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); 455addc5ddSAl Viro 46fa3536ccSEric Dumazet static struct list_head *mount_hashtable __read_mostly; 47e18b890bSChristoph Lameter static struct kmem_cache *mnt_cache __read_mostly; 48390c6843SRam Pai static struct rw_semaphore namespace_sem; 491da177e4SLinus Torvalds 50f87fd4c2SMiklos Szeredi /* /sys/fs */ 5100d26666SGreg Kroah-Hartman struct kobject *fs_kobj; 5200d26666SGreg Kroah-Hartman EXPORT_SYMBOL_GPL(fs_kobj); 53f87fd4c2SMiklos Szeredi 541da177e4SLinus Torvalds static inline unsigned long hash(struct vfsmount *mnt, struct dentry *dentry) 551da177e4SLinus Torvalds { 561da177e4SLinus Torvalds unsigned long tmp = ((unsigned long)mnt / L1_CACHE_BYTES); 571da177e4SLinus Torvalds tmp += ((unsigned long)dentry / L1_CACHE_BYTES); 5813f14b4dSEric Dumazet tmp = tmp + (tmp >> HASH_SHIFT); 5913f14b4dSEric Dumazet return tmp & (HASH_SIZE - 1); 601da177e4SLinus Torvalds } 611da177e4SLinus Torvalds 623d733633SDave Hansen #define MNT_WRITER_UNDERFLOW_LIMIT -(1<<16) 633d733633SDave Hansen 6473cd49ecSMiklos Szeredi /* allocation is serialized by namespace_sem */ 6573cd49ecSMiklos Szeredi static int mnt_alloc_id(struct vfsmount *mnt) 6673cd49ecSMiklos Szeredi { 6773cd49ecSMiklos Szeredi int res; 6873cd49ecSMiklos Szeredi 6973cd49ecSMiklos Szeredi retry: 7073cd49ecSMiklos Szeredi ida_pre_get(&mnt_id_ida, GFP_KERNEL); 7173cd49ecSMiklos Szeredi spin_lock(&vfsmount_lock); 7273cd49ecSMiklos Szeredi res = ida_get_new(&mnt_id_ida, &mnt->mnt_id); 7373cd49ecSMiklos Szeredi spin_unlock(&vfsmount_lock); 7473cd49ecSMiklos Szeredi if (res == -EAGAIN) 7573cd49ecSMiklos Szeredi goto retry; 7673cd49ecSMiklos Szeredi 7773cd49ecSMiklos Szeredi return res; 7873cd49ecSMiklos Szeredi } 7973cd49ecSMiklos Szeredi 8073cd49ecSMiklos Szeredi static void mnt_free_id(struct vfsmount *mnt) 8173cd49ecSMiklos Szeredi { 8273cd49ecSMiklos Szeredi spin_lock(&vfsmount_lock); 8373cd49ecSMiklos Szeredi ida_remove(&mnt_id_ida, mnt->mnt_id); 8473cd49ecSMiklos Szeredi spin_unlock(&vfsmount_lock); 8573cd49ecSMiklos Szeredi } 8673cd49ecSMiklos Szeredi 87719f5d7fSMiklos Szeredi /* 88719f5d7fSMiklos Szeredi * Allocate a new peer group ID 89719f5d7fSMiklos Szeredi * 90719f5d7fSMiklos Szeredi * mnt_group_ida is protected by namespace_sem 91719f5d7fSMiklos Szeredi */ 92719f5d7fSMiklos Szeredi static int mnt_alloc_group_id(struct vfsmount *mnt) 93719f5d7fSMiklos Szeredi { 94719f5d7fSMiklos Szeredi if (!ida_pre_get(&mnt_group_ida, GFP_KERNEL)) 95719f5d7fSMiklos Szeredi return -ENOMEM; 96719f5d7fSMiklos Szeredi 97719f5d7fSMiklos Szeredi return ida_get_new_above(&mnt_group_ida, 1, &mnt->mnt_group_id); 98719f5d7fSMiklos Szeredi } 99719f5d7fSMiklos Szeredi 100719f5d7fSMiklos Szeredi /* 101719f5d7fSMiklos Szeredi * Release a peer group ID 102719f5d7fSMiklos Szeredi */ 103719f5d7fSMiklos Szeredi void mnt_release_group_id(struct vfsmount *mnt) 104719f5d7fSMiklos Szeredi { 105719f5d7fSMiklos Szeredi ida_remove(&mnt_group_ida, mnt->mnt_group_id); 106719f5d7fSMiklos Szeredi mnt->mnt_group_id = 0; 107719f5d7fSMiklos Szeredi } 108719f5d7fSMiklos Szeredi 1091da177e4SLinus Torvalds struct vfsmount *alloc_vfsmnt(const char *name) 1101da177e4SLinus Torvalds { 111c3762229SRobert P. J. Day struct vfsmount *mnt = kmem_cache_zalloc(mnt_cache, GFP_KERNEL); 1121da177e4SLinus Torvalds if (mnt) { 11373cd49ecSMiklos Szeredi int err; 11473cd49ecSMiklos Szeredi 11573cd49ecSMiklos Szeredi err = mnt_alloc_id(mnt); 11688b38782SLi Zefan if (err) 11788b38782SLi Zefan goto out_free_cache; 11888b38782SLi Zefan 11988b38782SLi Zefan if (name) { 12088b38782SLi Zefan mnt->mnt_devname = kstrdup(name, GFP_KERNEL); 12188b38782SLi Zefan if (!mnt->mnt_devname) 12288b38782SLi Zefan goto out_free_id; 12373cd49ecSMiklos Szeredi } 12473cd49ecSMiklos Szeredi 1251da177e4SLinus Torvalds atomic_set(&mnt->mnt_count, 1); 1261da177e4SLinus Torvalds INIT_LIST_HEAD(&mnt->mnt_hash); 1271da177e4SLinus Torvalds INIT_LIST_HEAD(&mnt->mnt_child); 1281da177e4SLinus Torvalds INIT_LIST_HEAD(&mnt->mnt_mounts); 1291da177e4SLinus Torvalds INIT_LIST_HEAD(&mnt->mnt_list); 13055e700b9SMiklos Szeredi INIT_LIST_HEAD(&mnt->mnt_expire); 13103e06e68SRam Pai INIT_LIST_HEAD(&mnt->mnt_share); 132a58b0eb8SRam Pai INIT_LIST_HEAD(&mnt->mnt_slave_list); 133a58b0eb8SRam Pai INIT_LIST_HEAD(&mnt->mnt_slave); 134d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 135d3ef3d73Snpiggin@suse.de mnt->mnt_writers = alloc_percpu(int); 136d3ef3d73Snpiggin@suse.de if (!mnt->mnt_writers) 137d3ef3d73Snpiggin@suse.de goto out_free_devname; 138d3ef3d73Snpiggin@suse.de #else 139d3ef3d73Snpiggin@suse.de mnt->mnt_writers = 0; 140d3ef3d73Snpiggin@suse.de #endif 1411da177e4SLinus Torvalds } 1421da177e4SLinus Torvalds return mnt; 14388b38782SLi Zefan 144d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 145d3ef3d73Snpiggin@suse.de out_free_devname: 146d3ef3d73Snpiggin@suse.de kfree(mnt->mnt_devname); 147d3ef3d73Snpiggin@suse.de #endif 14888b38782SLi Zefan out_free_id: 14988b38782SLi Zefan mnt_free_id(mnt); 15088b38782SLi Zefan out_free_cache: 15188b38782SLi Zefan kmem_cache_free(mnt_cache, mnt); 15288b38782SLi Zefan return NULL; 1531da177e4SLinus Torvalds } 1541da177e4SLinus Torvalds 1558366025eSDave Hansen /* 1568366025eSDave Hansen * Most r/o checks on a fs are for operations that take 1578366025eSDave Hansen * discrete amounts of time, like a write() or unlink(). 1588366025eSDave Hansen * We must keep track of when those operations start 1598366025eSDave Hansen * (for permission checks) and when they end, so that 1608366025eSDave Hansen * we can determine when writes are able to occur to 1618366025eSDave Hansen * a filesystem. 1628366025eSDave Hansen */ 1633d733633SDave Hansen /* 1643d733633SDave Hansen * __mnt_is_readonly: check whether a mount is read-only 1653d733633SDave Hansen * @mnt: the mount to check for its write status 1663d733633SDave Hansen * 1673d733633SDave Hansen * This shouldn't be used directly ouside of the VFS. 1683d733633SDave Hansen * It does not guarantee that the filesystem will stay 1693d733633SDave Hansen * r/w, just that it is right *now*. This can not and 1703d733633SDave Hansen * should not be used in place of IS_RDONLY(inode). 1713d733633SDave Hansen * mnt_want/drop_write() will _keep_ the filesystem 1723d733633SDave Hansen * r/w. 1733d733633SDave Hansen */ 1743d733633SDave Hansen int __mnt_is_readonly(struct vfsmount *mnt) 1753d733633SDave Hansen { 1762e4b7fcdSDave Hansen if (mnt->mnt_flags & MNT_READONLY) 1772e4b7fcdSDave Hansen return 1; 1782e4b7fcdSDave Hansen if (mnt->mnt_sb->s_flags & MS_RDONLY) 1792e4b7fcdSDave Hansen return 1; 1802e4b7fcdSDave Hansen return 0; 1813d733633SDave Hansen } 1823d733633SDave Hansen EXPORT_SYMBOL_GPL(__mnt_is_readonly); 1833d733633SDave Hansen 184d3ef3d73Snpiggin@suse.de static inline void inc_mnt_writers(struct vfsmount *mnt) 1853d733633SDave Hansen { 186d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 187d3ef3d73Snpiggin@suse.de (*per_cpu_ptr(mnt->mnt_writers, smp_processor_id()))++; 188d3ef3d73Snpiggin@suse.de #else 189d3ef3d73Snpiggin@suse.de mnt->mnt_writers++; 190d3ef3d73Snpiggin@suse.de #endif 1913d733633SDave Hansen } 1923d733633SDave Hansen 193d3ef3d73Snpiggin@suse.de static inline void dec_mnt_writers(struct vfsmount *mnt) 1943d733633SDave Hansen { 195d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 196d3ef3d73Snpiggin@suse.de (*per_cpu_ptr(mnt->mnt_writers, smp_processor_id()))--; 197d3ef3d73Snpiggin@suse.de #else 198d3ef3d73Snpiggin@suse.de mnt->mnt_writers--; 199d3ef3d73Snpiggin@suse.de #endif 200d3ef3d73Snpiggin@suse.de } 201d3ef3d73Snpiggin@suse.de 202d3ef3d73Snpiggin@suse.de static unsigned int count_mnt_writers(struct vfsmount *mnt) 203d3ef3d73Snpiggin@suse.de { 204d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 205d3ef3d73Snpiggin@suse.de unsigned int count = 0; 2063d733633SDave Hansen int cpu; 2073d733633SDave Hansen 2083d733633SDave Hansen for_each_possible_cpu(cpu) { 209d3ef3d73Snpiggin@suse.de count += *per_cpu_ptr(mnt->mnt_writers, cpu); 2103d733633SDave Hansen } 2113d733633SDave Hansen 212d3ef3d73Snpiggin@suse.de return count; 213d3ef3d73Snpiggin@suse.de #else 214d3ef3d73Snpiggin@suse.de return mnt->mnt_writers; 215d3ef3d73Snpiggin@suse.de #endif 2163d733633SDave Hansen } 2173d733633SDave Hansen 2183d733633SDave Hansen /* 2193d733633SDave Hansen * Most r/o checks on a fs are for operations that take 2203d733633SDave Hansen * discrete amounts of time, like a write() or unlink(). 2213d733633SDave Hansen * We must keep track of when those operations start 2223d733633SDave Hansen * (for permission checks) and when they end, so that 2233d733633SDave Hansen * we can determine when writes are able to occur to 2243d733633SDave Hansen * a filesystem. 2253d733633SDave Hansen */ 2268366025eSDave Hansen /** 2278366025eSDave Hansen * mnt_want_write - get write access to a mount 2288366025eSDave Hansen * @mnt: the mount on which to take a write 2298366025eSDave Hansen * 2308366025eSDave Hansen * This tells the low-level filesystem that a write is 2318366025eSDave Hansen * about to be performed to it, and makes sure that 2328366025eSDave Hansen * writes are allowed before returning success. When 2338366025eSDave Hansen * the write operation is finished, mnt_drop_write() 2348366025eSDave Hansen * must be called. This is effectively a refcount. 2358366025eSDave Hansen */ 2368366025eSDave Hansen int mnt_want_write(struct vfsmount *mnt) 2378366025eSDave Hansen { 2383d733633SDave Hansen int ret = 0; 2393d733633SDave Hansen 240d3ef3d73Snpiggin@suse.de preempt_disable(); 241d3ef3d73Snpiggin@suse.de inc_mnt_writers(mnt); 242d3ef3d73Snpiggin@suse.de /* 243d3ef3d73Snpiggin@suse.de * The store to inc_mnt_writers must be visible before we pass 244d3ef3d73Snpiggin@suse.de * MNT_WRITE_HOLD loop below, so that the slowpath can see our 245d3ef3d73Snpiggin@suse.de * incremented count after it has set MNT_WRITE_HOLD. 246d3ef3d73Snpiggin@suse.de */ 247d3ef3d73Snpiggin@suse.de smp_mb(); 248d3ef3d73Snpiggin@suse.de while (mnt->mnt_flags & MNT_WRITE_HOLD) 249d3ef3d73Snpiggin@suse.de cpu_relax(); 250d3ef3d73Snpiggin@suse.de /* 251d3ef3d73Snpiggin@suse.de * After the slowpath clears MNT_WRITE_HOLD, mnt_is_readonly will 252d3ef3d73Snpiggin@suse.de * be set to match its requirements. So we must not load that until 253d3ef3d73Snpiggin@suse.de * MNT_WRITE_HOLD is cleared. 254d3ef3d73Snpiggin@suse.de */ 255d3ef3d73Snpiggin@suse.de smp_rmb(); 2563d733633SDave Hansen if (__mnt_is_readonly(mnt)) { 257d3ef3d73Snpiggin@suse.de dec_mnt_writers(mnt); 2583d733633SDave Hansen ret = -EROFS; 2593d733633SDave Hansen goto out; 2603d733633SDave Hansen } 2613d733633SDave Hansen out: 262d3ef3d73Snpiggin@suse.de preempt_enable(); 2633d733633SDave Hansen return ret; 2648366025eSDave Hansen } 2658366025eSDave Hansen EXPORT_SYMBOL_GPL(mnt_want_write); 2668366025eSDave Hansen 2678366025eSDave Hansen /** 26896029c4eSnpiggin@suse.de * mnt_clone_write - get write access to a mount 26996029c4eSnpiggin@suse.de * @mnt: the mount on which to take a write 27096029c4eSnpiggin@suse.de * 27196029c4eSnpiggin@suse.de * This is effectively like mnt_want_write, except 27296029c4eSnpiggin@suse.de * it must only be used to take an extra write reference 27396029c4eSnpiggin@suse.de * on a mountpoint that we already know has a write reference 27496029c4eSnpiggin@suse.de * on it. This allows some optimisation. 27596029c4eSnpiggin@suse.de * 27696029c4eSnpiggin@suse.de * After finished, mnt_drop_write must be called as usual to 27796029c4eSnpiggin@suse.de * drop the reference. 27896029c4eSnpiggin@suse.de */ 27996029c4eSnpiggin@suse.de int mnt_clone_write(struct vfsmount *mnt) 28096029c4eSnpiggin@suse.de { 28196029c4eSnpiggin@suse.de /* superblock may be r/o */ 28296029c4eSnpiggin@suse.de if (__mnt_is_readonly(mnt)) 28396029c4eSnpiggin@suse.de return -EROFS; 28496029c4eSnpiggin@suse.de preempt_disable(); 28596029c4eSnpiggin@suse.de inc_mnt_writers(mnt); 28696029c4eSnpiggin@suse.de preempt_enable(); 28796029c4eSnpiggin@suse.de return 0; 28896029c4eSnpiggin@suse.de } 28996029c4eSnpiggin@suse.de EXPORT_SYMBOL_GPL(mnt_clone_write); 29096029c4eSnpiggin@suse.de 29196029c4eSnpiggin@suse.de /** 29296029c4eSnpiggin@suse.de * mnt_want_write_file - get write access to a file's mount 29396029c4eSnpiggin@suse.de * @file: the file who's mount on which to take a write 29496029c4eSnpiggin@suse.de * 29596029c4eSnpiggin@suse.de * This is like mnt_want_write, but it takes a file and can 29696029c4eSnpiggin@suse.de * do some optimisations if the file is open for write already 29796029c4eSnpiggin@suse.de */ 29896029c4eSnpiggin@suse.de int mnt_want_write_file(struct file *file) 29996029c4eSnpiggin@suse.de { 30096029c4eSnpiggin@suse.de if (!(file->f_mode & FMODE_WRITE)) 30196029c4eSnpiggin@suse.de return mnt_want_write(file->f_path.mnt); 30296029c4eSnpiggin@suse.de else 30396029c4eSnpiggin@suse.de return mnt_clone_write(file->f_path.mnt); 30496029c4eSnpiggin@suse.de } 30596029c4eSnpiggin@suse.de EXPORT_SYMBOL_GPL(mnt_want_write_file); 30696029c4eSnpiggin@suse.de 30796029c4eSnpiggin@suse.de /** 3088366025eSDave Hansen * mnt_drop_write - give up write access to a mount 3098366025eSDave Hansen * @mnt: the mount on which to give up write access 3108366025eSDave Hansen * 3118366025eSDave Hansen * Tells the low-level filesystem that we are done 3128366025eSDave Hansen * performing writes to it. Must be matched with 3138366025eSDave Hansen * mnt_want_write() call above. 