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> 1499b7db7bSNick Piggin #include <linux/spinlock.h> 1599b7db7bSNick Piggin #include <linux/percpu.h> 161da177e4SLinus Torvalds #include <linux/init.h> 1715a67dd8SRandy Dunlap #include <linux/kernel.h> 181da177e4SLinus Torvalds #include <linux/acct.h> 1916f7e0feSRandy Dunlap #include <linux/capability.h> 203d733633SDave Hansen #include <linux/cpumask.h> 211da177e4SLinus Torvalds #include <linux/module.h> 22f20a9eadSAndrew Morton #include <linux/sysfs.h> 231da177e4SLinus Torvalds #include <linux/seq_file.h> 246b3286edSKirill Korotaev #include <linux/mnt_namespace.h> 251da177e4SLinus Torvalds #include <linux/namei.h> 26b43f3cbdSAlexey Dobriyan #include <linux/nsproxy.h> 271da177e4SLinus Torvalds #include <linux/security.h> 281da177e4SLinus Torvalds #include <linux/mount.h> 2907f3f05cSDavid Howells #include <linux/ramfs.h> 3013f14b4dSEric Dumazet #include <linux/log2.h> 3173cd49ecSMiklos Szeredi #include <linux/idr.h> 325ad4e53bSAl Viro #include <linux/fs_struct.h> 332504c5d6SAndreas Gruenbacher #include <linux/fsnotify.h> 341da177e4SLinus Torvalds #include <asm/uaccess.h> 351da177e4SLinus Torvalds #include <asm/unistd.h> 3607b20889SRam Pai #include "pnode.h" 37948730b0SAdrian Bunk #include "internal.h" 381da177e4SLinus Torvalds 3913f14b4dSEric Dumazet #define HASH_SHIFT ilog2(PAGE_SIZE / sizeof(struct list_head)) 4013f14b4dSEric Dumazet #define HASH_SIZE (1UL << HASH_SHIFT) 4113f14b4dSEric Dumazet 425addc5ddSAl Viro static int event; 4373cd49ecSMiklos Szeredi static DEFINE_IDA(mnt_id_ida); 44719f5d7fSMiklos Szeredi static DEFINE_IDA(mnt_group_ida); 4599b7db7bSNick Piggin static DEFINE_SPINLOCK(mnt_id_lock); 46f21f6220SAl Viro static int mnt_id_start = 0; 47f21f6220SAl Viro static int mnt_group_start = 1; 485addc5ddSAl Viro 49fa3536ccSEric Dumazet static struct list_head *mount_hashtable __read_mostly; 50e18b890bSChristoph Lameter static struct kmem_cache *mnt_cache __read_mostly; 51390c6843SRam Pai static struct rw_semaphore namespace_sem; 521da177e4SLinus Torvalds 53f87fd4c2SMiklos Szeredi /* /sys/fs */ 5400d26666SGreg Kroah-Hartman struct kobject *fs_kobj; 5500d26666SGreg Kroah-Hartman EXPORT_SYMBOL_GPL(fs_kobj); 56f87fd4c2SMiklos Szeredi 5799b7db7bSNick Piggin /* 5899b7db7bSNick Piggin * vfsmount lock may be taken for read to prevent changes to the 5999b7db7bSNick Piggin * vfsmount hash, ie. during mountpoint lookups or walking back 6099b7db7bSNick Piggin * up the tree. 6199b7db7bSNick Piggin * 6299b7db7bSNick Piggin * It should be taken for write in all cases where the vfsmount 6399b7db7bSNick Piggin * tree or hash is modified or when a vfsmount structure is modified. 6499b7db7bSNick Piggin */ 6599b7db7bSNick Piggin DEFINE_BRLOCK(vfsmount_lock); 6699b7db7bSNick Piggin 671da177e4SLinus Torvalds static inline unsigned long hash(struct vfsmount *mnt, struct dentry *dentry) 681da177e4SLinus Torvalds { 691da177e4SLinus Torvalds unsigned long tmp = ((unsigned long)mnt / L1_CACHE_BYTES); 701da177e4SLinus Torvalds tmp += ((unsigned long)dentry / L1_CACHE_BYTES); 7113f14b4dSEric Dumazet tmp = tmp + (tmp >> HASH_SHIFT); 7213f14b4dSEric Dumazet return tmp & (HASH_SIZE - 1); 731da177e4SLinus Torvalds } 741da177e4SLinus Torvalds 753d733633SDave Hansen #define MNT_WRITER_UNDERFLOW_LIMIT -(1<<16) 763d733633SDave Hansen 7799b7db7bSNick Piggin /* 7899b7db7bSNick Piggin * allocation is serialized by namespace_sem, but we need the spinlock to 7999b7db7bSNick Piggin * serialize with freeing. 8099b7db7bSNick Piggin */ 81b105e270SAl Viro static int mnt_alloc_id(struct mount *mnt) 8273cd49ecSMiklos Szeredi { 8373cd49ecSMiklos Szeredi int res; 8473cd49ecSMiklos Szeredi 8573cd49ecSMiklos Szeredi retry: 8673cd49ecSMiklos Szeredi ida_pre_get(&mnt_id_ida, GFP_KERNEL); 8799b7db7bSNick Piggin spin_lock(&mnt_id_lock); 88b105e270SAl Viro res = ida_get_new_above(&mnt_id_ida, mnt_id_start, &mnt->mnt.mnt_id); 89f21f6220SAl Viro if (!res) 90b105e270SAl Viro mnt_id_start = mnt->mnt.mnt_id + 1; 9199b7db7bSNick Piggin spin_unlock(&mnt_id_lock); 9273cd49ecSMiklos Szeredi if (res == -EAGAIN) 9373cd49ecSMiklos Szeredi goto retry; 9473cd49ecSMiklos Szeredi 9573cd49ecSMiklos Szeredi return res; 9673cd49ecSMiklos Szeredi } 9773cd49ecSMiklos Szeredi 98b105e270SAl Viro static void mnt_free_id(struct mount *mnt) 9973cd49ecSMiklos Szeredi { 100b105e270SAl Viro int id = mnt->mnt.mnt_id; 10199b7db7bSNick Piggin spin_lock(&mnt_id_lock); 102f21f6220SAl Viro ida_remove(&mnt_id_ida, id); 103f21f6220SAl Viro if (mnt_id_start > id) 104f21f6220SAl Viro mnt_id_start = id; 10599b7db7bSNick Piggin spin_unlock(&mnt_id_lock); 10673cd49ecSMiklos Szeredi } 10773cd49ecSMiklos Szeredi 108719f5d7fSMiklos Szeredi /* 109719f5d7fSMiklos Szeredi * Allocate a new peer group ID 110719f5d7fSMiklos Szeredi * 111719f5d7fSMiklos Szeredi * mnt_group_ida is protected by namespace_sem 112719f5d7fSMiklos Szeredi */ 1134b8b21f4SAl Viro static int mnt_alloc_group_id(struct mount *mnt) 114719f5d7fSMiklos Szeredi { 115f21f6220SAl Viro int res; 116f21f6220SAl Viro 117719f5d7fSMiklos Szeredi if (!ida_pre_get(&mnt_group_ida, GFP_KERNEL)) 118719f5d7fSMiklos Szeredi return -ENOMEM; 119719f5d7fSMiklos Szeredi 120f21f6220SAl Viro res = ida_get_new_above(&mnt_group_ida, 121f21f6220SAl Viro mnt_group_start, 1224b8b21f4SAl Viro &mnt->mnt.mnt_group_id); 123f21f6220SAl Viro if (!res) 1244b8b21f4SAl Viro mnt_group_start = mnt->mnt.mnt_group_id + 1; 125f21f6220SAl Viro 126f21f6220SAl Viro return res; 127719f5d7fSMiklos Szeredi } 128719f5d7fSMiklos Szeredi 129719f5d7fSMiklos Szeredi /* 130719f5d7fSMiklos Szeredi * Release a peer group ID 131719f5d7fSMiklos Szeredi */ 1324b8b21f4SAl Viro void mnt_release_group_id(struct mount *mnt) 133719f5d7fSMiklos Szeredi { 1344b8b21f4SAl Viro int id = mnt->mnt.mnt_group_id; 135f21f6220SAl Viro ida_remove(&mnt_group_ida, id); 136f21f6220SAl Viro if (mnt_group_start > id) 137f21f6220SAl Viro mnt_group_start = id; 1384b8b21f4SAl Viro mnt->mnt.mnt_group_id = 0; 139719f5d7fSMiklos Szeredi } 140719f5d7fSMiklos Szeredi 141b3e19d92SNick Piggin /* 142b3e19d92SNick Piggin * vfsmount lock must be held for read 143b3e19d92SNick Piggin */ 144b3e19d92SNick Piggin static inline void mnt_add_count(struct vfsmount *mnt, int n) 145b3e19d92SNick Piggin { 146b3e19d92SNick Piggin #ifdef CONFIG_SMP 147b3e19d92SNick Piggin this_cpu_add(mnt->mnt_pcp->mnt_count, n); 148b3e19d92SNick Piggin #else 149b3e19d92SNick Piggin preempt_disable(); 150b3e19d92SNick Piggin mnt->mnt_count += n; 151b3e19d92SNick Piggin preempt_enable(); 152b3e19d92SNick Piggin #endif 153b3e19d92SNick Piggin } 154b3e19d92SNick Piggin 155b3e19d92SNick Piggin /* 156b3e19d92SNick Piggin * vfsmount lock must be held for write 157b3e19d92SNick Piggin */ 158b3e19d92SNick Piggin unsigned int mnt_get_count(struct vfsmount *mnt) 159b3e19d92SNick Piggin { 160b3e19d92SNick Piggin #ifdef CONFIG_SMP 161f03c6599SAl Viro unsigned int count = 0; 162b3e19d92SNick Piggin int cpu; 163b3e19d92SNick Piggin 164b3e19d92SNick Piggin for_each_possible_cpu(cpu) { 165b3e19d92SNick Piggin count += per_cpu_ptr(mnt->mnt_pcp, cpu)->mnt_count; 166b3e19d92SNick Piggin } 167b3e19d92SNick Piggin 168b3e19d92SNick Piggin return count; 169b3e19d92SNick Piggin #else 170b3e19d92SNick Piggin return mnt->mnt_count; 171b3e19d92SNick Piggin #endif 172b3e19d92SNick Piggin } 173b3e19d92SNick Piggin 174b105e270SAl Viro static struct mount *alloc_vfsmnt(const char *name) 1751da177e4SLinus Torvalds { 1767d6fec45SAl Viro struct mount *p = kmem_cache_zalloc(mnt_cache, GFP_KERNEL); 1777d6fec45SAl Viro if (p) { 1787d6fec45SAl Viro struct vfsmount *mnt = &p->mnt; 17973cd49ecSMiklos Szeredi int err; 18073cd49ecSMiklos Szeredi 181b105e270SAl Viro err = mnt_alloc_id(p); 18288b38782SLi Zefan if (err) 18388b38782SLi Zefan goto out_free_cache; 18488b38782SLi Zefan 18588b38782SLi Zefan if (name) { 18688b38782SLi Zefan mnt->mnt_devname = kstrdup(name, GFP_KERNEL); 18788b38782SLi Zefan if (!mnt->mnt_devname) 18888b38782SLi Zefan goto out_free_id; 18973cd49ecSMiklos Szeredi } 19073cd49ecSMiklos Szeredi 191b3e19d92SNick Piggin #ifdef CONFIG_SMP 192b3e19d92SNick Piggin mnt->mnt_pcp = alloc_percpu(struct mnt_pcp); 193b3e19d92SNick Piggin if (!mnt->mnt_pcp) 194b3e19d92SNick Piggin goto out_free_devname; 195b3e19d92SNick Piggin 196f03c6599SAl Viro this_cpu_add(mnt->mnt_pcp->mnt_count, 1); 197b3e19d92SNick Piggin #else 198b3e19d92SNick Piggin mnt->mnt_count = 1; 199b3e19d92SNick Piggin mnt->mnt_writers = 0; 200b3e19d92SNick Piggin #endif 201b3e19d92SNick Piggin 2021da177e4SLinus Torvalds INIT_LIST_HEAD(&mnt->mnt_hash); 2031da177e4SLinus Torvalds INIT_LIST_HEAD(&mnt->mnt_child); 2041da177e4SLinus Torvalds INIT_LIST_HEAD(&mnt->mnt_mounts); 2051da177e4SLinus Torvalds INIT_LIST_HEAD(&mnt->mnt_list); 20655e700b9SMiklos Szeredi INIT_LIST_HEAD(&mnt->mnt_expire); 20703e06e68SRam Pai INIT_LIST_HEAD(&mnt->mnt_share); 208a58b0eb8SRam Pai INIT_LIST_HEAD(&mnt->mnt_slave_list); 209a58b0eb8SRam Pai INIT_LIST_HEAD(&mnt->mnt_slave); 2102504c5d6SAndreas Gruenbacher #ifdef CONFIG_FSNOTIFY 2112504c5d6SAndreas Gruenbacher INIT_HLIST_HEAD(&mnt->mnt_fsnotify_marks); 2122504c5d6SAndreas Gruenbacher #endif 2131da177e4SLinus Torvalds } 214b105e270SAl Viro return p; 21588b38782SLi Zefan 216d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 217d3ef3d73Snpiggin@suse.de out_free_devname: 2187d6fec45SAl Viro kfree(p->mnt.mnt_devname); 219d3ef3d73Snpiggin@suse.de #endif 22088b38782SLi Zefan out_free_id: 221b105e270SAl Viro mnt_free_id(p); 22288b38782SLi Zefan out_free_cache: 2237d6fec45SAl Viro kmem_cache_free(mnt_cache, p); 22488b38782SLi Zefan return NULL; 2251da177e4SLinus Torvalds } 2261da177e4SLinus Torvalds 2278366025eSDave Hansen /* 2288366025eSDave Hansen * Most r/o checks on a fs are for operations that take 2298366025eSDave Hansen * discrete amounts of time, like a write() or unlink(). 2308366025eSDave Hansen * We must keep track of when those operations start 2318366025eSDave Hansen * (for permission checks) and when they end, so that 2328366025eSDave Hansen * we can determine when writes are able to occur to 2338366025eSDave Hansen * a filesystem. 2348366025eSDave Hansen */ 2353d733633SDave Hansen /* 2363d733633SDave Hansen * __mnt_is_readonly: check whether a mount is read-only 2373d733633SDave Hansen * @mnt: the mount to check for its write status 2383d733633SDave Hansen * 2393d733633SDave Hansen * This shouldn't be used directly ouside of the VFS. 2403d733633SDave Hansen * It does not guarantee that the filesystem will stay 2413d733633SDave Hansen * r/w, just that it is right *now*. This can not and 2423d733633SDave Hansen * should not be used in place of IS_RDONLY(inode). 