsmount_lock); 4411c755af4SAl Viro if ((child_mnt = __lookup_mnt(path->mnt, path->dentry, 1))) 442a05964f3SRam Pai mntget(child_mnt); 443a05964f3SRam Pai spin_unlock(&vfsmount_lock); 444a05964f3SRam Pai return child_mnt; 445a05964f3SRam Pai } 446a05964f3SRam Pai 4471da177e4SLinus Torvalds static inline int check_mnt(struct vfsmount *mnt) 4481da177e4SLinus Torvalds { 4496b3286edSKirill Korotaev return mnt->mnt_ns == current->nsproxy->mnt_ns; 4501da177e4SLinus Torvalds } 4511da177e4SLinus Torvalds 4526b3286edSKirill Korotaev static void touch_mnt_namespace(struct mnt_namespace *ns) 4535addc5ddSAl Viro { 4545addc5ddSAl Viro if (ns) { 4555addc5ddSAl Viro ns->event = ++event; 4565addc5ddSAl Viro wake_up_interruptible(&ns->poll); 4575addc5ddSAl Viro } 4585addc5ddSAl Viro } 4595addc5ddSAl Viro 4606b3286edSKirill Korotaev static void __touch_mnt_namespace(struct mnt_namespace *ns) 4615addc5ddSAl Viro { 4625addc5ddSAl Viro if (ns && ns->event != event) { 4635addc5ddSAl Viro ns->event = event; 4645addc5ddSAl Viro wake_up_interruptible(&ns->poll); 4655addc5ddSAl Viro } 4665addc5ddSAl Viro } 4675addc5ddSAl Viro 4681a390689SAl Viro static void detach_mnt(struct vfsmount *mnt, struct path *old_path) 4691da177e4SLinus Torvalds { 4701a390689SAl Viro old_path->dentry = mnt->mnt_mountpoint; 4711a390689SAl Viro old_path->mnt = mnt->mnt_parent; 4721da177e4SLinus Torvalds mnt->mnt_parent = mnt; 4731da177e4SLinus Torvalds mnt->mnt_mountpoint = mnt->mnt_root; 4741da177e4SLinus Torvalds list_del_init(&mnt->mnt_child); 4751da177e4SLinus Torvalds list_del_init(&mnt->mnt_hash); 4761a390689SAl Viro old_path->dentry->d_mounted--; 4771da177e4SLinus Torvalds } 4781da177e4SLinus Torvalds 479b90fa9aeSRam Pai void mnt_set_mountpoint(struct vfsmount *mnt, struct dentry *dentry, 480b90fa9aeSRam Pai struct vfsmount *child_mnt) 481b90fa9aeSRam Pai { 482b90fa9aeSRam Pai child_mnt->mnt_parent = mntget(mnt); 483b90fa9aeSRam Pai child_mnt->mnt_mountpoint = dget(dentry); 484b90fa9aeSRam Pai dentry->d_mounted++; 485b90fa9aeSRam Pai } 486b90fa9aeSRam Pai 4871a390689SAl Viro static void attach_mnt(struct vfsmount *mnt, struct path *path) 4881da177e4SLinus Torvalds { 4891a390689SAl Viro mnt_set_mountpoint(path->mnt, path->dentry, mnt); 490b90fa9aeSRam Pai list_add_tail(&mnt->mnt_hash, mount_hashtable + 4911a390689SAl Viro hash(path->mnt, path->dentry)); 4921a390689SAl Viro list_add_tail(&mnt->mnt_child, &path->mnt->mnt_mounts); 493b90fa9aeSRam Pai } 494b90fa9aeSRam Pai 495b90fa9aeSRam Pai /* 496b90fa9aeSRam Pai * the caller must hold vfsmount_lock 497b90fa9aeSRam Pai */ 498b90fa9aeSRam Pai static void commit_tree(struct vfsmount *mnt) 499b90fa9aeSRam Pai { 500b90fa9aeSRam Pai struct vfsmount *parent = mnt->mnt_parent; 501b90fa9aeSRam Pai struct vfsmount *m; 502b90fa9aeSRam Pai LIST_HEAD(head); 5036b3286edSKirill Korotaev struct mnt_namespace *n = parent->mnt_ns; 504b90fa9aeSRam Pai 505b90fa9aeSRam Pai BUG_ON(parent == mnt); 506b90fa9aeSRam Pai 507b90fa9aeSRam Pai list_add_tail(&head, &mnt->mnt_list); 508b90fa9aeSRam Pai list_for_each_entry(m, &head, mnt_list) 5096b3286edSKirill Korotaev m->mnt_ns = n; 510b90fa9aeSRam Pai list_splice(&head, n->list.prev); 511b90fa9aeSRam Pai 512b90fa9aeSRam Pai list_add_tail(&mnt->mnt_hash, mount_hashtable + 513b90fa9aeSRam Pai hash(parent, mnt->mnt_mountpoint)); 514b90fa9aeSRam Pai list_add_tail(&mnt->mnt_child, &parent->mnt_mounts); 5156b3286edSKirill Korotaev touch_mnt_namespace(n); 5161da177e4SLinus Torvalds } 5171da177e4SLinus Torvalds 5181da177e4SLinus Torvalds static struct vfsmount *next_mnt(struct vfsmount *p, struct vfsmount *root) 5191da177e4SLinus Torvalds { 5201da177e4SLinus Torvalds struct list_head *next = p->mnt_mounts.next; 5211da177e4SLinus Torvalds if (next == &p->mnt_mounts) { 5221da177e4SLinus Torvalds while (1) { 5231da177e4SLinus Torvalds if (p == root) 5241da177e4SLinus Torvalds return NULL; 5251da177e4SLinus Torvalds next = p->mnt_child.next; 5261da177e4SLinus Torvalds if (next != &p->mnt_parent->mnt_mounts) 5271da177e4SLinus Torvalds break; 5281da177e4SLinus Torvalds p = p->mnt_parent; 5291da177e4SLinus Torvalds } 5301da177e4SLinus Torvalds } 5311da177e4SLinus Torvalds return list_entry(next, struct vfsmount, mnt_child); 5321da177e4SLinus Torvalds } 5331da177e4SLinus Torvalds 5349676f0c6SRam Pai static struct vfsmount *skip_mnt_tree(struct vfsmount *p) 5359676f0c6SRam Pai { 5369676f0c6SRam Pai struct list_head *prev = p->mnt_mounts.prev; 5379676f0c6SRam Pai while (prev != &p->mnt_mounts) { 5389676f0c6SRam Pai p = list_entry(prev, struct vfsmount, mnt_child); 5399676f0c6SRam Pai prev = p->mnt_mounts.prev; 5409676f0c6SRam Pai } 5419676f0c6SRam Pai return p; 5429676f0c6SRam Pai } 5439676f0c6SRam Pai 54436341f64SRam Pai static struct vfsmount *clone_mnt(struct vfsmount *old, struct dentry *root, 54536341f64SRam Pai int flag) 5461da177e4SLinus Torvalds { 5471da177e4SLinus Torvalds struct super_block *sb = old->mnt_sb; 5481da177e4SLinus Torvalds struct vfsmount *mnt = alloc_vfsmnt(old->mnt_devname); 5491da177e4SLinus Torvalds 5501da177e4SLinus Torvalds if (mnt) { 551719f5d7fSMiklos Szeredi if (flag & (CL_SLAVE | CL_PRIVATE)) 552719f5d7fSMiklos Szeredi mnt->mnt_group_id = 0; /* not a peer of original */ 553719f5d7fSMiklos Szeredi else 554719f5d7fSMiklos Szeredi mnt->mnt_group_id = old->mnt_group_id; 555719f5d7fSMiklos Szeredi 556719f5d7fSMiklos Szeredi if ((flag & CL_MAKE_SHARED) && !mnt->mnt_group_id) { 557719f5d7fSMiklos Szeredi int err = mnt_alloc_group_id(mnt); 558719f5d7fSMiklos Szeredi if (err) 559719f5d7fSMiklos Szeredi goto out_free; 560719f5d7fSMiklos Szeredi } 561719f5d7fSMiklos Szeredi 5621da177e4SLinus Torvalds mnt->mnt_flags = old->mnt_flags; 5631da177e4SLinus Torvalds atomic_inc(&sb->s_active); 5641da177e4SLinus Torvalds mnt->mnt_sb = sb; 5651da177e4SLinus Torvalds mnt->mnt_root = dget(root); 5661da177e4SLinus Torvalds mnt->mnt_mountpoint = mnt->mnt_root; 5671da177e4SLinus Torvalds mnt->mnt_parent = mnt; 568b90fa9aeSRam Pai 5695afe0022SRam Pai if (flag & CL_SLAVE) { 5705afe0022SRam Pai list_add(&mnt->mnt_slave, &old->mnt_slave_list); 5715afe0022SRam Pai mnt->mnt_master = old; 5725afe0022SRam Pai CLEAR_MNT_SHARED(mnt); 5738aec0809SAl Viro } else if (!(flag & CL_PRIVATE)) { 574b90fa9aeSRam Pai if ((flag & CL_PROPAGATION) || IS_MNT_SHARED(old)) 575b90fa9aeSRam Pai list_add(&mnt->mnt_share, &old->mnt_share); 5765afe0022SRam Pai if (IS_MNT_SLAVE(old)) 5775afe0022SRam Pai list_add(&mnt->mnt_slave, &old->mnt_slave); 5785afe0022SRam Pai mnt->mnt_master = old->mnt_master; 5795afe0022SRam Pai } 580b90fa9aeSRam Pai if (flag & CL_MAKE_SHARED) 581b90fa9aeSRam Pai set_mnt_shared(mnt); 5821da177e4SLinus Torvalds 5831da177e4SLinus Torvalds /* stick the duplicate mount on the same expiry list 5841da177e4SLinus Torvalds * as the original if that was on one */ 58536341f64SRam Pai if (flag & CL_EXPIRE) { 58655e700b9SMiklos Szeredi if (!list_empty(&old->mnt_expire)) 58755e700b9SMiklos Szeredi list_add(&mnt->mnt_expire, &old->mnt_expire); 5881da177e4SLinus Torvalds } 58936341f64SRam Pai } 5901da177e4SLinus Torvalds return mnt; 591719f5d7fSMiklos Szeredi 592719f5d7fSMiklos Szeredi out_free: 593719f5d7fSMiklos Szeredi free_vfsmnt(mnt); 594719f5d7fSMiklos Szeredi return NULL; 5951da177e4SLinus Torvalds } 5961da177e4SLinus Torvalds 5977b7b1aceSAl Viro static inline void __mntput(struct vfsmount *mnt) 5981da177e4SLinus Torvalds { 5991da177e4SLinus Torvalds struct super_block *sb = mnt->mnt_sb; 6003d733633SDave Hansen /* 6013d733633SDave Hansen * This probably indicates that somebody messed 6023d733633SDave Hansen * up a mnt_want/drop_write() pair. If this 6033d733633SDave Hansen * happens, the filesystem was probably unable 6043d733633SDave Hansen * to make r/w->r/o transitions. 6053d733633SDave Hansen */ 606d3ef3d73Snpiggin@suse.de /* 607d3ef3d73Snpiggin@suse.de * atomic_dec_and_lock() used to deal with ->mnt_count decrements 608d3ef3d73Snpiggin@suse.de * provides barriers, so count_mnt_writers() below is safe. AV 609d3ef3d73Snpiggin@suse.de */ 610d3ef3d73Snpiggin@suse.de WARN_ON(count_mnt_writers(mnt)); 6111da177e4SLinus Torvalds dput(mnt->mnt_root); 6121da177e4SLinus Torvalds free_vfsmnt(mnt); 6131da177e4SLinus Torvalds deactivate_super(sb); 6141da177e4SLinus Torvalds } 6151da177e4SLinus Torvalds 6167b7b1aceSAl Viro void mntput_no_expire(struct vfsmount *mnt) 6177b7b1aceSAl Viro { 6187b7b1aceSAl Viro repeat: 6197b7b1aceSAl Viro if (atomic_dec_and_lock(&mnt->mnt_count, &vfsmount_lock)) { 6207b7b1aceSAl Viro if (likely(!mnt->mnt_pinned)) { 6217b7b1aceSAl Viro spin_unlock(&vfsmount_lock); 6227b7b1aceSAl Viro __mntput(mnt); 6237b7b1aceSAl Viro return; 6247b7b1aceSAl Viro } 6257b7b1aceSAl Viro atomic_add(mnt->mnt_pinned + 1, &mnt->mnt_count); 6267b7b1aceSAl Viro mnt->mnt_pinned = 0; 6277b7b1aceSAl Viro spin_unlock(&vfsmount_lock); 6287b7b1aceSAl Viro acct_auto_close_mnt(mnt); 6297b7b1aceSAl Viro security_sb_umount_close(mnt); 6307b7b1aceSAl Viro goto repeat; 6317b7b1aceSAl Viro } 6327b7b1aceSAl Viro } 6337b7b1aceSAl Viro 6347b7b1aceSAl Viro EXPORT_SYMBOL(mntput_no_expire); 6357b7b1aceSAl Viro 6367b7b1aceSAl Viro void mnt_pin(struct vfsmount *mnt) 6377b7b1aceSAl Viro { 6387b7b1aceSAl Viro spin_lock(&vfsmount_lock); 6397b7b1aceSAl Viro mnt->mnt_pinned++; 6407b7b1aceSAl Viro spin_unlock(&vfsmount_lock); 6417b7b1aceSAl Viro } 6427b7b1aceSAl Viro 6437b7b1aceSAl Viro EXPORT_SYMBOL(mnt_pin); 6447b7b1aceSAl Viro 6457b7b1aceSAl Viro void mnt_unpin(struct vfsmount *mnt) 6467b7b1aceSAl Viro { 6477b7b1aceSAl Viro spin_lock(&vfsmount_lock); 6487b7b1aceSAl Viro if (mnt->mnt_pinned) { 6497b7b1aceSAl Viro atomic_inc(&mnt->mnt_count); 6507b7b1aceSAl Viro mnt->mnt_pinned--; 6517b7b1aceSAl Viro } 6527b7b1aceSAl Viro spin_unlock(&vfsmount_lock); 6537b7b1aceSAl Viro } 6547b7b1aceSAl Viro 6557b7b1aceSAl Viro EXPORT_SYMBOL(mnt_unpin); 6561da177e4SLinus Torvalds 657b3b304a2SMiklos Szeredi static inline void mangle(struct seq_file *m, const char *s) 658b3b304a2SMiklos Szeredi { 659b3b304a2SMiklos Szeredi seq_escape(m, s, " \t\n\\"); 660b3b304a2SMiklos Szeredi } 661b3b304a2SMiklos Szeredi 662b3b304a2SMiklos Szeredi /* 663b3b304a2SMiklos Szeredi * Simple .show_options callback for filesystems which don't want to 664b3b304a2SMiklos Szeredi * implement more complex mount option showing. 665b3b304a2SMiklos Szeredi * 666b3b304a2SMiklos Szeredi * See also save_mount_options(). 667b3b304a2SMiklos Szeredi */ 668b3b304a2SMiklos Szeredi int generic_show_options(struct seq_file *m, struct vfsmount *mnt) 669b3b304a2SMiklos Szeredi { 6702a32cebdSAl Viro const char *options; 6712a32cebdSAl Viro 6722a32cebdSAl Viro rcu_read_lock(); 6732a32cebdSAl Viro options = rcu_dereference(mnt->mnt_sb->s_options); 674b3b304a2SMiklos Szeredi 675b3b304a2SMiklos Szeredi if (options != NULL && options[0]) { 676b3b304a2SMiklos Szeredi seq_putc(m, ','); 677b3b304a2SMiklos Szeredi mangle(m, options); 678b3b304a2SMiklos Szeredi } 6792a32cebdSAl Viro rcu_read_unlock(); 680b3b304a2SMiklos Szeredi 681b3b304a2SMiklos Szeredi return 0; 682b3b304a2SMiklos Szeredi } 683b3b304a2SMiklos Szeredi EXPORT_SYMBOL(generic_show_options); 684b3b304a2SMiklos Szeredi 685b3b304a2SMiklos Szeredi /* 686b3b304a2SMiklos Szeredi * If filesystem uses generic_show_options(), this function should be 687b3b304a2SMiklos Szeredi * called from the fill_super() callback. 