3148366025eSDave Hansen */ 3158366025eSDave Hansen void mnt_drop_write(struct vfsmount *mnt) 3168366025eSDave Hansen { 317d3ef3d73Snpiggin@suse.de preempt_disable(); 318d3ef3d73Snpiggin@suse.de dec_mnt_writers(mnt); 319d3ef3d73Snpiggin@suse.de preempt_enable(); 3208366025eSDave Hansen } 3218366025eSDave Hansen EXPORT_SYMBOL_GPL(mnt_drop_write); 3228366025eSDave Hansen 3232e4b7fcdSDave Hansen static int mnt_make_readonly(struct vfsmount *mnt) 3248366025eSDave Hansen { 3253d733633SDave Hansen int ret = 0; 3263d733633SDave Hansen 3272e4b7fcdSDave Hansen spin_lock(&vfsmount_lock); 328d3ef3d73Snpiggin@suse.de mnt->mnt_flags |= MNT_WRITE_HOLD; 329d3ef3d73Snpiggin@suse.de /* 330d3ef3d73Snpiggin@suse.de * After storing MNT_WRITE_HOLD, we'll read the counters. This store 331d3ef3d73Snpiggin@suse.de * should be visible before we do. 332d3ef3d73Snpiggin@suse.de */ 333d3ef3d73Snpiggin@suse.de smp_mb(); 334d3ef3d73Snpiggin@suse.de 335d3ef3d73Snpiggin@suse.de /* 336d3ef3d73Snpiggin@suse.de * With writers on hold, if this value is zero, then there are 337d3ef3d73Snpiggin@suse.de * definitely no active writers (although held writers may subsequently 338d3ef3d73Snpiggin@suse.de * increment the count, they'll have to wait, and decrement it after 339d3ef3d73Snpiggin@suse.de * seeing MNT_READONLY). 340d3ef3d73Snpiggin@suse.de * 341d3ef3d73Snpiggin@suse.de * It is OK to have counter incremented on one CPU and decremented on 342d3ef3d73Snpiggin@suse.de * another: the sum will add up correctly. The danger would be when we 343d3ef3d73Snpiggin@suse.de * sum up each counter, if we read a counter before it is incremented, 344d3ef3d73Snpiggin@suse.de * but then read another CPU's count which it has been subsequently 345d3ef3d73Snpiggin@suse.de * decremented from -- we would see more decrements than we should. 346d3ef3d73Snpiggin@suse.de * MNT_WRITE_HOLD protects against this scenario, because 347d3ef3d73Snpiggin@suse.de * mnt_want_write first increments count, then smp_mb, then spins on 348d3ef3d73Snpiggin@suse.de * MNT_WRITE_HOLD, so it can't be decremented by another CPU while 349d3ef3d73Snpiggin@suse.de * we're counting up here. 350d3ef3d73Snpiggin@suse.de */ 351d3ef3d73Snpiggin@suse.de if (count_mnt_writers(mnt) > 0) 352d3ef3d73Snpiggin@suse.de ret = -EBUSY; 353d3ef3d73Snpiggin@suse.de else 3542e4b7fcdSDave Hansen mnt->mnt_flags |= MNT_READONLY; 355d3ef3d73Snpiggin@suse.de /* 356d3ef3d73Snpiggin@suse.de * MNT_READONLY must become visible before ~MNT_WRITE_HOLD, so writers 357d3ef3d73Snpiggin@suse.de * that become unheld will see MNT_READONLY. 358d3ef3d73Snpiggin@suse.de */ 359d3ef3d73Snpiggin@suse.de smp_wmb(); 360d3ef3d73Snpiggin@suse.de mnt->mnt_flags &= ~MNT_WRITE_HOLD; 3612e4b7fcdSDave Hansen spin_unlock(&vfsmount_lock); 3623d733633SDave Hansen return ret; 3633d733633SDave Hansen } 3648366025eSDave Hansen 3652e4b7fcdSDave Hansen static void __mnt_unmake_readonly(struct vfsmount *mnt) 3662e4b7fcdSDave Hansen { 3672e4b7fcdSDave Hansen spin_lock(&vfsmount_lock); 3682e4b7fcdSDave Hansen mnt->mnt_flags &= ~MNT_READONLY; 3692e4b7fcdSDave Hansen spin_unlock(&vfsmount_lock); 3702e4b7fcdSDave Hansen } 3712e4b7fcdSDave Hansen 372a3ec947cSSukadev Bhattiprolu void simple_set_mnt(struct vfsmount *mnt, struct super_block *sb) 373454e2398SDavid Howells { 374454e2398SDavid Howells mnt->mnt_sb = sb; 375454e2398SDavid Howells mnt->mnt_root = dget(sb->s_root); 376454e2398SDavid Howells } 377454e2398SDavid Howells 378454e2398SDavid Howells EXPORT_SYMBOL(simple_set_mnt); 379454e2398SDavid Howells 3801da177e4SLinus Torvalds void free_vfsmnt(struct vfsmount *mnt) 3811da177e4SLinus Torvalds { 3821da177e4SLinus Torvalds kfree(mnt->mnt_devname); 38373cd49ecSMiklos Szeredi mnt_free_id(mnt); 384d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 385d3ef3d73Snpiggin@suse.de free_percpu(mnt->mnt_writers); 386d3ef3d73Snpiggin@suse.de #endif 3871da177e4SLinus Torvalds kmem_cache_free(mnt_cache, mnt); 3881da177e4SLinus Torvalds } 3891da177e4SLinus Torvalds 3901da177e4SLinus Torvalds /* 391a05964f3SRam Pai * find the first or last mount at @dentry on vfsmount @mnt depending on 392a05964f3SRam Pai * @dir. If @dir is set return the first mount else return the last mount. 3931da177e4SLinus Torvalds */ 394a05964f3SRam Pai struct vfsmount *__lookup_mnt(struct vfsmount *mnt, struct dentry *dentry, 395a05964f3SRam Pai int dir) 3961da177e4SLinus Torvalds { 3971da177e4SLinus Torvalds struct list_head *head = mount_hashtable + hash(mnt, dentry); 3981da177e4SLinus Torvalds struct list_head *tmp = head; 3991da177e4SLinus Torvalds struct vfsmount *p, *found = NULL; 4001da177e4SLinus Torvalds 4011da177e4SLinus Torvalds for (;;) { 402a05964f3SRam Pai tmp = dir ? tmp->next : tmp->prev; 4031da177e4SLinus Torvalds p = NULL; 4041da177e4SLinus Torvalds if (tmp == head) 4051da177e4SLinus Torvalds break; 4061da177e4SLinus Torvalds p = list_entry(tmp, struct vfsmount, mnt_hash); 4071da177e4SLinus Torvalds if (p->mnt_parent == mnt && p->mnt_mountpoint == dentry) { 408a05964f3SRam Pai found = p; 4091da177e4SLinus Torvalds break; 4101da177e4SLinus Torvalds } 4111da177e4SLinus Torvalds } 4121da177e4SLinus Torvalds return found; 4131da177e4SLinus Torvalds } 4141da177e4SLinus Torvalds 415a05964f3SRam Pai /* 416a05964f3SRam Pai * lookup_mnt increments the ref count before returning 417a05964f3SRam Pai * the vfsmount struct. 418a05964f3SRam Pai */ 4191c755af4SAl Viro struct vfsmount *lookup_mnt(struct path *path) 420a05964f3SRam Pai { 421a05964f3SRam Pai struct vfsmount *child_mnt; 422a05964f3SRam Pai spin_lock(&vfsmount_lock); 4231c755af4SAl Viro if ((child_mnt = __lookup_mnt(path->mnt, path->dentry, 1))) 424a05964f3SRam Pai mntget(child_mnt); 425a05964f3SRam Pai spin_unlock(&vfsmount_lock); 426a05964f3SRam Pai return child_mnt; 427a05964f3SRam Pai } 428a05964f3SRam Pai 4291da177e4SLinus Torvalds static inline int check_mnt(struct vfsmount *mnt) 4301da177e4SLinus Torvalds { 4316b3286edSKirill Korotaev return mnt->mnt_ns == current->nsproxy->mnt_ns; 4321da177e4SLinus Torvalds } 4331da177e4SLinus Torvalds 4346b3286edSKirill Korotaev static void touch_mnt_namespace(struct mnt_namespace *ns) 4355addc5ddSAl Viro { 4365addc5ddSAl Viro if (ns) { 4375addc5ddSAl Viro ns->event = ++event; 4385addc5ddSAl Viro wake_up_interruptible(&ns->poll); 4395addc5ddSAl Viro } 4405addc5ddSAl Viro } 4415addc5ddSAl Viro 4426b3286edSKirill Korotaev static void __touch_mnt_namespace(struct mnt_namespace *ns) 4435addc5ddSAl Viro { 4445addc5ddSAl Viro if (ns && ns->event != event) { 4455addc5ddSAl Viro ns->event = event; 4465addc5ddSAl Viro wake_up_interruptible(&ns->poll); 4475addc5ddSAl Viro } 4485addc5ddSAl Viro } 4495addc5ddSAl Viro 4501a390689SAl Viro static void detach_mnt(struct vfsmount *mnt, struct path *old_path) 4511da177e4SLinus Torvalds { 4521a390689SAl Viro old_path->dentry = mnt->mnt_mountpoint; 4531a390689SAl Viro old_path->mnt = mnt->mnt_parent; 4541da177e4SLinus Torvalds mnt->mnt_parent = mnt; 4551da177e4SLinus Torvalds mnt->mnt_mountpoint = mnt->mnt_root; 4561da177e4SLinus Torvalds list_del_init(&mnt->mnt_child); 4571da177e4SLinus Torvalds list_del_init(&mnt->mnt_hash); 4581a390689SAl Viro old_path->dentry->d_mounted--; 4591da177e4SLinus Torvalds } 4601da177e4SLinus Torvalds 461b90fa9aeSRam Pai void mnt_set_mountpoint(struct vfsmount *mnt, struct dentry *dentry, 462b90fa9aeSRam Pai struct vfsmount *child_mnt) 463b90fa9aeSRam Pai { 464b90fa9aeSRam Pai child_mnt->mnt_parent = mntget(mnt); 465b90fa9aeSRam Pai child_mnt->mnt_mountpoint = dget(dentry); 466b90fa9aeSRam Pai dentry->d_mounted++; 467b90fa9aeSRam Pai } 468b90fa9aeSRam Pai 4691a390689SAl Viro static void attach_mnt(struct vfsmount *mnt, struct path *path) 4701da177e4SLinus Torvalds { 4711a390689SAl Viro mnt_set_mountpoint(path->mnt, path->dentry, mnt); 472b90fa9aeSRam Pai list_add_tail(&mnt->mnt_hash, mount_hashtable + 4731a390689SAl Viro hash(path->mnt, path->dentry)); 4741a390689SAl Viro list_add_tail(&mnt->mnt_child, &path->mnt->mnt_mounts); 475b90fa9aeSRam Pai } 476b90fa9aeSRam Pai 477b90fa9aeSRam Pai /* 478b90fa9aeSRam Pai * the caller must hold vfsmount_lock 479b90fa9aeSRam Pai */ 480b90fa9aeSRam Pai static void commit_tree(struct vfsmount *mnt) 481b90fa9aeSRam Pai { 482b90fa9aeSRam Pai struct vfsmount *parent = mnt->mnt_parent; 483b90fa9aeSRam Pai struct vfsmount *m; 484b90fa9aeSRam Pai LIST_HEAD(head); 4856b3286edSKirill Korotaev struct mnt_namespace *n = parent->mnt_ns; 486b90fa9aeSRam Pai 487b90fa9aeSRam Pai BUG_ON(parent == mnt); 488b90fa9aeSRam Pai 489b90fa9aeSRam Pai list_add_tail(&head, &mnt->mnt_list); 490b90fa9aeSRam Pai list_for_each_entry(m, &head, mnt_list) 4916b3286edSKirill Korotaev m->mnt_ns = n; 492b90fa9aeSRam Pai list_splice(&head, n->list.prev); 493b90fa9aeSRam Pai 494b90fa9aeSRam Pai list_add_tail(&mnt->mnt_hash, mount_hashtable + 495b90fa9aeSRam Pai hash(parent, mnt->mnt_mountpoint)); 496b90fa9aeSRam Pai list_add_tail(&mnt->mnt_child, &parent->mnt_mounts); 4976b3286edSKirill Korotaev touch_mnt_namespace(n); 4981da177e4SLinus Torvalds } 4991da177e4SLinus Torvalds 5001da177e4SLinus Torvalds static struct vfsmount *next_mnt(struct vfsmount *p, struct vfsmount *root) 5011da177e4SLinus Torvalds { 5021da177e4SLinus Torvalds struct list_head *next = p->mnt_mounts.next; 5031da177e4SLinus Torvalds if (next == &p->mnt_mounts) { 5041da177e4SLinus Torvalds while (1) { 5051da177e4SLinus Torvalds if (p == root) 5061da177e4SLinus Torvalds return NULL; 5071da177e4SLinus Torvalds next = p->mnt_child.next; 5081da177e4SLinus Torvalds if (next != &p->mnt_parent->mnt_mounts) 5091da177e4SLinus Torvalds break; 5101da177e4SLinus Torvalds p = p->mnt_parent; 5111da177e4SLinus Torvalds } 5121da177e4SLinus Torvalds } 5131da177e4SLinus Torvalds return list_entry(next, struct vfsmount, mnt_child); 5141da177e4SLinus Torvalds } 5151da177e4SLinus Torvalds 5169676f0c6SRam Pai static struct vfsmount *skip_mnt_tree(struct vfsmount *p) 5179676f0c6SRam Pai { 5189676f0c6SRam Pai struct list_head *prev = p->mnt_mounts.prev; 5199676f0c6SRam Pai while (prev != &p->mnt_mounts) { 5209676f0c6SRam Pai p = list_entry(prev, struct vfsmount, mnt_child); 5219676f0c6SRam Pai prev = p->mnt_mounts.prev; 5229676f0c6SRam Pai } 5239676f0c6SRam Pai return p; 5249676f0c6SRam Pai } 5259676f0c6SRam Pai 52636341f64SRam Pai static struct vfsmount *clone_mnt(struct vfsmount *old, struct dentry *root, 52736341f64SRam Pai int flag) 5281da177e4SLinus Torvalds { 5291da177e4SLinus Torvalds struct super_block *sb = old->mnt_sb; 5301da177e4SLinus Torvalds struct vfsmount *mnt = alloc_vfsmnt(old->mnt_devname); 5311da177e4SLinus Torvalds 5321da177e4SLinus Torvalds if (mnt) { 533719f5d7fSMiklos Szeredi if (flag & (CL_SLAVE | CL_PRIVATE)) 534719f5d7fSMiklos Szeredi mnt->mnt_group_id = 0; /* not a peer of original */ 535719f5d7fSMiklos Szeredi else 536719f5d7fSMiklos Szeredi mnt->mnt_group_id = old->mnt_group_id; 537719f5d7fSMiklos Szeredi 538719f5d7fSMiklos Szeredi if ((flag & CL_MAKE_SHARED) && !mnt->mnt_group_id) { 539719f5d7fSMiklos Szeredi int err = mnt_alloc_group_id(mnt); 540719f5d7fSMiklos Szeredi if (err) 541719f5d7fSMiklos Szeredi goto out_free; 542719f5d7fSMiklos Szeredi } 543719f5d7fSMiklos Szeredi 5441da177e4SLinus Torvalds mnt->mnt_flags = old->mnt_flags; 5451da177e4SLinus Torvalds atomic_inc(&sb->s_active); 5461da177e4SLinus Torvalds mnt->mnt_sb = sb; 5471da177e4SLinus Torvalds mnt->mnt_root = dget(root); 5481da177e4SLinus Torvalds mnt->mnt_mountpoint = mnt->mnt_root; 5491da177e4SLinus Torvalds mnt->mnt_parent = mnt; 550b90fa9aeSRam Pai 5515afe0022SRam Pai if (flag & CL_SLAVE) { 5525afe0022SRam Pai list_add(&mnt->mnt_slave, &old->mnt_slave_list); 5535afe0022SRam Pai mnt->mnt_master = old; 5545afe0022SRam Pai CLEAR_MNT_SHARED(mnt); 5558aec0809SAl Viro } else if (!(flag & CL_PRIVATE)) { 556b90fa9aeSRam Pai if ((flag & CL_PROPAGATION) || IS_MNT_SHARED(old)) 557b90fa9aeSRam Pai list_add(&mnt->mnt_share, &old->mnt_share); 5585afe0022SRam Pai if (IS_MNT_SLAVE(old)) 5595afe0022SRam Pai list_add(&mnt->mnt_slave, &old->mnt_slave); 5605afe0022SRam Pai mnt->mnt_master = old->mnt_master; 5615afe0022SRam Pai } 562b90fa9aeSRam Pai if (flag & CL_MAKE_SHARED) 563b90fa9aeSRam Pai set_mnt_shared(mnt); 5641da177e4SLinus Torvalds 5651da177e4SLinus Torvalds /* stick the duplicate mount on the same expiry list 5661da177e4SLinus Torvalds * as the original if that was on one */ 56736341f64SRam Pai if (flag & CL_EXPIRE) { 56855e700b9SMiklos Szeredi if (!list_empty(&old->mnt_expire)) 56955e700b9SMiklos Szeredi list_add(&mnt->mnt_expire, &old->mnt_expire); 5701da177e4SLinus Torvalds } 57136341f64SRam Pai } 5721da177e4SLinus Torvalds return mnt; 573719f5d7fSMiklos Szeredi 574719f5d7fSMiklos Szeredi out_free: 575719f5d7fSMiklos Szeredi free_vfsmnt(mnt); 576719f5d7fSMiklos Szeredi return NULL; 5771da177e4SLinus Torvalds } 5781da177e4SLinus Torvalds 5797b7b1aceSAl Viro static inline void __mntput(struct vfsmount *mnt) 5801da177e4SLinus Torvalds { 5811da177e4SLinus Torvalds struct super_block *sb = mnt->mnt_sb; 5823d733633SDave Hansen /* 5833d733633SDave Hansen * This probably indicates that somebody messed 5843d733633SDave Hansen * up a mnt_want/drop_write() pair. If this 5853d733633SDave Hansen * happens, the filesystem was probably unable 5863d733633SDave Hansen * to make r/w->r/o transitions. 