2433d733633SDave Hansen * mnt_want/drop_write() will _keep_ the filesystem 2443d733633SDave Hansen * r/w. 2453d733633SDave Hansen */ 2463d733633SDave Hansen int __mnt_is_readonly(struct vfsmount *mnt) 2473d733633SDave Hansen { 2482e4b7fcdSDave Hansen if (mnt->mnt_flags & MNT_READONLY) 2492e4b7fcdSDave Hansen return 1; 2502e4b7fcdSDave Hansen if (mnt->mnt_sb->s_flags & MS_RDONLY) 2512e4b7fcdSDave Hansen return 1; 2522e4b7fcdSDave Hansen return 0; 2533d733633SDave Hansen } 2543d733633SDave Hansen EXPORT_SYMBOL_GPL(__mnt_is_readonly); 2553d733633SDave Hansen 256c6653a83SNick Piggin static inline void mnt_inc_writers(struct vfsmount *mnt) 2573d733633SDave Hansen { 258d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 259b3e19d92SNick Piggin this_cpu_inc(mnt->mnt_pcp->mnt_writers); 260d3ef3d73Snpiggin@suse.de #else 261d3ef3d73Snpiggin@suse.de mnt->mnt_writers++; 262d3ef3d73Snpiggin@suse.de #endif 2633d733633SDave Hansen } 2643d733633SDave Hansen 265c6653a83SNick Piggin static inline void mnt_dec_writers(struct vfsmount *mnt) 2663d733633SDave Hansen { 267d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 268b3e19d92SNick Piggin this_cpu_dec(mnt->mnt_pcp->mnt_writers); 269d3ef3d73Snpiggin@suse.de #else 270d3ef3d73Snpiggin@suse.de mnt->mnt_writers--; 271d3ef3d73Snpiggin@suse.de #endif 272d3ef3d73Snpiggin@suse.de } 273d3ef3d73Snpiggin@suse.de 274c6653a83SNick Piggin static unsigned int mnt_get_writers(struct vfsmount *mnt) 275d3ef3d73Snpiggin@suse.de { 276d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 277d3ef3d73Snpiggin@suse.de unsigned int count = 0; 2783d733633SDave Hansen int cpu; 2793d733633SDave Hansen 2803d733633SDave Hansen for_each_possible_cpu(cpu) { 281b3e19d92SNick Piggin count += per_cpu_ptr(mnt->mnt_pcp, cpu)->mnt_writers; 2823d733633SDave Hansen } 2833d733633SDave Hansen 284d3ef3d73Snpiggin@suse.de return count; 285d3ef3d73Snpiggin@suse.de #else 286d3ef3d73Snpiggin@suse.de return mnt->mnt_writers; 287d3ef3d73Snpiggin@suse.de #endif 2883d733633SDave Hansen } 2893d733633SDave Hansen 2903d733633SDave Hansen /* 2913d733633SDave Hansen * Most r/o checks on a fs are for operations that take 2923d733633SDave Hansen * discrete amounts of time, like a write() or unlink(). 2933d733633SDave Hansen * We must keep track of when those operations start 2943d733633SDave Hansen * (for permission checks) and when they end, so that 2953d733633SDave Hansen * we can determine when writes are able to occur to 2963d733633SDave Hansen * a filesystem. 2973d733633SDave Hansen */ 2988366025eSDave Hansen /** 2998366025eSDave Hansen * mnt_want_write - get write access to a mount 3008366025eSDave Hansen * @mnt: the mount on which to take a write 3018366025eSDave Hansen * 3028366025eSDave Hansen * This tells the low-level filesystem that a write is 3038366025eSDave Hansen * about to be performed to it, and makes sure that 3048366025eSDave Hansen * writes are allowed before returning success. When 3058366025eSDave Hansen * the write operation is finished, mnt_drop_write() 3068366025eSDave Hansen * must be called. This is effectively a refcount. 3078366025eSDave Hansen */ 3088366025eSDave Hansen int mnt_want_write(struct vfsmount *mnt) 3098366025eSDave Hansen { 3103d733633SDave Hansen int ret = 0; 3113d733633SDave Hansen 312d3ef3d73Snpiggin@suse.de preempt_disable(); 313c6653a83SNick Piggin mnt_inc_writers(mnt); 314d3ef3d73Snpiggin@suse.de /* 315c6653a83SNick Piggin * The store to mnt_inc_writers must be visible before we pass 316d3ef3d73Snpiggin@suse.de * MNT_WRITE_HOLD loop below, so that the slowpath can see our 317d3ef3d73Snpiggin@suse.de * incremented count after it has set MNT_WRITE_HOLD. 318d3ef3d73Snpiggin@suse.de */ 319d3ef3d73Snpiggin@suse.de smp_mb(); 320d3ef3d73Snpiggin@suse.de while (mnt->mnt_flags & MNT_WRITE_HOLD) 321d3ef3d73Snpiggin@suse.de cpu_relax(); 322d3ef3d73Snpiggin@suse.de /* 323d3ef3d73Snpiggin@suse.de * After the slowpath clears MNT_WRITE_HOLD, mnt_is_readonly will 324d3ef3d73Snpiggin@suse.de * be set to match its requirements. So we must not load that until 325d3ef3d73Snpiggin@suse.de * MNT_WRITE_HOLD is cleared. 326d3ef3d73Snpiggin@suse.de */ 327d3ef3d73Snpiggin@suse.de smp_rmb(); 3283d733633SDave Hansen if (__mnt_is_readonly(mnt)) { 329c6653a83SNick Piggin mnt_dec_writers(mnt); 3303d733633SDave Hansen ret = -EROFS; 3313d733633SDave Hansen goto out; 3323d733633SDave Hansen } 3333d733633SDave Hansen out: 334d3ef3d73Snpiggin@suse.de preempt_enable(); 3353d733633SDave Hansen return ret; 3368366025eSDave Hansen } 3378366025eSDave Hansen EXPORT_SYMBOL_GPL(mnt_want_write); 3388366025eSDave Hansen 3398366025eSDave Hansen /** 34096029c4eSnpiggin@suse.de * mnt_clone_write - get write access to a mount 34196029c4eSnpiggin@suse.de * @mnt: the mount on which to take a write 34296029c4eSnpiggin@suse.de * 34396029c4eSnpiggin@suse.de * This is effectively like mnt_want_write, except 34496029c4eSnpiggin@suse.de * it must only be used to take an extra write reference 34596029c4eSnpiggin@suse.de * on a mountpoint that we already know has a write reference 34696029c4eSnpiggin@suse.de * on it. This allows some optimisation. 34796029c4eSnpiggin@suse.de * 34896029c4eSnpiggin@suse.de * After finished, mnt_drop_write must be called as usual to 34996029c4eSnpiggin@suse.de * drop the reference. 35096029c4eSnpiggin@suse.de */ 35196029c4eSnpiggin@suse.de int mnt_clone_write(struct vfsmount *mnt) 35296029c4eSnpiggin@suse.de { 35396029c4eSnpiggin@suse.de /* superblock may be r/o */ 35496029c4eSnpiggin@suse.de if (__mnt_is_readonly(mnt)) 35596029c4eSnpiggin@suse.de return -EROFS; 35696029c4eSnpiggin@suse.de preempt_disable(); 357c6653a83SNick Piggin mnt_inc_writers(mnt); 35896029c4eSnpiggin@suse.de preempt_enable(); 35996029c4eSnpiggin@suse.de return 0; 36096029c4eSnpiggin@suse.de } 36196029c4eSnpiggin@suse.de EXPORT_SYMBOL_GPL(mnt_clone_write); 36296029c4eSnpiggin@suse.de 36396029c4eSnpiggin@suse.de /** 36496029c4eSnpiggin@suse.de * mnt_want_write_file - get write access to a file's mount 36596029c4eSnpiggin@suse.de * @file: the file who's mount on which to take a write 36696029c4eSnpiggin@suse.de * 36796029c4eSnpiggin@suse.de * This is like mnt_want_write, but it takes a file and can 36896029c4eSnpiggin@suse.de * do some optimisations if the file is open for write already 36996029c4eSnpiggin@suse.de */ 37096029c4eSnpiggin@suse.de int mnt_want_write_file(struct file *file) 37196029c4eSnpiggin@suse.de { 3722d8dd38aSOGAWA Hirofumi struct inode *inode = file->f_dentry->d_inode; 3732d8dd38aSOGAWA Hirofumi if (!(file->f_mode & FMODE_WRITE) || special_file(inode->i_mode)) 37496029c4eSnpiggin@suse.de return mnt_want_write(file->f_path.mnt); 37596029c4eSnpiggin@suse.de else 37696029c4eSnpiggin@suse.de return mnt_clone_write(file->f_path.mnt); 37796029c4eSnpiggin@suse.de } 37896029c4eSnpiggin@suse.de EXPORT_SYMBOL_GPL(mnt_want_write_file); 37996029c4eSnpiggin@suse.de 38096029c4eSnpiggin@suse.de /** 3818366025eSDave Hansen * mnt_drop_write - give up write access to a mount 3828366025eSDave Hansen * @mnt: the mount on which to give up write access 3838366025eSDave Hansen * 3848366025eSDave Hansen * Tells the low-level filesystem that we are done 3858366025eSDave Hansen * performing writes to it. Must be matched with 3868366025eSDave Hansen * mnt_want_write() call above. 3878366025eSDave Hansen */ 3888366025eSDave Hansen void mnt_drop_write(struct vfsmount *mnt) 3898366025eSDave Hansen { 390d3ef3d73Snpiggin@suse.de preempt_disable(); 391c6653a83SNick Piggin mnt_dec_writers(mnt); 392d3ef3d73Snpiggin@suse.de preempt_enable(); 3938366025eSDave Hansen } 3948366025eSDave Hansen EXPORT_SYMBOL_GPL(mnt_drop_write); 3958366025eSDave Hansen 3962a79f17eSAl Viro void mnt_drop_write_file(struct file *file) 3972a79f17eSAl Viro { 3982a79f17eSAl Viro mnt_drop_write(file->f_path.mnt); 3992a79f17eSAl Viro } 4002a79f17eSAl Viro EXPORT_SYMBOL(mnt_drop_write_file); 4012a79f17eSAl Viro 4022e4b7fcdSDave Hansen static int mnt_make_readonly(struct vfsmount *mnt) 4038366025eSDave Hansen { 4043d733633SDave Hansen int ret = 0; 4053d733633SDave Hansen 40699b7db7bSNick Piggin br_write_lock(vfsmount_lock); 407d3ef3d73Snpiggin@suse.de mnt->mnt_flags |= MNT_WRITE_HOLD; 408d3ef3d73Snpiggin@suse.de /* 409d3ef3d73Snpiggin@suse.de * After storing MNT_WRITE_HOLD, we'll read the counters. This store 410d3ef3d73Snpiggin@suse.de * should be visible before we do. 411d3ef3d73Snpiggin@suse.de */ 412d3ef3d73Snpiggin@suse.de smp_mb(); 413d3ef3d73Snpiggin@suse.de 414d3ef3d73Snpiggin@suse.de /* 415d3ef3d73Snpiggin@suse.de * With writers on hold, if this value is zero, then there are 416d3ef3d73Snpiggin@suse.de * definitely no active writers (although held writers may subsequently 417d3ef3d73Snpiggin@suse.de * increment the count, they'll have to wait, and decrement it after 418d3ef3d73Snpiggin@suse.de * seeing MNT_READONLY). 419d3ef3d73Snpiggin@suse.de * 420d3ef3d73Snpiggin@suse.de * It is OK to have counter incremented on one CPU and decremented on 421d3ef3d73Snpiggin@suse.de * another: the sum will add up correctly. The danger would be when we 422d3ef3d73Snpiggin@suse.de * sum up each counter, if we read a counter before it is incremented, 423d3ef3d73Snpiggin@suse.de * but then read another CPU's count which it has been subsequently 424d3ef3d73Snpiggin@suse.de * decremented from -- we would see more decrements than we should. 425d3ef3d73Snpiggin@suse.de * MNT_WRITE_HOLD protects against this scenario, because 426d3ef3d73Snpiggin@suse.de * mnt_want_write first increments count, then smp_mb, then spins on 427d3ef3d73Snpiggin@suse.de * MNT_WRITE_HOLD, so it can't be decremented by another CPU while 428d3ef3d73Snpiggin@suse.de * we're counting up here. 429d3ef3d73Snpiggin@suse.de */ 430c6653a83SNick Piggin if (mnt_get_writers(mnt) > 0) 431d3ef3d73Snpiggin@suse.de ret = -EBUSY; 432d3ef3d73Snpiggin@suse.de else 4332e4b7fcdSDave Hansen mnt->mnt_flags |= MNT_READONLY; 434d3ef3d73Snpiggin@suse.de /* 435d3ef3d73Snpiggin@suse.de * MNT_READONLY must become visible before ~MNT_WRITE_HOLD, so writers 436d3ef3d73Snpiggin@suse.de * that become unheld will see MNT_READONLY. 437d3ef3d73Snpiggin@suse.de */ 438d3ef3d73Snpiggin@suse.de smp_wmb(); 439d3ef3d73Snpiggin@suse.de mnt->mnt_flags &= ~MNT_WRITE_HOLD; 44099b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 4413d733633SDave Hansen return ret; 4423d733633SDave Hansen } 4438366025eSDave Hansen 4442e4b7fcdSDave Hansen static void __mnt_unmake_readonly(struct vfsmount *mnt) 4452e4b7fcdSDave Hansen { 44699b7db7bSNick Piggin br_write_lock(vfsmount_lock); 4472e4b7fcdSDave Hansen mnt->mnt_flags &= ~MNT_READONLY; 44899b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 4492e4b7fcdSDave Hansen } 4502e4b7fcdSDave Hansen 451b105e270SAl Viro static void free_vfsmnt(struct mount *mnt) 4521da177e4SLinus Torvalds { 453b105e270SAl Viro kfree(mnt->mnt.mnt_devname); 45473cd49ecSMiklos Szeredi mnt_free_id(mnt); 455d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 456b105e270SAl Viro free_percpu(mnt->mnt.mnt_pcp); 457d3ef3d73Snpiggin@suse.de #endif 458b105e270SAl Viro kmem_cache_free(mnt_cache, mnt); 4591da177e4SLinus Torvalds } 4601da177e4SLinus Torvalds 4611da177e4SLinus Torvalds /* 462a05964f3SRam Pai * find the first or last mount at @dentry on vfsmount @mnt depending on 463a05964f3SRam Pai * @dir. If @dir is set return the first mount else return the last mount. 46499b7db7bSNick Piggin * vfsmount_lock must be held for read or write. 