688b3b304a2SMiklos Szeredi * 689b3b304a2SMiklos Szeredi * The .remount_fs callback usually needs to be handled in a special 690b3b304a2SMiklos Szeredi * way, to make sure, that previous options are not overwritten if the 691b3b304a2SMiklos Szeredi * remount fails. 692b3b304a2SMiklos Szeredi * 693b3b304a2SMiklos Szeredi * Also note, that if the filesystem's .remount_fs function doesn't 694b3b304a2SMiklos Szeredi * reset all options to their default value, but changes only newly 695b3b304a2SMiklos Szeredi * given options, then the displayed options will not reflect reality 696b3b304a2SMiklos Szeredi * any more. 697b3b304a2SMiklos Szeredi */ 698b3b304a2SMiklos Szeredi void save_mount_options(struct super_block *sb, char *options) 699b3b304a2SMiklos Szeredi { 7002a32cebdSAl Viro BUG_ON(sb->s_options); 7012a32cebdSAl Viro rcu_assign_pointer(sb->s_options, kstrdup(options, GFP_KERNEL)); 702b3b304a2SMiklos Szeredi } 703b3b304a2SMiklos Szeredi EXPORT_SYMBOL(save_mount_options); 704b3b304a2SMiklos Szeredi 7052a32cebdSAl Viro void replace_mount_options(struct super_block *sb, char *options) 7062a32cebdSAl Viro { 7072a32cebdSAl Viro char *old = sb->s_options; 7082a32cebdSAl Viro rcu_assign_pointer(sb->s_options, options); 7092a32cebdSAl Viro if (old) { 7102a32cebdSAl Viro synchronize_rcu(); 7112a32cebdSAl Viro kfree(old); 7122a32cebdSAl Viro } 7132a32cebdSAl Viro } 7142a32cebdSAl Viro EXPORT_SYMBOL(replace_mount_options); 7152a32cebdSAl Viro 716a1a2c409SMiklos Szeredi #ifdef CONFIG_PROC_FS 7171da177e4SLinus Torvalds /* iterator */ 7181da177e4SLinus Torvalds static void *m_start(struct seq_file *m, loff_t *pos) 7191da177e4SLinus Torvalds { 720a1a2c409SMiklos Szeredi struct proc_mounts *p = m->private; 7211da177e4SLinus Torvalds 722390c6843SRam Pai down_read(&namespace_sem); 723a1a2c409SMiklos Szeredi return seq_list_start(&p->ns->list, *pos); 7241da177e4SLinus Torvalds } 7251da177e4SLinus Torvalds 7261da177e4SLinus Torvalds static void *m_next(struct seq_file *m, void *v, loff_t *pos) 7271da177e4SLinus Torvalds { 728a1a2c409SMiklos Szeredi struct proc_mounts *p = m->private; 729b0765fb8SPavel Emelianov 730a1a2c409SMiklos Szeredi return seq_list_next(v, &p->ns->list, pos); 7311da177e4SLinus Torvalds } 7321da177e4SLinus Torvalds 7331da177e4SLinus Torvalds static void m_stop(struct seq_file *m, void *v) 7341da177e4SLinus Torvalds { 735390c6843SRam Pai up_read(&namespace_sem); 7361da177e4SLinus Torvalds } 7371da177e4SLinus Torvalds 7382d4d4864SRam Pai struct proc_fs_info { 7391da177e4SLinus Torvalds int flag; 7402d4d4864SRam Pai const char *str; 7412d4d4864SRam Pai }; 7422d4d4864SRam Pai 7432069f457SEric Paris static int show_sb_opts(struct seq_file *m, struct super_block *sb) 7442d4d4864SRam Pai { 7452d4d4864SRam Pai static const struct proc_fs_info fs_info[] = { 7461da177e4SLinus Torvalds { MS_SYNCHRONOUS, ",sync" }, 7471da177e4SLinus Torvalds { MS_DIRSYNC, ",dirsync" }, 7481da177e4SLinus Torvalds { MS_MANDLOCK, ",mand" }, 7491da177e4SLinus Torvalds { 0, NULL } 7501da177e4SLinus Torvalds }; 7512d4d4864SRam Pai const struct proc_fs_info *fs_infop; 7522d4d4864SRam Pai 7532d4d4864SRam Pai for (fs_infop = fs_info; fs_infop->flag; fs_infop++) { 7542d4d4864SRam Pai if (sb->s_flags & fs_infop->flag) 7552d4d4864SRam Pai seq_puts(m, fs_infop->str); 7562d4d4864SRam Pai } 7572069f457SEric Paris 7582069f457SEric Paris return security_sb_show_options(m, sb); 7592d4d4864SRam Pai } 7602d4d4864SRam Pai 7612d4d4864SRam Pai static void show_mnt_opts(struct seq_file *m, struct vfsmount *mnt) 7622d4d4864SRam Pai { 7632d4d4864SRam Pai static const struct proc_fs_info mnt_info[] = { 7641da177e4SLinus Torvalds { MNT_NOSUID, ",nosuid" }, 7651da177e4SLinus Torvalds { MNT_NODEV, ",nodev" }, 7661da177e4SLinus Torvalds { MNT_NOEXEC, ",noexec" }, 767fc33a7bbSChristoph Hellwig { MNT_NOATIME, ",noatime" }, 768fc33a7bbSChristoph Hellwig { MNT_NODIRATIME, ",nodiratime" }, 76947ae32d6SValerie Henson { MNT_RELATIME, ",relatime" }, 770d0adde57SMatthew Garrett { MNT_STRICTATIME, ",strictatime" }, 7711da177e4SLinus Torvalds { 0, NULL } 7721da177e4SLinus Torvalds }; 7732d4d4864SRam Pai const struct proc_fs_info *fs_infop; 7742d4d4864SRam Pai 7752d4d4864SRam Pai for (fs_infop = mnt_info; fs_infop->flag; fs_infop++) { 7762d4d4864SRam Pai if (mnt->mnt_flags & fs_infop->flag) 7772d4d4864SRam Pai seq_puts(m, fs_infop->str); 7782d4d4864SRam Pai } 7792d4d4864SRam Pai } 7802d4d4864SRam Pai 7812d4d4864SRam Pai static void show_type(struct seq_file *m, struct super_block *sb) 7822d4d4864SRam Pai { 7832d4d4864SRam Pai mangle(m, sb->s_type->name); 7842d4d4864SRam Pai if (sb->s_subtype && sb->s_subtype[0]) { 7852d4d4864SRam Pai seq_putc(m, '.'); 7862d4d4864SRam Pai mangle(m, sb->s_subtype); 7872d4d4864SRam Pai } 7882d4d4864SRam Pai } 7892d4d4864SRam Pai 7902d4d4864SRam Pai static int show_vfsmnt(struct seq_file *m, void *v) 7912d4d4864SRam Pai { 7922d4d4864SRam Pai struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); 7932d4d4864SRam Pai int err = 0; 794c32c2f63SJan Blunck struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; 7951da177e4SLinus Torvalds 7961da177e4SLinus Torvalds mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none"); 7971da177e4SLinus Torvalds seq_putc(m, ' '); 798c32c2f63SJan Blunck seq_path(m, &mnt_path, " \t\n\\"); 7991da177e4SLinus Torvalds seq_putc(m, ' '); 8002d4d4864SRam Pai show_type(m, mnt->mnt_sb); 8012e4b7fcdSDave Hansen seq_puts(m, __mnt_is_readonly(mnt) ? " ro" : " rw"); 8022069f457SEric Paris err = show_sb_opts(m, mnt->mnt_sb); 8032069f457SEric Paris if (err) 8042069f457SEric Paris goto out; 8052d4d4864SRam Pai show_mnt_opts(m, mnt); 8061da177e4SLinus Torvalds if (mnt->mnt_sb->s_op->show_options) 8071da177e4SLinus Torvalds err = mnt->mnt_sb->s_op->show_options(m, mnt); 8081da177e4SLinus Torvalds seq_puts(m, " 0 0\n"); 8092069f457SEric Paris out: 8101da177e4SLinus Torvalds return err; 8111da177e4SLinus Torvalds } 8121da177e4SLinus Torvalds 813a1a2c409SMiklos Szeredi const struct seq_operations mounts_op = { 8141da177e4SLinus Torvalds .start = m_start, 8151da177e4SLinus Torvalds .next = m_next, 8161da177e4SLinus Torvalds .stop = m_stop, 8171da177e4SLinus Torvalds .show = show_vfsmnt 8181da177e4SLinus Torvalds }; 8191da177e4SLinus Torvalds 8202d4d4864SRam Pai static int show_mountinfo(struct seq_file *m, void *v) 8212d4d4864SRam Pai { 8222d4d4864SRam Pai struct proc_mounts *p = m->private; 8232d4d4864SRam Pai struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); 8242d4d4864SRam Pai struct super_block *sb = mnt->mnt_sb; 8252d4d4864SRam Pai struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; 8262d4d4864SRam Pai struct path root = p->root; 8272d4d4864SRam Pai int err = 0; 8282d4d4864SRam Pai 8292d4d4864SRam Pai seq_printf(m, "%i %i %u:%u ", mnt->mnt_id, mnt->mnt_parent->mnt_id, 8302d4d4864SRam Pai MAJOR(sb->s_dev), MINOR(sb->s_dev)); 8312d4d4864SRam Pai seq_dentry(m, mnt->mnt_root, " \t\n\\"); 8322d4d4864SRam Pai seq_putc(m, ' '); 8332d4d4864SRam Pai seq_path_root(m, &mnt_path, &root, " \t\n\\"); 8342d4d4864SRam Pai if (root.mnt != p->root.mnt || root.dentry != p->root.dentry) { 8352d4d4864SRam Pai /* 8362d4d4864SRam Pai * Mountpoint is outside root, discard that one. Ugly, 8372d4d4864SRam Pai * but less so than trying to do that in iterator in a 8382d4d4864SRam Pai * race-free way (due to renames). 8392d4d4864SRam Pai */ 8402d4d4864SRam Pai return SEQ_SKIP; 8412d4d4864SRam Pai } 8422d4d4864SRam Pai seq_puts(m, mnt->mnt_flags & MNT_READONLY ? " ro" : " rw"); 8432d4d4864SRam Pai show_mnt_opts(m, mnt); 8442d4d4864SRam Pai 8452d4d4864SRam Pai /* Tagged fields ("foo:X" or "bar") */ 8462d4d4864SRam Pai if (IS_MNT_SHARED(mnt)) 8472d4d4864SRam Pai seq_printf(m, " shared:%i", mnt->mnt_group_id); 84897e7e0f7SMiklos Szeredi if (IS_MNT_SLAVE(mnt)) { 84997e7e0f7SMiklos Szeredi int master = mnt->mnt_master->mnt_group_id; 85097e7e0f7SMiklos Szeredi int dom = get_dominating_id(mnt, &p->root); 85197e7e0f7SMiklos Szeredi seq_printf(m, " master:%i", master); 85297e7e0f7SMiklos Szeredi if (dom && dom != master) 85397e7e0f7SMiklos Szeredi seq_printf(m, " propagate_from:%i", dom); 85497e7e0f7SMiklos Szeredi } 8552d4d4864SRam Pai if (IS_MNT_UNBINDABLE(mnt)) 8562d4d4864SRam Pai seq_puts(m, " unbindable"); 8572d4d4864SRam Pai 8582d4d4864SRam Pai /* Filesystem specific data */ 8592d4d4864SRam Pai seq_puts(m, " - "); 8602d4d4864SRam Pai show_type(m, sb); 8612d4d4864SRam Pai seq_putc(m, ' '); 8622d4d4864SRam Pai mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none"); 8632d4d4864SRam Pai seq_puts(m, sb->s_flags & MS_RDONLY ? " ro" : " rw"); 8642069f457SEric Paris err = show_sb_opts(m, sb); 8652069f457SEric Paris if (err) 8662069f457SEric Paris goto out; 8672d4d4864SRam Pai if (sb->s_op->show_options) 8682d4d4864SRam Pai err = sb->s_op->show_options(m, mnt); 8692d4d4864SRam Pai seq_putc(m, '\n'); 8702069f457SEric Paris out: 8712d4d4864SRam Pai return err; 8722d4d4864SRam Pai } 8732d4d4864SRam Pai 8742d4d4864SRam Pai const struct seq_operations mountinfo_op = { 8752d4d4864SRam Pai .start = m_start, 8762d4d4864SRam Pai .next = m_next, 8772d4d4864SRam Pai .stop = m_stop, 8782d4d4864SRam Pai .show = show_mountinfo, 8792d4d4864SRam Pai }; 8802d4d4864SRam Pai 881b4629fe2SChuck Lever static int show_vfsstat(struct seq_file *m, void *v) 882b4629fe2SChuck Lever { 883b0765fb8SPavel Emelianov struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); 884c32c2f63SJan Blunck struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; 885b4629fe2SChuck Lever int err = 0; 886b4629fe2SChuck Lever 887b4629fe2SChuck Lever /* device */ 888b4629fe2SChuck Lever if (mnt->mnt_devname) { 889b4629fe2SChuck Lever seq_puts(m, "device "); 890b4629fe2SChuck Lever mangle(m, mnt->mnt_devname); 891b4629fe2SChuck Lever } else 892b4629fe2SChuck Lever seq_puts(m, "no device"); 893b4629fe2SChuck Lever 894b4629fe2SChuck Lever /* mount point */ 895b4629fe2SChuck Lever seq_puts(m, " mounted on "); 896c32c2f63SJan Blunck seq_path(m, &mnt_path, " \t\n\\"); 897b4629fe2SChuck Lever seq_putc(m, ' '); 898b4629fe2SChuck Lever 899b4629fe2SChuck Lever /* file system type */ 900b4629fe2SChuck Lever seq_puts(m, "with fstype "); 9012d4d4864SRam Pai show_type(m, mnt->mnt_sb); 902b4629fe2SChuck Lever 903b4629fe2SChuck Lever /* optional statistics */ 904b4629fe2SChuck Lever if (mnt->mnt_sb->s_op->show_stats) { 905b4629fe2SChuck Lever seq_putc(m, ' '); 906b4629fe2SChuck Lever err = mnt->mnt_sb->s_op->show_stats(m, mnt); 907b4629fe2SChuck Lever } 908b4629fe2SChuck Lever 909b4629fe2SChuck Lever seq_putc(m, '\n'); 910b4629fe2SChuck Lever return err; 911b4629fe2SChuck Lever } 912b4629fe2SChuck Lever 913a1a2c409SMiklos Szeredi const struct seq_operations mountstats_op = { 914b4629fe2SChuck Lever .start = m_start, 915b4629fe2SChuck Lever .next = m_next, 916b4629fe2SChuck Lever .stop = m_stop, 917b4629fe2SChuck Lever .show = show_vfsstat, 918b4629fe2SChuck Lever }; 919a1a2c409SMiklos Szeredi #endif /* CONFIG_PROC_FS */ 920b4629fe2SChuck Lever 9211da177e4SLinus Torvalds /** 9221da177e4SLinus Torvalds * may_umount_tree - check if a mount tree is busy 9231da177e4SLinus Torvalds * @mnt: root of mount tree 9241da177e4SLinus Torvalds * 9251da177e4SLinus Torvalds * This is called to check if a tree of mounts has any 9261da177e4SLinus Torvalds * open files, pwds, chroots or sub mounts that are 9271da177e4SLinus Torvalds * busy. 9281da177e4SLinus Torvalds */ 9291da177e4SLinus Torvalds int may_umount_tree(struct vfsmount *mnt) 9301da177e4SLinus Torvalds { 93136341f64SRam Pai int actual_refs = 0; 93236341f64SRam Pai int minimum_refs = 0; 93336341f64SRam Pai struct vfsmount *p; 9341da177e4SLinus Torvalds 9351da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 93636341f64SRam Pai for (p = mnt; p; p = next_mnt(p, mnt)) { 9371da177e4SLinus Torvalds actual_refs += atomic_read(&p->mnt_count); 9381da177e4SLinus Torvalds minimum_refs += 2; 9391da177e4SLinus Torvalds } 9401da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 9411da177e4SLinus Torvalds 9421da177e4SLinus Torvalds if (actual_refs > minimum_refs) 9431da177e4SLinus Torvalds return 0; 944e3474a8eSIan Kent 945e3474a8eSIan Kent return 1; 9461da177e4SLinus Torvalds } 9471da177e4SLinus Torvalds 9481da177e4SLinus Torvalds EXPORT_SYMBOL(may_umount_tree); 9491da177e4SLinus Torvalds 9501da177e4SLinus Torvalds /** 9511da177e4SLinus Torvalds * may_umount - check if a mount point is busy 9521da177e4SLinus Torvalds * @mnt: root of mount 9531da177e4SLinus Torvalds * 9541da177e4SLinus Torvalds * This is called to check if a mount point has any 9551da177e4SLinus Torvalds * open files, pwds, chroots or sub mounts. If the 9561da177e4SLinus Torvalds * mount has sub mounts this will return busy 9571da177e4SLinus Torvalds * regardless of whether the sub mounts are busy. 9581da177e4SLinus Torvalds * 9591da177e4SLinus Torvalds * Doesn't take quota and stuff into account. IOW, in some cases it will 9601da177e4SLinus Torvalds * give false negatives. The main reason why it's here is that we need 9611da177e4SLinus Torvalds * a non-destructive way to look for easily umountable filesystems. 