5873d733633SDave Hansen */ 588d3ef3d73Snpiggin@suse.de /* 589d3ef3d73Snpiggin@suse.de * atomic_dec_and_lock() used to deal with ->mnt_count decrements 590d3ef3d73Snpiggin@suse.de * provides barriers, so count_mnt_writers() below is safe. AV 591d3ef3d73Snpiggin@suse.de */ 592d3ef3d73Snpiggin@suse.de WARN_ON(count_mnt_writers(mnt)); 5931da177e4SLinus Torvalds dput(mnt->mnt_root); 5941da177e4SLinus Torvalds free_vfsmnt(mnt); 5951da177e4SLinus Torvalds deactivate_super(sb); 5961da177e4SLinus Torvalds } 5971da177e4SLinus Torvalds 5987b7b1aceSAl Viro void mntput_no_expire(struct vfsmount *mnt) 5997b7b1aceSAl Viro { 6007b7b1aceSAl Viro repeat: 6017b7b1aceSAl Viro if (atomic_dec_and_lock(&mnt->mnt_count, &vfsmount_lock)) { 6027b7b1aceSAl Viro if (likely(!mnt->mnt_pinned)) { 6037b7b1aceSAl Viro spin_unlock(&vfsmount_lock); 6047b7b1aceSAl Viro __mntput(mnt); 6057b7b1aceSAl Viro return; 6067b7b1aceSAl Viro } 6077b7b1aceSAl Viro atomic_add(mnt->mnt_pinned + 1, &mnt->mnt_count); 6087b7b1aceSAl Viro mnt->mnt_pinned = 0; 6097b7b1aceSAl Viro spin_unlock(&vfsmount_lock); 6107b7b1aceSAl Viro acct_auto_close_mnt(mnt); 6117b7b1aceSAl Viro security_sb_umount_close(mnt); 6127b7b1aceSAl Viro goto repeat; 6137b7b1aceSAl Viro } 6147b7b1aceSAl Viro } 6157b7b1aceSAl Viro 6167b7b1aceSAl Viro EXPORT_SYMBOL(mntput_no_expire); 6177b7b1aceSAl Viro 6187b7b1aceSAl Viro void mnt_pin(struct vfsmount *mnt) 6197b7b1aceSAl Viro { 6207b7b1aceSAl Viro spin_lock(&vfsmount_lock); 6217b7b1aceSAl Viro mnt->mnt_pinned++; 6227b7b1aceSAl Viro spin_unlock(&vfsmount_lock); 6237b7b1aceSAl Viro } 6247b7b1aceSAl Viro 6257b7b1aceSAl Viro EXPORT_SYMBOL(mnt_pin); 6267b7b1aceSAl Viro 6277b7b1aceSAl Viro void mnt_unpin(struct vfsmount *mnt) 6287b7b1aceSAl Viro { 6297b7b1aceSAl Viro spin_lock(&vfsmount_lock); 6307b7b1aceSAl Viro if (mnt->mnt_pinned) { 6317b7b1aceSAl Viro atomic_inc(&mnt->mnt_count); 6327b7b1aceSAl Viro mnt->mnt_pinned--; 6337b7b1aceSAl Viro } 6347b7b1aceSAl Viro spin_unlock(&vfsmount_lock); 6357b7b1aceSAl Viro } 6367b7b1aceSAl Viro 6377b7b1aceSAl Viro EXPORT_SYMBOL(mnt_unpin); 6381da177e4SLinus Torvalds 639b3b304a2SMiklos Szeredi static inline void mangle(struct seq_file *m, const char *s) 640b3b304a2SMiklos Szeredi { 641b3b304a2SMiklos Szeredi seq_escape(m, s, " \t\n\\"); 642b3b304a2SMiklos Szeredi } 643b3b304a2SMiklos Szeredi 644b3b304a2SMiklos Szeredi /* 645b3b304a2SMiklos Szeredi * Simple .show_options callback for filesystems which don't want to 646b3b304a2SMiklos Szeredi * implement more complex mount option showing. 647b3b304a2SMiklos Szeredi * 648b3b304a2SMiklos Szeredi * See also save_mount_options(). 649b3b304a2SMiklos Szeredi */ 650b3b304a2SMiklos Szeredi int generic_show_options(struct seq_file *m, struct vfsmount *mnt) 651b3b304a2SMiklos Szeredi { 6522a32cebdSAl Viro const char *options; 6532a32cebdSAl Viro 6542a32cebdSAl Viro rcu_read_lock(); 6552a32cebdSAl Viro options = rcu_dereference(mnt->mnt_sb->s_options); 656b3b304a2SMiklos Szeredi 657b3b304a2SMiklos Szeredi if (options != NULL && options[0]) { 658b3b304a2SMiklos Szeredi seq_putc(m, ','); 659b3b304a2SMiklos Szeredi mangle(m, options); 660b3b304a2SMiklos Szeredi } 6612a32cebdSAl Viro rcu_read_unlock(); 662b3b304a2SMiklos Szeredi 663b3b304a2SMiklos Szeredi return 0; 664b3b304a2SMiklos Szeredi } 665b3b304a2SMiklos Szeredi EXPORT_SYMBOL(generic_show_options); 666b3b304a2SMiklos Szeredi 667b3b304a2SMiklos Szeredi /* 668b3b304a2SMiklos Szeredi * If filesystem uses generic_show_options(), this function should be 669b3b304a2SMiklos Szeredi * called from the fill_super() callback. 670b3b304a2SMiklos Szeredi * 671b3b304a2SMiklos Szeredi * The .remount_fs callback usually needs to be handled in a special 672b3b304a2SMiklos Szeredi * way, to make sure, that previous options are not overwritten if the 673b3b304a2SMiklos Szeredi * remount fails. 674b3b304a2SMiklos Szeredi * 675b3b304a2SMiklos Szeredi * Also note, that if the filesystem's .remount_fs function doesn't 676b3b304a2SMiklos Szeredi * reset all options to their default value, but changes only newly 677b3b304a2SMiklos Szeredi * given options, then the displayed options will not reflect reality 678b3b304a2SMiklos Szeredi * any more. 679b3b304a2SMiklos Szeredi */ 680b3b304a2SMiklos Szeredi void save_mount_options(struct super_block *sb, char *options) 681b3b304a2SMiklos Szeredi { 6822a32cebdSAl Viro BUG_ON(sb->s_options); 6832a32cebdSAl Viro rcu_assign_pointer(sb->s_options, kstrdup(options, GFP_KERNEL)); 684b3b304a2SMiklos Szeredi } 685b3b304a2SMiklos Szeredi EXPORT_SYMBOL(save_mount_options); 686b3b304a2SMiklos Szeredi 6872a32cebdSAl Viro void replace_mount_options(struct super_block *sb, char *options) 6882a32cebdSAl Viro { 6892a32cebdSAl Viro char *old = sb->s_options; 6902a32cebdSAl Viro rcu_assign_pointer(sb->s_options, options); 6912a32cebdSAl Viro if (old) { 6922a32cebdSAl Viro synchronize_rcu(); 6932a32cebdSAl Viro kfree(old); 6942a32cebdSAl Viro } 6952a32cebdSAl Viro } 6962a32cebdSAl Viro EXPORT_SYMBOL(replace_mount_options); 6972a32cebdSAl Viro 698a1a2c409SMiklos Szeredi #ifdef CONFIG_PROC_FS 6991da177e4SLinus Torvalds /* iterator */ 7001da177e4SLinus Torvalds static void *m_start(struct seq_file *m, loff_t *pos) 7011da177e4SLinus Torvalds { 702a1a2c409SMiklos Szeredi struct proc_mounts *p = m->private; 7031da177e4SLinus Torvalds 704390c6843SRam Pai down_read(&namespace_sem); 705a1a2c409SMiklos Szeredi return seq_list_start(&p->ns->list, *pos); 7061da177e4SLinus Torvalds } 7071da177e4SLinus Torvalds 7081da177e4SLinus Torvalds static void *m_next(struct seq_file *m, void *v, loff_t *pos) 7091da177e4SLinus Torvalds { 710a1a2c409SMiklos Szeredi struct proc_mounts *p = m->private; 711b0765fb8SPavel Emelianov 712a1a2c409SMiklos Szeredi return seq_list_next(v, &p->ns->list, pos); 7131da177e4SLinus Torvalds } 7141da177e4SLinus Torvalds 7151da177e4SLinus Torvalds static void m_stop(struct seq_file *m, void *v) 7161da177e4SLinus Torvalds { 717390c6843SRam Pai up_read(&namespace_sem); 7181da177e4SLinus Torvalds } 7191da177e4SLinus Torvalds 7202d4d4864SRam Pai struct proc_fs_info { 7211da177e4SLinus Torvalds int flag; 7222d4d4864SRam Pai const char *str; 7232d4d4864SRam Pai }; 7242d4d4864SRam Pai 7252069f457SEric Paris static int show_sb_opts(struct seq_file *m, struct super_block *sb) 7262d4d4864SRam Pai { 7272d4d4864SRam Pai static const struct proc_fs_info fs_info[] = { 7281da177e4SLinus Torvalds { MS_SYNCHRONOUS, ",sync" }, 7291da177e4SLinus Torvalds { MS_DIRSYNC, ",dirsync" }, 7301da177e4SLinus Torvalds { MS_MANDLOCK, ",mand" }, 7311da177e4SLinus Torvalds { 0, NULL } 7321da177e4SLinus Torvalds }; 7332d4d4864SRam Pai const struct proc_fs_info *fs_infop; 7342d4d4864SRam Pai 7352d4d4864SRam Pai for (fs_infop = fs_info; fs_infop->flag; fs_infop++) { 7362d4d4864SRam Pai if (sb->s_flags & fs_infop->flag) 7372d4d4864SRam Pai seq_puts(m, fs_infop->str); 7382d4d4864SRam Pai } 7392069f457SEric Paris 7402069f457SEric Paris return security_sb_show_options(m, sb); 7412d4d4864SRam Pai } 7422d4d4864SRam Pai 7432d4d4864SRam Pai static void show_mnt_opts(struct seq_file *m, struct vfsmount *mnt) 7442d4d4864SRam Pai { 7452d4d4864SRam Pai static const struct proc_fs_info mnt_info[] = { 7461da177e4SLinus Torvalds { MNT_NOSUID, ",nosuid" }, 7471da177e4SLinus Torvalds { MNT_NODEV, ",nodev" }, 7481da177e4SLinus Torvalds { MNT_NOEXEC, ",noexec" }, 749fc33a7bbSChristoph Hellwig { MNT_NOATIME, ",noatime" }, 750fc33a7bbSChristoph Hellwig { MNT_NODIRATIME, ",nodiratime" }, 75147ae32d6SValerie Henson { MNT_RELATIME, ",relatime" }, 752d0adde57SMatthew Garrett { MNT_STRICTATIME, ",strictatime" }, 7531da177e4SLinus Torvalds { 0, NULL } 7541da177e4SLinus Torvalds }; 7552d4d4864SRam Pai const struct proc_fs_info *fs_infop; 7562d4d4864SRam Pai 7572d4d4864SRam Pai for (fs_infop = mnt_info; fs_infop->flag; fs_infop++) { 7582d4d4864SRam Pai if (mnt->mnt_flags & fs_infop->flag) 7592d4d4864SRam Pai seq_puts(m, fs_infop->str); 7602d4d4864SRam Pai } 7612d4d4864SRam Pai } 7622d4d4864SRam Pai 7632d4d4864SRam Pai static void show_type(struct seq_file *m, struct super_block *sb) 7642d4d4864SRam Pai { 7652d4d4864SRam Pai mangle(m, sb->s_type->name); 7662d4d4864SRam Pai if (sb->s_subtype && sb->s_subtype[0]) { 7672d4d4864SRam Pai seq_putc(m, '.'); 7682d4d4864SRam Pai mangle(m, sb->s_subtype); 7692d4d4864SRam Pai } 7702d4d4864SRam Pai } 7712d4d4864SRam Pai 7722d4d4864SRam Pai static int show_vfsmnt(struct seq_file *m, void *v) 7732d4d4864SRam Pai { 7742d4d4864SRam Pai struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); 7752d4d4864SRam Pai int err = 0; 776c32c2f63SJan Blunck struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; 7771da177e4SLinus Torvalds 7781da177e4SLinus Torvalds mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none"); 7791da177e4SLinus Torvalds seq_putc(m, ' '); 780c32c2f63SJan Blunck seq_path(m, &mnt_path, " \t\n\\"); 7811da177e4SLinus Torvalds seq_putc(m, ' '); 7822d4d4864SRam Pai show_type(m, mnt->mnt_sb); 7832e4b7fcdSDave Hansen seq_puts(m, __mnt_is_readonly(mnt) ? " ro" : " rw"); 7842069f457SEric Paris err = show_sb_opts(m, mnt->mnt_sb); 7852069f457SEric Paris if (err) 7862069f457SEric Paris goto out; 7872d4d4864SRam Pai show_mnt_opts(m, mnt); 7881da177e4SLinus Torvalds if (mnt->mnt_sb->s_op->show_options) 7891da177e4SLinus Torvalds err = mnt->mnt_sb->s_op->show_options(m, mnt); 7901da177e4SLinus Torvalds seq_puts(m, " 0 0\n"); 7912069f457SEric Paris out: 7921da177e4SLinus Torvalds return err; 7931da177e4SLinus Torvalds } 7941da177e4SLinus Torvalds 795a1a2c409SMiklos Szeredi const struct seq_operations mounts_op = { 7961da177e4SLinus Torvalds .start = m_start, 7971da177e4SLinus Torvalds .next = m_next, 7981da177e4SLinus Torvalds .stop = m_stop, 7991da177e4SLinus Torvalds .show = show_vfsmnt 8001da177e4SLinus Torvalds }; 8011da177e4SLinus Torvalds 8022d4d4864SRam Pai static int show_mountinfo(struct seq_file *m, void *v) 8032d4d4864SRam Pai { 8042d4d4864SRam Pai struct proc_mounts *p = m->private; 8052d4d4864SRam Pai struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); 8062d4d4864SRam Pai struct super_block *sb = mnt->mnt_sb; 8072d4d4864SRam Pai struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; 8082d4d4864SRam Pai struct path root = p->root; 8092d4d4864SRam Pai int err = 0; 8102d4d4864SRam Pai 8112d4d4864SRam Pai seq_printf(m, "%i %i %u:%u ", mnt->mnt_id, mnt->mnt_parent->mnt_id, 8122d4d4864SRam Pai MAJOR(sb->s_dev), MINOR(sb->s_dev)); 8132d4d4864SRam Pai seq_dentry(m, mnt->mnt_root, " \t\n\\"); 8142d4d4864SRam Pai seq_putc(m, ' '); 8152d4d4864SRam Pai seq_path_root(m, &mnt_path, &root, " \t\n\\"); 8162d4d4864SRam Pai if (root.mnt != p->root.mnt || root.dentry != p->root.dentry) { 8172d4d4864SRam Pai /* 8182d4d4864SRam Pai * Mountpoint is outside root, discard that one. Ugly, 8192d4d4864SRam Pai * but less so than trying to do that in iterator in a 8202d4d4864SRam Pai * race-free way (due to renames). 8212d4d4864SRam Pai */ 8222d4d4864SRam Pai return SEQ_SKIP; 8232d4d4864SRam Pai } 8242d4d4864SRam Pai seq_puts(m, mnt->mnt_flags & MNT_READONLY ? " ro" : " rw"); 8252d4d4864SRam Pai show_mnt_opts(m, mnt); 8262d4d4864SRam Pai 8272d4d4864SRam Pai /* Tagged fields ("foo:X" or "bar") */ 8282d4d4864SRam Pai if (IS_MNT_SHARED(mnt)) 8292d4d4864SRam Pai seq_printf(m, " shared:%i", mnt->mnt_group_id); 83097e7e0f7SMiklos Szeredi if (IS_MNT_SLAVE(mnt)) { 83197e7e0f7SMiklos Szeredi int master = mnt->mnt_master->mnt_group_id; 83297e7e0f7SMiklos Szeredi int dom = get_dominating_id(mnt, &p->root); 83397e7e0f7SMiklos Szeredi seq_printf(m, " master:%i", master); 83497e7e0f7SMiklos Szeredi if (dom && dom != master) 83597e7e0f7SMiklos Szeredi seq_printf(m, " propagate_from:%i", dom); 83697e7e0f7SMiklos Szeredi } 8372d4d4864SRam Pai if (IS_MNT_UNBINDABLE(mnt)) 8382d4d4864SRam Pai seq_puts(m, " unbindable"); 8392d4d4864SRam Pai 8402d4d4864SRam Pai /* Filesystem specific data */ 8412d4d4864SRam Pai seq_puts(m, " - "); 8422d4d4864SRam Pai show_type(m, sb); 8432d4d4864SRam Pai seq_putc(m, ' '); 8442d4d4864SRam Pai mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none"); 8452d4d4864SRam Pai seq_puts(m, sb->s_flags & MS_RDONLY ? " ro" : " rw"); 8462069f457SEric Paris err = show_sb_opts(m, sb); 8472069f457SEric Paris if (err) 8482069f457SEric Paris goto out; 8492d4d4864SRam Pai if (sb->s_op->show_options) 8502d4d4864SRam Pai err = sb->s_op->show_options(m, mnt); 8512d4d4864SRam Pai seq_putc(m, '\n'); 8522069f457SEric Paris out: 8532d4d4864SRam Pai return err; 8542d4d4864SRam Pai } 8552d4d4864SRam Pai 8562d4d4864SRam Pai const struct seq_operations mountinfo_op = { 8572d4d4864SRam Pai .start = m_start, 8582d4d4864SRam Pai .next = m_next, 8592d4d4864SRam Pai .stop = m_stop, 8602d4d4864SRam Pai .show = show_mountinfo, 8612d4d4864SRam Pai }; 8622d4d4864SRam Pai 863b4629fe2SChuck Lever static int show_vfsstat(struct seq_file *m, void *v) 864b4629fe2SChuck Lever { 865b0765fb8SPavel Emelianov struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); 866c32c2f63SJan Blunck struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; 867b4629fe2SChuck Lever int err = 0; 868b4629fe2SChuck Lever 869b4629fe2SChuck Lever /* device */ 870b4629fe2SChuck Lever if (mnt->mnt_devname) { 871b4629fe2SChuck Lever seq_puts(m, "device "); 872b4629fe2SChuck Lever mangle(m, mnt->mnt_devname); 873b4629fe2SChuck Lever } else 874b4629fe2SChuck Lever seq_puts(m, "no device"); 875b4629fe2SChuck Lever 876b4629fe2SChuck Lever /* mount point */ 877b4629fe2SChuck Lever seq_puts(m, " mounted on "); 878c32c2f63SJan Blunck seq_path(m, &mnt_path, " \t\n\\"); 879b4629fe2SChuck Lever seq_putc(m, ' '); 880b4629fe2SChuck Lever 881b4629fe2SChuck Lever /* file system type */ 882b4629fe2SChuck Lever seq_puts(m, "with fstype "); 8832d4d4864SRam Pai show_type(m, mnt->mnt_sb); 884b4629fe2SChuck Lever 885b4629fe2SChuck Lever /* optional statistics */ 886b4629fe2SChuck Lever if (mnt->mnt_sb->s_op->show_stats) { 887b4629fe2SChuck Lever seq_putc(m, ' '); 888b4629fe2SChuck Lever err = mnt->mnt_sb->s_op->show_stats(m, mnt); 889b4629fe2SChuck Lever } 890b4629fe2SChuck Lever 891b4629fe2SChuck Lever seq_putc(m, '\n'); 892b4629fe2SChuck Lever return err; 893b4629fe2SChuck Lever } 894b4629fe2SChuck Lever 895a1a2c409SMiklos Szeredi const struct seq_operations mountstats_op = { 