4651da177e4SLinus Torvalds */ 466c7105365SAl Viro struct mount *__lookup_mnt(struct vfsmount *mnt, struct dentry *dentry, 467a05964f3SRam Pai int dir) 4681da177e4SLinus Torvalds { 4691da177e4SLinus Torvalds struct list_head *head = mount_hashtable + hash(mnt, dentry); 4701da177e4SLinus Torvalds struct list_head *tmp = head; 471c7105365SAl Viro struct mount *p, *found = NULL; 4721da177e4SLinus Torvalds 4731da177e4SLinus Torvalds for (;;) { 474a05964f3SRam Pai tmp = dir ? tmp->next : tmp->prev; 4751da177e4SLinus Torvalds p = NULL; 4761da177e4SLinus Torvalds if (tmp == head) 4771da177e4SLinus Torvalds break; 478c7105365SAl Viro p = list_entry(tmp, struct mount, mnt.mnt_hash); 479c7105365SAl Viro if (p->mnt.mnt_parent == mnt && p->mnt.mnt_mountpoint == dentry) { 480a05964f3SRam Pai found = p; 4811da177e4SLinus Torvalds break; 4821da177e4SLinus Torvalds } 4831da177e4SLinus Torvalds } 4841da177e4SLinus Torvalds return found; 4851da177e4SLinus Torvalds } 4861da177e4SLinus Torvalds 487a05964f3SRam Pai /* 488a05964f3SRam Pai * lookup_mnt increments the ref count before returning 489a05964f3SRam Pai * the vfsmount struct. 490a05964f3SRam Pai */ 4911c755af4SAl Viro struct vfsmount *lookup_mnt(struct path *path) 492a05964f3SRam Pai { 493c7105365SAl Viro struct mount *child_mnt; 49499b7db7bSNick Piggin 49599b7db7bSNick Piggin br_read_lock(vfsmount_lock); 496c7105365SAl Viro child_mnt = __lookup_mnt(path->mnt, path->dentry, 1); 497c7105365SAl Viro if (child_mnt) { 498c7105365SAl Viro mnt_add_count(child_mnt, 1); 49999b7db7bSNick Piggin br_read_unlock(vfsmount_lock); 500c7105365SAl Viro return &child_mnt->mnt; 501c7105365SAl Viro } else { 502c7105365SAl Viro br_read_unlock(vfsmount_lock); 503c7105365SAl Viro return NULL; 504c7105365SAl Viro } 505a05964f3SRam Pai } 506a05964f3SRam Pai 5071da177e4SLinus Torvalds static inline int check_mnt(struct vfsmount *mnt) 5081da177e4SLinus Torvalds { 5096b3286edSKirill Korotaev return mnt->mnt_ns == current->nsproxy->mnt_ns; 5101da177e4SLinus Torvalds } 5111da177e4SLinus Torvalds 51299b7db7bSNick Piggin /* 51399b7db7bSNick Piggin * vfsmount lock must be held for write 51499b7db7bSNick Piggin */ 5156b3286edSKirill Korotaev static void touch_mnt_namespace(struct mnt_namespace *ns) 5165addc5ddSAl Viro { 5175addc5ddSAl Viro if (ns) { 5185addc5ddSAl Viro ns->event = ++event; 5195addc5ddSAl Viro wake_up_interruptible(&ns->poll); 5205addc5ddSAl Viro } 5215addc5ddSAl Viro } 5225addc5ddSAl Viro 52399b7db7bSNick Piggin /* 52499b7db7bSNick Piggin * vfsmount lock must be held for write 52599b7db7bSNick Piggin */ 5266b3286edSKirill Korotaev static void __touch_mnt_namespace(struct mnt_namespace *ns) 5275addc5ddSAl Viro { 5285addc5ddSAl Viro if (ns && ns->event != event) { 5295addc5ddSAl Viro ns->event = event; 5305addc5ddSAl Viro wake_up_interruptible(&ns->poll); 5315addc5ddSAl Viro } 5325addc5ddSAl Viro } 5335addc5ddSAl Viro 53499b7db7bSNick Piggin /* 5355f57cbccSNick Piggin * Clear dentry's mounted state if it has no remaining mounts. 5365f57cbccSNick Piggin * vfsmount_lock must be held for write. 5375f57cbccSNick Piggin */ 538aa0a4cf0SAl Viro static void dentry_reset_mounted(struct dentry *dentry) 5395f57cbccSNick Piggin { 5405f57cbccSNick Piggin unsigned u; 5415f57cbccSNick Piggin 5425f57cbccSNick Piggin for (u = 0; u < HASH_SIZE; u++) { 5435f57cbccSNick Piggin struct vfsmount *p; 5445f57cbccSNick Piggin 5455f57cbccSNick Piggin list_for_each_entry(p, &mount_hashtable[u], mnt_hash) { 5465f57cbccSNick Piggin if (p->mnt_mountpoint == dentry) 5475f57cbccSNick Piggin return; 5485f57cbccSNick Piggin } 5495f57cbccSNick Piggin } 5505f57cbccSNick Piggin spin_lock(&dentry->d_lock); 5515f57cbccSNick Piggin dentry->d_flags &= ~DCACHE_MOUNTED; 5525f57cbccSNick Piggin spin_unlock(&dentry->d_lock); 5535f57cbccSNick Piggin } 5545f57cbccSNick Piggin 5555f57cbccSNick Piggin /* 55699b7db7bSNick Piggin * vfsmount lock must be held for write 55799b7db7bSNick Piggin */ 558419148daSAl Viro static void detach_mnt(struct mount *mnt, struct path *old_path) 5591da177e4SLinus Torvalds { 560419148daSAl Viro old_path->dentry = mnt->mnt.mnt_mountpoint; 561419148daSAl Viro old_path->mnt = mnt->mnt.mnt_parent; 562419148daSAl Viro mnt->mnt.mnt_parent = &mnt->mnt; 563419148daSAl Viro mnt->mnt.mnt_mountpoint = mnt->mnt.mnt_root; 564419148daSAl Viro list_del_init(&mnt->mnt.mnt_child); 565419148daSAl Viro list_del_init(&mnt->mnt.mnt_hash); 566aa0a4cf0SAl Viro dentry_reset_mounted(old_path->dentry); 5671da177e4SLinus Torvalds } 5681da177e4SLinus Torvalds 56999b7db7bSNick Piggin /* 57099b7db7bSNick Piggin * vfsmount lock must be held for write 57199b7db7bSNick Piggin */ 572b90fa9aeSRam Pai void mnt_set_mountpoint(struct vfsmount *mnt, struct dentry *dentry, 573b90fa9aeSRam Pai struct vfsmount *child_mnt) 574b90fa9aeSRam Pai { 575b90fa9aeSRam Pai child_mnt->mnt_parent = mntget(mnt); 576b90fa9aeSRam Pai child_mnt->mnt_mountpoint = dget(dentry); 5775f57cbccSNick Piggin spin_lock(&dentry->d_lock); 5785f57cbccSNick Piggin dentry->d_flags |= DCACHE_MOUNTED; 5795f57cbccSNick Piggin spin_unlock(&dentry->d_lock); 580b90fa9aeSRam Pai } 581b90fa9aeSRam Pai 58299b7db7bSNick Piggin /* 58399b7db7bSNick Piggin * vfsmount lock must be held for write 58499b7db7bSNick Piggin */ 585419148daSAl Viro static void attach_mnt(struct mount *mnt, struct path *path) 5861da177e4SLinus Torvalds { 587419148daSAl Viro mnt_set_mountpoint(path->mnt, path->dentry, &mnt->mnt); 588419148daSAl Viro list_add_tail(&mnt->mnt.mnt_hash, mount_hashtable + 5891a390689SAl Viro hash(path->mnt, path->dentry)); 590419148daSAl Viro list_add_tail(&mnt->mnt.mnt_child, &path->mnt->mnt_mounts); 591b90fa9aeSRam Pai } 592b90fa9aeSRam Pai 5937e3d0eb0SAl Viro static inline void __mnt_make_longterm(struct vfsmount *mnt) 5947e3d0eb0SAl Viro { 5957e3d0eb0SAl Viro #ifdef CONFIG_SMP 5967e3d0eb0SAl Viro atomic_inc(&mnt->mnt_longterm); 5977e3d0eb0SAl Viro #endif 5987e3d0eb0SAl Viro } 5997e3d0eb0SAl Viro 6007e3d0eb0SAl Viro /* needs vfsmount lock for write */ 6017e3d0eb0SAl Viro static inline void __mnt_make_shortterm(struct vfsmount *mnt) 6027e3d0eb0SAl Viro { 6037e3d0eb0SAl Viro #ifdef CONFIG_SMP 6047e3d0eb0SAl Viro atomic_dec(&mnt->mnt_longterm); 6057e3d0eb0SAl Viro #endif 6067e3d0eb0SAl Viro } 6077e3d0eb0SAl Viro 608b90fa9aeSRam Pai /* 60999b7db7bSNick Piggin * vfsmount lock must be held for write 610b90fa9aeSRam Pai */ 6114b2619a5SAl Viro static void commit_tree(struct mount *mnt) 612b90fa9aeSRam Pai { 6134b2619a5SAl Viro struct vfsmount *parent = mnt->mnt.mnt_parent; 614b90fa9aeSRam Pai struct vfsmount *m; 615b90fa9aeSRam Pai LIST_HEAD(head); 6166b3286edSKirill Korotaev struct mnt_namespace *n = parent->mnt_ns; 617b90fa9aeSRam Pai 6184b2619a5SAl Viro BUG_ON(parent == &mnt->mnt); 619b90fa9aeSRam Pai 6204b2619a5SAl Viro list_add_tail(&head, &mnt->mnt.mnt_list); 621f03c6599SAl Viro list_for_each_entry(m, &head, mnt_list) { 6226b3286edSKirill Korotaev m->mnt_ns = n; 6237e3d0eb0SAl Viro __mnt_make_longterm(m); 624f03c6599SAl Viro } 625f03c6599SAl Viro 626b90fa9aeSRam Pai list_splice(&head, n->list.prev); 627b90fa9aeSRam Pai 6284b2619a5SAl Viro list_add_tail(&mnt->mnt.mnt_hash, mount_hashtable + 6294b2619a5SAl Viro hash(parent, mnt->mnt.mnt_mountpoint)); 6304b2619a5SAl Viro list_add_tail(&mnt->mnt.mnt_child, &parent->mnt_mounts); 6316b3286edSKirill Korotaev touch_mnt_namespace(n); 6321da177e4SLinus Torvalds } 6331da177e4SLinus Torvalds 634315fc83eSAl Viro static struct mount *next_mnt(struct mount *p, struct vfsmount *root) 6351da177e4SLinus Torvalds { 636315fc83eSAl Viro struct list_head *next = p->mnt.mnt_mounts.next; 637315fc83eSAl Viro if (next == &p->mnt.mnt_mounts) { 6381da177e4SLinus Torvalds while (1) { 639315fc83eSAl Viro if (&p->mnt == root) 6401da177e4SLinus Torvalds return NULL; 641315fc83eSAl Viro next = p->mnt.mnt_child.next; 642315fc83eSAl Viro if (next != &p->mnt.mnt_parent->mnt_mounts) 6431da177e4SLinus Torvalds break; 644315fc83eSAl Viro p = real_mount(p->mnt.mnt_parent); 6451da177e4SLinus Torvalds } 6461da177e4SLinus Torvalds } 647315fc83eSAl Viro return list_entry(next, struct mount, mnt.mnt_child); 6481da177e4SLinus Torvalds } 6491da177e4SLinus Torvalds 650315fc83eSAl Viro static struct mount *skip_mnt_tree(struct mount *p) 6519676f0c6SRam Pai { 652315fc83eSAl Viro struct list_head *prev = p->mnt.mnt_mounts.prev; 653315fc83eSAl Viro while (prev != &p->mnt.mnt_mounts) { 654315fc83eSAl Viro p = list_entry(prev, struct mount, mnt.mnt_child); 655315fc83eSAl Viro prev = p->mnt.mnt_mounts.prev; 6569676f0c6SRam Pai } 6579676f0c6SRam Pai return p; 6589676f0c6SRam Pai } 6599676f0c6SRam Pai 6609d412a43SAl Viro struct vfsmount * 6619d412a43SAl Viro vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data) 6629d412a43SAl Viro { 663b105e270SAl Viro struct mount *mnt; 6649d412a43SAl Viro struct dentry *root; 6659d412a43SAl Viro 6669d412a43SAl Viro if (!type) 6679d412a43SAl Viro return ERR_PTR(-ENODEV); 6689d412a43SAl Viro 6699d412a43SAl Viro mnt = alloc_vfsmnt(name); 6709d412a43SAl Viro if (!mnt) 6719d412a43SAl Viro return ERR_PTR(-ENOMEM); 6729d412a43SAl Viro 6739d412a43SAl Viro if (flags & MS_KERNMOUNT) 674b105e270SAl Viro mnt->mnt.mnt_flags = MNT_INTERNAL; 6759d412a43SAl Viro 6769d412a43SAl Viro root = mount_fs(type, flags, name, data); 6779d412a43SAl Viro if (IS_ERR(root)) { 6789d412a43SAl Viro free_vfsmnt(mnt); 6799d412a43SAl Viro return ERR_CAST(root); 6809d412a43SAl Viro } 6819d412a43SAl Viro 682b105e270SAl Viro mnt->mnt.mnt_root = root; 683b105e270SAl Viro mnt->mnt.mnt_sb = root->d_sb; 684b105e270SAl Viro mnt->mnt.mnt_mountpoint = mnt->mnt.mnt_root; 685b105e270SAl Viro mnt->mnt.mnt_parent = &mnt->mnt; 686b105e270SAl Viro return &mnt->mnt; 6879d412a43SAl Viro } 6889d412a43SAl Viro EXPORT_SYMBOL_GPL(vfs_kern_mount); 6899d412a43SAl Viro 690cb338d06SAl Viro static struct mount *clone_mnt(struct vfsmount *old, struct dentry *root, 69136341f64SRam Pai int flag) 6921da177e4SLinus Torvalds { 6931da177e4SLinus Torvalds struct super_block *sb = old->mnt_sb; 694b105e270SAl Viro struct mount *mnt = alloc_vfsmnt(old->mnt_devname); 6951da177e4SLinus Torvalds 6961da177e4SLinus Torvalds if (mnt) { 697719f5d7fSMiklos Szeredi if (flag & (CL_SLAVE | CL_PRIVATE)) 698b105e270SAl Viro mnt->mnt.mnt_group_id = 0; /* not a peer of original */ 699719f5d7fSMiklos Szeredi else 700b105e270SAl Viro mnt->mnt.mnt_group_id = old->mnt_group_id; 701719f5d7fSMiklos Szeredi 702b105e270SAl Viro if ((flag & CL_MAKE_SHARED) && !mnt->mnt.mnt_group_id) { 703b105e270SAl Viro int err = mnt_alloc_group_id(mnt); 704719f5d7fSMiklos Szeredi if (err) 705719f5d7fSMiklos Szeredi goto out_free; 706719f5d7fSMiklos Szeredi } 707719f5d7fSMiklos Szeredi 708b105e270SAl Viro mnt->mnt.mnt_flags = old->mnt_flags & ~MNT_WRITE_HOLD; 7091da177e4SLinus Torvalds atomic_inc(&sb->s_active); 710b105e270SAl Viro mnt->mnt.mnt_sb = sb; 711b105e270SAl Viro mnt->mnt.mnt_root = dget(root); 712b105e270SAl Viro mnt->mnt.mnt_mountpoint = mnt->mnt.mnt_root; 713b105e270SAl Viro mnt->mnt.mnt_parent = &mnt->mnt; 714b90fa9aeSRam Pai 7155afe0022SRam Pai if (flag & CL_SLAVE) { 716b105e270SAl Viro list_add(&mnt->mnt.mnt_slave, &old->mnt_slave_list); 717b105e270SAl Viro mnt->mnt.mnt_master = old; 718b105e270SAl Viro CLEAR_MNT_SHARED(&mnt->mnt); 7198aec0809SAl Viro } else if (!