9621da177e4SLinus Torvalds */ 9631da177e4SLinus Torvalds int may_umount(struct vfsmount *mnt) 9641da177e4SLinus Torvalds { 965e3474a8eSIan Kent int ret = 1; 966a05964f3SRam Pai spin_lock(&vfsmount_lock); 967a05964f3SRam Pai if (propagate_mount_busy(mnt, 2)) 968e3474a8eSIan Kent ret = 0; 969a05964f3SRam Pai spin_unlock(&vfsmount_lock); 970a05964f3SRam Pai return ret; 9711da177e4SLinus Torvalds } 9721da177e4SLinus Torvalds 9731da177e4SLinus Torvalds EXPORT_SYMBOL(may_umount); 9741da177e4SLinus Torvalds 975b90fa9aeSRam Pai void release_mounts(struct list_head *head) 9761da177e4SLinus Torvalds { 97770fbcdf4SRam Pai struct vfsmount *mnt; 97870fbcdf4SRam Pai while (!list_empty(head)) { 979b5e61818SPavel Emelianov mnt = list_first_entry(head, struct vfsmount, mnt_hash); 98070fbcdf4SRam Pai list_del_init(&mnt->mnt_hash); 98170fbcdf4SRam Pai if (mnt->mnt_parent != mnt) { 98270fbcdf4SRam Pai struct dentry *dentry; 98370fbcdf4SRam Pai struct vfsmount *m; 98470fbcdf4SRam Pai spin_lock(&vfsmount_lock); 98570fbcdf4SRam Pai dentry = mnt->mnt_mountpoint; 98670fbcdf4SRam Pai m = mnt->mnt_parent; 98770fbcdf4SRam Pai mnt->mnt_mountpoint = mnt->mnt_root; 98870fbcdf4SRam Pai mnt->mnt_parent = mnt; 9897c4b93d8SAl Viro m->mnt_ghosts--; 9901da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 99170fbcdf4SRam Pai dput(dentry); 99270fbcdf4SRam Pai mntput(m); 9931da177e4SLinus Torvalds } 9941da177e4SLinus Torvalds mntput(mnt); 99570fbcdf4SRam Pai } 99670fbcdf4SRam Pai } 99770fbcdf4SRam Pai 998a05964f3SRam Pai void umount_tree(struct vfsmount *mnt, int propagate, struct list_head *kill) 99970fbcdf4SRam Pai { 100070fbcdf4SRam Pai struct vfsmount *p; 100170fbcdf4SRam Pai 10021bfba4e8SAkinobu Mita for (p = mnt; p; p = next_mnt(p, mnt)) 10031bfba4e8SAkinobu Mita list_move(&p->mnt_hash, kill); 100470fbcdf4SRam Pai 1005a05964f3SRam Pai if (propagate) 1006a05964f3SRam Pai propagate_umount(kill); 1007a05964f3SRam Pai 100870fbcdf4SRam Pai list_for_each_entry(p, kill, mnt_hash) { 100970fbcdf4SRam Pai list_del_init(&p->mnt_expire); 101070fbcdf4SRam Pai list_del_init(&p->mnt_list); 10116b3286edSKirill Korotaev __touch_mnt_namespace(p->mnt_ns); 10126b3286edSKirill Korotaev p->mnt_ns = NULL; 101370fbcdf4SRam Pai list_del_init(&p->mnt_child); 10147c4b93d8SAl Viro if (p->mnt_parent != p) { 10157c4b93d8SAl Viro p->mnt_parent->mnt_ghosts++; 1016f30ac319SAl Viro p->mnt_mountpoint->d_mounted--; 10177c4b93d8SAl Viro } 1018a05964f3SRam Pai change_mnt_propagation(p, MS_PRIVATE); 10191da177e4SLinus Torvalds } 10201da177e4SLinus Torvalds } 10211da177e4SLinus Torvalds 1022c35038beSAl Viro static void shrink_submounts(struct vfsmount *mnt, struct list_head *umounts); 1023c35038beSAl Viro 10241da177e4SLinus Torvalds static int do_umount(struct vfsmount *mnt, int flags) 10251da177e4SLinus Torvalds { 10261da177e4SLinus Torvalds struct super_block *sb = mnt->mnt_sb; 10271da177e4SLinus Torvalds int retval; 102870fbcdf4SRam Pai LIST_HEAD(umount_list); 10291da177e4SLinus Torvalds 10301da177e4SLinus Torvalds retval = security_sb_umount(mnt, flags); 10311da177e4SLinus Torvalds if (retval) 10321da177e4SLinus Torvalds return retval; 10331da177e4SLinus Torvalds 10341da177e4SLinus Torvalds /* 10351da177e4SLinus Torvalds * Allow userspace to request a mountpoint be expired rather than 10361da177e4SLinus Torvalds * unmounting unconditionally. Unmount only happens if: 10371da177e4SLinus Torvalds * (1) the mark is already set (the mark is cleared by mntput()) 10381da177e4SLinus Torvalds * (2) the usage count == 1 [parent vfsmount] + 1 [sys_umount] 10391da177e4SLinus Torvalds */ 10401da177e4SLinus Torvalds if (flags & MNT_EXPIRE) { 10416ac08c39SJan Blunck if (mnt == current->fs->root.mnt || 10421da177e4SLinus Torvalds flags & (MNT_FORCE | MNT_DETACH)) 10431da177e4SLinus Torvalds return -EINVAL; 10441da177e4SLinus Torvalds 10451da177e4SLinus Torvalds if (atomic_read(&mnt->mnt_count) != 2) 10461da177e4SLinus Torvalds return -EBUSY; 10471da177e4SLinus Torvalds 10481da177e4SLinus Torvalds if (!xchg(&mnt->mnt_expiry_mark, 1)) 10491da177e4SLinus Torvalds return -EAGAIN; 10501da177e4SLinus Torvalds } 10511da177e4SLinus Torvalds 10521da177e4SLinus Torvalds /* 10531da177e4SLinus Torvalds * If we may have to abort operations to get out of this 10541da177e4SLinus Torvalds * mount, and they will themselves hold resources we must 10551da177e4SLinus Torvalds * allow the fs to do things. In the Unix tradition of 10561da177e4SLinus Torvalds * 'Gee thats tricky lets do it in userspace' the umount_begin 10571da177e4SLinus Torvalds * might fail to complete on the first run through as other tasks 10581da177e4SLinus Torvalds * must return, and the like. Thats for the mount program to worry 10591da177e4SLinus Torvalds * about for the moment. 10601da177e4SLinus Torvalds */ 10611da177e4SLinus Torvalds 106242faad99SAl Viro if (flags & MNT_FORCE && sb->s_op->umount_begin) { 106342faad99SAl Viro sb->s_op->umount_begin(sb); 106442faad99SAl Viro } 10651da177e4SLinus Torvalds 10661da177e4SLinus Torvalds /* 10671da177e4SLinus Torvalds * No sense to grab the lock for this test, but test itself looks 10681da177e4SLinus Torvalds * somewhat bogus. Suggestions for better replacement? 10691da177e4SLinus Torvalds * Ho-hum... In principle, we might treat that as umount + switch 10701da177e4SLinus Torvalds * to rootfs. GC would eventually take care of the old vfsmount. 10711da177e4SLinus Torvalds * Actually it makes sense, especially if rootfs would contain a 10721da177e4SLinus Torvalds * /reboot - static binary that would close all descriptors and 10731da177e4SLinus Torvalds * call reboot(9). Then init(8) could umount root and exec /reboot. 10741da177e4SLinus Torvalds */ 10756ac08c39SJan Blunck if (mnt == current->fs->root.mnt && !(flags & MNT_DETACH)) { 10761da177e4SLinus Torvalds /* 10771da177e4SLinus Torvalds * Special case for "unmounting" root ... 10781da177e4SLinus Torvalds * we just try to remount it readonly. 10791da177e4SLinus Torvalds */ 10801da177e4SLinus Torvalds down_write(&sb->s_umount); 10814aa98cf7SAl Viro if (!(sb->s_flags & MS_RDONLY)) 10821da177e4SLinus Torvalds retval = do_remount_sb(sb, MS_RDONLY, NULL, 0); 10831da177e4SLinus Torvalds up_write(&sb->s_umount); 10841da177e4SLinus Torvalds return retval; 10851da177e4SLinus Torvalds } 10861da177e4SLinus Torvalds 1087390c6843SRam Pai down_write(&namespace_sem); 10881da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 10895addc5ddSAl Viro event++; 10901da177e4SLinus Torvalds 1091c35038beSAl Viro if (!(flags & MNT_DETACH)) 1092c35038beSAl Viro shrink_submounts(mnt, &umount_list); 1093c35038beSAl Viro 10941da177e4SLinus Torvalds retval = -EBUSY; 1095a05964f3SRam Pai if (flags & MNT_DETACH || !propagate_mount_busy(mnt, 2)) { 10961da177e4SLinus Torvalds if (!list_empty(&mnt->mnt_list)) 1097a05964f3SRam Pai umount_tree(mnt, 1, &umount_list); 10981da177e4SLinus Torvalds retval = 0; 10991da177e4SLinus Torvalds } 11001da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 11011da177e4SLinus Torvalds if (retval) 11021da177e4SLinus Torvalds security_sb_umount_busy(mnt); 1103390c6843SRam Pai up_write(&namespace_sem); 110470fbcdf4SRam Pai release_mounts(&umount_list); 11051da177e4SLinus Torvalds return retval; 11061da177e4SLinus Torvalds } 11071da177e4SLinus Torvalds 11081da177e4SLinus Torvalds /* 11091da177e4SLinus Torvalds * Now umount can handle mount points as well as block devices. 11101da177e4SLinus Torvalds * This is important for filesystems which use unnamed block devices. 11111da177e4SLinus Torvalds * 11121da177e4SLinus Torvalds * We now support a flag for forced unmount like the other 'big iron' 11131da177e4SLinus Torvalds * unixes. Our API is identical to OSF/1 to avoid making a mess of AMD 11141da177e4SLinus Torvalds */ 11151da177e4SLinus Torvalds 1116bdc480e3SHeiko Carstens SYSCALL_DEFINE2(umount, char __user *, name, int, flags) 11171da177e4SLinus Torvalds { 11182d8f3038SAl Viro struct path path; 11191da177e4SLinus Torvalds int retval; 11201da177e4SLinus Torvalds 11212d8f3038SAl Viro retval = user_path(name, &path); 11221da177e4SLinus Torvalds if (retval) 11231da177e4SLinus Torvalds goto out; 11241da177e4SLinus Torvalds retval = -EINVAL; 11252d8f3038SAl Viro if (path.dentry != path.mnt->mnt_root) 11261da177e4SLinus Torvalds goto dput_and_out; 11272d8f3038SAl Viro if (!check_mnt(path.mnt)) 11281da177e4SLinus Torvalds goto dput_and_out; 11291da177e4SLinus Torvalds 11301da177e4SLinus Torvalds retval = -EPERM; 11311da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 11321da177e4SLinus Torvalds goto dput_and_out; 11331da177e4SLinus Torvalds 11342d8f3038SAl Viro retval = do_umount(path.mnt, flags); 11351da177e4SLinus Torvalds dput_and_out: 1136429731b1SJan Blunck /* we mustn't call path_put() as that would clear mnt_expiry_mark */ 11372d8f3038SAl Viro dput(path.dentry); 11382d8f3038SAl Viro mntput_no_expire(path.mnt); 11391da177e4SLinus Torvalds out: 11401da177e4SLinus Torvalds return retval; 11411da177e4SLinus Torvalds } 11421da177e4SLinus Torvalds 11431da177e4SLinus Torvalds #ifdef __ARCH_WANT_SYS_OLDUMOUNT 11441da177e4SLinus Torvalds 11451da177e4SLinus Torvalds /* 11461da177e4SLinus Torvalds * The 2.0 compatible umount. No flags. 