896b4629fe2SChuck Lever .start = m_start, 897b4629fe2SChuck Lever .next = m_next, 898b4629fe2SChuck Lever .stop = m_stop, 899b4629fe2SChuck Lever .show = show_vfsstat, 900b4629fe2SChuck Lever }; 901a1a2c409SMiklos Szeredi #endif /* CONFIG_PROC_FS */ 902b4629fe2SChuck Lever 9031da177e4SLinus Torvalds /** 9041da177e4SLinus Torvalds * may_umount_tree - check if a mount tree is busy 9051da177e4SLinus Torvalds * @mnt: root of mount tree 9061da177e4SLinus Torvalds * 9071da177e4SLinus Torvalds * This is called to check if a tree of mounts has any 9081da177e4SLinus Torvalds * open files, pwds, chroots or sub mounts that are 9091da177e4SLinus Torvalds * busy. 9101da177e4SLinus Torvalds */ 9111da177e4SLinus Torvalds int may_umount_tree(struct vfsmount *mnt) 9121da177e4SLinus Torvalds { 91336341f64SRam Pai int actual_refs = 0; 91436341f64SRam Pai int minimum_refs = 0; 91536341f64SRam Pai struct vfsmount *p; 9161da177e4SLinus Torvalds 9171da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 91836341f64SRam Pai for (p = mnt; p; p = next_mnt(p, mnt)) { 9191da177e4SLinus Torvalds actual_refs += atomic_read(&p->mnt_count); 9201da177e4SLinus Torvalds minimum_refs += 2; 9211da177e4SLinus Torvalds } 9221da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 9231da177e4SLinus Torvalds 9241da177e4SLinus Torvalds if (actual_refs > minimum_refs) 9251da177e4SLinus Torvalds return 0; 926e3474a8eSIan Kent 927e3474a8eSIan Kent return 1; 9281da177e4SLinus Torvalds } 9291da177e4SLinus Torvalds 9301da177e4SLinus Torvalds EXPORT_SYMBOL(may_umount_tree); 9311da177e4SLinus Torvalds 9321da177e4SLinus Torvalds /** 9331da177e4SLinus Torvalds * may_umount - check if a mount point is busy 9341da177e4SLinus Torvalds * @mnt: root of mount 9351da177e4SLinus Torvalds * 9361da177e4SLinus Torvalds * This is called to check if a mount point has any 9371da177e4SLinus Torvalds * open files, pwds, chroots or sub mounts. If the 9381da177e4SLinus Torvalds * mount has sub mounts this will return busy 9391da177e4SLinus Torvalds * regardless of whether the sub mounts are busy. 9401da177e4SLinus Torvalds * 9411da177e4SLinus Torvalds * Doesn't take quota and stuff into account. IOW, in some cases it will 9421da177e4SLinus Torvalds * give false negatives. The main reason why it's here is that we need 9431da177e4SLinus Torvalds * a non-destructive way to look for easily umountable filesystems. 9441da177e4SLinus Torvalds */ 9451da177e4SLinus Torvalds int may_umount(struct vfsmount *mnt) 9461da177e4SLinus Torvalds { 947e3474a8eSIan Kent int ret = 1; 948a05964f3SRam Pai spin_lock(&vfsmount_lock); 949a05964f3SRam Pai if (propagate_mount_busy(mnt, 2)) 950e3474a8eSIan Kent ret = 0; 951a05964f3SRam Pai spin_unlock(&vfsmount_lock); 952a05964f3SRam Pai return ret; 9531da177e4SLinus Torvalds } 9541da177e4SLinus Torvalds 9551da177e4SLinus Torvalds EXPORT_SYMBOL(may_umount); 9561da177e4SLinus Torvalds 957b90fa9aeSRam Pai void release_mounts(struct list_head *head) 9581da177e4SLinus Torvalds { 95970fbcdf4SRam Pai struct vfsmount *mnt; 96070fbcdf4SRam Pai while (!list_empty(head)) { 961b5e61818SPavel Emelianov mnt = list_first_entry(head, struct vfsmount, mnt_hash); 96270fbcdf4SRam Pai list_del_init(&mnt->mnt_hash); 96370fbcdf4SRam Pai if (mnt->mnt_parent != mnt) { 96470fbcdf4SRam Pai struct dentry *dentry; 96570fbcdf4SRam Pai struct vfsmount *m; 96670fbcdf4SRam Pai spin_lock(&vfsmount_lock); 96770fbcdf4SRam Pai dentry = mnt->mnt_mountpoint; 96870fbcdf4SRam Pai m = mnt->mnt_parent; 96970fbcdf4SRam Pai mnt->mnt_mountpoint = mnt->mnt_root; 97070fbcdf4SRam Pai mnt->mnt_parent = mnt; 9717c4b93d8SAl Viro m->mnt_ghosts--; 9721da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 97370fbcdf4SRam Pai dput(dentry); 97470fbcdf4SRam Pai mntput(m); 9751da177e4SLinus Torvalds } 9761da177e4SLinus Torvalds mntput(mnt); 97770fbcdf4SRam Pai } 97870fbcdf4SRam Pai } 97970fbcdf4SRam Pai 980a05964f3SRam Pai void umount_tree(struct vfsmount *mnt, int propagate, struct list_head *kill) 98170fbcdf4SRam Pai { 98270fbcdf4SRam Pai struct vfsmount *p; 98370fbcdf4SRam Pai 9841bfba4e8SAkinobu Mita for (p = mnt; p; p = next_mnt(p, mnt)) 9851bfba4e8SAkinobu Mita list_move(&p->mnt_hash, kill); 98670fbcdf4SRam Pai 987a05964f3SRam Pai if (propagate) 988a05964f3SRam Pai propagate_umount(kill); 989a05964f3SRam Pai 99070fbcdf4SRam Pai list_for_each_entry(p, kill, mnt_hash) { 99170fbcdf4SRam Pai list_del_init(&p->mnt_expire); 99270fbcdf4SRam Pai list_del_init(&p->mnt_list); 9936b3286edSKirill Korotaev __touch_mnt_namespace(p->mnt_ns); 9946b3286edSKirill Korotaev p->mnt_ns = NULL; 99570fbcdf4SRam Pai list_del_init(&p->mnt_child); 9967c4b93d8SAl Viro if (p->mnt_parent != p) { 9977c4b93d8SAl Viro p->mnt_parent->mnt_ghosts++; 998f30ac319SAl Viro p->mnt_mountpoint->d_mounted--; 9997c4b93d8SAl Viro } 1000a05964f3SRam Pai change_mnt_propagation(p, MS_PRIVATE); 10011da177e4SLinus Torvalds } 10021da177e4SLinus Torvalds } 10031da177e4SLinus Torvalds 1004c35038beSAl Viro static void shrink_submounts(struct vfsmount *mnt, struct list_head *umounts); 1005c35038beSAl Viro 10061da177e4SLinus Torvalds static int do_umount(struct vfsmount *mnt, int flags) 10071da177e4SLinus Torvalds { 10081da177e4SLinus Torvalds struct super_block *sb = mnt->mnt_sb; 10091da177e4SLinus Torvalds int retval; 101070fbcdf4SRam Pai LIST_HEAD(umount_list); 10111da177e4SLinus Torvalds 10121da177e4SLinus Torvalds retval = security_sb_umount(mnt, flags); 10131da177e4SLinus Torvalds if (retval) 10141da177e4SLinus Torvalds return retval; 10151da177e4SLinus Torvalds 10161da177e4SLinus Torvalds /* 10171da177e4SLinus Torvalds * Allow userspace to request a mountpoint be expired rather than 10181da177e4SLinus Torvalds * unmounting unconditionally. Unmount only happens if: 10191da177e4SLinus Torvalds * (1) the mark is already set (the mark is cleared by mntput()) 10201da177e4SLinus Torvalds * (2) the usage count == 1 [parent vfsmount] + 1 [sys_umount] 10211da177e4SLinus Torvalds */ 10221da177e4SLinus Torvalds if (flags & MNT_EXPIRE) { 10236ac08c39SJan Blunck if (mnt == current->fs->root.mnt || 10241da177e4SLinus Torvalds flags & (MNT_FORCE | MNT_DETACH)) 10251da177e4SLinus Torvalds return -EINVAL; 10261da177e4SLinus Torvalds 10271da177e4SLinus Torvalds if (atomic_read(&mnt->mnt_count) != 2) 10281da177e4SLinus Torvalds return -EBUSY; 10291da177e4SLinus Torvalds 10301da177e4SLinus Torvalds if (!xchg(&mnt->mnt_expiry_mark, 1)) 10311da177e4SLinus Torvalds return -EAGAIN; 10321da177e4SLinus Torvalds } 10331da177e4SLinus Torvalds 10341da177e4SLinus Torvalds /* 10351da177e4SLinus Torvalds * If we may have to abort operations to get out of this 10361da177e4SLinus Torvalds * mount, and they will themselves hold resources we must 10371da177e4SLinus Torvalds * allow the fs to do things. In the Unix tradition of 10381da177e4SLinus Torvalds * 'Gee thats tricky lets do it in userspace' the umount_begin 10391da177e4SLinus Torvalds * might fail to complete on the first run through as other tasks 10401da177e4SLinus Torvalds * must return, and the like. Thats for the mount program to worry 10411da177e4SLinus Torvalds * about for the moment. 10421da177e4SLinus Torvalds */ 10431da177e4SLinus Torvalds 104442faad99SAl Viro if (flags & MNT_FORCE && sb->s_op->umount_begin) { 104542faad99SAl Viro sb->s_op->umount_begin(sb); 104642faad99SAl Viro } 10471da177e4SLinus Torvalds 10481da177e4SLinus Torvalds /* 10491da177e4SLinus Torvalds * No sense to grab the lock for this test, but test itself looks 10501da177e4SLinus Torvalds * somewhat bogus. Suggestions for better replacement? 10511da177e4SLinus Torvalds * Ho-hum... In principle, we might treat that as umount + switch 10521da177e4SLinus Torvalds * to rootfs. GC would eventually take care of the old vfsmount. 10531da177e4SLinus Torvalds * Actually it makes sense, especially if rootfs would contain a 10541da177e4SLinus Torvalds * /reboot - static binary that would close all descriptors and 10551da177e4SLinus Torvalds * call reboot(9). Then init(8) could umount root and exec /reboot. 10561da177e4SLinus Torvalds */ 10576ac08c39SJan Blunck if (mnt == current->fs->root.mnt && !(flags & MNT_DETACH)) { 10581da177e4SLinus Torvalds /* 10591da177e4SLinus Torvalds * Special case for "unmounting" root ... 10601da177e4SLinus Torvalds * we just try to remount it readonly. 10611da177e4SLinus Torvalds */ 10621da177e4SLinus Torvalds down_write(&sb->s_umount); 10631da177e4SLinus Torvalds if (!(sb->s_flags & MS_RDONLY)) { 10641da177e4SLinus Torvalds lock_kernel(); 10651da177e4SLinus Torvalds retval = do_remount_sb(sb, MS_RDONLY, NULL, 0); 10661da177e4SLinus Torvalds unlock_kernel(); 10671da177e4SLinus Torvalds } 10681da177e4SLinus Torvalds up_write(&sb->s_umount); 10691da177e4SLinus Torvalds return retval; 10701da177e4SLinus Torvalds } 10711da177e4SLinus Torvalds 1072390c6843SRam Pai down_write(&namespace_sem); 10731da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 10745addc5ddSAl Viro event++; 10751da177e4SLinus Torvalds 1076c35038beSAl Viro if (!(flags & MNT_DETACH)) 1077c35038beSAl Viro shrink_submounts(mnt, &umount_list); 1078c35038beSAl Viro 10791da177e4SLinus Torvalds retval = -EBUSY; 1080a05964f3SRam Pai if (flags & MNT_DETACH || !propagate_mount_busy(mnt, 2)) { 10811da177e4SLinus Torvalds if (!list_empty(&mnt->mnt_list)) 1082a05964f3SRam Pai umount_tree(mnt, 1, &umount_list); 10831da177e4SLinus Torvalds retval = 0; 10841da177e4SLinus Torvalds } 10851da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 10861da177e4SLinus Torvalds if (retval) 10871da177e4SLinus Torvalds security_sb_umount_busy(mnt); 1088390c6843SRam Pai up_write(&namespace_sem); 108970fbcdf4SRam Pai release_mounts(&umount_list); 10901da177e4SLinus Torvalds return retval; 10911da177e4SLinus Torvalds } 10921da177e4SLinus Torvalds 10931da177e4SLinus Torvalds /* 10941da177e4SLinus Torvalds * Now umount can handle mount points as well as block devices. 10951da177e4SLinus Torvalds * This is important for filesystems which use unnamed block devices. 10961da177e4SLinus Torvalds * 10971da177e4SLinus Torvalds * We now support a flag for forced unmount like the other 'big iron' 10981da177e4SLinus Torvalds * unixes. Our API is identical to OSF/1 to avoid making a mess of AMD 10991da177e4SLinus Torvalds */ 11001da177e4SLinus Torvalds 1101bdc480e3SHeiko Carstens SYSCALL_DEFINE2(umount, char __user *, name, int, flags) 11021da177e4SLinus Torvalds { 11032d8f3038SAl Viro struct path path; 11041da177e4SLinus Torvalds int retval; 11051da177e4SLinus Torvalds 11062d8f3038SAl Viro retval = user_path(name, &path); 11071da177e4SLinus Torvalds if (retval) 11081da177e4SLinus Torvalds goto out; 11091da177e4SLinus Torvalds retval = -EINVAL; 11102d8f3038SAl Viro if (path.dentry != path.mnt->mnt_root) 11111da177e4SLinus Torvalds goto dput_and_out; 11122d8f3038SAl Viro if (!check_mnt(path.mnt)) 11131da177e4SLinus Torvalds goto dput_and_out; 11141da177e4SLinus Torvalds 11151da177e4SLinus Torvalds retval = -EPERM; 11161da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 11171da177e4SLinus Torvalds goto dput_and_out; 11181da177e4SLinus Torvalds 11192d8f3038SAl Viro retval = do_umount(path.mnt, flags); 11201da177e4SLinus Torvalds dput_and_out: 1121429731b1SJan Blunck /* we mustn't call path_put() as that would clear mnt_expiry_mark */ 11222d8f3038SAl Viro dput(path.dentry); 11232d8f3038SAl Viro mntput_no_expire(path.mnt); 11241da177e4SLinus Torvalds out: 11251da177e4SLinus Torvalds return retval; 11261da177e4SLinus Torvalds } 11271da177e4SLinus Torvalds 11281da177e4SLinus Torvalds #ifdef __ARCH_WANT_SYS_OLDUMOUNT 11291da177e4SLinus Torvalds 11301da177e4SLinus Torvalds /* 11311da177e4SLinus Torvalds * The 2.0 compatible umount. No flags. 