(flag & CL_PRIVATE)) { 720796a6b52SAl Viro if ((flag & CL_MAKE_SHARED) || IS_MNT_SHARED(old)) 721b105e270SAl Viro list_add(&mnt->mnt.mnt_share, &old->mnt_share); 7225afe0022SRam Pai if (IS_MNT_SLAVE(old)) 723b105e270SAl Viro list_add(&mnt->mnt.mnt_slave, &old->mnt_slave); 724b105e270SAl Viro mnt->mnt.mnt_master = old->mnt_master; 7255afe0022SRam Pai } 726b90fa9aeSRam Pai if (flag & CL_MAKE_SHARED) 7270f0afb1dSAl Viro set_mnt_shared(mnt); 7281da177e4SLinus Torvalds 7291da177e4SLinus Torvalds /* stick the duplicate mount on the same expiry list 7301da177e4SLinus Torvalds * as the original if that was on one */ 73136341f64SRam Pai if (flag & CL_EXPIRE) { 73255e700b9SMiklos Szeredi if (!list_empty(&old->mnt_expire)) 733b105e270SAl Viro list_add(&mnt->mnt.mnt_expire, &old->mnt_expire); 7341da177e4SLinus Torvalds } 73536341f64SRam Pai } 736cb338d06SAl Viro return mnt; 737719f5d7fSMiklos Szeredi 738719f5d7fSMiklos Szeredi out_free: 739719f5d7fSMiklos Szeredi free_vfsmnt(mnt); 740719f5d7fSMiklos Szeredi return NULL; 7411da177e4SLinus Torvalds } 7421da177e4SLinus Torvalds 743b3e19d92SNick Piggin static inline void mntfree(struct vfsmount *mnt) 7441da177e4SLinus Torvalds { 7451da177e4SLinus Torvalds struct super_block *sb = mnt->mnt_sb; 746b3e19d92SNick Piggin 7473d733633SDave Hansen /* 7483d733633SDave Hansen * This probably indicates that somebody messed 7493d733633SDave Hansen * up a mnt_want/drop_write() pair. If this 7503d733633SDave Hansen * happens, the filesystem was probably unable 7513d733633SDave Hansen * to make r/w->r/o transitions. 7523d733633SDave Hansen */ 753d3ef3d73Snpiggin@suse.de /* 754b3e19d92SNick Piggin * The locking used to deal with mnt_count decrement provides barriers, 755b3e19d92SNick Piggin * so mnt_get_writers() below is safe. 756d3ef3d73Snpiggin@suse.de */ 757c6653a83SNick Piggin WARN_ON(mnt_get_writers(mnt)); 758ca9c726eSAndreas Gruenbacher fsnotify_vfsmount_delete(mnt); 7591da177e4SLinus Torvalds dput(mnt->mnt_root); 760b105e270SAl Viro free_vfsmnt(real_mount(mnt)); 7611da177e4SLinus Torvalds deactivate_super(sb); 7621da177e4SLinus Torvalds } 7631da177e4SLinus Torvalds 764f03c6599SAl Viro static void mntput_no_expire(struct vfsmount *mnt) 7657b7b1aceSAl Viro { 766b3e19d92SNick Piggin put_again: 767f03c6599SAl Viro #ifdef CONFIG_SMP 768b3e19d92SNick Piggin br_read_lock(vfsmount_lock); 769f03c6599SAl Viro if (likely(atomic_read(&mnt->mnt_longterm))) { 770aa9c0e07SAl Viro mnt_add_count(mnt, -1); 771b3e19d92SNick Piggin br_read_unlock(vfsmount_lock); 77299b7db7bSNick Piggin return; 773b3e19d92SNick Piggin } 774b3e19d92SNick Piggin br_read_unlock(vfsmount_lock); 775b3e19d92SNick Piggin 77699b7db7bSNick Piggin br_write_lock(vfsmount_lock); 777aa9c0e07SAl Viro mnt_add_count(mnt, -1); 778b3e19d92SNick Piggin if (mnt_get_count(mnt)) { 77999b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 78099b7db7bSNick Piggin return; 78199b7db7bSNick Piggin } 782b3e19d92SNick Piggin #else 783aa9c0e07SAl Viro mnt_add_count(mnt, -1); 784b3e19d92SNick Piggin if (likely(mnt_get_count(mnt))) 785b3e19d92SNick Piggin return; 786b3e19d92SNick Piggin br_write_lock(vfsmount_lock); 787f03c6599SAl Viro #endif 788b3e19d92SNick Piggin if (unlikely(mnt->mnt_pinned)) { 789b3e19d92SNick Piggin mnt_add_count(mnt, mnt->mnt_pinned + 1); 790b3e19d92SNick Piggin mnt->mnt_pinned = 0; 791b3e19d92SNick Piggin br_write_unlock(vfsmount_lock); 792b3e19d92SNick Piggin acct_auto_close_mnt(mnt); 793b3e19d92SNick Piggin goto put_again; 794b3e19d92SNick Piggin } 795b3e19d92SNick Piggin br_write_unlock(vfsmount_lock); 796b3e19d92SNick Piggin mntfree(mnt); 797b3e19d92SNick Piggin } 798b3e19d92SNick Piggin 799b3e19d92SNick Piggin void mntput(struct vfsmount *mnt) 800b3e19d92SNick Piggin { 801b3e19d92SNick Piggin if (mnt) { 802b3e19d92SNick Piggin /* avoid cacheline pingpong, hope gcc doesn't get "smart" */ 803b3e19d92SNick Piggin if (unlikely(mnt->mnt_expiry_mark)) 804b3e19d92SNick Piggin mnt->mnt_expiry_mark = 0; 805f03c6599SAl Viro mntput_no_expire(mnt); 806b3e19d92SNick Piggin } 807b3e19d92SNick Piggin } 808b3e19d92SNick Piggin EXPORT_SYMBOL(mntput); 809b3e19d92SNick Piggin 810b3e19d92SNick Piggin struct vfsmount *mntget(struct vfsmount *mnt) 811b3e19d92SNick Piggin { 812b3e19d92SNick Piggin if (mnt) 813aa9c0e07SAl Viro mnt_add_count(mnt, 1); 814b3e19d92SNick Piggin return mnt; 815b3e19d92SNick Piggin } 816b3e19d92SNick Piggin EXPORT_SYMBOL(mntget); 817b3e19d92SNick Piggin 8187b7b1aceSAl Viro void mnt_pin(struct vfsmount *mnt) 8197b7b1aceSAl Viro { 82099b7db7bSNick Piggin br_write_lock(vfsmount_lock); 8217b7b1aceSAl Viro mnt->mnt_pinned++; 82299b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 8237b7b1aceSAl Viro } 8247b7b1aceSAl Viro EXPORT_SYMBOL(mnt_pin); 8257b7b1aceSAl Viro 8267b7b1aceSAl Viro void mnt_unpin(struct vfsmount *mnt) 8277b7b1aceSAl Viro { 82899b7db7bSNick Piggin br_write_lock(vfsmount_lock); 8297b7b1aceSAl Viro if (mnt->mnt_pinned) { 830aa9c0e07SAl Viro mnt_add_count(mnt, 1); 8317b7b1aceSAl Viro mnt->mnt_pinned--; 8327b7b1aceSAl Viro } 83399b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 8347b7b1aceSAl Viro } 8357b7b1aceSAl Viro EXPORT_SYMBOL(mnt_unpin); 8361da177e4SLinus Torvalds 837b3b304a2SMiklos Szeredi static inline void mangle(struct seq_file *m, const char *s) 838b3b304a2SMiklos Szeredi { 839b3b304a2SMiklos Szeredi seq_escape(m, s, " \t\n\\"); 840b3b304a2SMiklos Szeredi } 841b3b304a2SMiklos Szeredi 842b3b304a2SMiklos Szeredi /* 843b3b304a2SMiklos Szeredi * Simple .show_options callback for filesystems which don't want to 844b3b304a2SMiklos Szeredi * implement more complex mount option showing. 845b3b304a2SMiklos Szeredi * 846b3b304a2SMiklos Szeredi * See also save_mount_options(). 847b3b304a2SMiklos Szeredi */ 848b3b304a2SMiklos Szeredi int generic_show_options(struct seq_file *m, struct vfsmount *mnt) 849b3b304a2SMiklos Szeredi { 8502a32cebdSAl Viro const char *options; 8512a32cebdSAl Viro 8522a32cebdSAl Viro rcu_read_lock(); 8532a32cebdSAl Viro options = rcu_dereference(mnt->mnt_sb->s_options); 854b3b304a2SMiklos Szeredi 855b3b304a2SMiklos Szeredi if (options != NULL && options[0]) { 856b3b304a2SMiklos Szeredi seq_putc(m, ','); 857b3b304a2SMiklos Szeredi mangle(m, options); 858b3b304a2SMiklos Szeredi } 8592a32cebdSAl Viro rcu_read_unlock(); 860b3b304a2SMiklos Szeredi 861b3b304a2SMiklos Szeredi return 0; 862b3b304a2SMiklos Szeredi } 863b3b304a2SMiklos Szeredi EXPORT_SYMBOL(generic_show_options); 864b3b304a2SMiklos Szeredi 865b3b304a2SMiklos Szeredi /* 866b3b304a2SMiklos Szeredi * If filesystem uses generic_show_options(), this function should be 867b3b304a2SMiklos Szeredi * called from the fill_super() callback. 868b3b304a2SMiklos Szeredi * 869b3b304a2SMiklos Szeredi * The .remount_fs callback usually needs to be handled in a special 870b3b304a2SMiklos Szeredi * way, to make sure, that previous options are not overwritten if the 871b3b304a2SMiklos Szeredi * remount fails. 872b3b304a2SMiklos Szeredi * 873b3b304a2SMiklos Szeredi * Also note, that if the filesystem's .remount_fs function doesn't 874b3b304a2SMiklos Szeredi * reset all options to their default value, but changes only newly 875b3b304a2SMiklos Szeredi * given options, then the displayed options will not reflect reality 876b3b304a2SMiklos Szeredi * any more. 877b3b304a2SMiklos Szeredi */ 878b3b304a2SMiklos Szeredi void save_mount_options(struct super_block *sb, char *options) 879b3b304a2SMiklos Szeredi { 8802a32cebdSAl Viro BUG_ON(sb->s_options); 8812a32cebdSAl Viro rcu_assign_pointer(sb->s_options, kstrdup(options, GFP_KERNEL)); 882b3b304a2SMiklos Szeredi } 883b3b304a2SMiklos Szeredi EXPORT_SYMBOL(save_mount_options); 884b3b304a2SMiklos Szeredi 8852a32cebdSAl Viro void replace_mount_options(struct super_block *sb, char *options) 8862a32cebdSAl Viro { 8872a32cebdSAl Viro char *old = sb->s_options; 8882a32cebdSAl Viro rcu_assign_pointer(sb->s_options, options); 8892a32cebdSAl Viro if (old) { 8902a32cebdSAl Viro synchronize_rcu(); 8912a32cebdSAl Viro kfree(old); 8922a32cebdSAl Viro } 8932a32cebdSAl Viro } 8942a32cebdSAl Viro EXPORT_SYMBOL(replace_mount_options); 8952a32cebdSAl Viro 896a1a2c409SMiklos Szeredi #ifdef CONFIG_PROC_FS 8971da177e4SLinus Torvalds /* iterator */ 8981da177e4SLinus Torvalds static void *m_start(struct seq_file *m, loff_t *pos) 8991da177e4SLinus Torvalds { 900a1a2c409SMiklos Szeredi struct proc_mounts *p = m->private; 9011da177e4SLinus Torvalds 902390c6843SRam Pai down_read(&namespace_sem); 903a1a2c409SMiklos Szeredi return seq_list_start(&p->ns->list, *pos); 9041da177e4SLinus Torvalds } 9051da177e4SLinus Torvalds 9061da177e4SLinus Torvalds static void *m_next(struct seq_file *m, void *v, loff_t *pos) 9071da177e4SLinus Torvalds { 908a1a2c409SMiklos Szeredi struct proc_mounts *p = m->private; 909b0765fb8SPavel Emelianov 910a1a2c409SMiklos Szeredi return seq_list_next(v, &p->ns->list, pos); 9111da177e4SLinus Torvalds } 9121da177e4SLinus Torvalds 9131da177e4SLinus Torvalds static void m_stop(struct seq_file *m, void *v) 9141da177e4SLinus Torvalds { 915390c6843SRam Pai up_read(&namespace_sem); 9161da177e4SLinus Torvalds } 9171da177e4SLinus Torvalds 9189f5596afSAl Viro int mnt_had_events(struct proc_mounts *p) 9199f5596afSAl Viro { 9209f5596afSAl Viro struct mnt_namespace *ns = p->ns; 9219f5596afSAl Viro int res = 0; 9229f5596afSAl Viro 92399b7db7bSNick Piggin br_read_lock(vfsmount_lock); 924f1514638SKay Sievers if (p->m.poll_event != ns->event) { 925f1514638SKay Sievers p->m.poll_event = ns->event; 9269f5596afSAl Viro res = 1; 9279f5596afSAl Viro } 92899b7db7bSNick Piggin br_read_unlock(vfsmount_lock); 9299f5596afSAl Viro 9309f5596afSAl Viro return res; 9319f5596afSAl Viro } 9329f5596afSAl Viro 9332d4d4864SRam Pai struct proc_fs_info { 9341da177e4SLinus Torvalds int flag; 9352d4d4864SRam Pai const char *str; 9362d4d4864SRam Pai }; 9372d4d4864SRam Pai 9382069f457SEric Paris static int show_sb_opts(struct seq_file *m, struct super_block *sb) 9392d4d4864SRam Pai { 9402d4d4864SRam Pai static const struct proc_fs_info fs_info[] = { 9411da177e4SLinus Torvalds { MS_SYNCHRONOUS, ",sync" }, 9421da177e4SLinus Torvalds { MS_DIRSYNC, ",dirsync" }, 9431da177e4SLinus Torvalds { MS_MANDLOCK, ",mand" }, 9441da177e4SLinus Torvalds { 0, NULL } 9451da177e4SLinus Torvalds }; 9462d4d4864SRam Pai const struct proc_fs_info *fs_infop; 9472d4d4864SRam Pai 9482d4d4864SRam Pai for (fs_infop = fs_info; fs_infop->flag; fs_infop++) { 9492d4d4864SRam Pai if (sb->s_flags & fs_infop->flag) 9502d4d4864SRam Pai seq_puts(m, fs_infop->str); 9512d4d4864SRam Pai } 9522069f457SEric Paris 9532069f457SEric Paris return security_sb_show_options(m, sb); 9542d4d4864SRam Pai } 9552d4d4864SRam Pai 9562d4d4864SRam Pai static void show_mnt_opts(struct seq_file *m, struct vfsmount *mnt) 9572d4d4864SRam Pai { 9582d4d4864SRam Pai static const struct proc_fs_info mnt_info[] = { 9591da177e4SLinus Torvalds { MNT_NOSUID, ",nosuid" }, 9601da177e4SLinus Torvalds { MNT_NODEV, ",nodev" }, 9611da177e4SLinus Torvalds { MNT_NOEXEC, ",noexec" }, 962fc33a7bbSChristoph Hellwig { MNT_NOATIME, ",noatime" }, 963fc33a7bbSChristoph Hellwig { MNT_NODIRATIME, ",nodiratime" }, 96447ae32d6SValerie Henson { MNT_RELATIME, ",relatime" }, 9651da177e4SLinus Torvalds { 0, NULL } 9661da177e4SLinus Torvalds }; 9672d4d4864SRam Pai const struct proc_fs_info *fs_infop; 9682d4d4864SRam Pai 9692d4d4864SRam Pai for (fs_infop = mnt_info; fs_infop->flag; fs_infop++) { 9702d4d4864SRam Pai if (mnt->mnt_flags & fs_infop->flag) 9712d4d4864SRam Pai seq_puts(m, fs_infop->str); 9722d4d4864SRam Pai } 9732d4d4864SRam Pai } 9742d4d4864SRam Pai 9752d4d4864SRam Pai static void show_type(struct seq_file *m, struct super_block *sb) 9762d4d4864SRam Pai { 9772d4d4864SRam Pai mangle(m, sb->s_type->name); 9782d4d4864SRam Pai if (sb->s_subtype && sb->s_subtype[0]) { 9792d4d4864SRam Pai