11471da177e4SLinus Torvalds */ 1148bdc480e3SHeiko Carstens SYSCALL_DEFINE1(oldumount, char __user *, name) 11491da177e4SLinus Torvalds { 11501da177e4SLinus Torvalds return sys_umount(name, 0); 11511da177e4SLinus Torvalds } 11521da177e4SLinus Torvalds 11531da177e4SLinus Torvalds #endif 11541da177e4SLinus Torvalds 11552d92ab3cSAl Viro static int mount_is_safe(struct path *path) 11561da177e4SLinus Torvalds { 11571da177e4SLinus Torvalds if (capable(CAP_SYS_ADMIN)) 11581da177e4SLinus Torvalds return 0; 11591da177e4SLinus Torvalds return -EPERM; 11601da177e4SLinus Torvalds #ifdef notyet 11612d92ab3cSAl Viro if (S_ISLNK(path->dentry->d_inode->i_mode)) 11621da177e4SLinus Torvalds return -EPERM; 11632d92ab3cSAl Viro if (path->dentry->d_inode->i_mode & S_ISVTX) { 1164da9592edSDavid Howells if (current_uid() != path->dentry->d_inode->i_uid) 11651da177e4SLinus Torvalds return -EPERM; 11661da177e4SLinus Torvalds } 11672d92ab3cSAl Viro if (inode_permission(path->dentry->d_inode, MAY_WRITE)) 11681da177e4SLinus Torvalds return -EPERM; 11691da177e4SLinus Torvalds return 0; 11701da177e4SLinus Torvalds #endif 11711da177e4SLinus Torvalds } 11721da177e4SLinus Torvalds 1173b90fa9aeSRam Pai struct vfsmount *copy_tree(struct vfsmount *mnt, struct dentry *dentry, 117436341f64SRam Pai int flag) 11751da177e4SLinus Torvalds { 11761da177e4SLinus Torvalds struct vfsmount *res, *p, *q, *r, *s; 11771a390689SAl Viro struct path path; 11781da177e4SLinus Torvalds 11799676f0c6SRam Pai if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(mnt)) 11809676f0c6SRam Pai return NULL; 11819676f0c6SRam Pai 118236341f64SRam Pai res = q = clone_mnt(mnt, dentry, flag); 11831da177e4SLinus Torvalds if (!q) 11841da177e4SLinus Torvalds goto Enomem; 11851da177e4SLinus Torvalds q->mnt_mountpoint = mnt->mnt_mountpoint; 11861da177e4SLinus Torvalds 11871da177e4SLinus Torvalds p = mnt; 1188fdadd65fSDomen Puncer list_for_each_entry(r, &mnt->mnt_mounts, mnt_child) { 11897ec02ef1SJan Blunck if (!is_subdir(r->mnt_mountpoint, dentry)) 11901da177e4SLinus Torvalds continue; 11911da177e4SLinus Torvalds 11921da177e4SLinus Torvalds for (s = r; s; s = next_mnt(s, r)) { 11939676f0c6SRam Pai if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(s)) { 11949676f0c6SRam Pai s = skip_mnt_tree(s); 11959676f0c6SRam Pai continue; 11969676f0c6SRam Pai } 11971da177e4SLinus Torvalds while (p != s->mnt_parent) { 11981da177e4SLinus Torvalds p = p->mnt_parent; 11991da177e4SLinus Torvalds q = q->mnt_parent; 12001da177e4SLinus Torvalds } 12011da177e4SLinus Torvalds p = s; 12021a390689SAl Viro path.mnt = q; 12031a390689SAl Viro path.dentry = p->mnt_mountpoint; 120436341f64SRam Pai q = clone_mnt(p, p->mnt_root, flag); 12051da177e4SLinus Torvalds if (!q) 12061da177e4SLinus Torvalds goto Enomem; 12071da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 12081da177e4SLinus Torvalds list_add_tail(&q->mnt_list, &res->mnt_list); 12091a390689SAl Viro attach_mnt(q, &path); 12101da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 12111da177e4SLinus Torvalds } 12121da177e4SLinus Torvalds } 12131da177e4SLinus Torvalds return res; 12141da177e4SLinus Torvalds Enomem: 12151da177e4SLinus Torvalds if (res) { 121670fbcdf4SRam Pai LIST_HEAD(umount_list); 12171da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 1218a05964f3SRam Pai umount_tree(res, 0, &umount_list); 12191da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 122070fbcdf4SRam Pai release_mounts(&umount_list); 12211da177e4SLinus Torvalds } 12221da177e4SLinus Torvalds return NULL; 12231da177e4SLinus Torvalds } 12241da177e4SLinus Torvalds 1225589ff870SAl Viro struct vfsmount *collect_mounts(struct path *path) 12268aec0809SAl Viro { 12278aec0809SAl Viro struct vfsmount *tree; 12281a60a280SAl Viro down_write(&namespace_sem); 1229589ff870SAl Viro tree = copy_tree(path->mnt, path->dentry, CL_COPY_ALL | CL_PRIVATE); 12301a60a280SAl Viro up_write(&namespace_sem); 12318aec0809SAl Viro return tree; 12328aec0809SAl Viro } 12338aec0809SAl Viro 12348aec0809SAl Viro void drop_collected_mounts(struct vfsmount *mnt) 12358aec0809SAl Viro { 12368aec0809SAl Viro LIST_HEAD(umount_list); 12371a60a280SAl Viro down_write(&namespace_sem); 12388aec0809SAl Viro spin_lock(&vfsmount_lock); 12398aec0809SAl Viro umount_tree(mnt, 0, &umount_list); 12408aec0809SAl Viro spin_unlock(&vfsmount_lock); 12411a60a280SAl Viro up_write(&namespace_sem); 12428aec0809SAl Viro release_mounts(&umount_list); 12438aec0809SAl Viro } 12448aec0809SAl Viro 1245719f5d7fSMiklos Szeredi static void cleanup_group_ids(struct vfsmount *mnt, struct vfsmount *end) 1246719f5d7fSMiklos Szeredi { 1247719f5d7fSMiklos Szeredi struct vfsmount *p; 1248719f5d7fSMiklos Szeredi 1249719f5d7fSMiklos Szeredi for (p = mnt; p != end; p = next_mnt(p, mnt)) { 1250719f5d7fSMiklos Szeredi if (p->mnt_group_id && !IS_MNT_SHARED(p)) 1251719f5d7fSMiklos Szeredi mnt_release_group_id(p); 1252719f5d7fSMiklos Szeredi } 1253719f5d7fSMiklos Szeredi } 1254719f5d7fSMiklos Szeredi 1255719f5d7fSMiklos Szeredi static int invent_group_ids(struct vfsmount *mnt, bool recurse) 1256719f5d7fSMiklos Szeredi { 1257719f5d7fSMiklos Szeredi struct vfsmount *p; 1258719f5d7fSMiklos Szeredi 1259719f5d7fSMiklos Szeredi for (p = mnt; p; p = recurse ? next_mnt(p, mnt) : NULL) { 1260719f5d7fSMiklos Szeredi if (!p->mnt_group_id && !IS_MNT_SHARED(p)) { 1261719f5d7fSMiklos Szeredi int err = mnt_alloc_group_id(p); 1262719f5d7fSMiklos Szeredi if (err) { 1263719f5d7fSMiklos Szeredi cleanup_group_ids(mnt, p); 1264719f5d7fSMiklos Szeredi return err; 1265719f5d7fSMiklos Szeredi } 1266719f5d7fSMiklos Szeredi } 1267719f5d7fSMiklos Szeredi } 1268719f5d7fSMiklos Szeredi 1269719f5d7fSMiklos Szeredi return 0; 1270719f5d7fSMiklos Szeredi } 1271719f5d7fSMiklos Szeredi 1272b90fa9aeSRam Pai /* 1273b90fa9aeSRam Pai * @source_mnt : mount tree to be attached 1274b90fa9aeSRam Pai * @nd : place the mount tree @source_mnt is attached 127521444403SRam Pai * @parent_nd : if non-null, detach the source_mnt from its parent and 127621444403SRam Pai * store the parent mount and mountpoint dentry. 127721444403SRam Pai * (done when source_mnt is moved) 1278b90fa9aeSRam Pai * 1279b90fa9aeSRam Pai * NOTE: in the table below explains the semantics when a source mount 1280b90fa9aeSRam Pai * of a given type is attached to a destination mount of a given type. 12819676f0c6SRam Pai * --------------------------------------------------------------------------- 1282b90fa9aeSRam Pai * | BIND MOUNT OPERATION | 12839676f0c6SRam Pai * |************************************************************************** 12849676f0c6SRam Pai * | source-->| shared | private | slave | unbindable | 12859676f0c6SRam Pai * | dest | | | | | 12869676f0c6SRam Pai * | | | | | | | 12879676f0c6SRam Pai * | v | | | | | 12889676f0c6SRam Pai * |************************************************************************** 12899676f0c6SRam Pai * | shared | shared (++) | shared (+) | shared(+++)| invalid | 12905afe0022SRam Pai * | | | | | | 12919676f0c6SRam Pai * |non-shared| shared (+) | private | slave (*) | invalid | 12929676f0c6SRam Pai * *************************************************************************** 1293b90fa9aeSRam Pai * A bind operation clones the source mount and mounts the clone on the 1294b90fa9aeSRam Pai * destination mount. 1295b90fa9aeSRam Pai * 1296b90fa9aeSRam Pai * (++) the cloned mount is propagated to all the mounts in the propagation 1297b90fa9aeSRam Pai * tree of the destination mount and the cloned mount is added to 1298b90fa9aeSRam Pai * the peer group of the source mount. 1299b90fa9aeSRam Pai * (+) the cloned mount is created under the destination mount and is marked 1300b90fa9aeSRam Pai * as shared. The cloned mount is added to the peer group of the source 1301b90fa9aeSRam Pai * mount. 13025afe0022SRam Pai * (+++) the mount is propagated to all the mounts in the propagation tree 13035afe0022SRam Pai * of the destination mount and the cloned mount is made slave 13045afe0022SRam Pai * of the same master as that of the source mount. The cloned mount 13055afe0022SRam Pai * is marked as 'shared and slave'. 13065afe0022SRam Pai * (*) the cloned mount is made a slave of the same master as that of the 13075afe0022SRam Pai * source mount. 13085afe0022SRam Pai * 13099676f0c6SRam Pai * --------------------------------------------------------------------------- 131021444403SRam Pai * | MOVE MOUNT OPERATION | 13119676f0c6SRam Pai * |************************************************************************** 13129676f0c6SRam Pai * | source-->| shared | private | slave | unbindable | 13139676f0c6SRam Pai * | dest | | | | | 13149676f0c6SRam Pai * | | | | | | | 13159676f0c6SRam Pai * | v | | | | | 13169676f0c6SRam Pai * |************************************************************************** 13179676f0c6SRam Pai * | shared | shared (+) | shared (+) | shared(+++) | invalid | 13185afe0022SRam Pai * | | | | | | 13199676f0c6SRam Pai * |non-shared| shared (+*) | private | slave (*) | unbindable | 13209676f0c6SRam Pai * *************************************************************************** 13215afe0022SRam Pai * 13225afe0022SRam Pai * (+) the mount is moved to the destination. And is then propagated to 13235afe0022SRam Pai * all the mounts in the propagation tree of the destination mount. 132421444403SRam Pai * (+*) the mount is moved to the destination. 13255afe0022SRam Pai * (+++) the mount is moved to the destination and is then propagated to 13265afe0022SRam Pai * all the mounts belonging to the destination mount's propagation tree. 13275afe0022SRam Pai * the mount is marked as 'shared and slave'. 13285afe0022SRam Pai * (*) the mount continues to be a slave at the new location. 1329b90fa9aeSRam Pai * 1330b90fa9aeSRam Pai * if the source mount is a tree, the operations explained above is 1331b90fa9aeSRam Pai * applied to each mount in the tree. 1332b90fa9aeSRam Pai * Must be called without spinlocks held, since this function can sleep 1333b90fa9aeSRam Pai * in allocations. 1334b90fa9aeSRam Pai */ 1335b90fa9aeSRam Pai static int attach_recursive_mnt(struct vfsmount *source_mnt, 13361a390689SAl Viro struct path *path, struct path *parent_path) 1337b90fa9aeSRam Pai { 1338b90fa9aeSRam Pai LIST_HEAD(tree_list); 13391a390689SAl Viro struct vfsmount *dest_mnt = path->mnt; 13401a390689SAl Viro struct dentry *dest_dentry = path->dentry; 1341b90fa9aeSRam Pai struct vfsmount *child, *p; 1342719f5d7fSMiklos Szeredi int err; 1343b90fa9aeSRam Pai 1344719f5d7fSMiklos Szeredi if (IS_MNT_SHARED(dest_mnt)) { 1345719f5d7fSMiklos Szeredi err = invent_group_ids(source_mnt, true); 1346719f5d7fSMiklos Szeredi if (err) 1347719f5d7fSMiklos Szeredi goto out; 1348719f5d7fSMiklos Szeredi } 1349719f5d7fSMiklos Szeredi err = propagate_mnt(dest_mnt, dest_dentry, source_mnt, &tree_list); 1350719f5d7fSMiklos Szeredi if (err) 1351719f5d7fSMiklos Szeredi goto out_cleanup_ids; 1352b90fa9aeSRam Pai 1353b90fa9aeSRam Pai if (IS_MNT_SHARED(dest_mnt)) { 1354b90fa9aeSRam Pai for (p = source_mnt; p; p = next_mnt(p, source_mnt)) 1355b90fa9aeSRam Pai set_mnt_shared(p); 1356b90fa9aeSRam Pai } 1357b90fa9aeSRam Pai 1358b90fa9aeSRam Pai spin_lock(&vfsmount_lock); 13591a390689SAl Viro if (parent_path) { 13601a390689SAl Viro detach_mnt(source_mnt, parent_path); 13611a390689SAl Viro attach_mnt(source_mnt, path); 1362e5d67f07SAl Viro touch_mnt_namespace(parent_path->mnt->mnt_ns); 136321444403SRam Pai } else { 1364b90fa9aeSRam Pai mnt_set_mountpoint(dest_mnt, dest_dentry, source_mnt); 1365b90fa9aeSRam Pai commit_tree(source_mnt); 136621444403SRam Pai } 1367b90fa9aeSRam Pai 1368b90fa9aeSRam Pai list_for_each_entry_safe(child, p, &tree_list, mnt_hash) { 1369b90fa9aeSRam Pai list_del_init(&child->mnt_hash); 1370b90fa9aeSRam Pai commit_tree(child); 1371b90fa9aeSRam Pai } 1372b90fa9aeSRam Pai spin_unlock(&vfsmount_lock); 1373b90fa9aeSRam Pai return 0; 1374719f5d7fSMiklos Szeredi 1375719f5d7fSMiklos Szeredi out_cleanup_ids: 1376719f5d7fSMiklos Szeredi if (IS_MNT_SHARED(dest_mnt)) 1377719f5d7fSMiklos Szeredi cleanup_group_ids(source_mnt, NULL); 1378719f5d7fSMiklos Szeredi out: 1379719f5d7fSMiklos Szeredi return err; 1380b90fa9aeSRam Pai } 1381b90fa9aeSRam Pai 13828c3ee42eSAl Viro static int graft_tree(struct vfsmount *mnt, struct path *path) 13831da177e4SLinus Torvalds { 13841da177e4SLinus Torvalds int err; 13851da177e4SLinus Torvalds if (mnt->mnt_sb->s_flags & MS_NOUSER) 13861da177e4SLinus Torvalds return -EINVAL; 13871da177e4SLinus Torvalds 13888c3ee42eSAl Viro if (S_ISDIR(path->dentry->d_inode->i_mode) != 13891da177e4SLinus Torvalds S_ISDIR(mnt->mnt_root->d_inode->i_mode)) 13901da177e4SLinus Torvalds return -ENOTDIR; 13911da177e4SLinus Torvalds 13921da177e4SLinus Torvalds err = -ENOENT; 13938c3ee42eSAl Viro mutex_lock(&path->dentry->d_inode->i_mutex); 13948c3ee42eSAl Viro if (IS_DEADDIR(path->dentry->d_inode)) 13951da177e4SLinus Torvalds goto out_unlock; 13961da177e4SLinus Torvalds 13978c3ee42eSAl Viro err = security_sb_check_sb(mnt, path); 13981da177e4SLinus Torvalds if (err) 13991da177e4SLinus Torvalds goto out_unlock; 14001da177e4SLinus Torvalds 14011da177e4SLinus Torvalds err = -ENOENT; 1402f3da392eSAlexey Dobriyan if (!d_unlinked(path->dentry)) 14038c3ee42eSAl Viro err = attach_recursive_mnt(mnt, path, NULL); 14041da177e4SLinus Torvalds out_unlock: 14058c3ee42eSAl Viro mutex_unlock(&path->dentry->d_inode->i_mutex); 14061da177e4SLinus Torvalds if (!err) 14078c3ee42eSAl Viro security_sb_post_addmount(mnt, path); 14081da177e4SLinus Torvalds return err; 14091da177e4SLinus Torvalds } 14101da177e4SLinus Torvalds 14111da177e4SLinus Torvalds /* 141207b20889SRam Pai * recursively change the type of the mountpoint. 