11321da177e4SLinus Torvalds */ 1133bdc480e3SHeiko Carstens SYSCALL_DEFINE1(oldumount, char __user *, name) 11341da177e4SLinus Torvalds { 11351da177e4SLinus Torvalds return sys_umount(name, 0); 11361da177e4SLinus Torvalds } 11371da177e4SLinus Torvalds 11381da177e4SLinus Torvalds #endif 11391da177e4SLinus Torvalds 11402d92ab3cSAl Viro static int mount_is_safe(struct path *path) 11411da177e4SLinus Torvalds { 11421da177e4SLinus Torvalds if (capable(CAP_SYS_ADMIN)) 11431da177e4SLinus Torvalds return 0; 11441da177e4SLinus Torvalds return -EPERM; 11451da177e4SLinus Torvalds #ifdef notyet 11462d92ab3cSAl Viro if (S_ISLNK(path->dentry->d_inode->i_mode)) 11471da177e4SLinus Torvalds return -EPERM; 11482d92ab3cSAl Viro if (path->dentry->d_inode->i_mode & S_ISVTX) { 1149da9592edSDavid Howells if (current_uid() != path->dentry->d_inode->i_uid) 11501da177e4SLinus Torvalds return -EPERM; 11511da177e4SLinus Torvalds } 11522d92ab3cSAl Viro if (inode_permission(path->dentry->d_inode, MAY_WRITE)) 11531da177e4SLinus Torvalds return -EPERM; 11541da177e4SLinus Torvalds return 0; 11551da177e4SLinus Torvalds #endif 11561da177e4SLinus Torvalds } 11571da177e4SLinus Torvalds 1158b90fa9aeSRam Pai struct vfsmount *copy_tree(struct vfsmount *mnt, struct dentry *dentry, 115936341f64SRam Pai int flag) 11601da177e4SLinus Torvalds { 11611da177e4SLinus Torvalds struct vfsmount *res, *p, *q, *r, *s; 11621a390689SAl Viro struct path path; 11631da177e4SLinus Torvalds 11649676f0c6SRam Pai if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(mnt)) 11659676f0c6SRam Pai return NULL; 11669676f0c6SRam Pai 116736341f64SRam Pai res = q = clone_mnt(mnt, dentry, flag); 11681da177e4SLinus Torvalds if (!q) 11691da177e4SLinus Torvalds goto Enomem; 11701da177e4SLinus Torvalds q->mnt_mountpoint = mnt->mnt_mountpoint; 11711da177e4SLinus Torvalds 11721da177e4SLinus Torvalds p = mnt; 1173fdadd65fSDomen Puncer list_for_each_entry(r, &mnt->mnt_mounts, mnt_child) { 11747ec02ef1SJan Blunck if (!is_subdir(r->mnt_mountpoint, dentry)) 11751da177e4SLinus Torvalds continue; 11761da177e4SLinus Torvalds 11771da177e4SLinus Torvalds for (s = r; s; s = next_mnt(s, r)) { 11789676f0c6SRam Pai if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(s)) { 11799676f0c6SRam Pai s = skip_mnt_tree(s); 11809676f0c6SRam Pai continue; 11819676f0c6SRam Pai } 11821da177e4SLinus Torvalds while (p != s->mnt_parent) { 11831da177e4SLinus Torvalds p = p->mnt_parent; 11841da177e4SLinus Torvalds q = q->mnt_parent; 11851da177e4SLinus Torvalds } 11861da177e4SLinus Torvalds p = s; 11871a390689SAl Viro path.mnt = q; 11881a390689SAl Viro path.dentry = p->mnt_mountpoint; 118936341f64SRam Pai q = clone_mnt(p, p->mnt_root, flag); 11901da177e4SLinus Torvalds if (!q) 11911da177e4SLinus Torvalds goto Enomem; 11921da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 11931da177e4SLinus Torvalds list_add_tail(&q->mnt_list, &res->mnt_list); 11941a390689SAl Viro attach_mnt(q, &path); 11951da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 11961da177e4SLinus Torvalds } 11971da177e4SLinus Torvalds } 11981da177e4SLinus Torvalds return res; 11991da177e4SLinus Torvalds Enomem: 12001da177e4SLinus Torvalds if (res) { 120170fbcdf4SRam Pai LIST_HEAD(umount_list); 12021da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 1203a05964f3SRam Pai umount_tree(res, 0, &umount_list); 12041da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 120570fbcdf4SRam Pai release_mounts(&umount_list); 12061da177e4SLinus Torvalds } 12071da177e4SLinus Torvalds return NULL; 12081da177e4SLinus Torvalds } 12091da177e4SLinus Torvalds 1210589ff870SAl Viro struct vfsmount *collect_mounts(struct path *path) 12118aec0809SAl Viro { 12128aec0809SAl Viro struct vfsmount *tree; 12131a60a280SAl Viro down_write(&namespace_sem); 1214589ff870SAl Viro tree = copy_tree(path->mnt, path->dentry, CL_COPY_ALL | CL_PRIVATE); 12151a60a280SAl Viro up_write(&namespace_sem); 12168aec0809SAl Viro return tree; 12178aec0809SAl Viro } 12188aec0809SAl Viro 12198aec0809SAl Viro void drop_collected_mounts(struct vfsmount *mnt) 12208aec0809SAl Viro { 12218aec0809SAl Viro LIST_HEAD(umount_list); 12221a60a280SAl Viro down_write(&namespace_sem); 12238aec0809SAl Viro spin_lock(&vfsmount_lock); 12248aec0809SAl Viro umount_tree(mnt, 0, &umount_list); 12258aec0809SAl Viro spin_unlock(&vfsmount_lock); 12261a60a280SAl Viro up_write(&namespace_sem); 12278aec0809SAl Viro release_mounts(&umount_list); 12288aec0809SAl Viro } 12298aec0809SAl Viro 1230719f5d7fSMiklos Szeredi static void cleanup_group_ids(struct vfsmount *mnt, struct vfsmount *end) 1231719f5d7fSMiklos Szeredi { 1232719f5d7fSMiklos Szeredi struct vfsmount *p; 1233719f5d7fSMiklos Szeredi 1234719f5d7fSMiklos Szeredi for (p = mnt; p != end; p = next_mnt(p, mnt)) { 1235719f5d7fSMiklos Szeredi if (p->mnt_group_id && !IS_MNT_SHARED(p)) 1236719f5d7fSMiklos Szeredi mnt_release_group_id(p); 1237719f5d7fSMiklos Szeredi } 1238719f5d7fSMiklos Szeredi } 1239719f5d7fSMiklos Szeredi 1240719f5d7fSMiklos Szeredi static int invent_group_ids(struct vfsmount *mnt, bool recurse) 1241719f5d7fSMiklos Szeredi { 1242719f5d7fSMiklos Szeredi struct vfsmount *p; 1243719f5d7fSMiklos Szeredi 1244719f5d7fSMiklos Szeredi for (p = mnt; p; p = recurse ? next_mnt(p, mnt) : NULL) { 1245719f5d7fSMiklos Szeredi if (!p->mnt_group_id && !IS_MNT_SHARED(p)) { 1246719f5d7fSMiklos Szeredi int err = mnt_alloc_group_id(p); 1247719f5d7fSMiklos Szeredi if (err) { 1248719f5d7fSMiklos Szeredi cleanup_group_ids(mnt, p); 1249719f5d7fSMiklos Szeredi return err; 1250719f5d7fSMiklos Szeredi } 1251719f5d7fSMiklos Szeredi } 1252719f5d7fSMiklos Szeredi } 1253719f5d7fSMiklos Szeredi 1254719f5d7fSMiklos Szeredi return 0; 1255719f5d7fSMiklos Szeredi } 1256719f5d7fSMiklos Szeredi 1257b90fa9aeSRam Pai /* 1258b90fa9aeSRam Pai * @source_mnt : mount tree to be attached 1259b90fa9aeSRam Pai * @nd : place the mount tree @source_mnt is attached 126021444403SRam Pai * @parent_nd : if non-null, detach the source_mnt from its parent and 126121444403SRam Pai * store the parent mount and mountpoint dentry. 126221444403SRam Pai * (done when source_mnt is moved) 1263b90fa9aeSRam Pai * 1264b90fa9aeSRam Pai * NOTE: in the table below explains the semantics when a source mount 1265b90fa9aeSRam Pai * of a given type is attached to a destination mount of a given type. 12669676f0c6SRam Pai * --------------------------------------------------------------------------- 1267b90fa9aeSRam Pai * | BIND MOUNT OPERATION | 12689676f0c6SRam Pai * |************************************************************************** 12699676f0c6SRam Pai * | source-->| shared | private | slave | unbindable | 12709676f0c6SRam Pai * | dest | | | | | 12719676f0c6SRam Pai * | | | | | | | 12729676f0c6SRam Pai * | v | | | | | 12739676f0c6SRam Pai * |************************************************************************** 12749676f0c6SRam Pai * | shared | shared (++) | shared (+) | shared(+++)| invalid | 12755afe0022SRam Pai * | | | | | | 12769676f0c6SRam Pai * |non-shared| shared (+) | private | slave (*) | invalid | 12779676f0c6SRam Pai * *************************************************************************** 1278b90fa9aeSRam Pai * A bind operation clones the source mount and mounts the clone on the 1279b90fa9aeSRam Pai * destination mount. 1280b90fa9aeSRam Pai * 1281b90fa9aeSRam Pai * (++) the cloned mount is propagated to all the mounts in the propagation 1282b90fa9aeSRam Pai * tree of the destination mount and the cloned mount is added to 1283b90fa9aeSRam Pai * the peer group of the source mount. 1284b90fa9aeSRam Pai * (+) the cloned mount is created under the destination mount and is marked 1285b90fa9aeSRam Pai * as shared. The cloned mount is added to the peer group of the source 1286b90fa9aeSRam Pai * mount. 12875afe0022SRam Pai * (+++) the mount is propagated to all the mounts in the propagation tree 12885afe0022SRam Pai * of the destination mount and the cloned mount is made slave 12895afe0022SRam Pai * of the same master as that of the source mount. The cloned mount 12905afe0022SRam Pai * is marked as 'shared and slave'. 12915afe0022SRam Pai * (*) the cloned mount is made a slave of the same master as that of the 12925afe0022SRam Pai * source mount. 12935afe0022SRam Pai * 12949676f0c6SRam Pai * --------------------------------------------------------------------------- 129521444403SRam Pai * | MOVE MOUNT OPERATION | 12969676f0c6SRam Pai * |************************************************************************** 12979676f0c6SRam Pai * | source-->| shared | private | slave | unbindable | 12989676f0c6SRam Pai * | dest | | | | | 12999676f0c6SRam Pai * | | | | | | | 13009676f0c6SRam Pai * | v | | | | | 13019676f0c6SRam Pai * |************************************************************************** 13029676f0c6SRam Pai * | shared | shared (+) | shared (+) | shared(+++) | invalid | 13035afe0022SRam Pai * | | | | | | 13049676f0c6SRam Pai * |non-shared| shared (+*) | private | slave (*) | unbindable | 13059676f0c6SRam Pai * *************************************************************************** 13065afe0022SRam Pai * 13075afe0022SRam Pai * (+) the mount is moved to the destination. And is then propagated to 13085afe0022SRam Pai * all the mounts in the propagation tree of the destination mount. 130921444403SRam Pai * (+*) the mount is moved to the destination. 13105afe0022SRam Pai * (+++) the mount is moved to the destination and is then propagated to 13115afe0022SRam Pai * all the mounts belonging to the destination mount's propagation tree. 13125afe0022SRam Pai * the mount is marked as 'shared and slave'. 13135afe0022SRam Pai * (*) the mount continues to be a slave at the new location. 1314b90fa9aeSRam Pai * 1315b90fa9aeSRam Pai * if the source mount is a tree, the operations explained above is 1316b90fa9aeSRam Pai * applied to each mount in the tree. 1317b90fa9aeSRam Pai * Must be called without spinlocks held, since this function can sleep 1318b90fa9aeSRam Pai * in allocations. 1319b90fa9aeSRam Pai */ 1320b90fa9aeSRam Pai static int attach_recursive_mnt(struct vfsmount *source_mnt, 13211a390689SAl Viro struct path *path, struct path *parent_path) 1322b90fa9aeSRam Pai { 1323b90fa9aeSRam Pai LIST_HEAD(tree_list); 13241a390689SAl Viro struct vfsmount *dest_mnt = path->mnt; 13251a390689SAl Viro struct dentry *dest_dentry = path->dentry; 1326b90fa9aeSRam Pai struct vfsmount *child, *p; 1327719f5d7fSMiklos Szeredi int err; 1328b90fa9aeSRam Pai 1329719f5d7fSMiklos Szeredi if (IS_MNT_SHARED(dest_mnt)) { 1330719f5d7fSMiklos Szeredi err = invent_group_ids(source_mnt, true); 1331719f5d7fSMiklos Szeredi if (err) 1332719f5d7fSMiklos Szeredi goto out; 1333719f5d7fSMiklos Szeredi } 1334719f5d7fSMiklos Szeredi err = propagate_mnt(dest_mnt, dest_dentry, source_mnt, &tree_list); 1335719f5d7fSMiklos Szeredi if (err) 1336719f5d7fSMiklos Szeredi goto out_cleanup_ids; 1337b90fa9aeSRam Pai 1338b90fa9aeSRam Pai if (IS_MNT_SHARED(dest_mnt)) { 1339b90fa9aeSRam Pai for (p = source_mnt; p; p = next_mnt(p, source_mnt)) 1340b90fa9aeSRam Pai set_mnt_shared(p); 1341b90fa9aeSRam Pai } 1342b90fa9aeSRam Pai 1343b90fa9aeSRam Pai spin_lock(&vfsmount_lock); 13441a390689SAl Viro if (parent_path) { 13451a390689SAl Viro detach_mnt(source_mnt, parent_path); 13461a390689SAl Viro attach_mnt(source_mnt, path); 1347e5d67f07SAl Viro touch_mnt_namespace(parent_path->mnt->mnt_ns); 134821444403SRam Pai } else { 1349b90fa9aeSRam Pai mnt_set_mountpoint(dest_mnt, dest_dentry, source_mnt); 1350b90fa9aeSRam Pai commit_tree(source_mnt); 135121444403SRam Pai } 1352b90fa9aeSRam Pai 1353b90fa9aeSRam Pai list_for_each_entry_safe(child, p, &tree_list, mnt_hash) { 1354b90fa9aeSRam Pai list_del_init(&child->mnt_hash); 1355b90fa9aeSRam Pai commit_tree(child); 1356b90fa9aeSRam Pai } 1357b90fa9aeSRam Pai spin_unlock(&vfsmount_lock); 1358b90fa9aeSRam Pai return 0; 1359719f5d7fSMiklos Szeredi 1360719f5d7fSMiklos Szeredi out_cleanup_ids: 1361719f5d7fSMiklos Szeredi if (IS_MNT_SHARED(dest_mnt)) 1362719f5d7fSMiklos Szeredi cleanup_group_ids(source_mnt, NULL); 1363719f5d7fSMiklos Szeredi out: 1364719f5d7fSMiklos Szeredi return err; 1365b90fa9aeSRam Pai } 1366b90fa9aeSRam Pai 13678c3ee42eSAl Viro static int graft_tree(struct vfsmount *mnt, struct path *path) 13681da177e4SLinus Torvalds { 13691da177e4SLinus Torvalds int err; 13701da177e4SLinus Torvalds if (mnt->mnt_sb->s_flags & MS_NOUSER) 13711da177e4SLinus Torvalds return -EINVAL; 13721da177e4SLinus Torvalds 13738c3ee42eSAl Viro if (S_ISDIR(path->dentry->d_inode->i_mode) != 13741da177e4SLinus Torvalds S_ISDIR(mnt->mnt_root->d_inode->i_mode)) 13751da177e4SLinus Torvalds return -ENOTDIR; 13761da177e4SLinus Torvalds 13771da177e4SLinus Torvalds err = -ENOENT; 13788c3ee42eSAl Viro mutex_lock(&path->dentry->d_inode->i_mutex); 13798c3ee42eSAl Viro if (IS_DEADDIR(path->dentry->d_inode)) 13801da177e4SLinus Torvalds goto out_unlock; 13811da177e4SLinus Torvalds 13828c3ee42eSAl Viro err = security_sb_check_sb(mnt, path); 13831da177e4SLinus Torvalds if (err) 13841da177e4SLinus Torvalds goto out_unlock; 13851da177e4SLinus Torvalds 13861da177e4SLinus Torvalds err = -ENOENT; 1387f3da392eSAlexey Dobriyan if (!d_unlinked(path->dentry)) 13888c3ee42eSAl Viro err = attach_recursive_mnt(mnt, path, NULL); 13891da177e4SLinus Torvalds out_unlock: 13908c3ee42eSAl Viro mutex_unlock(&path->dentry->d_inode->i_mutex); 13911da177e4SLinus Torvalds if (!err) 13928c3ee42eSAl Viro security_sb_post_addmount(mnt, path); 13931da177e4SLinus Torvalds return err; 13941da177e4SLinus Torvalds } 13951da177e4SLinus Torvalds 13961da177e4SLinus Torvalds /* 139707b20889SRam Pai * recursively change the type of the mountpoint. 139807b20889SRam Pai */ 13990a0d8a46SAl Viro static int do_change_type(struct path *path, int flag) 140007b20889SRam Pai { 14012d92ab3cSAl Viro struct vfsmount *m, *mnt = path->mnt; 140207b20889SRam Pai int recurse = flag & MS_REC; 140307b20889SRam Pai int type = flag & ~MS_REC; 1404719f5d7fSMiklos Szeredi int err = 0; 140507b20889SRam Pai 1406ee6f9582SMiklos Szeredi if (!capable(CAP_SYS_ADMIN)) 1407ee6f9582SMiklos Szeredi return -EPERM; 1408ee6f9582SMiklos Szeredi 14092d92ab3cSAl Viro if (path->dentry != path->mnt->mnt_root) 141007b20889SRam Pai return -EINVAL; 141107b20889SRam Pai 141207b20889SRam Pai down_write(&namespace_sem); 1413719f5d7fSMiklos Szeredi if (type == MS_SHARED) { 1414719f5d7fSMiklos Szeredi err = invent_group_ids(mnt, recurse); 1415719f5d7fSMiklos Szeredi if (err) 1416719f5d7fSMiklos Szeredi goto out_unlock; 1417719f5d7fSMiklos Szeredi } 1418719f5d7fSMiklos Szeredi 141907b20889SRam Pai spin_lock(&vfsmount_lock); 142007b20889SRam Pai for (m = mnt; m; m = (recurse ? next_mnt(m, mnt) : NULL)) 142107b20889SRam Pai change_mnt_propagation(m, type); 142207b20889SRam Pai spin_unlock(&vfsmount_lock); 1423719f5d7fSMiklos Szeredi 1424719f5d7fSMiklos Szeredi out_unlock: 142507b20889SRam Pai up_write(&namespace_sem); 1426719f5d7fSMiklos Szeredi return err; 142707b20889SRam Pai } 142807b20889SRam Pai 142907b20889SRam Pai /* 14301da177e4SLinus Torvalds * do loopback mount. 14311da177e4SLinus Torvalds */ 14320a0d8a46SAl Viro static int do_loopback(struct path *path, char *old_name, 14332dafe1c4SEric Sandeen int recurse) 14341da177e4SLinus Torvalds { 14352d92ab3cSAl Viro struct path old_path; 14361da177e4SLinus Torvalds struct vfsmount *mnt = NULL; 14372d92ab3cSAl Viro int err = mount_is_safe(path); 14381da177e4SLinus Torvalds if (err) 14391da177e4SLinus Torvalds return err; 14401da177e4SLinus Torvalds if (!old_name || !