seq_putc(m, '.'); 9802d4d4864SRam Pai mangle(m, sb->s_subtype); 9812d4d4864SRam Pai } 9822d4d4864SRam Pai } 9832d4d4864SRam Pai 9842d4d4864SRam Pai static int show_vfsmnt(struct seq_file *m, void *v) 9852d4d4864SRam Pai { 9862d4d4864SRam Pai struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); 9872d4d4864SRam Pai int err = 0; 988c32c2f63SJan Blunck struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; 9891da177e4SLinus Torvalds 990c7f404b4SAl Viro if (mnt->mnt_sb->s_op->show_devname) { 991c7f404b4SAl Viro err = mnt->mnt_sb->s_op->show_devname(m, mnt); 992c7f404b4SAl Viro if (err) 993c7f404b4SAl Viro goto out; 994c7f404b4SAl Viro } else { 9951da177e4SLinus Torvalds mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none"); 996c7f404b4SAl Viro } 9971da177e4SLinus Torvalds seq_putc(m, ' '); 998c32c2f63SJan Blunck seq_path(m, &mnt_path, " \t\n\\"); 9991da177e4SLinus Torvalds seq_putc(m, ' '); 10002d4d4864SRam Pai show_type(m, mnt->mnt_sb); 10012e4b7fcdSDave Hansen seq_puts(m, __mnt_is_readonly(mnt) ? " ro" : " rw"); 10022069f457SEric Paris err = show_sb_opts(m, mnt->mnt_sb); 10032069f457SEric Paris if (err) 10042069f457SEric Paris goto out; 10052d4d4864SRam Pai show_mnt_opts(m, mnt); 10061da177e4SLinus Torvalds if (mnt->mnt_sb->s_op->show_options) 10071da177e4SLinus Torvalds err = mnt->mnt_sb->s_op->show_options(m, mnt); 10081da177e4SLinus Torvalds seq_puts(m, " 0 0\n"); 10092069f457SEric Paris out: 10101da177e4SLinus Torvalds return err; 10111da177e4SLinus Torvalds } 10121da177e4SLinus Torvalds 1013a1a2c409SMiklos Szeredi const struct seq_operations mounts_op = { 10141da177e4SLinus Torvalds .start = m_start, 10151da177e4SLinus Torvalds .next = m_next, 10161da177e4SLinus Torvalds .stop = m_stop, 10171da177e4SLinus Torvalds .show = show_vfsmnt 10181da177e4SLinus Torvalds }; 10191da177e4SLinus Torvalds 10202d4d4864SRam Pai static int show_mountinfo(struct seq_file *m, void *v) 10212d4d4864SRam Pai { 10222d4d4864SRam Pai struct proc_mounts *p = m->private; 10232d4d4864SRam Pai struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); 10242d4d4864SRam Pai struct super_block *sb = mnt->mnt_sb; 10252d4d4864SRam Pai struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; 10262d4d4864SRam Pai struct path root = p->root; 10272d4d4864SRam Pai int err = 0; 10282d4d4864SRam Pai 10292d4d4864SRam Pai seq_printf(m, "%i %i %u:%u ", mnt->mnt_id, mnt->mnt_parent->mnt_id, 10302d4d4864SRam Pai MAJOR(sb->s_dev), MINOR(sb->s_dev)); 1031c7f404b4SAl Viro if (sb->s_op->show_path) 1032c7f404b4SAl Viro err = sb->s_op->show_path(m, mnt); 1033c7f404b4SAl Viro else 10342d4d4864SRam Pai seq_dentry(m, mnt->mnt_root, " \t\n\\"); 1035c7f404b4SAl Viro if (err) 1036c7f404b4SAl Viro goto out; 10372d4d4864SRam Pai seq_putc(m, ' '); 103802125a82SAl Viro 103902125a82SAl Viro /* mountpoints outside of chroot jail will give SEQ_SKIP on this */ 104002125a82SAl Viro err = seq_path_root(m, &mnt_path, &root, " \t\n\\"); 104102125a82SAl Viro if (err) 104202125a82SAl Viro goto out; 104302125a82SAl Viro 10442d4d4864SRam Pai seq_puts(m, mnt->mnt_flags & MNT_READONLY ? " ro" : " rw"); 10452d4d4864SRam Pai show_mnt_opts(m, mnt); 10462d4d4864SRam Pai 10472d4d4864SRam Pai /* Tagged fields ("foo:X" or "bar") */ 10482d4d4864SRam Pai if (IS_MNT_SHARED(mnt)) 10492d4d4864SRam Pai seq_printf(m, " shared:%i", mnt->mnt_group_id); 105097e7e0f7SMiklos Szeredi if (IS_MNT_SLAVE(mnt)) { 105197e7e0f7SMiklos Szeredi int master = mnt->mnt_master->mnt_group_id; 105297e7e0f7SMiklos Szeredi int dom = get_dominating_id(mnt, &p->root); 105397e7e0f7SMiklos Szeredi seq_printf(m, " master:%i", master); 105497e7e0f7SMiklos Szeredi if (dom && dom != master) 105597e7e0f7SMiklos Szeredi seq_printf(m, " propagate_from:%i", dom); 105697e7e0f7SMiklos Szeredi } 10572d4d4864SRam Pai if (IS_MNT_UNBINDABLE(mnt)) 10582d4d4864SRam Pai seq_puts(m, " unbindable"); 10592d4d4864SRam Pai 10602d4d4864SRam Pai /* Filesystem specific data */ 10612d4d4864SRam Pai seq_puts(m, " - "); 10622d4d4864SRam Pai show_type(m, sb); 10632d4d4864SRam Pai seq_putc(m, ' '); 1064c7f404b4SAl Viro if (sb->s_op->show_devname) 1065c7f404b4SAl Viro err = sb->s_op->show_devname(m, mnt); 1066c7f404b4SAl Viro else 10672d4d4864SRam Pai mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none"); 1068c7f404b4SAl Viro if (err) 1069c7f404b4SAl Viro goto out; 10702d4d4864SRam Pai seq_puts(m, sb->s_flags & MS_RDONLY ? " ro" : " rw"); 10712069f457SEric Paris err = show_sb_opts(m, sb); 10722069f457SEric Paris if (err) 10732069f457SEric Paris goto out; 10742d4d4864SRam Pai if (sb->s_op->show_options) 10752d4d4864SRam Pai err = sb->s_op->show_options(m, mnt); 10762d4d4864SRam Pai seq_putc(m, '\n'); 10772069f457SEric Paris out: 10782d4d4864SRam Pai return err; 10792d4d4864SRam Pai } 10802d4d4864SRam Pai 10812d4d4864SRam Pai const struct seq_operations mountinfo_op = { 10822d4d4864SRam Pai .start = m_start, 10832d4d4864SRam Pai .next = m_next, 10842d4d4864SRam Pai .stop = m_stop, 10852d4d4864SRam Pai .show = show_mountinfo, 10862d4d4864SRam Pai }; 10872d4d4864SRam Pai 1088b4629fe2SChuck Lever static int show_vfsstat(struct seq_file *m, void *v) 1089b4629fe2SChuck Lever { 1090b0765fb8SPavel Emelianov struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); 1091c32c2f63SJan Blunck struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; 1092b4629fe2SChuck Lever int err = 0; 1093b4629fe2SChuck Lever 1094b4629fe2SChuck Lever /* device */ 1095c7f404b4SAl Viro if (mnt->mnt_sb->s_op->show_devname) { 1096a877ee03SBryan Schumaker seq_puts(m, "device "); 1097c7f404b4SAl Viro err = mnt->mnt_sb->s_op->show_devname(m, mnt); 1098c7f404b4SAl Viro } else { 1099b4629fe2SChuck Lever if (mnt->mnt_devname) { 1100b4629fe2SChuck Lever seq_puts(m, "device "); 1101b4629fe2SChuck Lever mangle(m, mnt->mnt_devname); 1102b4629fe2SChuck Lever } else 1103b4629fe2SChuck Lever seq_puts(m, "no device"); 1104c7f404b4SAl Viro } 1105b4629fe2SChuck Lever 1106b4629fe2SChuck Lever /* mount point */ 1107b4629fe2SChuck Lever seq_puts(m, " mounted on "); 1108c32c2f63SJan Blunck seq_path(m, &mnt_path, " \t\n\\"); 1109b4629fe2SChuck Lever seq_putc(m, ' '); 1110b4629fe2SChuck Lever 1111b4629fe2SChuck Lever /* file system type */ 1112b4629fe2SChuck Lever seq_puts(m, "with fstype "); 11132d4d4864SRam Pai show_type(m, mnt->mnt_sb); 1114b4629fe2SChuck Lever 1115b4629fe2SChuck Lever /* optional statistics */ 1116b4629fe2SChuck Lever if (mnt->mnt_sb->s_op->show_stats) { 1117b4629fe2SChuck Lever seq_putc(m, ' '); 1118c7f404b4SAl Viro if (!err) 1119b4629fe2SChuck Lever err = mnt->mnt_sb->s_op->show_stats(m, mnt); 1120b4629fe2SChuck Lever } 1121b4629fe2SChuck Lever 1122b4629fe2SChuck Lever seq_putc(m, '\n'); 1123b4629fe2SChuck Lever return err; 1124b4629fe2SChuck Lever } 1125b4629fe2SChuck Lever 1126a1a2c409SMiklos Szeredi const struct seq_operations mountstats_op = { 1127b4629fe2SChuck Lever .start = m_start, 1128b4629fe2SChuck Lever .next = m_next, 1129b4629fe2SChuck Lever .stop = m_stop, 1130b4629fe2SChuck Lever .show = show_vfsstat, 1131b4629fe2SChuck Lever }; 1132a1a2c409SMiklos Szeredi #endif /* CONFIG_PROC_FS */ 1133b4629fe2SChuck Lever 11341da177e4SLinus Torvalds /** 11351da177e4SLinus Torvalds * may_umount_tree - check if a mount tree is busy 11361da177e4SLinus Torvalds * @mnt: root of mount tree 11371da177e4SLinus Torvalds * 11381da177e4SLinus Torvalds * This is called to check if a tree of mounts has any 11391da177e4SLinus Torvalds * open files, pwds, chroots or sub mounts that are 11401da177e4SLinus Torvalds * busy. 11411da177e4SLinus Torvalds */ 11421da177e4SLinus Torvalds int may_umount_tree(struct vfsmount *mnt) 11431da177e4SLinus Torvalds { 114436341f64SRam Pai int actual_refs = 0; 114536341f64SRam Pai int minimum_refs = 0; 1146315fc83eSAl Viro struct mount *p; 1147315fc83eSAl Viro BUG_ON(!mnt); 11481da177e4SLinus Torvalds 1149b3e19d92SNick Piggin /* write lock needed for mnt_get_count */ 1150b3e19d92SNick Piggin br_write_lock(vfsmount_lock); 1151315fc83eSAl Viro for (p = real_mount(mnt); p; p = next_mnt(p, mnt)) { 1152315fc83eSAl Viro actual_refs += mnt_get_count(&p->mnt); 11531da177e4SLinus Torvalds minimum_refs += 2; 11541da177e4SLinus Torvalds } 1155b3e19d92SNick Piggin br_write_unlock(vfsmount_lock); 11561da177e4SLinus Torvalds 11571da177e4SLinus Torvalds if (actual_refs > minimum_refs) 11581da177e4SLinus Torvalds return 0; 1159e3474a8eSIan Kent 1160e3474a8eSIan Kent return 1; 11611da177e4SLinus Torvalds } 11621da177e4SLinus Torvalds 11631da177e4SLinus Torvalds EXPORT_SYMBOL(may_umount_tree); 11641da177e4SLinus Torvalds 11651da177e4SLinus Torvalds /** 11661da177e4SLinus Torvalds * may_umount - check if a mount point is busy 11671da177e4SLinus Torvalds * @mnt: root of mount 11681da177e4SLinus Torvalds * 11691da177e4SLinus Torvalds * This is called to check if a mount point has any 11701da177e4SLinus Torvalds * open files, pwds, chroots or sub mounts. If the 11711da177e4SLinus Torvalds * mount has sub mounts this will return busy 11721da177e4SLinus Torvalds * regardless of whether the sub mounts are busy. 11731da177e4SLinus Torvalds * 11741da177e4SLinus Torvalds * Doesn't take quota and stuff into account. IOW, in some cases it will 11751da177e4SLinus Torvalds * give false negatives. The main reason why it's here is that we need 11761da177e4SLinus Torvalds * a non-destructive way to look for easily umountable filesystems. 11771da177e4SLinus Torvalds */ 11781da177e4SLinus Torvalds int may_umount(struct vfsmount *mnt) 11791da177e4SLinus Torvalds { 1180e3474a8eSIan Kent int ret = 1; 11818ad08d8aSAl Viro down_read(&namespace_sem); 1182b3e19d92SNick Piggin br_write_lock(vfsmount_lock); 1183a05964f3SRam Pai if (propagate_mount_busy(mnt, 2)) 1184e3474a8eSIan Kent ret = 0; 1185b3e19d92SNick Piggin br_write_unlock(vfsmount_lock); 11868ad08d8aSAl Viro up_read(&namespace_sem); 1187a05964f3SRam Pai return ret; 11881da177e4SLinus Torvalds } 11891da177e4SLinus Torvalds 11901da177e4SLinus Torvalds EXPORT_SYMBOL(may_umount); 11911da177e4SLinus Torvalds 1192b90fa9aeSRam Pai void release_mounts(struct list_head *head) 11931da177e4SLinus Torvalds { 119470fbcdf4SRam Pai struct vfsmount *mnt; 119570fbcdf4SRam Pai while (!list_empty(head)) { 1196b5e61818SPavel Emelianov mnt = list_first_entry(head, struct vfsmount, mnt_hash); 119770fbcdf4SRam Pai list_del_init(&mnt->mnt_hash); 1198b2dba1afSAl Viro if (mnt_has_parent(mnt)) { 119970fbcdf4SRam Pai struct dentry *dentry; 120070fbcdf4SRam Pai struct vfsmount *m; 120199b7db7bSNick Piggin 120299b7db7bSNick Piggin br_write_lock(vfsmount_lock); 120370fbcdf4SRam Pai dentry = mnt->mnt_mountpoint; 120470fbcdf4SRam Pai m = mnt->mnt_parent; 120570fbcdf4SRam Pai mnt->mnt_mountpoint = mnt->mnt_root; 120670fbcdf4SRam Pai mnt->mnt_parent = mnt; 12077c4b93d8SAl Viro m->mnt_ghosts--; 120899b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 120970fbcdf4SRam Pai dput(dentry); 121070fbcdf4SRam Pai mntput(m); 12111da177e4SLinus Torvalds } 1212f03c6599SAl Viro mntput(mnt); 121370fbcdf4SRam Pai } 121470fbcdf4SRam Pai } 121570fbcdf4SRam Pai 121699b7db7bSNick Piggin /* 121799b7db7bSNick Piggin * vfsmount lock must be held for write 121899b7db7bSNick Piggin * namespace_sem must be held for write 121999b7db7bSNick Piggin */ 1220a05964f3SRam Pai void umount_tree(struct vfsmount *mnt, int propagate, struct