141307b20889SRam Pai */ 14140a0d8a46SAl Viro static int do_change_type(struct path *path, int flag) 141507b20889SRam Pai { 14162d92ab3cSAl Viro struct vfsmount *m, *mnt = path->mnt; 141707b20889SRam Pai int recurse = flag & MS_REC; 141807b20889SRam Pai int type = flag & ~MS_REC; 1419719f5d7fSMiklos Szeredi int err = 0; 142007b20889SRam Pai 1421ee6f9582SMiklos Szeredi if (!capable(CAP_SYS_ADMIN)) 1422ee6f9582SMiklos Szeredi return -EPERM; 1423ee6f9582SMiklos Szeredi 14242d92ab3cSAl Viro if (path->dentry != path->mnt->mnt_root) 142507b20889SRam Pai return -EINVAL; 142607b20889SRam Pai 142707b20889SRam Pai down_write(&namespace_sem); 1428719f5d7fSMiklos Szeredi if (type == MS_SHARED) { 1429719f5d7fSMiklos Szeredi err = invent_group_ids(mnt, recurse); 1430719f5d7fSMiklos Szeredi if (err) 1431719f5d7fSMiklos Szeredi goto out_unlock; 1432719f5d7fSMiklos Szeredi } 1433719f5d7fSMiklos Szeredi 143407b20889SRam Pai spin_lock(&vfsmount_lock); 143507b20889SRam Pai for (m = mnt; m; m = (recurse ? next_mnt(m, mnt) : NULL)) 143607b20889SRam Pai change_mnt_propagation(m, type); 143707b20889SRam Pai spin_unlock(&vfsmount_lock); 1438719f5d7fSMiklos Szeredi 1439719f5d7fSMiklos Szeredi out_unlock: 144007b20889SRam Pai up_write(&namespace_sem); 1441719f5d7fSMiklos Szeredi return err; 144207b20889SRam Pai } 144307b20889SRam Pai 144407b20889SRam Pai /* 14451da177e4SLinus Torvalds * do loopback mount. 14461da177e4SLinus Torvalds */ 14470a0d8a46SAl Viro static int do_loopback(struct path *path, char *old_name, 14482dafe1c4SEric Sandeen int recurse) 14491da177e4SLinus Torvalds { 14502d92ab3cSAl Viro struct path old_path; 14511da177e4SLinus Torvalds struct vfsmount *mnt = NULL; 14522d92ab3cSAl Viro int err = mount_is_safe(path); 14531da177e4SLinus Torvalds if (err) 14541da177e4SLinus Torvalds return err; 14551da177e4SLinus Torvalds if (!old_name || !*old_name) 14561da177e4SLinus Torvalds return -EINVAL; 14572d92ab3cSAl Viro err = kern_path(old_name, LOOKUP_FOLLOW, &old_path); 14581da177e4SLinus Torvalds if (err) 14591da177e4SLinus Torvalds return err; 14601da177e4SLinus Torvalds 1461390c6843SRam Pai down_write(&namespace_sem); 14621da177e4SLinus Torvalds err = -EINVAL; 14632d92ab3cSAl Viro if (IS_MNT_UNBINDABLE(old_path.mnt)) 14649676f0c6SRam Pai goto out; 14659676f0c6SRam Pai 14662d92ab3cSAl Viro if (!check_mnt(path->mnt) || !check_mnt(old_path.mnt)) 1467ccd48bc7SAl Viro goto out; 1468ccd48bc7SAl Viro 14691da177e4SLinus Torvalds err = -ENOMEM; 14701da177e4SLinus Torvalds if (recurse) 14712d92ab3cSAl Viro mnt = copy_tree(old_path.mnt, old_path.dentry, 0); 14721da177e4SLinus Torvalds else 14732d92ab3cSAl Viro mnt = clone_mnt(old_path.mnt, old_path.dentry, 0); 14741da177e4SLinus Torvalds 1475ccd48bc7SAl Viro if (!mnt) 1476ccd48bc7SAl Viro goto out; 1477ccd48bc7SAl Viro 14782d92ab3cSAl Viro err = graft_tree(mnt, path); 14791da177e4SLinus Torvalds if (err) { 148070fbcdf4SRam Pai LIST_HEAD(umount_list); 14811da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 1482a05964f3SRam Pai umount_tree(mnt, 0, &umount_list); 14831da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 148470fbcdf4SRam Pai release_mounts(&umount_list); 14855b83d2c5SRam Pai } 14861da177e4SLinus Torvalds 1487ccd48bc7SAl Viro out: 1488390c6843SRam Pai up_write(&namespace_sem); 14892d92ab3cSAl Viro path_put(&old_path); 14901da177e4SLinus Torvalds return err; 14911da177e4SLinus Torvalds } 14921da177e4SLinus Torvalds 14932e4b7fcdSDave Hansen static int change_mount_flags(struct vfsmount *mnt, int ms_flags) 14942e4b7fcdSDave Hansen { 14952e4b7fcdSDave Hansen int error = 0; 14962e4b7fcdSDave Hansen int readonly_request = 0; 14972e4b7fcdSDave Hansen 14982e4b7fcdSDave Hansen if (ms_flags & MS_RDONLY) 14992e4b7fcdSDave Hansen readonly_request = 1; 15002e4b7fcdSDave Hansen if (readonly_request == __mnt_is_readonly(mnt)) 15012e4b7fcdSDave Hansen return 0; 15022e4b7fcdSDave Hansen 15032e4b7fcdSDave Hansen if (readonly_request) 15042e4b7fcdSDave Hansen error = mnt_make_readonly(mnt); 15052e4b7fcdSDave Hansen else 15062e4b7fcdSDave Hansen __mnt_unmake_readonly(mnt); 15072e4b7fcdSDave Hansen return error; 15082e4b7fcdSDave Hansen } 15092e4b7fcdSDave Hansen 15101da177e4SLinus Torvalds /* 15111da177e4SLinus Torvalds * change filesystem flags. dir should be a physical root of filesystem. 15121da177e4SLinus Torvalds * If you've mounted a non-root directory somewhere and want to do remount 15131da177e4SLinus Torvalds * on it - tough luck. 15141da177e4SLinus Torvalds */ 15150a0d8a46SAl Viro static int do_remount(struct path *path, int flags, int mnt_flags, 15161da177e4SLinus Torvalds void *data) 15171da177e4SLinus Torvalds { 15181da177e4SLinus Torvalds int err; 15192d92ab3cSAl Viro struct super_block *sb = path->mnt->mnt_sb; 15201da177e4SLinus Torvalds 15211da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 15221da177e4SLinus Torvalds return -EPERM; 15231da177e4SLinus Torvalds 15242d92ab3cSAl Viro if (!check_mnt(path->mnt)) 15251da177e4SLinus Torvalds return -EINVAL; 15261da177e4SLinus Torvalds 15272d92ab3cSAl Viro if (path->dentry != path->mnt->mnt_root) 15281da177e4SLinus Torvalds return -EINVAL; 15291da177e4SLinus Torvalds 15301da177e4SLinus Torvalds down_write(&sb->s_umount); 15312e4b7fcdSDave Hansen if (flags & MS_BIND) 15322d92ab3cSAl Viro err = change_mount_flags(path->mnt, flags); 15334aa98cf7SAl Viro else 15341da177e4SLinus Torvalds err = do_remount_sb(sb, flags, data, 0); 15351da177e4SLinus Torvalds if (!err) 15362d92ab3cSAl Viro path->mnt->mnt_flags = mnt_flags; 15371da177e4SLinus Torvalds up_write(&sb->s_umount); 15380e55a7ccSDan Williams if (!err) { 15392d92ab3cSAl Viro security_sb_post_remount(path->mnt, flags, data); 15400e55a7ccSDan Williams 15410e55a7ccSDan Williams spin_lock(&vfsmount_lock); 15420e55a7ccSDan Williams touch_mnt_namespace(path->mnt->mnt_ns); 15430e55a7ccSDan Williams spin_unlock(&vfsmount_lock); 15440e55a7ccSDan Williams } 15451da177e4SLinus Torvalds return err; 15461da177e4SLinus Torvalds } 15471da177e4SLinus Torvalds 15489676f0c6SRam Pai static inline int tree_contains_unbindable(struct vfsmount *mnt) 15499676f0c6SRam Pai { 15509676f0c6SRam Pai struct vfsmount *p; 15519676f0c6SRam Pai for (p = mnt; p; p = next_mnt(p, mnt)) { 15529676f0c6SRam Pai if (IS_MNT_UNBINDABLE(p)) 15539676f0c6SRam Pai return 1; 15549676f0c6SRam Pai } 15559676f0c6SRam Pai return 0; 15569676f0c6SRam Pai } 15579676f0c6SRam Pai 15580a0d8a46SAl Viro static int do_move_mount(struct path *path, char *old_name) 15591da177e4SLinus Torvalds { 15602d92ab3cSAl Viro struct path old_path, parent_path; 15611da177e4SLinus Torvalds struct vfsmount *p; 15621da177e4SLinus Torvalds int err = 0; 15631da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 15641da177e4SLinus Torvalds return -EPERM; 15651da177e4SLinus Torvalds if (!old_name || !*old_name) 15661da177e4SLinus Torvalds return -EINVAL; 15672d92ab3cSAl Viro err = kern_path(old_name, LOOKUP_FOLLOW, &old_path); 15681da177e4SLinus Torvalds if (err) 15691da177e4SLinus Torvalds return err; 15701da177e4SLinus Torvalds 1571390c6843SRam Pai down_write(&namespace_sem); 15722d92ab3cSAl Viro while (d_mountpoint(path->dentry) && 15739393bd07SAl Viro follow_down(path)) 15741da177e4SLinus Torvalds ; 15751da177e4SLinus Torvalds err = -EINVAL; 15762d92ab3cSAl Viro if (!check_mnt(path->mnt) || !check_mnt(old_path.mnt)) 15771da177e4SLinus Torvalds goto out; 15781da177e4SLinus Torvalds 15791da177e4SLinus Torvalds err = -ENOENT; 15802d92ab3cSAl Viro mutex_lock(&path->dentry->d_inode->i_mutex); 15812d92ab3cSAl Viro if (IS_DEADDIR(path->dentry->d_inode)) 15821da177e4SLinus Torvalds goto out1; 15831da177e4SLinus Torvalds 1584f3da392eSAlexey Dobriyan if (d_unlinked(path->dentry)) 158521444403SRam Pai goto out1; 15861da177e4SLinus Torvalds 15871da177e4SLinus Torvalds err = -EINVAL; 15882d92ab3cSAl Viro if (old_path.dentry != old_path.mnt->mnt_root) 158921444403SRam Pai goto out1; 15901da177e4SLinus Torvalds 15912d92ab3cSAl Viro if (old_path.mnt == old_path.mnt->mnt_parent) 159221444403SRam Pai goto out1; 15931da177e4SLinus Torvalds 15942d92ab3cSAl Viro if (S_ISDIR(path->dentry->d_inode->i_mode) != 15952d92ab3cSAl Viro S_ISDIR(old_path.dentry->d_inode->i_mode)) 159621444403SRam Pai goto out1; 159721444403SRam Pai /* 159821444403SRam Pai * Don't move a mount residing in a shared parent. 159921444403SRam Pai */ 16002d92ab3cSAl Viro if (old_path.mnt->mnt_parent && 16012d92ab3cSAl Viro IS_MNT_SHARED(old_path.mnt->mnt_parent)) 160221444403SRam Pai goto out1; 16039676f0c6SRam Pai /* 16049676f0c6SRam Pai * Don't move a mount tree containing unbindable mounts to a destination 16059676f0c6SRam Pai * mount which is shared. 