*old_name) 14411da177e4SLinus Torvalds return -EINVAL; 14422d92ab3cSAl Viro err = kern_path(old_name, LOOKUP_FOLLOW, &old_path); 14431da177e4SLinus Torvalds if (err) 14441da177e4SLinus Torvalds return err; 14451da177e4SLinus Torvalds 1446390c6843SRam Pai down_write(&namespace_sem); 14471da177e4SLinus Torvalds err = -EINVAL; 14482d92ab3cSAl Viro if (IS_MNT_UNBINDABLE(old_path.mnt)) 14499676f0c6SRam Pai goto out; 14509676f0c6SRam Pai 14512d92ab3cSAl Viro if (!check_mnt(path->mnt) || !check_mnt(old_path.mnt)) 1452ccd48bc7SAl Viro goto out; 1453ccd48bc7SAl Viro 14541da177e4SLinus Torvalds err = -ENOMEM; 14551da177e4SLinus Torvalds if (recurse) 14562d92ab3cSAl Viro mnt = copy_tree(old_path.mnt, old_path.dentry, 0); 14571da177e4SLinus Torvalds else 14582d92ab3cSAl Viro mnt = clone_mnt(old_path.mnt, old_path.dentry, 0); 14591da177e4SLinus Torvalds 1460ccd48bc7SAl Viro if (!mnt) 1461ccd48bc7SAl Viro goto out; 1462ccd48bc7SAl Viro 14632d92ab3cSAl Viro err = graft_tree(mnt, path); 14641da177e4SLinus Torvalds if (err) { 146570fbcdf4SRam Pai LIST_HEAD(umount_list); 14661da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 1467a05964f3SRam Pai umount_tree(mnt, 0, &umount_list); 14681da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 146970fbcdf4SRam Pai release_mounts(&umount_list); 14705b83d2c5SRam Pai } 14711da177e4SLinus Torvalds 1472ccd48bc7SAl Viro out: 1473390c6843SRam Pai up_write(&namespace_sem); 14742d92ab3cSAl Viro path_put(&old_path); 14751da177e4SLinus Torvalds return err; 14761da177e4SLinus Torvalds } 14771da177e4SLinus Torvalds 14782e4b7fcdSDave Hansen static int change_mount_flags(struct vfsmount *mnt, int ms_flags) 14792e4b7fcdSDave Hansen { 14802e4b7fcdSDave Hansen int error = 0; 14812e4b7fcdSDave Hansen int readonly_request = 0; 14822e4b7fcdSDave Hansen 14832e4b7fcdSDave Hansen if (ms_flags & MS_RDONLY) 14842e4b7fcdSDave Hansen readonly_request = 1; 14852e4b7fcdSDave Hansen if (readonly_request == __mnt_is_readonly(mnt)) 14862e4b7fcdSDave Hansen return 0; 14872e4b7fcdSDave Hansen 14882e4b7fcdSDave Hansen if (readonly_request) 14892e4b7fcdSDave Hansen error = mnt_make_readonly(mnt); 14902e4b7fcdSDave Hansen else 14912e4b7fcdSDave Hansen __mnt_unmake_readonly(mnt); 14922e4b7fcdSDave Hansen return error; 14932e4b7fcdSDave Hansen } 14942e4b7fcdSDave Hansen 14951da177e4SLinus Torvalds /* 14961da177e4SLinus Torvalds * change filesystem flags. dir should be a physical root of filesystem. 14971da177e4SLinus Torvalds * If you've mounted a non-root directory somewhere and want to do remount 14981da177e4SLinus Torvalds * on it - tough luck. 14991da177e4SLinus Torvalds */ 15000a0d8a46SAl Viro static int do_remount(struct path *path, int flags, int mnt_flags, 15011da177e4SLinus Torvalds void *data) 15021da177e4SLinus Torvalds { 15031da177e4SLinus Torvalds int err; 15042d92ab3cSAl Viro struct super_block *sb = path->mnt->mnt_sb; 15051da177e4SLinus Torvalds 15061da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 15071da177e4SLinus Torvalds return -EPERM; 15081da177e4SLinus Torvalds 15092d92ab3cSAl Viro if (!check_mnt(path->mnt)) 15101da177e4SLinus Torvalds return -EINVAL; 15111da177e4SLinus Torvalds 15122d92ab3cSAl Viro if (path->dentry != path->mnt->mnt_root) 15131da177e4SLinus Torvalds return -EINVAL; 15141da177e4SLinus Torvalds 15151da177e4SLinus Torvalds down_write(&sb->s_umount); 15162e4b7fcdSDave Hansen if (flags & MS_BIND) 15172d92ab3cSAl Viro err = change_mount_flags(path->mnt, flags); 1518*7f78d4cdSAl Viro else { 1519*7f78d4cdSAl Viro lock_kernel(); 15201da177e4SLinus Torvalds err = do_remount_sb(sb, flags, data, 0); 1521*7f78d4cdSAl Viro unlock_kernel(); 1522*7f78d4cdSAl Viro } 15231da177e4SLinus Torvalds if (!err) 15242d92ab3cSAl Viro path->mnt->mnt_flags = mnt_flags; 15251da177e4SLinus Torvalds up_write(&sb->s_umount); 15260e55a7ccSDan Williams if (!err) { 15272d92ab3cSAl Viro security_sb_post_remount(path->mnt, flags, data); 15280e55a7ccSDan Williams 15290e55a7ccSDan Williams spin_lock(&vfsmount_lock); 15300e55a7ccSDan Williams touch_mnt_namespace(path->mnt->mnt_ns); 15310e55a7ccSDan Williams spin_unlock(&vfsmount_lock); 15320e55a7ccSDan Williams } 15331da177e4SLinus Torvalds return err; 15341da177e4SLinus Torvalds } 15351da177e4SLinus Torvalds 15369676f0c6SRam Pai static inline int tree_contains_unbindable(struct vfsmount *mnt) 15379676f0c6SRam Pai { 15389676f0c6SRam Pai struct vfsmount *p; 15399676f0c6SRam Pai for (p = mnt; p; p = next_mnt(p, mnt)) { 15409676f0c6SRam Pai if (IS_MNT_UNBINDABLE(p)) 15419676f0c6SRam Pai return 1; 15429676f0c6SRam Pai } 15439676f0c6SRam Pai return 0; 15449676f0c6SRam Pai } 15459676f0c6SRam Pai 15460a0d8a46SAl Viro static int do_move_mount(struct path *path, char *old_name) 15471da177e4SLinus Torvalds { 15482d92ab3cSAl Viro struct path old_path, parent_path; 15491da177e4SLinus Torvalds struct vfsmount *p; 15501da177e4SLinus Torvalds int err = 0; 15511da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 15521da177e4SLinus Torvalds return -EPERM; 15531da177e4SLinus Torvalds if (!old_name || !*old_name) 15541da177e4SLinus Torvalds return -EINVAL; 15552d92ab3cSAl Viro err = kern_path(old_name, LOOKUP_FOLLOW, &old_path); 15561da177e4SLinus Torvalds if (err) 15571da177e4SLinus Torvalds return err; 15581da177e4SLinus Torvalds 1559390c6843SRam Pai down_write(&namespace_sem); 15602d92ab3cSAl Viro while (d_mountpoint(path->dentry) && 15619393bd07SAl Viro follow_down(path)) 15621da177e4SLinus Torvalds ; 15631da177e4SLinus Torvalds err = -EINVAL; 15642d92ab3cSAl Viro if (!check_mnt(path->mnt) || !check_mnt(old_path.mnt)) 15651da177e4SLinus Torvalds goto out; 15661da177e4SLinus Torvalds 15671da177e4SLinus Torvalds err = -ENOENT; 15682d92ab3cSAl Viro mutex_lock(&path->dentry->d_inode->i_mutex); 15692d92ab3cSAl Viro if (IS_DEADDIR(path->dentry->d_inode)) 15701da177e4SLinus Torvalds goto out1; 15711da177e4SLinus Torvalds 1572f3da392eSAlexey Dobriyan if (d_unlinked(path->dentry)) 157321444403SRam Pai goto out1; 15741da177e4SLinus Torvalds 15751da177e4SLinus Torvalds err = -EINVAL; 15762d92ab3cSAl Viro if (old_path.dentry != old_path.mnt->mnt_root) 157721444403SRam Pai goto out1; 15781da177e4SLinus Torvalds 15792d92ab3cSAl Viro if (old_path.mnt == old_path.mnt->mnt_parent) 158021444403SRam Pai goto out1; 15811da177e4SLinus Torvalds 15822d92ab3cSAl Viro if (S_ISDIR(path->dentry->d_inode->i_mode) != 15832d92ab3cSAl Viro S_ISDIR(old_path.dentry->d_inode->i_mode)) 158421444403SRam Pai goto out1; 158521444403SRam Pai /* 158621444403SRam Pai * Don't move a mount residing in a shared parent. 158721444403SRam Pai */ 15882d92ab3cSAl Viro if (old_path.mnt->mnt_parent && 15892d92ab3cSAl Viro IS_MNT_SHARED(old_path.mnt->mnt_parent)) 159021444403SRam Pai goto out1; 15919676f0c6SRam Pai /* 15929676f0c6SRam Pai * Don't move a mount tree containing unbindable mounts to a destination 15939676f0c6SRam Pai * mount which is shared. 15949676f0c6SRam Pai */ 15952d92ab3cSAl Viro if (IS_MNT_SHARED(path->mnt) && 15962d92ab3cSAl Viro tree_contains_unbindable(old_path.mnt)) 15979676f0c6SRam Pai goto out1; 15981da177e4SLinus Torvalds err = -ELOOP; 15992d92ab3cSAl Viro for (p = path->mnt; p->mnt_parent != p; p = p->mnt_parent) 16002d92ab3cSAl Viro if (p == old_path.mnt) 160121444403SRam Pai goto out1; 16021da177e4SLinus Torvalds 16032d92ab3cSAl Viro err = attach_recursive_mnt(old_path.mnt, path, &parent_path); 16044ac91378SJan Blunck if (err) 160521444403SRam Pai goto out1; 16061da177e4SLinus Torvalds 16071da177e4SLinus Torvalds /* if the mount is moved, it should no longer be expire 16081da177e4SLinus Torvalds * automatically */ 16092d92ab3cSAl Viro list_del_init(&old_path.mnt->mnt_expire); 16101da177e4SLinus Torvalds out1: 16112d92ab3cSAl Viro mutex_unlock(&path->dentry->d_inode->i_mutex); 16121da177e4SLinus Torvalds out: 1613390c6843SRam Pai up_write(&namespace_sem); 16141da177e4SLinus Torvalds if (!err) 16151a390689SAl Viro path_put(&parent_path); 16162d92ab3cSAl Viro path_put(&old_path); 16171da177e4SLinus Torvalds return err; 16181da177e4SLinus Torvalds } 16191da177e4SLinus Torvalds 16201da177e4SLinus Torvalds /* 16211da177e4SLinus Torvalds * create a new mount for userspace and request it to be added into the 16221da177e4SLinus Torvalds * namespace's tree 16231da177e4SLinus Torvalds */ 16240a0d8a46SAl Viro static int do_new_mount(struct path *path, char *type, int flags, 16251da177e4SLinus Torvalds int mnt_flags, char *name, void *data) 16261da177e4SLinus Torvalds { 16271da177e4SLinus Torvalds struct vfsmount *mnt; 16281da177e4SLinus Torvalds 16291da177e4SLinus Torvalds if (!type || !memchr(type, 0, PAGE_SIZE)) 16301da177e4SLinus Torvalds return -EINVAL; 16311da177e4SLinus Torvalds 16321da177e4SLinus Torvalds /* we need capabilities... */ 16331da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 16341da177e4SLinus Torvalds return -EPERM; 16351da177e4SLinus Torvalds 1636*7f78d4cdSAl Viro lock_kernel(); 16371da177e4SLinus Torvalds mnt = do_kern_mount(type, flags, name, data); 1638*7f78d4cdSAl Viro unlock_kernel(); 16391da177e4SLinus Torvalds if (IS_ERR(mnt)) 16401da177e4SLinus Torvalds return PTR_ERR(mnt); 16411da177e4SLinus Torvalds 16422d92ab3cSAl Viro return do_add_mount(mnt, path, mnt_flags, NULL); 16431da177e4SLinus Torvalds } 16441da177e4SLinus Torvalds 16451da177e4SLinus Torvalds /* 16461da177e4SLinus Torvalds * add a mount into a namespace's mount tree 16471da177e4SLinus Torvalds * - provide the option of adding the new mount to an expiration list 16481da177e4SLinus Torvalds */ 16498d66bf54SAl Viro int do_add_mount(struct vfsmount *newmnt, struct path *path, 16501da177e4SLinus Torvalds int mnt_flags, struct list_head *fslist) 16511da177e4SLinus Torvalds { 16521da177e4SLinus Torvalds int err; 16531da177e4SLinus Torvalds 1654390c6843SRam Pai down_write(&namespace_sem); 16551da177e4SLinus Torvalds /* Something was mounted here while we slept */ 16568d66bf54SAl Viro while (d_mountpoint(path->dentry) && 16579393bd07SAl Viro follow_down(path)) 16581da177e4SLinus Torvalds ; 16591da177e4SLinus Torvalds err = -EINVAL; 1660dd5cae6eSAl Viro if (!(mnt_flags & MNT_SHRINKABLE) && !check_mnt(path->mnt)) 16611da177e4SLinus Torvalds goto unlock; 16621da177e4SLinus Torvalds 16631da177e4SLinus Torvalds /* Refuse the same filesystem on the same mount point */ 16641da177e4SLinus Torvalds err = -EBUSY; 16658d66bf54SAl Viro if (path->mnt->mnt_sb == newmnt->mnt_sb && 16668d66bf54SAl Viro path->mnt->mnt_root == path->dentry) 16671da177e4SLinus Torvalds goto unlock; 16681da177e4SLinus Torvalds 16691da177e4SLinus Torvalds err = -EINVAL; 16701da177e4SLinus Torvalds if (S_ISLNK(newmnt->mnt_root->d_inode->i_mode)) 16711da177e4SLinus Torvalds goto unlock; 16721da177e4SLinus Torvalds 16731da177e4SLinus Torvalds newmnt->mnt_flags = mnt_flags; 16748d66bf54SAl Viro if ((err = graft_tree(newmnt, path))) 16755b83d2c5SRam Pai goto unlock; 16761da177e4SLinus Torvalds 16776758f953SAl Viro if (fslist) /* add to the specified expiration list */ 167855e700b9SMiklos Szeredi list_add_tail(&newmnt->mnt_expire, fslist); 16796758f953SAl Viro 1680390c6843SRam Pai up_write(&namespace_sem); 16815b83d2c5SRam Pai return 0; 16821da177e4SLinus Torvalds 16831da177e4SLinus Torvalds unlock: 1684390c6843SRam Pai up_write(&namespace_sem); 16851da177e4SLinus Torvalds mntput(newmnt); 16861da177e4SLinus Torvalds return err; 16871da177e4SLinus Torvalds } 16881da177e4SLinus Torvalds 16891da177e4SLinus Torvalds EXPORT_SYMBOL_GPL(do_add_mount); 16901da177e4SLinus Torvalds 16915528f911STrond Myklebust /* 16921da177e4SLinus Torvalds * process a list of expirable mountpoints with the intent of discarding any 16931da177e4SLinus Torvalds * mountpoints that aren't in use and haven't been touched since last we came 16941da177e4SLinus Torvalds * here 16951da177e4SLinus Torvalds */ 16961da177e4SLinus Torvalds void mark_mounts_for_expiry(struct list_head *mounts) 16971da177e4SLinus Torvalds { 16981da177e4SLinus Torvalds struct vfsmount *mnt, *next; 16991da177e4SLinus Torvalds LIST_HEAD(graveyard); 1700bcc5c7d2SAl Viro LIST_HEAD(umounts); 17011da177e4SLinus Torvalds 17021da177e4SLinus Torvalds if (list_empty(mounts)) 17031da177e4SLinus Torvalds return; 17041da177e4SLinus Torvalds 1705bcc5c7d2SAl Viro down_write(&namespace_sem); 17061da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 17071da177e4SLinus Torvalds 17081da177e4SLinus Torvalds /* extract from the expiration list every vfsmount that matches the 17091da177e4SLinus Torvalds * following criteria: 17101da177e4SLinus Torvalds * - only referenced by its parent vfsmount 17111da177e4SLinus Torvalds * - still marked for expiry (marked on the last call here; marks are 17121da177e4SLinus Torvalds * cleared by mntput()) 17131da177e4SLinus Torvalds */ 171455e700b9SMiklos Szeredi list_for_each_entry_safe(mnt, next, mounts, mnt_expire) { 17151da177e4SLinus Torvalds if (!xchg(&mnt->mnt_expiry_mark, 1) || 1716bcc5c7d2SAl Viro propagate_mount_busy(mnt, 1)) 17171da177e4SLinus Torvalds continue; 171855e700b9SMiklos Szeredi list_move(&mnt->mnt_expire, &graveyard); 17191da177e4SLinus Torvalds } 1720bcc5c7d2SAl Viro while (!list_empty(&graveyard)) { 1721bcc5c7d2SAl Viro mnt = list_first_entry(&graveyard, struct vfsmount, mnt_expire); 1722bcc5c7d2SAl Viro touch_mnt_namespace(mnt->mnt_ns); 1723bcc5c7d2SAl Viro umount_tree(mnt, 1, &umounts); 1724bcc5c7d2SAl Viro } 17251da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 1726bcc5c7d2SAl Viro up_write(&namespace_sem); 1727bcc5c7d2SAl Viro 1728bcc5c7d2SAl Viro release_mounts(&umounts); 17291da177e4SLinus Torvalds } 17301da177e4SLinus Torvalds 17311da177e4SLinus Torvalds EXPORT_SYMBOL_GPL(mark_mounts_for_expiry); 17321da177e4SLinus Torvalds 17331da177e4SLinus Torvalds /* 17345528f911STrond Myklebust * Ripoff of 'select_parent()' 17355528f911STrond Myklebust * 17365528f911STrond Myklebust * search the list of submounts for a given mountpoint, and move any 17375528f911STrond Myklebust * shrinkable submounts to the 'graveyard' list. 17385528f911STrond Myklebust */ 17395528f911STrond Myklebust static int select_submounts(struct vfsmount *parent, struct list_head *graveyard) 17405528f911STrond Myklebust { 17415528f911STrond Myklebust struct vfsmount *this_parent = parent; 17425528f911STrond Myklebust struct list_head *next; 17435528f911STrond Myklebust int found = 0; 17445528f911STrond Myklebust 17455528f911STrond Myklebust repeat: 17465528f911STrond Myklebust next = this_parent->mnt_mounts.next; 17475528f911STrond Myklebust resume: 17485528f911STrond Myklebust while (next != &this_parent->mnt_mounts) { 17495528f911STrond Myklebust struct list_head *tmp = next; 17505528f911STrond Myklebust struct vfsmount *mnt = list_entry(tmp, struct vfsmount, mnt_child); 17515528f911STrond Myklebust 17525528f911STrond Myklebust next = tmp->next; 17535528f911STrond Myklebust if (!