list_head *kill) 122170fbcdf4SRam Pai { 12227b8a53fdSAl Viro LIST_HEAD(tmp_list); 1223315fc83eSAl Viro struct mount *p; 122470fbcdf4SRam Pai 1225315fc83eSAl Viro for (p = real_mount(mnt); p; p = next_mnt(p, mnt)) 1226315fc83eSAl Viro list_move(&p->mnt.mnt_hash, &tmp_list); 122770fbcdf4SRam Pai 1228a05964f3SRam Pai if (propagate) 12297b8a53fdSAl Viro propagate_umount(&tmp_list); 1230a05964f3SRam Pai 1231315fc83eSAl Viro list_for_each_entry(p, &tmp_list, mnt.mnt_hash) { 1232315fc83eSAl Viro list_del_init(&p->mnt.mnt_expire); 1233315fc83eSAl Viro list_del_init(&p->mnt.mnt_list); 1234315fc83eSAl Viro __touch_mnt_namespace(p->mnt.mnt_ns); 1235315fc83eSAl Viro p->mnt.mnt_ns = NULL; 1236315fc83eSAl Viro __mnt_make_shortterm(&p->mnt); 1237315fc83eSAl Viro list_del_init(&p->mnt.mnt_child); 1238315fc83eSAl Viro if (mnt_has_parent(&p->mnt)) { 1239315fc83eSAl Viro p->mnt.mnt_parent->mnt_ghosts++; 1240315fc83eSAl Viro dentry_reset_mounted(p->mnt.mnt_mountpoint); 12417c4b93d8SAl Viro } 12420f0afb1dSAl Viro change_mnt_propagation(p, MS_PRIVATE); 12431da177e4SLinus Torvalds } 12447b8a53fdSAl Viro list_splice(&tmp_list, kill); 12451da177e4SLinus Torvalds } 12461da177e4SLinus Torvalds 1247c35038beSAl Viro static void shrink_submounts(struct vfsmount *mnt, struct list_head *umounts); 1248c35038beSAl Viro 12491da177e4SLinus Torvalds static int do_umount(struct vfsmount *mnt, int flags) 12501da177e4SLinus Torvalds { 12511da177e4SLinus Torvalds struct super_block *sb = mnt->mnt_sb; 12521da177e4SLinus Torvalds int retval; 125370fbcdf4SRam Pai LIST_HEAD(umount_list); 12541da177e4SLinus Torvalds 12551da177e4SLinus Torvalds retval = security_sb_umount(mnt, flags); 12561da177e4SLinus Torvalds if (retval) 12571da177e4SLinus Torvalds return retval; 12581da177e4SLinus Torvalds 12591da177e4SLinus Torvalds /* 12601da177e4SLinus Torvalds * Allow userspace to request a mountpoint be expired rather than 12611da177e4SLinus Torvalds * unmounting unconditionally. Unmount only happens if: 12621da177e4SLinus Torvalds * (1) the mark is already set (the mark is cleared by mntput()) 12631da177e4SLinus Torvalds * (2) the usage count == 1 [parent vfsmount] + 1 [sys_umount] 12641da177e4SLinus Torvalds */ 12651da177e4SLinus Torvalds if (flags & MNT_EXPIRE) { 12666ac08c39SJan Blunck if (mnt == current->fs->root.mnt || 12671da177e4SLinus Torvalds flags & (MNT_FORCE | MNT_DETACH)) 12681da177e4SLinus Torvalds return -EINVAL; 12691da177e4SLinus Torvalds 1270b3e19d92SNick Piggin /* 1271b3e19d92SNick Piggin * probably don't strictly need the lock here if we examined 1272b3e19d92SNick Piggin * all race cases, but it's a slowpath. 1273b3e19d92SNick Piggin */ 1274b3e19d92SNick Piggin br_write_lock(vfsmount_lock); 1275b3e19d92SNick Piggin if (mnt_get_count(mnt) != 2) { 1276bf9faa2aSJ. R. Okajima br_write_unlock(vfsmount_lock); 12771da177e4SLinus Torvalds return -EBUSY; 1278b3e19d92SNick Piggin } 1279b3e19d92SNick Piggin br_write_unlock(vfsmount_lock); 12801da177e4SLinus Torvalds 12811da177e4SLinus Torvalds if (!xchg(&mnt->mnt_expiry_mark, 1)) 12821da177e4SLinus Torvalds return -EAGAIN; 12831da177e4SLinus Torvalds } 12841da177e4SLinus Torvalds 12851da177e4SLinus Torvalds /* 12861da177e4SLinus Torvalds * If we may have to abort operations to get out of this 12871da177e4SLinus Torvalds * mount, and they will themselves hold resources we must 12881da177e4SLinus Torvalds * allow the fs to do things. In the Unix tradition of 12891da177e4SLinus Torvalds * 'Gee thats tricky lets do it in userspace' the umount_begin 12901da177e4SLinus Torvalds * might fail to complete on the first run through as other tasks 12911da177e4SLinus Torvalds * must return, and the like. Thats for the mount program to worry 12921da177e4SLinus Torvalds * about for the moment. 12931da177e4SLinus Torvalds */ 12941da177e4SLinus Torvalds 129542faad99SAl Viro if (flags & MNT_FORCE && sb->s_op->umount_begin) { 129642faad99SAl Viro sb->s_op->umount_begin(sb); 129742faad99SAl Viro } 12981da177e4SLinus Torvalds 12991da177e4SLinus Torvalds /* 13001da177e4SLinus Torvalds * No sense to grab the lock for this test, but test itself looks 13011da177e4SLinus Torvalds * somewhat bogus. Suggestions for better replacement? 13021da177e4SLinus Torvalds * Ho-hum... In principle, we might treat that as umount + switch 13031da177e4SLinus Torvalds * to rootfs. GC would eventually take care of the old vfsmount. 13041da177e4SLinus Torvalds * Actually it makes sense, especially if rootfs would contain a 13051da177e4SLinus Torvalds * /reboot - static binary that would close all descriptors and 13061da177e4SLinus Torvalds * call reboot(9). Then init(8) could umount root and exec /reboot. 13071da177e4SLinus Torvalds */ 13086ac08c39SJan Blunck if (mnt == current->fs->root.mnt && !(flags & MNT_DETACH)) { 13091da177e4SLinus Torvalds /* 13101da177e4SLinus Torvalds * Special case for "unmounting" root ... 13111da177e4SLinus Torvalds * we just try to remount it readonly. 13121da177e4SLinus Torvalds */ 13131da177e4SLinus Torvalds down_write(&sb->s_umount); 13144aa98cf7SAl Viro if (!(sb->s_flags & MS_RDONLY)) 13151da177e4SLinus Torvalds retval = do_remount_sb(sb, MS_RDONLY, NULL, 0); 13161da177e4SLinus Torvalds up_write(&sb->s_umount); 13171da177e4SLinus Torvalds return retval; 13181da177e4SLinus Torvalds } 13191da177e4SLinus Torvalds 1320390c6843SRam Pai down_write(&namespace_sem); 132199b7db7bSNick Piggin br_write_lock(vfsmount_lock); 13225addc5ddSAl Viro event++; 13231da177e4SLinus Torvalds 1324c35038beSAl Viro if (!(flags & MNT_DETACH)) 1325c35038beSAl Viro shrink_submounts(mnt, &umount_list); 1326c35038beSAl Viro 13271da177e4SLinus Torvalds retval = -EBUSY; 1328a05964f3SRam Pai if (flags & MNT_DETACH || !propagate_mount_busy(mnt, 2)) { 13291da177e4SLinus Torvalds if (!list_empty(&mnt->mnt_list)) 1330a05964f3SRam Pai umount_tree(mnt, 1, &umount_list); 13311da177e4SLinus Torvalds retval = 0; 13321da177e4SLinus Torvalds } 133399b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 1334390c6843SRam Pai up_write(&namespace_sem); 133570fbcdf4SRam Pai release_mounts(&umount_list); 13361da177e4SLinus Torvalds return retval; 13371da177e4SLinus Torvalds } 13381da177e4SLinus Torvalds 13391da177e4SLinus Torvalds /* 13401da177e4SLinus Torvalds * Now umount can handle mount points as well as block devices. 13411da177e4SLinus Torvalds * This is important for filesystems which use unnamed block devices. 13421da177e4SLinus Torvalds * 13431da177e4SLinus Torvalds * We now support a flag for forced unmount like the other 'big iron' 13441da177e4SLinus Torvalds * unixes. Our API is identical to OSF/1 to avoid making a mess of AMD 13451da177e4SLinus Torvalds */ 13461da177e4SLinus Torvalds 1347bdc480e3SHeiko Carstens SYSCALL_DEFINE2(umount, char __user *, name, int, flags) 13481da177e4SLinus Torvalds { 13492d8f3038SAl Viro struct path path; 13501da177e4SLinus Torvalds int retval; 1351db1f05bbSMiklos Szeredi int lookup_flags = 0; 13521da177e4SLinus Torvalds 1353db1f05bbSMiklos Szeredi if (flags & ~(MNT_FORCE | MNT_DETACH | MNT_EXPIRE | UMOUNT_NOFOLLOW)) 1354db1f05bbSMiklos Szeredi return -EINVAL; 1355db1f05bbSMiklos Szeredi 1356db1f05bbSMiklos Szeredi if (!(flags & UMOUNT_NOFOLLOW)) 1357db1f05bbSMiklos Szeredi lookup_flags |= LOOKUP_FOLLOW; 1358db1f05bbSMiklos Szeredi 1359db1f05bbSMiklos Szeredi retval = user_path_at(AT_FDCWD, name, lookup_flags, &path); 13601da177e4SLinus Torvalds if (retval) 13611da177e4SLinus Torvalds goto out; 13621da177e4SLinus Torvalds retval = -EINVAL; 13632d8f3038SAl Viro if (path.dentry != path.mnt->mnt_root) 13641da177e4SLinus Torvalds goto dput_and_out; 13652d8f3038SAl Viro if (!check_mnt(path.mnt)) 13661da177e4SLinus Torvalds goto dput_and_out; 13671da177e4SLinus Torvalds 13681da177e4SLinus Torvalds retval = -EPERM; 13691da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 13701da177e4SLinus Torvalds goto dput_and_out; 13711da177e4SLinus Torvalds 13722d8f3038SAl Viro retval = do_umount(path.mnt, flags); 13731da177e4SLinus Torvalds dput_and_out: 1374429731b1SJan Blunck /* we mustn't call path_put() as that would clear mnt_expiry_mark */ 13752d8f3038SAl Viro dput(path.dentry); 13762d8f3038SAl Viro mntput_no_expire(path.mnt); 13771da177e4SLinus Torvalds out: 13781da177e4SLinus Torvalds return retval; 13791da177e4SLinus Torvalds } 13801da177e4SLinus Torvalds 13811da177e4SLinus Torvalds #ifdef __ARCH_WANT_SYS_OLDUMOUNT 13821da177e4SLinus Torvalds 13831da177e4SLinus Torvalds /* 13841da177e4SLinus Torvalds * The 2.0 compatible umount. No flags. 13851da177e4SLinus Torvalds */ 1386bdc480e3SHeiko Carstens SYSCALL_DEFINE1(oldumount, char __user *, name) 13871da177e4SLinus Torvalds { 13881da177e4SLinus Torvalds return sys_umount(name, 0); 13891da177e4SLinus Torvalds } 13901da177e4SLinus Torvalds 13911da177e4SLinus Torvalds #endif 13921da177e4SLinus Torvalds 13932d92ab3cSAl Viro static int mount_is_safe(struct path *path) 13941da177e4SLinus Torvalds { 13951da177e4SLinus Torvalds if (capable(CAP_SYS_ADMIN)) 13961da177e4SLinus Torvalds return 0; 13971da177e4SLinus Torvalds return -EPERM; 13981da177e4SLinus Torvalds #ifdef notyet 13992d92ab3cSAl Viro if (S_ISLNK(path->dentry->d_inode->i_mode)) 14001da177e4SLinus Torvalds return -EPERM; 14012d92ab3cSAl Viro if (path->dentry->d_inode->i_mode & S_ISVTX) { 1402da9592edSDavid Howells if (current_uid() != path->dentry->d_inode->i_uid) 14031da177e4SLinus Torvalds return -EPERM; 14041da177e4SLinus Torvalds } 14052d92ab3cSAl Viro if (inode_permission(path->dentry->d_inode, MAY_WRITE)) 14061da177e4SLinus Torvalds return -EPERM; 14071da177e4SLinus Torvalds return 0; 14081da177e4SLinus Torvalds #endif 14091da177e4SLinus Torvalds } 14101da177e4SLinus Torvalds 1411cb338d06SAl Viro struct mount *copy_tree(struct vfsmount *mnt, struct dentry *dentry, 141236341f64SRam Pai int flag) 14131da177e4SLinus Torvalds { 1414cb338d06SAl Viro struct mount *res, *q; 1415cb338d06SAl Viro struct vfsmount *p, *r; 14161a390689SAl Viro struct path path; 14171da177e4SLinus Torvalds 14189676f0c6SRam Pai if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(mnt)) 14199676f0c6SRam Pai return NULL; 14209676f0c6SRam Pai 142136341f64SRam Pai res = q = clone_mnt(mnt, dentry, flag); 14221da177e4SLinus Torvalds if (!q) 14231da177e4SLinus Torvalds goto Enomem; 1424cb338d06SAl Viro q->mnt.mnt_mountpoint = mnt->mnt_mountpoint; 14251da177e4SLinus Torvalds 14261da177e4SLinus Torvalds p = mnt; 1427fdadd65fSDomen Puncer list_for_each_entry(r, &mnt->mnt_mounts, mnt_child) { 1428315fc83eSAl Viro struct mount *s; 14297ec02ef1SJan Blunck if (!is_subdir(r->mnt_mountpoint, dentry)) 14301da177e4SLinus Torvalds continue; 14311da177e4SLinus Torvalds 1432315fc83eSAl Viro for (s = real_mount(r); s; s = next_mnt(s, r)) { 1433315fc83eSAl Viro if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(&s->mnt)) { 14349676f0c6SRam Pai s = skip_mnt_tree(s); 14359676f0c6SRam Pai continue; 14369676f0c6SRam Pai } 1437315fc83eSAl Viro while (p != s->mnt.mnt_parent) { 14381da177e4SLinus Torvalds p = p->mnt_parent; 1439cb338d06SAl Viro q = real_mount(q->mnt.mnt_parent); 14401da177e4SLinus Torvalds } 1441315fc83eSAl Viro p = &s->mnt; 1442cb338d06SAl Viro path.mnt = &q->mnt; 14431a390689SAl Viro path.dentry = p->mnt_mountpoint; 144436341f64SRam Pai q = clone_mnt(p, p->mnt_root, flag); 14451da177e4SLinus Torvalds