16069676f0c6SRam Pai */ 16072d92ab3cSAl Viro if (IS_MNT_SHARED(path->mnt) && 16082d92ab3cSAl Viro tree_contains_unbindable(old_path.mnt)) 16099676f0c6SRam Pai goto out1; 16101da177e4SLinus Torvalds err = -ELOOP; 16112d92ab3cSAl Viro for (p = path->mnt; p->mnt_parent != p; p = p->mnt_parent) 16122d92ab3cSAl Viro if (p == old_path.mnt) 161321444403SRam Pai goto out1; 16141da177e4SLinus Torvalds 16152d92ab3cSAl Viro err = attach_recursive_mnt(old_path.mnt, path, &parent_path); 16164ac91378SJan Blunck if (err) 161721444403SRam Pai goto out1; 16181da177e4SLinus Torvalds 16191da177e4SLinus Torvalds /* if the mount is moved, it should no longer be expire 16201da177e4SLinus Torvalds * automatically */ 16212d92ab3cSAl Viro list_del_init(&old_path.mnt->mnt_expire); 16221da177e4SLinus Torvalds out1: 16232d92ab3cSAl Viro mutex_unlock(&path->dentry->d_inode->i_mutex); 16241da177e4SLinus Torvalds out: 1625390c6843SRam Pai up_write(&namespace_sem); 16261da177e4SLinus Torvalds if (!err) 16271a390689SAl Viro path_put(&parent_path); 16282d92ab3cSAl Viro path_put(&old_path); 16291da177e4SLinus Torvalds return err; 16301da177e4SLinus Torvalds } 16311da177e4SLinus Torvalds 16321da177e4SLinus Torvalds /* 16331da177e4SLinus Torvalds * create a new mount for userspace and request it to be added into the 16341da177e4SLinus Torvalds * namespace's tree 16351da177e4SLinus Torvalds */ 16360a0d8a46SAl Viro static int do_new_mount(struct path *path, char *type, int flags, 16371da177e4SLinus Torvalds int mnt_flags, char *name, void *data) 16381da177e4SLinus Torvalds { 16391da177e4SLinus Torvalds struct vfsmount *mnt; 16401da177e4SLinus Torvalds 16411da177e4SLinus Torvalds if (!type || !memchr(type, 0, PAGE_SIZE)) 16421da177e4SLinus Torvalds return -EINVAL; 16431da177e4SLinus Torvalds 16441da177e4SLinus Torvalds /* we need capabilities... */ 16451da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 16461da177e4SLinus Torvalds return -EPERM; 16471da177e4SLinus Torvalds 16487f78d4cdSAl Viro lock_kernel(); 16491da177e4SLinus Torvalds mnt = do_kern_mount(type, flags, name, data); 16507f78d4cdSAl Viro unlock_kernel(); 16511da177e4SLinus Torvalds if (IS_ERR(mnt)) 16521da177e4SLinus Torvalds return PTR_ERR(mnt); 16531da177e4SLinus Torvalds 16542d92ab3cSAl Viro return do_add_mount(mnt, path, mnt_flags, NULL); 16551da177e4SLinus Torvalds } 16561da177e4SLinus Torvalds 16571da177e4SLinus Torvalds /* 16581da177e4SLinus Torvalds * add a mount into a namespace's mount tree 16591da177e4SLinus Torvalds * - provide the option of adding the new mount to an expiration list 16601da177e4SLinus Torvalds */ 16618d66bf54SAl Viro int do_add_mount(struct vfsmount *newmnt, struct path *path, 16621da177e4SLinus Torvalds int mnt_flags, struct list_head *fslist) 16631da177e4SLinus Torvalds { 16641da177e4SLinus Torvalds int err; 16651da177e4SLinus Torvalds 1666390c6843SRam Pai down_write(&namespace_sem); 16671da177e4SLinus Torvalds /* Something was mounted here while we slept */ 16688d66bf54SAl Viro while (d_mountpoint(path->dentry) && 16699393bd07SAl Viro follow_down(path)) 16701da177e4SLinus Torvalds ; 16711da177e4SLinus Torvalds err = -EINVAL; 1672dd5cae6eSAl Viro if (!(mnt_flags & MNT_SHRINKABLE) && !check_mnt(path->mnt)) 16731da177e4SLinus Torvalds goto unlock; 16741da177e4SLinus Torvalds 16751da177e4SLinus Torvalds /* Refuse the same filesystem on the same mount point */ 16761da177e4SLinus Torvalds err = -EBUSY; 16778d66bf54SAl Viro if (path->mnt->mnt_sb == newmnt->mnt_sb && 16788d66bf54SAl Viro path->mnt->mnt_root == path->dentry) 16791da177e4SLinus Torvalds goto unlock; 16801da177e4SLinus Torvalds 16811da177e4SLinus Torvalds err = -EINVAL; 16821da177e4SLinus Torvalds if (S_ISLNK(newmnt->mnt_root->d_inode->i_mode)) 16831da177e4SLinus Torvalds goto unlock; 16841da177e4SLinus Torvalds 16851da177e4SLinus Torvalds newmnt->mnt_flags = mnt_flags; 16868d66bf54SAl Viro if ((err = graft_tree(newmnt, path))) 16875b83d2c5SRam Pai goto unlock; 16881da177e4SLinus Torvalds 16896758f953SAl Viro if (fslist) /* add to the specified expiration list */ 169055e700b9SMiklos Szeredi list_add_tail(&newmnt->mnt_expire, fslist); 16916758f953SAl Viro 1692390c6843SRam Pai up_write(&namespace_sem); 16935b83d2c5SRam Pai return 0; 16941da177e4SLinus Torvalds 16951da177e4SLinus Torvalds unlock: 1696390c6843SRam Pai up_write(&namespace_sem); 16971da177e4SLinus Torvalds mntput(newmnt); 16981da177e4SLinus Torvalds return err; 16991da177e4SLinus Torvalds } 17001da177e4SLinus Torvalds 17011da177e4SLinus Torvalds EXPORT_SYMBOL_GPL(do_add_mount); 17021da177e4SLinus Torvalds 17035528f911STrond Myklebust /* 17041da177e4SLinus Torvalds * process a list of expirable mountpoints with the intent of discarding any 17051da177e4SLinus Torvalds * mountpoints that aren't in use and haven't been touched since last we came 17061da177e4SLinus Torvalds * here 17071da177e4SLinus Torvalds */ 17081da177e4SLinus Torvalds void mark_mounts_for_expiry(struct list_head *mounts) 17091da177e4SLinus Torvalds { 17101da177e4SLinus Torvalds struct vfsmount *mnt, *next; 17111da177e4SLinus Torvalds LIST_HEAD(graveyard); 1712bcc5c7d2SAl Viro LIST_HEAD(umounts); 17131da177e4SLinus Torvalds 17141da177e4SLinus Torvalds if (list_empty(mounts)) 17151da177e4SLinus Torvalds return; 17161da177e4SLinus Torvalds 1717bcc5c7d2SAl Viro down_write(&namespace_sem); 17181da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 17191da177e4SLinus Torvalds 17201da177e4SLinus Torvalds /* extract from the expiration list every vfsmount that matches the 17211da177e4SLinus Torvalds * following criteria: 17221da177e4SLinus Torvalds * - only referenced by its parent vfsmount 17231da177e4SLinus Torvalds * - still marked for expiry (marked on the last call here; marks are 17241da177e4SLinus Torvalds * cleared by mntput()) 17251da177e4SLinus Torvalds */ 172655e700b9SMiklos Szeredi list_for_each_entry_safe(mnt, next, mounts, mnt_expire) { 17271da177e4SLinus Torvalds if (!xchg(&mnt->mnt_expiry_mark, 1) || 1728bcc5c7d2SAl Viro propagate_mount_busy(mnt, 1)) 17291da177e4SLinus Torvalds continue; 173055e700b9SMiklos Szeredi list_move(&mnt->mnt_expire, &graveyard); 17311da177e4SLinus Torvalds } 1732bcc5c7d2SAl Viro while (!list_empty(&graveyard)) { 1733bcc5c7d2SAl Viro mnt = list_first_entry(&graveyard, struct vfsmount, mnt_expire); 1734bcc5c7d2SAl Viro touch_mnt_namespace(mnt->mnt_ns); 1735bcc5c7d2SAl Viro umount_tree(mnt, 1, &umounts); 1736bcc5c7d2SAl Viro } 17371da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 1738bcc5c7d2SAl Viro up_write(&namespace_sem); 1739bcc5c7d2SAl Viro 1740bcc5c7d2SAl Viro release_mounts(&umounts); 17411da177e4SLinus Torvalds } 17421da177e4SLinus Torvalds 17431da177e4SLinus Torvalds EXPORT_SYMBOL_GPL(mark_mounts_for_expiry); 17441da177e4SLinus Torvalds 17451da177e4SLinus Torvalds /* 17465528f911STrond Myklebust * Ripoff of 'select_parent()' 17475528f911STrond Myklebust * 17485528f911STrond Myklebust * search the list of submounts for a given mountpoint, and move any 17495528f911STrond Myklebust * shrinkable submounts to the 'graveyard' list. 17505528f911STrond Myklebust */ 17515528f911STrond Myklebust static int select_submounts(struct vfsmount *parent, struct list_head *graveyard) 17525528f911STrond Myklebust { 17535528f911STrond Myklebust struct vfsmount *this_parent = parent; 17545528f911STrond Myklebust struct list_head *next; 17555528f911STrond Myklebust int found = 0; 17565528f911STrond Myklebust 17575528f911STrond Myklebust repeat: 17585528f911STrond Myklebust next = this_parent->mnt_mounts.next; 17595528f911STrond Myklebust resume: 17605528f911STrond Myklebust while (next != &this_parent->mnt_mounts) { 17615528f911STrond Myklebust struct list_head *tmp = next; 17625528f911STrond Myklebust struct vfsmount *mnt = list_entry(tmp, struct vfsmount, mnt_child); 17635528f911STrond Myklebust 17645528f911STrond Myklebust next = tmp->next; 17655528f911STrond Myklebust if (!(mnt->mnt_flags & MNT_SHRINKABLE)) 17665528f911STrond Myklebust continue; 17675528f911STrond Myklebust /* 17685528f911STrond Myklebust * Descend a level if the d_mounts list is non-empty. 17695528f911STrond Myklebust */ 17705528f911STrond Myklebust if (!list_empty(&mnt->mnt_mounts)) { 17715528f911STrond Myklebust this_parent = mnt; 17725528f911STrond Myklebust goto repeat; 17735528f911STrond Myklebust } 17745528f911STrond Myklebust 17755528f911STrond Myklebust if (!propagate_mount_busy(mnt, 1)) { 17765528f911STrond Myklebust list_move_tail(&mnt->mnt_expire, graveyard); 17775528f911STrond Myklebust found++; 17785528f911STrond Myklebust } 17795528f911STrond Myklebust } 17805528f911STrond Myklebust /* 17815528f911STrond Myklebust * All done at this level ... ascend and resume the search 17825528f911STrond Myklebust */ 17835528f911STrond Myklebust if (this_parent != parent) { 17845528f911STrond Myklebust next = this_parent->mnt_child.next; 17855528f911STrond Myklebust this_parent = this_parent->mnt_parent; 17865528f911STrond Myklebust goto resume; 17875528f911STrond Myklebust } 17885528f911STrond Myklebust return found; 17895528f911STrond Myklebust } 17905528f911STrond Myklebust 17915528f911STrond Myklebust /* 17925528f911STrond Myklebust * process a list of expirable mountpoints with the intent of discarding any 17935528f911STrond Myklebust * submounts of a specific parent mountpoint 17945528f911STrond Myklebust */ 1795c35038beSAl Viro static void shrink_submounts(struct vfsmount *mnt, struct list_head *umounts) 17965528f911STrond Myklebust { 17975528f911STrond Myklebust LIST_HEAD(graveyard); 1798c35038beSAl Viro struct vfsmount *m; 17995528f911STrond Myklebust 18005528f911STrond Myklebust /* extract submounts of 'mountpoint' from the expiration list */ 1801c35038beSAl Viro while (select_submounts(mnt, &graveyard)) { 1802bcc5c7d2SAl Viro while (!list_empty(&graveyard)) { 1803c35038beSAl Viro m = list_first_entry(&graveyard, struct vfsmount, 1804bcc5c7d2SAl Viro mnt_expire); 1805afef80b3SEric W. Biederman touch_mnt_namespace(m->mnt_ns); 1806afef80b3SEric W. Biederman umount_tree(m, 1, umounts); 1807bcc5c7d2SAl Viro } 1808bcc5c7d2SAl Viro } 18095528f911STrond Myklebust } 18105528f911STrond Myklebust 18115528f911STrond Myklebust /* 18121da177e4SLinus Torvalds * Some copy_from_user() implementations do not return the exact number of 18131da177e4SLinus Torvalds * bytes remaining to copy on a fault. But copy_mount_options() requires that. 18141da177e4SLinus Torvalds * Note that this function differs from copy_from_user() in that it will oops 18151da177e4SLinus Torvalds * on bad values of `to', rather than returning a short copy. 18161da177e4SLinus Torvalds */ 1817b58fed8bSRam Pai static long exact_copy_from_user(void *to, const void __user * from, 1818b58fed8bSRam Pai unsigned long n) 18191da177e4SLinus Torvalds { 18201da177e4SLinus Torvalds char *t = to; 18211da177e4SLinus Torvalds const char __user *f = from; 18221da177e4SLinus Torvalds char c; 18231da177e4SLinus Torvalds 18241da177e4SLinus Torvalds if (!access_ok(VERIFY_READ, from, n)) 18251da177e4SLinus Torvalds return n; 18261da177e4SLinus Torvalds 18271da177e4SLinus Torvalds while (n) { 18281da177e4SLinus Torvalds if (__get_user(c, f)) { 18291da177e4SLinus Torvalds memset(t, 0, n); 18301da177e4SLinus Torvalds break; 18311da177e4SLinus Torvalds } 18321da177e4SLinus Torvalds *t++ = c; 18331da177e4SLinus Torvalds f++; 18341da177e4SLinus Torvalds n--; 18351da177e4SLinus Torvalds } 18361da177e4SLinus Torvalds return n; 18371da177e4SLinus Torvalds } 18381da177e4SLinus Torvalds 18391da177e4SLinus Torvalds int copy_mount_options(const void __user * data, unsigned long *where) 18401da177e4SLinus Torvalds { 18411da177e4SLinus Torvalds int i; 18421da177e4SLinus Torvalds unsigned long page; 18431da177e4SLinus Torvalds unsigned long size; 18441da177e4SLinus Torvalds 18451da177e4SLinus Torvalds *where = 0; 18461da177e4SLinus Torvalds if (!data) 18471da177e4SLinus Torvalds return 0; 18481da177e4SLinus Torvalds 18491da177e4SLinus Torvalds if (!(page = __get_free_page(GFP_KERNEL))) 18501da177e4SLinus Torvalds return -ENOMEM; 18511da177e4SLinus Torvalds 18521da177e4SLinus Torvalds /* We only care that *some* data at the address the user 18531da177e4SLinus Torvalds * gave us is valid. Just in case, we'll zero 18541da177e4SLinus Torvalds * the remainder of the page. 18551da177e4SLinus Torvalds */ 18561da177e4SLinus Torvalds /* copy_from_user cannot cross TASK_SIZE ! */ 18571da177e4SLinus Torvalds size = TASK_SIZE - (unsigned long)data; 18581da177e4SLinus Torvalds if (size > PAGE_SIZE) 18591da177e4SLinus Torvalds size = PAGE_SIZE; 18601da177e4SLinus Torvalds 18611da177e4SLinus Torvalds i = size - exact_copy_from_user((void *)page, data, size); 18621da177e4SLinus Torvalds if (!i) { 18631da177e4SLinus Torvalds free_page(page); 18641da177e4SLinus Torvalds return -EFAULT; 18651da177e4SLinus Torvalds } 18661da177e4SLinus Torvalds if (i != PAGE_SIZE) 18671da177e4SLinus Torvalds memset((char *)page + i, 0, PAGE_SIZE - i); 18681da177e4SLinus Torvalds *where = page; 18691da177e4SLinus Torvalds return 0; 18701da177e4SLinus Torvalds } 18711da177e4SLinus Torvalds 18721da177e4SLinus Torvalds /* 18731da177e4SLinus Torvalds * Flags is a 32-bit value that allows up to 31 non-fs dependent flags to 18741da177e4SLinus Torvalds * be given to the mount() call (ie: read-only, no-dev, no-suid etc). 