(mnt->mnt_flags & MNT_SHRINKABLE)) 17545528f911STrond Myklebust continue; 17555528f911STrond Myklebust /* 17565528f911STrond Myklebust * Descend a level if the d_mounts list is non-empty. 17575528f911STrond Myklebust */ 17585528f911STrond Myklebust if (!list_empty(&mnt->mnt_mounts)) { 17595528f911STrond Myklebust this_parent = mnt; 17605528f911STrond Myklebust goto repeat; 17615528f911STrond Myklebust } 17625528f911STrond Myklebust 17635528f911STrond Myklebust if (!propagate_mount_busy(mnt, 1)) { 17645528f911STrond Myklebust list_move_tail(&mnt->mnt_expire, graveyard); 17655528f911STrond Myklebust found++; 17665528f911STrond Myklebust } 17675528f911STrond Myklebust } 17685528f911STrond Myklebust /* 17695528f911STrond Myklebust * All done at this level ... ascend and resume the search 17705528f911STrond Myklebust */ 17715528f911STrond Myklebust if (this_parent != parent) { 17725528f911STrond Myklebust next = this_parent->mnt_child.next; 17735528f911STrond Myklebust this_parent = this_parent->mnt_parent; 17745528f911STrond Myklebust goto resume; 17755528f911STrond Myklebust } 17765528f911STrond Myklebust return found; 17775528f911STrond Myklebust } 17785528f911STrond Myklebust 17795528f911STrond Myklebust /* 17805528f911STrond Myklebust * process a list of expirable mountpoints with the intent of discarding any 17815528f911STrond Myklebust * submounts of a specific parent mountpoint 17825528f911STrond Myklebust */ 1783c35038beSAl Viro static void shrink_submounts(struct vfsmount *mnt, struct list_head *umounts) 17845528f911STrond Myklebust { 17855528f911STrond Myklebust LIST_HEAD(graveyard); 1786c35038beSAl Viro struct vfsmount *m; 17875528f911STrond Myklebust 17885528f911STrond Myklebust /* extract submounts of 'mountpoint' from the expiration list */ 1789c35038beSAl Viro while (select_submounts(mnt, &graveyard)) { 1790bcc5c7d2SAl Viro while (!list_empty(&graveyard)) { 1791c35038beSAl Viro m = list_first_entry(&graveyard, struct vfsmount, 1792bcc5c7d2SAl Viro mnt_expire); 1793afef80b3SEric W. Biederman touch_mnt_namespace(m->mnt_ns); 1794afef80b3SEric W. Biederman umount_tree(m, 1, umounts); 1795bcc5c7d2SAl Viro } 1796bcc5c7d2SAl Viro } 17975528f911STrond Myklebust } 17985528f911STrond Myklebust 17995528f911STrond Myklebust /* 18001da177e4SLinus Torvalds * Some copy_from_user() implementations do not return the exact number of 18011da177e4SLinus Torvalds * bytes remaining to copy on a fault. But copy_mount_options() requires that. 18021da177e4SLinus Torvalds * Note that this function differs from copy_from_user() in that it will oops 18031da177e4SLinus Torvalds * on bad values of `to', rather than returning a short copy. 18041da177e4SLinus Torvalds */ 1805b58fed8bSRam Pai static long exact_copy_from_user(void *to, const void __user * from, 1806b58fed8bSRam Pai unsigned long n) 18071da177e4SLinus Torvalds { 18081da177e4SLinus Torvalds char *t = to; 18091da177e4SLinus Torvalds const char __user *f = from; 18101da177e4SLinus Torvalds char c; 18111da177e4SLinus Torvalds 18121da177e4SLinus Torvalds if (!access_ok(VERIFY_READ, from, n)) 18131da177e4SLinus Torvalds return n; 18141da177e4SLinus Torvalds 18151da177e4SLinus Torvalds while (n) { 18161da177e4SLinus Torvalds if (__get_user(c, f)) { 18171da177e4SLinus Torvalds memset(t, 0, n); 18181da177e4SLinus Torvalds break; 18191da177e4SLinus Torvalds } 18201da177e4SLinus Torvalds *t++ = c; 18211da177e4SLinus Torvalds f++; 18221da177e4SLinus Torvalds n--; 18231da177e4SLinus Torvalds } 18241da177e4SLinus Torvalds return n; 18251da177e4SLinus Torvalds } 18261da177e4SLinus Torvalds 18271da177e4SLinus Torvalds int copy_mount_options(const void __user * data, unsigned long *where) 18281da177e4SLinus Torvalds { 18291da177e4SLinus Torvalds int i; 18301da177e4SLinus Torvalds unsigned long page; 18311da177e4SLinus Torvalds unsigned long size; 18321da177e4SLinus Torvalds 18331da177e4SLinus Torvalds *where = 0; 18341da177e4SLinus Torvalds if (!data) 18351da177e4SLinus Torvalds return 0; 18361da177e4SLinus Torvalds 18371da177e4SLinus Torvalds if (!(page = __get_free_page(GFP_KERNEL))) 18381da177e4SLinus Torvalds return -ENOMEM; 18391da177e4SLinus Torvalds 18401da177e4SLinus Torvalds /* We only care that *some* data at the address the user 18411da177e4SLinus Torvalds * gave us is valid. Just in case, we'll zero 18421da177e4SLinus Torvalds * the remainder of the page. 18431da177e4SLinus Torvalds */ 18441da177e4SLinus Torvalds /* copy_from_user cannot cross TASK_SIZE ! */ 18451da177e4SLinus Torvalds size = TASK_SIZE - (unsigned long)data; 18461da177e4SLinus Torvalds if (size > PAGE_SIZE) 18471da177e4SLinus Torvalds size = PAGE_SIZE; 18481da177e4SLinus Torvalds 18491da177e4SLinus Torvalds i = size - exact_copy_from_user((void *)page, data, size); 18501da177e4SLinus Torvalds if (!i) { 18511da177e4SLinus Torvalds free_page(page); 18521da177e4SLinus Torvalds return -EFAULT; 18531da177e4SLinus Torvalds } 18541da177e4SLinus Torvalds if (i != PAGE_SIZE) 18551da177e4SLinus Torvalds memset((char *)page + i, 0, PAGE_SIZE - i); 18561da177e4SLinus Torvalds *where = page; 18571da177e4SLinus Torvalds return 0; 18581da177e4SLinus Torvalds } 18591da177e4SLinus Torvalds 18601da177e4SLinus Torvalds /* 18611da177e4SLinus Torvalds * Flags is a 32-bit value that allows up to 31 non-fs dependent flags to 18621da177e4SLinus Torvalds * be given to the mount() call (ie: read-only, no-dev, no-suid etc). 18631da177e4SLinus Torvalds * 18641da177e4SLinus Torvalds * data is a (void *) that can point to any structure up to 18651da177e4SLinus Torvalds * PAGE_SIZE-1 bytes, which can contain arbitrary fs-dependent 18661da177e4SLinus Torvalds * information (or be NULL). 18671da177e4SLinus Torvalds * 18681da177e4SLinus Torvalds * Pre-0.97 versions of mount() didn't have a flags word. 18691da177e4SLinus Torvalds * When the flags word was introduced its top half was required 18701da177e4SLinus Torvalds * to have the magic value 0xC0ED, and this remained so until 2.4.0-test9. 18711da177e4SLinus Torvalds * Therefore, if this magic number is present, it carries no information 18721da177e4SLinus Torvalds * and must be discarded. 18731da177e4SLinus Torvalds */ 18741da177e4SLinus Torvalds long do_mount(char *dev_name, char *dir_name, char *type_page, 18751da177e4SLinus Torvalds unsigned long flags, void *data_page) 18761da177e4SLinus Torvalds { 18772d92ab3cSAl Viro struct path path; 18781da177e4SLinus Torvalds int retval = 0; 18791da177e4SLinus Torvalds int mnt_flags = 0; 18801da177e4SLinus Torvalds 18811da177e4SLinus Torvalds /* Discard magic */ 18821da177e4SLinus Torvalds if ((flags & MS_MGC_MSK) == MS_MGC_VAL) 18831da177e4SLinus Torvalds flags &= ~MS_MGC_MSK; 18841da177e4SLinus Torvalds 18851da177e4SLinus Torvalds /* Basic sanity checks */ 18861da177e4SLinus Torvalds 18871da177e4SLinus Torvalds if (!dir_name || !*dir_name || !memchr(dir_name, 0, PAGE_SIZE)) 18881da177e4SLinus Torvalds return -EINVAL; 18891da177e4SLinus Torvalds if (dev_name && !memchr(dev_name, 0, PAGE_SIZE)) 18901da177e4SLinus Torvalds return -EINVAL; 18911da177e4SLinus Torvalds 18921da177e4SLinus Torvalds if (data_page) 18931da177e4SLinus Torvalds ((char *)data_page)[PAGE_SIZE - 1] = 0; 18941da177e4SLinus Torvalds 1895613cbe3dSAndi Kleen /* Default to relatime unless overriden */ 1896613cbe3dSAndi Kleen if (!(flags & MS_NOATIME)) 18970a1c01c9SMatthew Garrett mnt_flags |= MNT_RELATIME; 18980a1c01c9SMatthew Garrett 18991da177e4SLinus Torvalds /* Separate the per-mountpoint flags */ 19001da177e4SLinus Torvalds if (flags & MS_NOSUID) 19011da177e4SLinus Torvalds mnt_flags |= MNT_NOSUID; 19021da177e4SLinus Torvalds if (flags & MS_NODEV) 19031da177e4SLinus Torvalds mnt_flags |= MNT_NODEV; 19041da177e4SLinus Torvalds if (flags & MS_NOEXEC) 19051da177e4SLinus Torvalds mnt_flags |= MNT_NOEXEC; 1906fc33a7bbSChristoph Hellwig if (flags & MS_NOATIME) 1907fc33a7bbSChristoph Hellwig mnt_flags |= MNT_NOATIME; 1908fc33a7bbSChristoph Hellwig if (flags & MS_NODIRATIME) 1909fc33a7bbSChristoph Hellwig mnt_flags |= MNT_NODIRATIME; 1910d0adde57SMatthew Garrett if (flags & MS_STRICTATIME) 1911d0adde57SMatthew Garrett mnt_flags &= ~(MNT_RELATIME | MNT_NOATIME); 19122e4b7fcdSDave Hansen if (flags & MS_RDONLY) 19132e4b7fcdSDave Hansen mnt_flags |= MNT_READONLY; 1914fc33a7bbSChristoph Hellwig 1915fc33a7bbSChristoph Hellwig flags &= ~(MS_NOSUID | MS_NOEXEC | MS_NODEV | MS_ACTIVE | 1916d0adde57SMatthew Garrett MS_NOATIME | MS_NODIRATIME | MS_RELATIME| MS_KERNMOUNT | 1917d0adde57SMatthew Garrett MS_STRICTATIME); 19181da177e4SLinus Torvalds 19191da177e4SLinus Torvalds /* ... and get the mountpoint */ 19202d92ab3cSAl Viro retval = kern_path(dir_name, LOOKUP_FOLLOW, &path); 19211da177e4SLinus Torvalds if (retval) 19221da177e4SLinus Torvalds return retval; 19231da177e4SLinus Torvalds 19242d92ab3cSAl Viro retval = security_sb_mount(dev_name, &path, 1925b5266eb4SAl Viro type_page, flags, data_page); 19261da177e4SLinus Torvalds if (retval) 19271da177e4SLinus Torvalds goto dput_out; 19281da177e4SLinus Torvalds 19291da177e4SLinus Torvalds if (flags & MS_REMOUNT) 19302d92ab3cSAl Viro retval = do_remount(&path, flags & ~MS_REMOUNT, mnt_flags, 19311da177e4SLinus Torvalds data_page); 19321da177e4SLinus Torvalds else if (flags & MS_BIND) 19332d92ab3cSAl Viro retval = do_loopback(&path, dev_name, flags & MS_REC); 19349676f0c6SRam Pai else if (flags & (MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE)) 19352d92ab3cSAl Viro retval = do_change_type(&path, flags); 19361da177e4SLinus Torvalds else if (flags & MS_MOVE) 19372d92ab3cSAl Viro retval = do_move_mount(&path, dev_name); 19381da177e4SLinus Torvalds else 19392d92ab3cSAl Viro retval = do_new_mount(&path, type_page, flags, mnt_flags, 19401da177e4SLinus Torvalds dev_name, data_page); 19411da177e4SLinus Torvalds dput_out: 19422d92ab3cSAl Viro path_put(&path); 19431da177e4SLinus Torvalds return retval; 19441da177e4SLinus Torvalds } 19451da177e4SLinus Torvalds 1946741a2951SJANAK DESAI /* 1947741a2951SJANAK DESAI * Allocate a new namespace structure and populate it with contents 1948741a2951SJANAK DESAI * copied from the namespace of the passed in task structure. 1949741a2951SJANAK DESAI */ 1950e3222c4eSBadari Pulavarty static struct mnt_namespace *dup_mnt_ns(struct mnt_namespace *mnt_ns, 19516b3286edSKirill Korotaev struct fs_struct *fs) 19521da177e4SLinus Torvalds { 19536b3286edSKirill Korotaev struct mnt_namespace *new_ns; 19547f2da1e7SAl Viro struct vfsmount *rootmnt = NULL, *pwdmnt = NULL; 19551da177e4SLinus Torvalds struct vfsmount *p, *q; 19561da177e4SLinus Torvalds 19576b3286edSKirill Korotaev new_ns = kmalloc(sizeof(struct mnt_namespace), GFP_KERNEL); 19581da177e4SLinus Torvalds if (!new_ns) 1959467e9f4bSCedric Le Goater return ERR_PTR(-ENOMEM); 19601da177e4SLinus Torvalds 19611da177e4SLinus Torvalds atomic_set(&new_ns->count, 1); 19621da177e4SLinus Torvalds INIT_LIST_HEAD(&new_ns->list); 19635addc5ddSAl Viro init_waitqueue_head(&new_ns->poll); 19645addc5ddSAl Viro new_ns->event = 0; 19651da177e4SLinus Torvalds 1966390c6843SRam Pai down_write(&namespace_sem); 19671da177e4SLinus Torvalds /* First pass: copy the tree topology */ 19686b3286edSKirill Korotaev new_ns->root = copy_tree(mnt_ns->root, mnt_ns->root->mnt_root, 19699676f0c6SRam Pai CL_COPY_ALL | CL_EXPIRE); 19701da177e4SLinus Torvalds if (!new_ns->root) { 1971390c6843SRam Pai up_write(&namespace_sem); 19721da177e4SLinus Torvalds kfree(new_ns); 19735cc4a034SJulia Lawall return ERR_PTR(-ENOMEM); 19741da177e4SLinus Torvalds } 19751da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 19761da177e4SLinus Torvalds list_add_tail(&new_ns->list, &new_ns->root->mnt_list); 19771da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 19781da177e4SLinus Torvalds 19791da177e4SLinus Torvalds /* 19801da177e4SLinus Torvalds * Second pass: switch the tsk->fs->* elements and mark new vfsmounts 19811da177e4SLinus Torvalds * as belonging to new namespace. We have already acquired a private 19821da177e4SLinus Torvalds * fs_struct, so tsk->fs->lock is not needed. 