if (!q) 14461da177e4SLinus Torvalds goto Enomem; 144799b7db7bSNick Piggin br_write_lock(vfsmount_lock); 1448cb338d06SAl Viro list_add_tail(&q->mnt.mnt_list, &res->mnt.mnt_list); 1449cb338d06SAl Viro attach_mnt(q, &path); 145099b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 14511da177e4SLinus Torvalds } 14521da177e4SLinus Torvalds } 14531da177e4SLinus Torvalds return res; 14541da177e4SLinus Torvalds Enomem: 14551da177e4SLinus Torvalds if (res) { 145670fbcdf4SRam Pai LIST_HEAD(umount_list); 145799b7db7bSNick Piggin br_write_lock(vfsmount_lock); 1458cb338d06SAl Viro umount_tree(&res->mnt, 0, &umount_list); 145999b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 146070fbcdf4SRam Pai release_mounts(&umount_list); 14611da177e4SLinus Torvalds } 14621da177e4SLinus Torvalds return NULL; 14631da177e4SLinus Torvalds } 14641da177e4SLinus Torvalds 1465589ff870SAl Viro struct vfsmount *collect_mounts(struct path *path) 14668aec0809SAl Viro { 1467cb338d06SAl Viro struct mount *tree; 14681a60a280SAl Viro down_write(&namespace_sem); 1469589ff870SAl Viro tree = copy_tree(path->mnt, path->dentry, CL_COPY_ALL | CL_PRIVATE); 14701a60a280SAl Viro up_write(&namespace_sem); 1471cb338d06SAl Viro return tree ? &tree->mnt : NULL; 14728aec0809SAl Viro } 14738aec0809SAl Viro 14748aec0809SAl Viro void drop_collected_mounts(struct vfsmount *mnt) 14758aec0809SAl Viro { 14768aec0809SAl Viro LIST_HEAD(umount_list); 14771a60a280SAl Viro down_write(&namespace_sem); 147899b7db7bSNick Piggin br_write_lock(vfsmount_lock); 14798aec0809SAl Viro umount_tree(mnt, 0, &umount_list); 148099b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 14811a60a280SAl Viro up_write(&namespace_sem); 14828aec0809SAl Viro release_mounts(&umount_list); 14838aec0809SAl Viro } 14848aec0809SAl Viro 14851f707137SAl Viro int iterate_mounts(int (*f)(struct vfsmount *, void *), void *arg, 14861f707137SAl Viro struct vfsmount *root) 14871f707137SAl Viro { 14881f707137SAl Viro struct vfsmount *mnt; 14891f707137SAl Viro int res = f(root, arg); 14901f707137SAl Viro if (res) 14911f707137SAl Viro return res; 14921f707137SAl Viro list_for_each_entry(mnt, &root->mnt_list, mnt_list) { 14931f707137SAl Viro res = f(mnt, arg); 14941f707137SAl Viro if (res) 14951f707137SAl Viro return res; 14961f707137SAl Viro } 14971f707137SAl Viro return 0; 14981f707137SAl Viro } 14991f707137SAl Viro 15004b8b21f4SAl Viro static void cleanup_group_ids(struct mount *mnt, struct mount *end) 1501719f5d7fSMiklos Szeredi { 1502315fc83eSAl Viro struct mount *p; 1503719f5d7fSMiklos Szeredi 15044b8b21f4SAl Viro for (p = mnt; p != end; p = next_mnt(p, &mnt->mnt)) { 1505315fc83eSAl Viro if (p->mnt.mnt_group_id && !IS_MNT_SHARED(&p->mnt)) 15064b8b21f4SAl Viro mnt_release_group_id(p); 1507719f5d7fSMiklos Szeredi } 1508719f5d7fSMiklos Szeredi } 1509719f5d7fSMiklos Szeredi 15104b8b21f4SAl Viro static int invent_group_ids(struct mount *mnt, bool recurse) 1511719f5d7fSMiklos Szeredi { 1512315fc83eSAl Viro struct mount *p; 1513719f5d7fSMiklos Szeredi 15144b8b21f4SAl Viro for (p = mnt; p; p = recurse ? next_mnt(p, &mnt->mnt) : NULL) { 1515315fc83eSAl Viro if (!p->mnt.mnt_group_id && !IS_MNT_SHARED(&p->mnt)) { 15164b8b21f4SAl Viro int err = mnt_alloc_group_id(p); 1517719f5d7fSMiklos Szeredi if (err) { 15184b8b21f4SAl Viro cleanup_group_ids(mnt, p); 1519719f5d7fSMiklos Szeredi return err; 1520719f5d7fSMiklos Szeredi } 1521719f5d7fSMiklos Szeredi } 1522719f5d7fSMiklos Szeredi } 1523719f5d7fSMiklos Szeredi 1524719f5d7fSMiklos Szeredi return 0; 1525719f5d7fSMiklos Szeredi } 1526719f5d7fSMiklos Szeredi 1527b90fa9aeSRam Pai /* 1528b90fa9aeSRam Pai * @source_mnt : mount tree to be attached 1529b90fa9aeSRam Pai * @nd : place the mount tree @source_mnt is attached 153021444403SRam Pai * @parent_nd : if non-null, detach the source_mnt from its parent and 153121444403SRam Pai * store the parent mount and mountpoint dentry. 153221444403SRam Pai * (done when source_mnt is moved) 1533b90fa9aeSRam Pai * 1534b90fa9aeSRam Pai * NOTE: in the table below explains the semantics when a source mount 1535b90fa9aeSRam Pai * of a given type is attached to a destination mount of a given type. 15369676f0c6SRam Pai * --------------------------------------------------------------------------- 1537b90fa9aeSRam Pai * | BIND MOUNT OPERATION | 15389676f0c6SRam Pai * |************************************************************************** 15399676f0c6SRam Pai * | source-->| shared | private | slave | unbindable | 15409676f0c6SRam Pai * | dest | | | | | 15419676f0c6SRam Pai * | | | | | | | 15429676f0c6SRam Pai * | v | | | | | 15439676f0c6SRam Pai * |************************************************************************** 15449676f0c6SRam Pai * | shared | shared (++) | shared (+) | shared(+++)| invalid | 15455afe0022SRam Pai * | | | | | | 15469676f0c6SRam Pai * |non-shared| shared (+) | private | slave (*) | invalid | 15479676f0c6SRam Pai * *************************************************************************** 1548b90fa9aeSRam Pai * A bind operation clones the source mount and mounts the clone on the 1549b90fa9aeSRam Pai * destination mount. 1550b90fa9aeSRam Pai * 1551b90fa9aeSRam Pai * (++) the cloned mount is propagated to all the mounts in the propagation 1552b90fa9aeSRam Pai * tree of the destination mount and the cloned mount is added to 1553b90fa9aeSRam Pai * the peer group of the source mount. 1554b90fa9aeSRam Pai * (+) the cloned mount is created under the destination mount and is marked 1555b90fa9aeSRam Pai * as shared. The cloned mount is added to the peer group of the source 1556b90fa9aeSRam Pai * mount. 15575afe0022SRam Pai * (+++) the mount is propagated to all the mounts in the propagation tree 15585afe0022SRam Pai * of the destination mount and the cloned mount is made slave 15595afe0022SRam Pai * of the same master as that of the source mount. The cloned mount 15605afe0022SRam Pai * is marked as 'shared and slave'. 15615afe0022SRam Pai * (*) the cloned mount is made a slave of the same master as that of the 15625afe0022SRam Pai * source mount. 15635afe0022SRam Pai * 15649676f0c6SRam Pai * --------------------------------------------------------------------------- 156521444403SRam Pai * | MOVE MOUNT OPERATION | 15669676f0c6SRam Pai * |************************************************************************** 15679676f0c6SRam Pai * | source-->| shared | private | slave | unbindable | 15689676f0c6SRam Pai * | dest | | | | | 15699676f0c6SRam Pai * | | | | | | | 15709676f0c6SRam Pai * | v | | | | | 15719676f0c6SRam Pai * |************************************************************************** 15729676f0c6SRam Pai * | shared | shared (+) | shared (+) | shared(+++) | invalid | 15735afe0022SRam Pai * | | | | | | 15749676f0c6SRam Pai * |non-shared| shared (+*) | private | slave (*) | unbindable | 15759676f0c6SRam Pai * *************************************************************************** 15765afe0022SRam Pai * 15775afe0022SRam Pai * (+) the mount is moved to the destination. And is then propagated to 15785afe0022SRam Pai * all the mounts in the propagation tree of the destination mount. 157921444403SRam Pai * (+*) the mount is moved to the destination. 15805afe0022SRam Pai * (+++) the mount is moved to the destination and is then propagated to 15815afe0022SRam Pai * all the mounts belonging to the destination mount's propagation tree. 15825afe0022SRam Pai * the mount is marked as 'shared and slave'. 15835afe0022SRam Pai * (*) the mount continues to be a slave at the new location. 1584b90fa9aeSRam Pai * 1585b90fa9aeSRam Pai * if the source mount is a tree, the operations explained above is 1586b90fa9aeSRam Pai * applied to each mount in the tree. 1587b90fa9aeSRam Pai * Must be called without spinlocks held, since this function can sleep 1588b90fa9aeSRam Pai * in allocations. 1589b90fa9aeSRam Pai */ 15900fb54e50SAl Viro static int attach_recursive_mnt(struct mount *source_mnt, 15911a390689SAl Viro struct path *path, struct path *parent_path) 1592b90fa9aeSRam Pai { 1593b90fa9aeSRam Pai LIST_HEAD(tree_list); 15941a390689SAl Viro struct vfsmount *dest_mnt = path->mnt; 15951a390689SAl Viro struct dentry *dest_dentry = path->dentry; 1596315fc83eSAl Viro struct mount *child, *p; 1597719f5d7fSMiklos Szeredi int err; 1598b90fa9aeSRam Pai 1599719f5d7fSMiklos Szeredi if (IS_MNT_SHARED(dest_mnt)) { 16000fb54e50SAl Viro err = invent_group_ids(source_mnt, true); 1601719f5d7fSMiklos Szeredi if (err) 1602719f5d7fSMiklos Szeredi goto out; 1603719f5d7fSMiklos Szeredi } 16040fb54e50SAl Viro err = propagate_mnt(dest_mnt, dest_dentry, &source_mnt->mnt, &tree_list); 1605719f5d7fSMiklos Szeredi if (err) 1606719f5d7fSMiklos Szeredi goto out_cleanup_ids; 1607b90fa9aeSRam Pai 160899b7db7bSNick Piggin br_write_lock(vfsmount_lock); 1609df1a1ad2SAl Viro 1610b90fa9aeSRam Pai if (IS_MNT_SHARED(dest_mnt)) { 16110fb54e50SAl Viro for (p = source_mnt; p; p = next_mnt(p, &source_mnt->mnt)) 16120f0afb1dSAl Viro set_mnt_shared(p); 1613b90fa9aeSRam Pai } 16141a390689SAl Viro if (parent_path) { 16150fb54e50SAl Viro detach_mnt(source_mnt, parent_path); 16160fb54e50SAl Viro attach_mnt(source_mnt, path); 1617e5d67f07SAl Viro touch_mnt_namespace(parent_path->mnt->mnt_ns); 161821444403SRam Pai } else { 16190fb54e50SAl Viro mnt_set_mountpoint(dest_mnt, dest_dentry, &source_mnt->mnt); 16200fb54e50SAl Viro commit_tree(source_mnt); 162121444403SRam Pai } 1622b90fa9aeSRam Pai 1623315fc83eSAl Viro list_for_each_entry_safe(child, p, &tree_list, mnt.mnt_hash) { 1624315fc83eSAl Viro list_del_init(&child->mnt.mnt_hash); 16254b2619a5SAl Viro commit_tree(child); 1626b90fa9aeSRam Pai } 162799b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 162899b7db7bSNick Piggin 1629b90fa9aeSRam Pai return 0; 1630719f5d7fSMiklos Szeredi 1631719f5d7fSMiklos Szeredi out_cleanup_ids: 1632719f5d7fSMiklos Szeredi if (IS_MNT_SHARED(dest_mnt)) 16330fb54e50SAl Viro cleanup_group_ids(source_mnt, NULL); 1634719f5d7fSMiklos Szeredi out: 1635719f5d7fSMiklos Szeredi return err; 1636b90fa9aeSRam Pai } 1637b90fa9aeSRam Pai 1638b12cea91SAl Viro static int lock_mount(struct path *path) 1639b12cea91SAl Viro { 1640b12cea91SAl Viro struct vfsmount *mnt; 1641b12cea91SAl Viro retry: 1642b12cea91SAl Viro mutex_lock(&path->dentry->d_inode->i_mutex); 1643b12cea91SAl Viro if (unlikely(cant_mount(path->dentry))) { 1644b12cea91SAl Viro mutex_unlock(&path->dentry->d_inode->i_mutex); 1645b12cea91SAl Viro return -ENOENT; 1646b12cea91SAl Viro } 1647b12cea91SAl Viro down_write(&namespace_sem); 1648b12cea91SAl Viro mnt = lookup_mnt(path); 1649b12cea91SAl Viro if (likely(!mnt)) 1650b12cea91SAl Viro return 0; 1651b12cea91SAl Viro up_write(&namespace_sem); 1652b12cea91SAl Viro mutex_unlock(&path->dentry->d_inode->i_mutex); 1653b12cea91SAl Viro path_put(path); 1654b12cea91SAl Viro path->mnt = mnt; 1655b12cea91SAl Viro path->dentry = dget(mnt->mnt_root); 1656b12cea91SAl Viro goto retry; 1657b12cea91SAl Viro } 1658b12cea91SAl Viro 1659b12cea91SAl Viro static void unlock_mount(struct path *path) 1660b12cea91SAl Viro { 1661b12cea91SAl Viro up_write(&namespace_sem); 1662b12cea91SAl Viro mutex_unlock(&path->dentry->d_inode->i_mutex); 1663b12cea91SAl Viro } 1664b12cea91SAl Viro 16658c3ee42eSAl Viro static int graft_tree(struct vfsmount *mnt, struct path *path) 16661da177e4SLinus Torvalds { 16671da177e4SLinus Torvalds if (mnt->mnt_sb->s_flags & MS_NOUSER) 16681da177e4SLinus Torvalds return -EINVAL; 16691da177e4SLinus Torvalds 16708c3ee42eSAl Viro if (S_ISDIR(path->dentry->d_inode->i_mode) != 16711da177e4SLinus Torvalds S_ISDIR(mnt->mnt_root->d_inode->i_mode)) 16721da177e4SLinus Torvalds return -ENOTDIR; 16731da177e4SLinus Torvalds 1674b12cea91SAl Viro if (d_unlinked(path->dentry)) 1675b12cea91SAl Viro return -ENOENT; 16761da177e4SLinus Torvalds 16770fb54e50SAl Viro return attach_recursive_mnt(real_mount(mnt), path, NULL); 16781da177e4SLinus Torvalds } 16791da177e4SLinus Torvalds 16801da177e4SLinus Torvalds /* 16817a2e8a8fSValerie Aurora * Sanity check the flags to change_mnt_propagation. 