18751da177e4SLinus Torvalds * 18761da177e4SLinus Torvalds * data is a (void *) that can point to any structure up to 18771da177e4SLinus Torvalds * PAGE_SIZE-1 bytes, which can contain arbitrary fs-dependent 18781da177e4SLinus Torvalds * information (or be NULL). 18791da177e4SLinus Torvalds * 18801da177e4SLinus Torvalds * Pre-0.97 versions of mount() didn't have a flags word. 18811da177e4SLinus Torvalds * When the flags word was introduced its top half was required 18821da177e4SLinus Torvalds * to have the magic value 0xC0ED, and this remained so until 2.4.0-test9. 18831da177e4SLinus Torvalds * Therefore, if this magic number is present, it carries no information 18841da177e4SLinus Torvalds * and must be discarded. 18851da177e4SLinus Torvalds */ 18861da177e4SLinus Torvalds long do_mount(char *dev_name, char *dir_name, char *type_page, 18871da177e4SLinus Torvalds unsigned long flags, void *data_page) 18881da177e4SLinus Torvalds { 18892d92ab3cSAl Viro struct path path; 18901da177e4SLinus Torvalds int retval = 0; 18911da177e4SLinus Torvalds int mnt_flags = 0; 18921da177e4SLinus Torvalds 18931da177e4SLinus Torvalds /* Discard magic */ 18941da177e4SLinus Torvalds if ((flags & MS_MGC_MSK) == MS_MGC_VAL) 18951da177e4SLinus Torvalds flags &= ~MS_MGC_MSK; 18961da177e4SLinus Torvalds 18971da177e4SLinus Torvalds /* Basic sanity checks */ 18981da177e4SLinus Torvalds 18991da177e4SLinus Torvalds if (!dir_name || !*dir_name || !memchr(dir_name, 0, PAGE_SIZE)) 19001da177e4SLinus Torvalds return -EINVAL; 19011da177e4SLinus Torvalds if (dev_name && !memchr(dev_name, 0, PAGE_SIZE)) 19021da177e4SLinus Torvalds return -EINVAL; 19031da177e4SLinus Torvalds 19041da177e4SLinus Torvalds if (data_page) 19051da177e4SLinus Torvalds ((char *)data_page)[PAGE_SIZE - 1] = 0; 19061da177e4SLinus Torvalds 1907613cbe3dSAndi Kleen /* Default to relatime unless overriden */ 1908613cbe3dSAndi Kleen if (!(flags & MS_NOATIME)) 19090a1c01c9SMatthew Garrett mnt_flags |= MNT_RELATIME; 19100a1c01c9SMatthew Garrett 19111da177e4SLinus Torvalds /* Separate the per-mountpoint flags */ 19121da177e4SLinus Torvalds if (flags & MS_NOSUID) 19131da177e4SLinus Torvalds mnt_flags |= MNT_NOSUID; 19141da177e4SLinus Torvalds if (flags & MS_NODEV) 19151da177e4SLinus Torvalds mnt_flags |= MNT_NODEV; 19161da177e4SLinus Torvalds if (flags & MS_NOEXEC) 19171da177e4SLinus Torvalds mnt_flags |= MNT_NOEXEC; 1918fc33a7bbSChristoph Hellwig if (flags & MS_NOATIME) 1919fc33a7bbSChristoph Hellwig mnt_flags |= MNT_NOATIME; 1920fc33a7bbSChristoph Hellwig if (flags & MS_NODIRATIME) 1921fc33a7bbSChristoph Hellwig mnt_flags |= MNT_NODIRATIME; 1922d0adde57SMatthew Garrett if (flags & MS_STRICTATIME) 1923d0adde57SMatthew Garrett mnt_flags &= ~(MNT_RELATIME | MNT_NOATIME); 19242e4b7fcdSDave Hansen if (flags & MS_RDONLY) 19252e4b7fcdSDave Hansen mnt_flags |= MNT_READONLY; 1926fc33a7bbSChristoph Hellwig 1927fc33a7bbSChristoph Hellwig flags &= ~(MS_NOSUID | MS_NOEXEC | MS_NODEV | MS_ACTIVE | 1928d0adde57SMatthew Garrett MS_NOATIME | MS_NODIRATIME | MS_RELATIME| MS_KERNMOUNT | 1929d0adde57SMatthew Garrett MS_STRICTATIME); 19301da177e4SLinus Torvalds 19311da177e4SLinus Torvalds /* ... and get the mountpoint */ 19322d92ab3cSAl Viro retval = kern_path(dir_name, LOOKUP_FOLLOW, &path); 19331da177e4SLinus Torvalds if (retval) 19341da177e4SLinus Torvalds return retval; 19351da177e4SLinus Torvalds 19362d92ab3cSAl Viro retval = security_sb_mount(dev_name, &path, 1937b5266eb4SAl Viro type_page, flags, data_page); 19381da177e4SLinus Torvalds if (retval) 19391da177e4SLinus Torvalds goto dput_out; 19401da177e4SLinus Torvalds 19411da177e4SLinus Torvalds if (flags & MS_REMOUNT) 19422d92ab3cSAl Viro retval = do_remount(&path, flags & ~MS_REMOUNT, mnt_flags, 19431da177e4SLinus Torvalds data_page); 19441da177e4SLinus Torvalds else if (flags & MS_BIND) 19452d92ab3cSAl Viro retval = do_loopback(&path, dev_name, flags & MS_REC); 19469676f0c6SRam Pai else if (flags & (MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE)) 19472d92ab3cSAl Viro retval = do_change_type(&path, flags); 19481da177e4SLinus Torvalds else if (flags & MS_MOVE) 19492d92ab3cSAl Viro retval = do_move_mount(&path, dev_name); 19501da177e4SLinus Torvalds else 19512d92ab3cSAl Viro retval = do_new_mount(&path, type_page, flags, mnt_flags, 19521da177e4SLinus Torvalds dev_name, data_page); 19531da177e4SLinus Torvalds dput_out: 19542d92ab3cSAl Viro path_put(&path); 19551da177e4SLinus Torvalds return retval; 19561da177e4SLinus Torvalds } 19571da177e4SLinus Torvalds 1958cf8d2c11STrond Myklebust static struct mnt_namespace *alloc_mnt_ns(void) 1959cf8d2c11STrond Myklebust { 1960cf8d2c11STrond Myklebust struct mnt_namespace *new_ns; 1961cf8d2c11STrond Myklebust 1962cf8d2c11STrond Myklebust new_ns = kmalloc(sizeof(struct mnt_namespace), GFP_KERNEL); 1963cf8d2c11STrond Myklebust if (!new_ns) 1964cf8d2c11STrond Myklebust return ERR_PTR(-ENOMEM); 1965cf8d2c11STrond Myklebust atomic_set(&new_ns->count, 1); 1966cf8d2c11STrond Myklebust new_ns->root = NULL; 1967cf8d2c11STrond Myklebust INIT_LIST_HEAD(&new_ns->list); 1968cf8d2c11STrond Myklebust init_waitqueue_head(&new_ns->poll); 1969cf8d2c11STrond Myklebust new_ns->event = 0; 1970cf8d2c11STrond Myklebust return new_ns; 1971cf8d2c11STrond Myklebust } 1972cf8d2c11STrond Myklebust 1973741a2951SJANAK DESAI /* 1974741a2951SJANAK DESAI * Allocate a new namespace structure and populate it with contents 1975741a2951SJANAK DESAI * copied from the namespace of the passed in task structure. 1976741a2951SJANAK DESAI */ 1977e3222c4eSBadari Pulavarty static struct mnt_namespace *dup_mnt_ns(struct mnt_namespace *mnt_ns, 19786b3286edSKirill Korotaev struct fs_struct *fs) 19791da177e4SLinus Torvalds { 19806b3286edSKirill Korotaev struct mnt_namespace *new_ns; 19817f2da1e7SAl Viro struct vfsmount *rootmnt = NULL, *pwdmnt = NULL; 19821da177e4SLinus Torvalds struct vfsmount *p, *q; 19831da177e4SLinus Torvalds 1984cf8d2c11STrond Myklebust new_ns = alloc_mnt_ns(); 1985cf8d2c11STrond Myklebust if (IS_ERR(new_ns)) 1986cf8d2c11STrond Myklebust return new_ns; 19871da177e4SLinus Torvalds 1988390c6843SRam Pai down_write(&namespace_sem); 19891da177e4SLinus Torvalds /* First pass: copy the tree topology */ 19906b3286edSKirill Korotaev new_ns->root = copy_tree(mnt_ns->root, mnt_ns->root->mnt_root, 19919676f0c6SRam Pai CL_COPY_ALL | CL_EXPIRE); 19921da177e4SLinus Torvalds if (!new_ns->root) { 1993390c6843SRam Pai up_write(&namespace_sem); 19941da177e4SLinus Torvalds kfree(new_ns); 19955cc4a034SJulia Lawall return ERR_PTR(-ENOMEM); 19961da177e4SLinus Torvalds } 19971da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 19981da177e4SLinus Torvalds list_add_tail(&new_ns->list, &new_ns->root->mnt_list); 19991da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 20001da177e4SLinus Torvalds 20011da177e4SLinus Torvalds /* 20021da177e4SLinus Torvalds * Second pass: switch the tsk->fs->* elements and mark new vfsmounts 20031da177e4SLinus Torvalds * as belonging to new namespace. We have already acquired a private 20041da177e4SLinus Torvalds * fs_struct, so tsk->fs->lock is not needed. 20051da177e4SLinus Torvalds */ 20066b3286edSKirill Korotaev p = mnt_ns->root; 20071da177e4SLinus Torvalds q = new_ns->root; 20081da177e4SLinus Torvalds while (p) { 20096b3286edSKirill Korotaev q->mnt_ns = new_ns; 20101da177e4SLinus Torvalds if (fs) { 20116ac08c39SJan Blunck if (p == fs->root.mnt) { 20121da177e4SLinus Torvalds rootmnt = p; 20136ac08c39SJan Blunck fs->root.mnt = mntget(q); 20141da177e4SLinus Torvalds } 20156ac08c39SJan Blunck if (p == fs->pwd.mnt) { 20161da177e4SLinus Torvalds pwdmnt = p; 20176ac08c39SJan Blunck fs->pwd.mnt = mntget(q); 20181da177e4SLinus Torvalds } 20191da177e4SLinus Torvalds } 20206b3286edSKirill Korotaev p = next_mnt(p, mnt_ns->root); 20211da177e4SLinus Torvalds q = next_mnt(q, new_ns->root); 20221da177e4SLinus Torvalds } 2023390c6843SRam Pai up_write(&namespace_sem); 20241da177e4SLinus Torvalds 20251da177e4SLinus Torvalds if (rootmnt) 20261da177e4SLinus Torvalds mntput(rootmnt); 20271da177e4SLinus Torvalds if (pwdmnt) 20281da177e4SLinus Torvalds mntput(pwdmnt); 20291da177e4SLinus Torvalds 2030741a2951SJANAK DESAI return new_ns; 2031741a2951SJANAK DESAI } 2032741a2951SJANAK DESAI 2033213dd266SEric W. Biederman struct mnt_namespace *copy_mnt_ns(unsigned long flags, struct mnt_namespace *ns, 2034e3222c4eSBadari Pulavarty struct fs_struct *new_fs) 2035741a2951SJANAK DESAI { 20366b3286edSKirill Korotaev struct mnt_namespace *new_ns; 2037741a2951SJANAK DESAI 2038e3222c4eSBadari Pulavarty BUG_ON(!ns); 20396b3286edSKirill Korotaev get_mnt_ns(ns); 2040741a2951SJANAK DESAI 2041741a2951SJANAK DESAI if (!(flags & CLONE_NEWNS)) 2042e3222c4eSBadari Pulavarty return ns; 2043741a2951SJANAK DESAI 2044e3222c4eSBadari Pulavarty new_ns = dup_mnt_ns(ns, new_fs); 2045741a2951SJANAK DESAI 20466b3286edSKirill Korotaev put_mnt_ns(ns); 2047e3222c4eSBadari Pulavarty return new_ns; 20481da177e4SLinus Torvalds } 20491da177e4SLinus Torvalds 2050cf8d2c11STrond Myklebust /** 2051cf8d2c11STrond Myklebust * create_mnt_ns - creates a private namespace and adds a root filesystem 2052cf8d2c11STrond Myklebust * @mnt: pointer to the new root filesystem mountpoint 2053cf8d2c11STrond Myklebust */ 2054cf8d2c11STrond Myklebust struct mnt_namespace *create_mnt_ns(struct vfsmount *mnt) 2055cf8d2c11STrond Myklebust { 2056cf8d2c11STrond Myklebust struct mnt_namespace *new_ns; 2057cf8d2c11STrond Myklebust 2058cf8d2c11STrond Myklebust new_ns = alloc_mnt_ns(); 2059cf8d2c11STrond Myklebust if (!IS_ERR(new_ns)) { 2060cf8d2c11STrond Myklebust mnt->mnt_ns = new_ns; 2061cf8d2c11STrond Myklebust new_ns->root = mnt; 2062cf8d2c11STrond Myklebust list_add(&new_ns->list, &new_ns->root->mnt_list); 2063cf8d2c11STrond Myklebust } 2064cf8d2c11STrond Myklebust return new_ns; 2065cf8d2c11STrond Myklebust } 2066cf8d2c11STrond Myklebust EXPORT_SYMBOL(create_mnt_ns); 2067cf8d2c11STrond Myklebust 2068bdc480e3SHeiko Carstens SYSCALL_DEFINE5(mount, char __user *, dev_name, char __user *, dir_name, 2069bdc480e3SHeiko Carstens char __user *, type, unsigned long, flags, void __user *, data) 20701da177e4SLinus Torvalds { 20711da177e4SLinus Torvalds int retval; 20721da177e4SLinus Torvalds unsigned long data_page; 20731da177e4SLinus Torvalds unsigned long type_page; 20741da177e4SLinus Torvalds unsigned long dev_page; 20751da177e4SLinus Torvalds char *dir_page; 20761da177e4SLinus Torvalds 20771da177e4SLinus Torvalds retval = copy_mount_options(type, &type_page); 20781da177e4SLinus Torvalds if (retval < 0) 20791da177e4SLinus Torvalds return retval; 20801da177e4SLinus Torvalds 20811da177e4SLinus Torvalds dir_page = getname(dir_name); 20821da177e4SLinus Torvalds retval = PTR_ERR(dir_page); 20831da177e4SLinus Torvalds if (IS_ERR(dir_page)) 20841da177e4SLinus Torvalds goto out1; 