19831da177e4SLinus Torvalds */ 19846b3286edSKirill Korotaev p = mnt_ns->root; 19851da177e4SLinus Torvalds q = new_ns->root; 19861da177e4SLinus Torvalds while (p) { 19876b3286edSKirill Korotaev q->mnt_ns = new_ns; 19881da177e4SLinus Torvalds if (fs) { 19896ac08c39SJan Blunck if (p == fs->root.mnt) { 19901da177e4SLinus Torvalds rootmnt = p; 19916ac08c39SJan Blunck fs->root.mnt = mntget(q); 19921da177e4SLinus Torvalds } 19936ac08c39SJan Blunck if (p == fs->pwd.mnt) { 19941da177e4SLinus Torvalds pwdmnt = p; 19956ac08c39SJan Blunck fs->pwd.mnt = mntget(q); 19961da177e4SLinus Torvalds } 19971da177e4SLinus Torvalds } 19986b3286edSKirill Korotaev p = next_mnt(p, mnt_ns->root); 19991da177e4SLinus Torvalds q = next_mnt(q, new_ns->root); 20001da177e4SLinus Torvalds } 2001390c6843SRam Pai up_write(&namespace_sem); 20021da177e4SLinus Torvalds 20031da177e4SLinus Torvalds if (rootmnt) 20041da177e4SLinus Torvalds mntput(rootmnt); 20051da177e4SLinus Torvalds if (pwdmnt) 20061da177e4SLinus Torvalds mntput(pwdmnt); 20071da177e4SLinus Torvalds 2008741a2951SJANAK DESAI return new_ns; 2009741a2951SJANAK DESAI } 2010741a2951SJANAK DESAI 2011213dd266SEric W. Biederman struct mnt_namespace *copy_mnt_ns(unsigned long flags, struct mnt_namespace *ns, 2012e3222c4eSBadari Pulavarty struct fs_struct *new_fs) 2013741a2951SJANAK DESAI { 20146b3286edSKirill Korotaev struct mnt_namespace *new_ns; 2015741a2951SJANAK DESAI 2016e3222c4eSBadari Pulavarty BUG_ON(!ns); 20176b3286edSKirill Korotaev get_mnt_ns(ns); 2018741a2951SJANAK DESAI 2019741a2951SJANAK DESAI if (!(flags & CLONE_NEWNS)) 2020e3222c4eSBadari Pulavarty return ns; 2021741a2951SJANAK DESAI 2022e3222c4eSBadari Pulavarty new_ns = dup_mnt_ns(ns, new_fs); 2023741a2951SJANAK DESAI 20246b3286edSKirill Korotaev put_mnt_ns(ns); 2025e3222c4eSBadari Pulavarty return new_ns; 20261da177e4SLinus Torvalds } 20271da177e4SLinus Torvalds 2028bdc480e3SHeiko Carstens SYSCALL_DEFINE5(mount, char __user *, dev_name, char __user *, dir_name, 2029bdc480e3SHeiko Carstens char __user *, type, unsigned long, flags, void __user *, data) 20301da177e4SLinus Torvalds { 20311da177e4SLinus Torvalds int retval; 20321da177e4SLinus Torvalds unsigned long data_page; 20331da177e4SLinus Torvalds unsigned long type_page; 20341da177e4SLinus Torvalds unsigned long dev_page; 20351da177e4SLinus Torvalds char *dir_page; 20361da177e4SLinus Torvalds 20371da177e4SLinus Torvalds retval = copy_mount_options(type, &type_page); 20381da177e4SLinus Torvalds if (retval < 0) 20391da177e4SLinus Torvalds return retval; 20401da177e4SLinus Torvalds 20411da177e4SLinus Torvalds dir_page = getname(dir_name); 20421da177e4SLinus Torvalds retval = PTR_ERR(dir_page); 20431da177e4SLinus Torvalds if (IS_ERR(dir_page)) 20441da177e4SLinus Torvalds goto out1; 20451da177e4SLinus Torvalds 20461da177e4SLinus Torvalds retval = copy_mount_options(dev_name, &dev_page); 20471da177e4SLinus Torvalds if (retval < 0) 20481da177e4SLinus Torvalds goto out2; 20491da177e4SLinus Torvalds 20501da177e4SLinus Torvalds retval = copy_mount_options(data, &data_page); 20511da177e4SLinus Torvalds if (retval < 0) 20521da177e4SLinus Torvalds goto out3; 20531da177e4SLinus Torvalds 20541da177e4SLinus Torvalds retval = do_mount((char *)dev_page, dir_page, (char *)type_page, 20551da177e4SLinus Torvalds flags, (void *)data_page); 20561da177e4SLinus Torvalds free_page(data_page); 20571da177e4SLinus Torvalds 20581da177e4SLinus Torvalds out3: 20591da177e4SLinus Torvalds free_page(dev_page); 20601da177e4SLinus Torvalds out2: 20611da177e4SLinus Torvalds putname(dir_page); 20621da177e4SLinus Torvalds out1: 20631da177e4SLinus Torvalds free_page(type_page); 20641da177e4SLinus Torvalds return retval; 20651da177e4SLinus Torvalds } 20661da177e4SLinus Torvalds 20671da177e4SLinus Torvalds /* 20681da177e4SLinus Torvalds * pivot_root Semantics: 20691da177e4SLinus Torvalds * Moves the root file system of the current process to the directory put_old, 20701da177e4SLinus Torvalds * makes new_root as the new root file system of the current process, and sets 20711da177e4SLinus Torvalds * root/cwd of all processes which had them on the current root to new_root. 20721da177e4SLinus Torvalds * 20731da177e4SLinus Torvalds * Restrictions: 20741da177e4SLinus Torvalds * The new_root and put_old must be directories, and must not be on the 20751da177e4SLinus Torvalds * same file system as the current process root. The put_old must be 20761da177e4SLinus Torvalds * underneath new_root, i.e. adding a non-zero number of /.. to the string 20771da177e4SLinus Torvalds * pointed to by put_old must yield the same directory as new_root. No other 20781da177e4SLinus Torvalds * file system may be mounted on put_old. After all, new_root is a mountpoint. 20791da177e4SLinus Torvalds * 20804a0d11faSNeil Brown * Also, the current root cannot be on the 'rootfs' (initial ramfs) filesystem. 20814a0d11faSNeil Brown * See Documentation/filesystems/ramfs-rootfs-initramfs.txt for alternatives 20824a0d11faSNeil Brown * in this situation. 20834a0d11faSNeil Brown * 20841da177e4SLinus Torvalds * Notes: 20851da177e4SLinus Torvalds * - we don't move root/cwd if they are not at the root (reason: if something 20861da177e4SLinus Torvalds * cared enough to change them, it's probably wrong to force them elsewhere) 20871da177e4SLinus Torvalds * - it's okay to pick a root that isn't the root of a file system, e.g. 20881da177e4SLinus Torvalds * /nfs/my_root where /nfs is the mount point. It must be a mountpoint, 20891da177e4SLinus Torvalds * though, so you may need to say mount --bind /nfs/my_root /nfs/my_root 20901da177e4SLinus Torvalds * first. 20911da177e4SLinus Torvalds */ 20923480b257SHeiko Carstens SYSCALL_DEFINE2(pivot_root, const char __user *, new_root, 20933480b257SHeiko Carstens const char __user *, put_old) 20941da177e4SLinus Torvalds { 20951da177e4SLinus Torvalds struct vfsmount *tmp; 20962d8f3038SAl Viro struct path new, old, parent_path, root_parent, root; 20971da177e4SLinus Torvalds int error; 20981da177e4SLinus Torvalds 20991da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 21001da177e4SLinus Torvalds return -EPERM; 21011da177e4SLinus Torvalds 21022d8f3038SAl Viro error = user_path_dir(new_root, &new); 21031da177e4SLinus Torvalds if (error) 21041da177e4SLinus Torvalds goto out0; 21051da177e4SLinus Torvalds error = -EINVAL; 21062d8f3038SAl Viro if (!check_mnt(new.mnt)) 21071da177e4SLinus Torvalds goto out1; 21081da177e4SLinus Torvalds 21092d8f3038SAl Viro error = user_path_dir(put_old, &old); 21101da177e4SLinus Torvalds if (error) 21111da177e4SLinus Torvalds goto out1; 21121da177e4SLinus Torvalds 21132d8f3038SAl Viro error = security_sb_pivotroot(&old, &new); 21141da177e4SLinus Torvalds if (error) { 21152d8f3038SAl Viro path_put(&old); 21161da177e4SLinus Torvalds goto out1; 21171da177e4SLinus Torvalds } 21181da177e4SLinus Torvalds 21191da177e4SLinus Torvalds read_lock(¤t->fs->lock); 21208c3ee42eSAl Viro root = current->fs->root; 21216ac08c39SJan Blunck path_get(¤t->fs->root); 21221da177e4SLinus Torvalds read_unlock(¤t->fs->lock); 2123390c6843SRam Pai down_write(&namespace_sem); 21242d8f3038SAl Viro mutex_lock(&old.dentry->d_inode->i_mutex); 21251da177e4SLinus Torvalds error = -EINVAL; 21262d8f3038SAl Viro if (IS_MNT_SHARED(old.mnt) || 21272d8f3038SAl Viro IS_MNT_SHARED(new.mnt->mnt_parent) || 21288c3ee42eSAl Viro IS_MNT_SHARED(root.mnt->mnt_parent)) 212921444403SRam Pai goto out2; 21308c3ee42eSAl Viro if (!check_mnt(root.mnt)) 21311da177e4SLinus Torvalds goto out2; 21321da177e4SLinus Torvalds error = -ENOENT; 21332d8f3038SAl Viro if (IS_DEADDIR(new.dentry->d_inode)) 21341da177e4SLinus Torvalds goto out2; 2135f3da392eSAlexey Dobriyan if (d_unlinked(new.dentry)) 21361da177e4SLinus Torvalds goto out2; 2137f3da392eSAlexey Dobriyan if (d_unlinked(old.dentry)) 21381da177e4SLinus Torvalds goto out2; 21391da177e4SLinus Torvalds error = -EBUSY; 21402d8f3038SAl Viro if (new.mnt == root.mnt || 21412d8f3038SAl Viro old.mnt == root.mnt) 21421da177e4SLinus Torvalds goto out2; /* loop, on the same file system */ 21431da177e4SLinus Torvalds error = -EINVAL; 21448c3ee42eSAl Viro if (root.mnt->mnt_root != root.dentry) 21451da177e4SLinus Torvalds goto out2; /* not a mountpoint */ 21468c3ee42eSAl Viro if (root.mnt->mnt_parent == root.mnt) 21470bb6fcc1SMiklos Szeredi goto out2; /* not attached */ 21482d8f3038SAl Viro if (new.mnt->mnt_root != new.dentry) 21491da177e4SLinus Torvalds goto out2; /* not a mountpoint */ 21502d8f3038SAl Viro if (new.mnt->mnt_parent == new.mnt) 21510bb6fcc1SMiklos Szeredi goto out2; /* not attached */ 21524ac91378SJan Blunck /* make sure we can reach put_old from new_root */ 21532d8f3038SAl Viro tmp = old.mnt; 21541da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 21552d8f3038SAl Viro if (tmp != new.mnt) { 21561da177e4SLinus Torvalds for (;;) { 21571da177e4SLinus Torvalds if (tmp->mnt_parent == tmp) 21581da177e4SLinus Torvalds goto out3; /* already mounted on put_old */ 21592d8f3038SAl Viro if (tmp->mnt_parent == new.mnt) 21601da177e4SLinus Torvalds break; 21611da177e4SLinus Torvalds tmp = tmp->mnt_parent; 21621da177e4SLinus Torvalds } 21632d8f3038SAl Viro if (!is_subdir(tmp->mnt_mountpoint, new.dentry)) 21641da177e4SLinus Torvalds goto out3; 21652d8f3038SAl Viro } else if (!is_subdir(old.dentry, new.dentry)) 21661da177e4SLinus Torvalds goto out3; 21672d8f3038SAl Viro detach_mnt(new.mnt, &parent_path); 21688c3ee42eSAl Viro detach_mnt(root.mnt, &root_parent); 21694ac91378SJan Blunck /* mount old root on put_old */ 21702d8f3038SAl Viro attach_mnt(root.mnt, &old); 21714ac91378SJan Blunck /* mount new_root on / */ 21722d8f3038SAl Viro attach_mnt(new.mnt, &root_parent); 21736b3286edSKirill Korotaev touch_mnt_namespace(current->nsproxy->mnt_ns); 21741da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 21752d8f3038SAl Viro chroot_fs_refs(&root, &new); 21762d8f3038SAl Viro security_sb_post_pivotroot(&root, &new); 21771da177e4SLinus Torvalds error = 0; 21781a390689SAl Viro path_put(&root_parent); 21791a390689SAl Viro path_put(&parent_path); 21801da177e4SLinus Torvalds out2: 21812d8f3038SAl Viro mutex_unlock(&old.dentry->d_inode->i_mutex); 2182390c6843SRam Pai up_write(&namespace_sem); 21838c3ee42eSAl Viro path_put(&root); 21842d8f3038SAl Viro path_put(&old); 21851da177e4SLinus Torvalds out1: 21862d8f3038SAl Viro path_put(&new); 21871da177e4SLinus Torvalds out0: 21881da177e4SLinus Torvalds return error; 21891da177e4SLinus Torvalds out3: 21901da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 21911da177e4SLinus Torvalds goto out2; 21921da177e4SLinus Torvalds } 21931da177e4SLinus Torvalds 21941da177e4SLinus Torvalds static void __init init_mount_tree(void) 21951da177e4SLinus Torvalds { 21961da177e4SLinus Torvalds struct vfsmount *mnt; 21976b3286edSKirill Korotaev struct mnt_namespace *ns; 2198ac748a09SJan Blunck struct path root; 21991da177e4SLinus Torvalds 22001da177e4SLinus Torvalds mnt = do_kern_mount("rootfs", 0, "rootfs", NULL); 22011da177e4SLinus Torvalds if (IS_ERR(mnt)) 22021da177e4SLinus Torvalds panic("Can't create rootfs"); 22036b3286edSKirill Korotaev ns = kmalloc(sizeof(*ns), GFP_KERNEL); 22046b3286edSKirill Korotaev if (!ns) 22051da177e4SLinus Torvalds panic("Can't allocate initial namespace"); 22066b3286edSKirill Korotaev atomic_set(&ns->count, 1); 22076b3286edSKirill Korotaev INIT_LIST_HEAD(&ns->list); 22086b3286edSKirill Korotaev init_waitqueue_head(&ns->poll); 22096b3286edSKirill Korotaev ns->event = 0; 22106b3286edSKirill Korotaev list_add(&mnt->mnt_list, &ns->list); 22116b3286edSKirill Korotaev ns->root = mnt; 22126b3286edSKirill Korotaev mnt->mnt_ns = ns; 22131da177e4SLinus Torvalds 22146b3286edSKirill Korotaev init_task.nsproxy->mnt_ns = ns; 22156b3286edSKirill Korotaev get_mnt_ns(ns); 22161da177e4SLinus Torvalds 2217ac748a09SJan Blunck root.mnt = ns->root; 2218ac748a09SJan Blunck root.dentry = ns->root->mnt_root; 2219ac748a09SJan Blunck 2220ac748a09SJan Blunck set_fs_pwd(current->fs, &root); 2221ac748a09SJan Blunck set_fs_root(current->fs, &root); 22221da177e4SLinus Torvalds } 22231da177e4SLinus Torvalds 222474bf17cfSDenis Cheng void __init mnt_init(void) 22251da177e4SLinus Torvalds { 222613f14b4dSEric Dumazet unsigned u; 222715a67dd8SRandy Dunlap int err; 22281da177e4SLinus Torvalds 2229390c6843SRam Pai init_rwsem(&namespace_sem); 2230390c6843SRam Pai 22311da177e4SLinus Torvalds mnt_cache = kmem_cache_create("mnt_cache", sizeof(struct vfsmount), 223220c2df83SPaul Mundt 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC, NULL); 22331da177e4SLinus Torvalds 2234b58fed8bSRam Pai mount_hashtable = (struct list_head *)__get_free_page(GFP_ATOMIC); 22351da177e4SLinus Torvalds 22361da177e4SLinus Torvalds if (!mount_hashtable) 22371da177e4SLinus Torvalds panic("Failed to allocate mount hash table\n"); 22381da177e4SLinus Torvalds 223913f14b4dSEric Dumazet printk("Mount-cache hash table entries: %lu\n", HASH_SIZE); 22401da177e4SLinus Torvalds 224113f14b4dSEric Dumazet for (u = 0; u < HASH_SIZE; u++) 224213f14b4dSEric Dumazet INIT_LIST_HEAD(&mount_hashtable[u]); 22431da177e4SLinus Torvalds 224415a67dd8SRandy Dunlap err = sysfs_init(); 224515a67dd8SRandy Dunlap if (err) 224615a67dd8SRandy Dunlap printk(KERN_WARNING "%s: sysfs_init error: %d\n", 22478e24eea7SHarvey Harrison __func__, err); 224800d26666SGreg Kroah-Hartman fs_kobj = kobject_create_and_add("fs", NULL); 224900d26666SGreg Kroah-Hartman if (!fs_kobj) 22508e24eea7SHarvey Harrison printk(KERN_WARNING "%s: kobj create error\n", __func__); 22511da177e4SLinus Torvalds init_rootfs(); 22521da177e4SLinus Torvalds init_mount_tree(); 22531da177e4SLinus Torvalds } 22541da177e4SLinus Torvalds 22556b3286edSKirill Korotaev void __put_mnt_ns(struct mnt_namespace *ns) 22561da177e4SLinus Torvalds { 22576b3286edSKirill Korotaev struct vfsmount *root = ns->root; 225870fbcdf4SRam Pai LIST_HEAD(umount_list); 22596b3286edSKirill Korotaev ns->root = NULL; 22601ce88cf4SMiklos Szeredi spin_unlock(&vfsmount_lock); 2261390c6843SRam Pai down_write(&namespace_sem); 22621da177e4SLinus Torvalds spin_lock(&vfsmount_lock); 2263a05964f3SRam Pai umount_tree(root, 0, &umount_list); 22641da177e4SLinus Torvalds spin_unlock(&vfsmount_lock); 2265390c6843SRam Pai up_write(&namespace_sem); 226670fbcdf4SRam Pai release_mounts(&umount_list); 22676b3286edSKirill Korotaev kfree(ns); 22681da177e4SLinus Torvalds } 2269