20851da177e4SLinus Torvalds 20861da177e4SLinus Torvalds retval = copy_mount_options(dev_name, &dev_page); 20871da177e4SLinus Torvalds if (retval < 0) 20881da177e4SLinus Torvalds goto out2; 20891da177e4SLinus Torvalds 20901da177e4SLinus Torvalds retval = copy_mount_options(data, &data_page); 20911da177e4SLinus Torvalds if (retval < 0) 20921da177e4SLinus Torvalds goto out3; 20931da177e4SLinus Torvalds 20941da177e4SLinus Torvalds retval = do_mount((char *)dev_page, dir_page, (char *)type_page, 20951da177e4SLinus Torvalds flags, (void *)data_page); 20961da177e4SLinus Torvalds free_page(data_page); 20971da177e4SLinus Torvalds 20981da177e4SLinus Torvalds out3: 20991da177e4SLinus Torvalds free_page(dev_page); 21001da177e4SLinus Torvalds out2: 21011da177e4SLinus Torvalds putname(dir_page); 21021da177e4SLinus Torvalds out1: 21031da177e4SLinus Torvalds free_page(type_page); 21041da177e4SLinus Torvalds return retval; 21051da177e4SLinus Torvalds } 21061da177e4SLinus Torvalds 21071da177e4SLinus Torvalds /* 21081da177e4SLinus Torvalds * pivot_root Semantics: 21091da177e4SLinus Torvalds * Moves the root file system of the current process to the directory put_old, 21101da177e4SLinus Torvalds * makes new_root as the new root file system of the current process, and sets 21111da177e4SLinus Torvalds * root/cwd of all processes which had them on the current root to new_root. 21121da177e4SLinus Torvalds * 21131da177e4SLinus Torvalds * Restrictions: 21141da177e4SLinus Torvalds * The new_root and put_old must be directories, and must not be on the 21151da177e4SLinus Torvalds * same file system as the current process root. The put_old must be 21161da177e4SLinus Torvalds * underneath new_root, i.e. adding a non-zero number of /.. to the string 21171da177e4SLinus Torvalds * pointed to by put_old must yield the same directory as new_root. No other 21181da177e4SLinus Torvalds * file system may be mounted on put_old. After all, new_root is a mountpoint. 21191da177e4SLinus Torvalds * 21204a0d11faSNeil Brown * Also, the current root cannot be on the 'rootfs' (initial ramfs) filesystem. 21214a0d11faSNeil Brown * See Documentation/filesystems/ramfs-rootfs-initramfs.txt for alternatives 21224a0d11faSNeil Brown * in this situation. 21234a0d11faSNeil Brown * 21241da177e4SLinus Torvalds * Notes: 21251da177e4SLinus Torvalds * - we don't move root/cwd if they are not at the root (reason: if something 21261da177e4SLinus Torvalds * cared enough to change them, it's probably wrong to force them elsewhere) 21271da177e4SLinus Torvalds * - it's okay to pick a root that isn't the root of a file system, e.g. 21281da177e4SLinus Torvalds * /nfs/my_root where /nfs is the mount point. It must be a mountpoint, 21291da177e4SLinus Torvalds * though, so you may need to say mount --bind /nfs/my_root /nfs/my_root 21301da177e4SLinus Torvalds * first. 21311da177e4SLinus Torvalds */ 21323480b257SHeiko Carstens SYSCALL_DEFINE2(pivot_root, const char __user *, new_root, 21333480b257SHeiko Carstens const char __user *, put_old) 21341da177e4SLinus Torvalds { 21351da177e4SLinus Torvalds struct vfsmount *tmp; 21362d8f3038SAl Viro struct path new, old, parent_path, root_parent, root; 21371da177e4SLinus Torvalds int error; 21381da177e4SLinus Torvalds 21391da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 21401da177e4SLinus Torvalds return -EPERM; 21411da177e4SLinus Torvalds 21422d8f3038SAl Viro error = user_path_dir(new_root, &new); 21431da177e4SLinus Torvalds if (error) 21441da177e4SLinus Torvalds goto out0; 21451da177e4SLinus Torvalds error = -EINVAL; 21462d8f3038SAl Viro if (!check_mnt(new.mnt)) 21471da177e4SLinus Torvalds goto out1; 21481da177e4SLinus Torvalds 21492d8f3038SAl Viro error = user_path_dir(put_old, &old); 21501da177e4SLinus Torvalds if (error) 21511da177e4SLinus Torvalds goto out1; 21521da177e4SLinus Torvalds 21532d8f3038SAl Viro error = security_sb_pivotroot(&old, &new); 21541da177e4SLinus Torvalds if (error) { 21552d8f3038SAl Viro path_put(&old); 21561da177e4SLinus Torvalds goto out1; 21571da177e4SLinus Torvalds } 21581da177e4SLinus Torvalds 21591da177e4SLinus Torvalds read_lock(¤t->fs->lock); 21608c3ee42eSAl Viro root = current->fs->root; 21616ac08c39SJan Blunck path_get(¤t->fs->root); 21621da177e4SLinus Torvalds read_unlock(¤t->fs->lock); 2163390c6843SRam Pai down_write(&namespace_sem); 21642d8f3038SAl Viro mutex_lock(&old.dentry->d_inode->i_mutex); 21651da177e4SLinus Torvalds error = -EINVAL; 21662d8f3038SAl Viro if (IS_MNT_SHARED(old.mnt) || 21672d8f3038SAl Viro IS_MNT_SHARED(new.mnt->mnt_parent) || 21688c3ee42eSAl Viro IS_MNT_SHARED(root.mnt->mnt_parent)) 216921444403SRam Pai goto out2; 21708c3ee42eSAl Viro if (!check_mnt(root.mnt)) 21711da177e4SLinus Torvalds goto out2; 21721da177e4SLinus Torvalds error = -ENOENT; 21732d8f3038SAl Viro if (IS_DEADDIR(new.dentry->d_inode)) 21741da177e4SLinus Torvalds goto out2; 2175f3da392eSAlexey Dobriyan if (d_unlinked(new.dentry)) 21761da177e4SLinus Torvalds goto out2; 2177f3da392eSAlexey Dobriyan if (d_unlinked(old.dentry)) 21781da177e4SLinus Torvalds goto out2; 21791da177e4SLinus Torvalds error = -EBUSY; 21802d8f3038SAl Viro if (new.mnt == root.mnt || 21812d8f3038SAl Viro old.mnt == root.mnt) 21821da177e4SLinus Torvalds goto out2; /* loop, on the same file system */ 21831da177e4SLinus Torvalds error = -EINVAL; 21848c3ee42eSAl Viro if (root.mnt->mnt_root != root.dentry) 21851da177e4SLinus Torvalds goto out2; /* not a mountpoint */ 21868c3ee42eSAl Viro if (root.mnt->mnt_parent == root.mnt) 21870bb6fcc1SMiklos Szeredi goto out2; /* not attached */ 21882d8f3038SAl Viro if (new.mnt->mnt_root != new.dentry) 21891da177e4SLinus Torvalds goto out2; /* not a mountpoint */ 21902d8f3038SAl Viro if (new.mnt->mnt_parent == new.mnt) 21910bb6fcc1SMiklos Szeredi goto out2; /* not attached */ 21924ac91378SJan Blunck /* make sure we can reach put_old from new_root */ 21932d8f3038SAl Viro tmp = old.mnt; 21941da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 21952d8f3038SAl Viro if (tmp != new.mnt) { 21961da177e4SLinus Torvalds for (;;) { 21971da177e4SLinus Torvalds if (tmp->mnt_parent == tmp) 21981da177e4SLinus Torvalds goto out3; /* already mounted on put_old */ 21992d8f3038SAl Viro if (tmp->mnt_parent == new.mnt) 22001da177e4SLinus Torvalds break; 22011da177e4SLinus Torvalds tmp = tmp->mnt_parent; 22021da177e4SLinus Torvalds } 22032d8f3038SAl Viro if (!is_subdir(tmp->mnt_mountpoint, new.dentry)) 22041da177e4SLinus Torvalds goto out3; 22052d8f3038SAl Viro } else if (!is_subdir(old.dentry, new.dentry)) 22061da177e4SLinus Torvalds goto out3; 22072d8f3038SAl Viro detach_mnt(new.mnt, &parent_path); 22088c3ee42eSAl Viro detach_mnt(root.mnt, &root_parent); 22094ac91378SJan Blunck /* mount old root on put_old */ 22102d8f3038SAl Viro attach_mnt(root.mnt, &old); 22114ac91378SJan Blunck /* mount new_root on / */ 22122d8f3038SAl Viro attach_mnt(new.mnt, &root_parent); 22136b3286edSKirill Korotaev touch_mnt_namespace(current->nsproxy->mnt_ns); 22141da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 22152d8f3038SAl Viro chroot_fs_refs(&root, &new); 22162d8f3038SAl Viro security_sb_post_pivotroot(&root, &new); 22171da177e4SLinus Torvalds error = 0; 22181a390689SAl Viro path_put(&root_parent); 22191a390689SAl Viro path_put(&parent_path); 22201da177e4SLinus Torvalds out2: 22212d8f3038SAl Viro mutex_unlock(&old.dentry->d_inode->i_mutex); 2222390c6843SRam Pai up_write(&namespace_sem); 22238c3ee42eSAl Viro path_put(&root); 22242d8f3038SAl Viro path_put(&old); 22251da177e4SLinus Torvalds out1: 22262d8f3038SAl Viro path_put(&new); 22271da177e4SLinus Torvalds out0: 22281da177e4SLinus Torvalds return error; 22291da177e4SLinus Torvalds out3: 22301da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 22311da177e4SLinus Torvalds goto out2; 22321da177e4SLinus Torvalds } 22331da177e4SLinus Torvalds 22341da177e4SLinus Torvalds static void __init init_mount_tree(void) 22351da177e4SLinus Torvalds { 22361da177e4SLinus Torvalds struct vfsmount *mnt; 22376b3286edSKirill Korotaev struct mnt_namespace *ns; 2238ac748a09SJan Blunck struct path root; 22391da177e4SLinus Torvalds 22401da177e4SLinus Torvalds mnt = do_kern_mount("rootfs", 0, "rootfs", NULL); 22411da177e4SLinus Torvalds if (IS_ERR(mnt)) 22421da177e4SLinus Torvalds panic("Can't create rootfs"); 22433b22edc5STrond Myklebust ns = create_mnt_ns(mnt); 22443b22edc5STrond Myklebust if (IS_ERR(ns)) 22451da177e4SLinus Torvalds panic("Can't allocate initial namespace"); 22461da177e4SLinus Torvalds 22476b3286edSKirill Korotaev init_task.nsproxy->mnt_ns = ns; 22486b3286edSKirill Korotaev get_mnt_ns(ns); 22491da177e4SLinus Torvalds 2250ac748a09SJan Blunck root.mnt = ns->root; 2251ac748a09SJan Blunck root.dentry = ns->root->mnt_root; 2252ac748a09SJan Blunck 2253ac748a09SJan Blunck set_fs_pwd(current->fs, &root); 2254ac748a09SJan Blunck set_fs_root(current->fs, &root); 22551da177e4SLinus Torvalds } 22561da177e4SLinus Torvalds 225774bf17cfSDenis Cheng void __init mnt_init(void) 22581da177e4SLinus Torvalds { 225913f14b4dSEric Dumazet unsigned u; 226015a67dd8SRandy Dunlap int err; 22611da177e4SLinus Torvalds 2262390c6843SRam Pai init_rwsem(&namespace_sem); 2263390c6843SRam Pai 22641da177e4SLinus Torvalds mnt_cache = kmem_cache_create("mnt_cache", sizeof(struct vfsmount), 226520c2df83SPaul Mundt 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC, NULL); 22661da177e4SLinus Torvalds 2267b58fed8bSRam Pai mount_hashtable = (struct list_head *)__get_free_page(GFP_ATOMIC); 22681da177e4SLinus Torvalds 22691da177e4SLinus Torvalds if (!mount_hashtable) 22701da177e4SLinus Torvalds panic("Failed to allocate mount hash table\n"); 22711da177e4SLinus Torvalds 227213f14b4dSEric Dumazet printk("Mount-cache hash table entries: %lu\n", HASH_SIZE); 22731da177e4SLinus Torvalds 227413f14b4dSEric Dumazet for (u = 0; u < HASH_SIZE; u++) 227513f14b4dSEric Dumazet INIT_LIST_HEAD(&mount_hashtable[u]); 22761da177e4SLinus Torvalds 227715a67dd8SRandy Dunlap err = sysfs_init(); 227815a67dd8SRandy Dunlap if (err) 227915a67dd8SRandy Dunlap printk(KERN_WARNING "%s: sysfs_init error: %d\n", 22808e24eea7SHarvey Harrison __func__, err); 228100d26666SGreg Kroah-Hartman fs_kobj = kobject_create_and_add("fs", NULL); 228200d26666SGreg Kroah-Hartman if (!fs_kobj) 22838e24eea7SHarvey Harrison printk(KERN_WARNING "%s: kobj create error\n", __func__); 22841da177e4SLinus Torvalds init_rootfs(); 22851da177e4SLinus Torvalds init_mount_tree(); 22861da177e4SLinus Torvalds } 22871da177e4SLinus Torvalds 2288616511d0STrond Myklebust void put_mnt_ns(struct mnt_namespace *ns) 22891da177e4SLinus Torvalds { 2290616511d0STrond Myklebust struct vfsmount *root; 229170fbcdf4SRam Pai LIST_HEAD(umount_list); 2292616511d0STrond Myklebust 2293616511d0STrond Myklebust if (!atomic_dec_and_lock(&ns->count, &vfsmount_lock)) 2294616511d0STrond Myklebust return; 2295616511d0STrond Myklebust root = ns->root; 22966b3286edSKirill Korotaev ns->root = NULL; 22971ce88cf4SMiklos Szeredi spin_unlock(&vfsmount_lock); 2298390c6843SRam Pai down_write(&namespace_sem); 22991da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 2300a05964f3SRam Pai umount_tree(root, 0, &umount_list); 23011da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 2302390c6843SRam Pai up_write(&namespace_sem); 230370fbcdf4SRam Pai release_mounts(&umount_list); 23046b3286edSKirill Korotaev kfree(ns); 23051da177e4SLinus Torvalds } 2306cf8d2c11STrond Myklebust EXPORT_SYMBOL(put_mnt_ns); 2307