16827a2e8a8fSValerie Aurora */ 16837a2e8a8fSValerie Aurora 16847a2e8a8fSValerie Aurora static int flags_to_propagation_type(int flags) 16857a2e8a8fSValerie Aurora { 16867c6e984dSRoman Borisov int type = flags & ~(MS_REC | MS_SILENT); 16877a2e8a8fSValerie Aurora 16887a2e8a8fSValerie Aurora /* Fail if any non-propagation flags are set */ 16897a2e8a8fSValerie Aurora if (type & ~(MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE)) 16907a2e8a8fSValerie Aurora return 0; 16917a2e8a8fSValerie Aurora /* Only one propagation flag should be set */ 16927a2e8a8fSValerie Aurora if (!is_power_of_2(type)) 16937a2e8a8fSValerie Aurora return 0; 16947a2e8a8fSValerie Aurora return type; 16957a2e8a8fSValerie Aurora } 16967a2e8a8fSValerie Aurora 16977a2e8a8fSValerie Aurora /* 169807b20889SRam Pai * recursively change the type of the mountpoint. 169907b20889SRam Pai */ 17000a0d8a46SAl Viro static int do_change_type(struct path *path, int flag) 170107b20889SRam Pai { 1702315fc83eSAl Viro struct mount *m; 17034b8b21f4SAl Viro struct mount *mnt = real_mount(path->mnt); 170407b20889SRam Pai int recurse = flag & MS_REC; 17057a2e8a8fSValerie Aurora int type; 1706719f5d7fSMiklos Szeredi int err = 0; 170707b20889SRam Pai 1708ee6f9582SMiklos Szeredi if (!capable(CAP_SYS_ADMIN)) 1709ee6f9582SMiklos Szeredi return -EPERM; 1710ee6f9582SMiklos Szeredi 17112d92ab3cSAl Viro if (path->dentry != path->mnt->mnt_root) 171207b20889SRam Pai return -EINVAL; 171307b20889SRam Pai 17147a2e8a8fSValerie Aurora type = flags_to_propagation_type(flag); 17157a2e8a8fSValerie Aurora if (!type) 17167a2e8a8fSValerie Aurora return -EINVAL; 17177a2e8a8fSValerie Aurora 171807b20889SRam Pai down_write(&namespace_sem); 1719719f5d7fSMiklos Szeredi if (type == MS_SHARED) { 1720719f5d7fSMiklos Szeredi err = invent_group_ids(mnt, recurse); 1721719f5d7fSMiklos Szeredi if (err) 1722719f5d7fSMiklos Szeredi goto out_unlock; 1723719f5d7fSMiklos Szeredi } 1724719f5d7fSMiklos Szeredi 172599b7db7bSNick Piggin br_write_lock(vfsmount_lock); 17264b8b21f4SAl Viro for (m = mnt; m; m = (recurse ? next_mnt(m, &mnt->mnt) : NULL)) 17270f0afb1dSAl Viro change_mnt_propagation(m, type); 172899b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 1729719f5d7fSMiklos Szeredi 1730719f5d7fSMiklos Szeredi out_unlock: 173107b20889SRam Pai up_write(&namespace_sem); 1732719f5d7fSMiklos Szeredi return err; 173307b20889SRam Pai } 173407b20889SRam Pai 173507b20889SRam Pai /* 17361da177e4SLinus Torvalds * do loopback mount. 17371da177e4SLinus Torvalds */ 17380a0d8a46SAl Viro static int do_loopback(struct path *path, char *old_name, 17392dafe1c4SEric Sandeen int recurse) 17401da177e4SLinus Torvalds { 1741b12cea91SAl Viro LIST_HEAD(umount_list); 17422d92ab3cSAl Viro struct path old_path; 1743cb338d06SAl Viro struct mount *mnt = NULL; 17442d92ab3cSAl Viro int err = mount_is_safe(path); 17451da177e4SLinus Torvalds if (err) 17461da177e4SLinus Torvalds return err; 17471da177e4SLinus Torvalds if (!old_name || !*old_name) 17481da177e4SLinus Torvalds return -EINVAL; 1749815d405cSTrond Myklebust err = kern_path(old_name, LOOKUP_FOLLOW|LOOKUP_AUTOMOUNT, &old_path); 17501da177e4SLinus Torvalds if (err) 17511da177e4SLinus Torvalds return err; 17521da177e4SLinus Torvalds 1753b12cea91SAl Viro err = lock_mount(path); 1754b12cea91SAl Viro if (err) 17559676f0c6SRam Pai goto out; 17569676f0c6SRam Pai 1757b12cea91SAl Viro err = -EINVAL; 1758b12cea91SAl Viro if (IS_MNT_UNBINDABLE(old_path.mnt)) 1759b12cea91SAl Viro goto out2; 1760b12cea91SAl Viro 17612d92ab3cSAl Viro if (!check_mnt(path->mnt) || !check_mnt(old_path.mnt)) 1762b12cea91SAl Viro goto out2; 1763ccd48bc7SAl Viro 17641da177e4SLinus Torvalds err = -ENOMEM; 17651da177e4SLinus Torvalds if (recurse) 17662d92ab3cSAl Viro mnt = copy_tree(old_path.mnt, old_path.dentry, 0); 17671da177e4SLinus Torvalds else 17682d92ab3cSAl Viro mnt = clone_mnt(old_path.mnt, old_path.dentry, 0); 17691da177e4SLinus Torvalds 1770ccd48bc7SAl Viro if (!mnt) 1771b12cea91SAl Viro goto out2; 1772ccd48bc7SAl Viro 1773cb338d06SAl Viro err = graft_tree(&mnt->mnt, path); 17741da177e4SLinus Torvalds if (err) { 177599b7db7bSNick Piggin br_write_lock(vfsmount_lock); 1776cb338d06SAl Viro umount_tree(&mnt->mnt, 0, &umount_list); 177799b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 17785b83d2c5SRam Pai } 1779b12cea91SAl Viro out2: 1780b12cea91SAl Viro unlock_mount(path); 1781b12cea91SAl Viro release_mounts(&umount_list); 1782ccd48bc7SAl Viro out: 17832d92ab3cSAl Viro path_put(&old_path); 17841da177e4SLinus Torvalds return err; 17851da177e4SLinus Torvalds } 17861da177e4SLinus Torvalds 17872e4b7fcdSDave Hansen static int change_mount_flags(struct vfsmount *mnt, int ms_flags) 17882e4b7fcdSDave Hansen { 17892e4b7fcdSDave Hansen int error = 0; 17902e4b7fcdSDave Hansen int readonly_request = 0; 17912e4b7fcdSDave Hansen 17922e4b7fcdSDave Hansen if (ms_flags & MS_RDONLY) 17932e4b7fcdSDave Hansen readonly_request = 1; 17942e4b7fcdSDave Hansen if (readonly_request == __mnt_is_readonly(mnt)) 17952e4b7fcdSDave Hansen return 0; 17962e4b7fcdSDave Hansen 17972e4b7fcdSDave Hansen if (readonly_request) 17982e4b7fcdSDave Hansen error = mnt_make_readonly(mnt); 17992e4b7fcdSDave Hansen else 18002e4b7fcdSDave Hansen __mnt_unmake_readonly(mnt); 18012e4b7fcdSDave Hansen return error; 18022e4b7fcdSDave Hansen } 18032e4b7fcdSDave Hansen 18041da177e4SLinus Torvalds /* 18051da177e4SLinus Torvalds * change filesystem flags. dir should be a physical root of filesystem. 18061da177e4SLinus Torvalds * If you've mounted a non-root directory somewhere and want to do remount 18071da177e4SLinus Torvalds * on it - tough luck. 18081da177e4SLinus Torvalds */ 18090a0d8a46SAl Viro static int do_remount(struct path *path, int flags, int mnt_flags, 18101da177e4SLinus Torvalds void *data) 18111da177e4SLinus Torvalds { 18121da177e4SLinus Torvalds int err; 18132d92ab3cSAl Viro struct super_block *sb = path->mnt->mnt_sb; 18141da177e4SLinus Torvalds 18151da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 18161da177e4SLinus Torvalds return -EPERM; 18171da177e4SLinus Torvalds 18182d92ab3cSAl Viro if (!check_mnt(path->mnt)) 18191da177e4SLinus Torvalds return -EINVAL; 18201da177e4SLinus Torvalds 18212d92ab3cSAl Viro if (path->dentry != path->mnt->mnt_root) 18221da177e4SLinus Torvalds return -EINVAL; 18231da177e4SLinus Torvalds 1824ff36fe2cSEric Paris err = security_sb_remount(sb, data); 1825ff36fe2cSEric Paris if (err) 1826ff36fe2cSEric Paris return err; 1827ff36fe2cSEric Paris 18281da177e4SLinus Torvalds down_write(&sb->s_umount); 18292e4b7fcdSDave Hansen if (flags & MS_BIND) 18302d92ab3cSAl Viro err = change_mount_flags(path->mnt, flags); 18314aa98cf7SAl Viro else 18321da177e4SLinus Torvalds err = do_remount_sb(sb, flags, data, 0); 18337b43a79fSAl Viro if (!err) { 183499b7db7bSNick Piggin br_write_lock(vfsmount_lock); 1835495d6c9cSValerie Aurora mnt_flags |= path->mnt->mnt_flags & MNT_PROPAGATION_MASK; 18362d92ab3cSAl Viro path->mnt->mnt_flags = mnt_flags; 183799b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 18387b43a79fSAl Viro } 18391da177e4SLinus Torvalds up_write(&sb->s_umount); 18400e55a7ccSDan Williams if (!err) { 184199b7db7bSNick Piggin br_write_lock(vfsmount_lock); 18420e55a7ccSDan Williams touch_mnt_namespace(path->mnt->mnt_ns); 184399b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 18440e55a7ccSDan Williams } 18451da177e4SLinus Torvalds return err; 18461da177e4SLinus Torvalds } 18471da177e4SLinus Torvalds 1848cbbe362cSAl Viro static inline int tree_contains_unbindable(struct mount *mnt) 18499676f0c6SRam Pai { 1850315fc83eSAl Viro struct mount *p; 1851cbbe362cSAl Viro for (p = mnt; p; p = next_mnt(p, &mnt->mnt)) { 1852315fc83eSAl Viro if (IS_MNT_UNBINDABLE(&p->mnt)) 18539676f0c6SRam Pai return 1; 18549676f0c6SRam Pai } 18559676f0c6SRam Pai return 0; 18569676f0c6SRam Pai } 18579676f0c6SRam Pai 18580a0d8a46SAl Viro static int do_move_mount(struct path *path, char *old_name) 18591da177e4SLinus Torvalds { 18602d92ab3cSAl Viro struct path old_path, parent_path; 18611da177e4SLinus Torvalds struct vfsmount *p; 18620fb54e50SAl Viro struct mount *old; 18631da177e4SLinus Torvalds int err = 0; 18641da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 18651da177e4SLinus Torvalds return -EPERM; 18661da177e4SLinus Torvalds if (!old_name || !*old_name) 18671da177e4SLinus Torvalds return -EINVAL; 18682d92ab3cSAl Viro err = kern_path(old_name, LOOKUP_FOLLOW, &old_path); 18691da177e4SLinus Torvalds if (err) 18701da177e4SLinus Torvalds return err; 18711da177e4SLinus Torvalds 1872b12cea91SAl Viro err = lock_mount(path); 1873cc53ce53SDavid Howells if (err < 0) 1874cc53ce53SDavid Howells goto out; 1875cc53ce53SDavid Howells 18761da177e4SLinus Torvalds err = -EINVAL; 18772d92ab3cSAl Viro if (!check_mnt(path->mnt) || !check_mnt(old_path.mnt)) 18781da177e4SLinus Torvalds goto out1; 18791da177e4SLinus Torvalds 1880f3da392eSAlexey Dobriyan if (d_unlinked(path->dentry)) 188121444403SRam Pai goto out1; 18821da177e4SLinus Torvalds 18831da177e4SLinus Torvalds err = -EINVAL; 18842d92ab3cSAl Viro if (old_path.dentry != old_path.mnt->mnt_root) 188521444403SRam Pai goto out1; 18861da177e4SLinus Torvalds 18870fb54e50SAl Viro old = real_mount(old_path.mnt); 18880fb54e50SAl Viro 1889b2dba1afSAl Viro if (!mnt_has_parent(old_path.mnt)) 189021444403SRam Pai goto out1; 18911da177e4SLinus Torvalds 18922d92ab3cSAl Viro if (S_ISDIR(path->dentry->d_inode->i_mode) != 18932d92ab3cSAl Viro S_ISDIR(old_path.dentry->d_inode->i_mode)) 189421444403SRam Pai goto out1; 189521444403SRam Pai /* 189621444403SRam Pai * Don't move a mount residing in a shared parent. 189721444403SRam Pai */ 1898afac7cbaSAl Viro if (IS_MNT_SHARED(old_path.mnt->mnt_parent)) 189921444403SRam Pai goto out1; 19009676f0c6SRam Pai /* 19019676f0c6SRam Pai * Don't move a mount tree containing unbindable mounts to a destination 19029676f0c6SRam Pai * mount which is shared. 19039676f0c6SRam Pai */ 19042d92ab3cSAl Viro if (IS_MNT_SHARED(path->mnt) && 1905cbbe362cSAl Viro tree_contains_unbindable(old)) 19069676f0c6SRam Pai goto out1; 19071da177e4SLinus Torvalds err = -ELOOP; 1908b2dba1afSAl Viro for (p = path->mnt; mnt_has_parent(p); p = p->mnt_parent) 19092d92ab3cSAl Viro if (p == old_path.mnt) 191021444403SRam Pai goto out1; 19111da177e4SLinus Torvalds 19120fb54e50SAl Viro err = attach_recursive_mnt(old, path, &parent_path); 19134ac91378SJan Blunck if (err) 191421444403SRam Pai goto out1; 19151da177e4SLinus Torvalds 19161da177e4SLinus Torvalds /* if the mount is moved, it should no longer be expire 19171da177e4SLinus Torvalds * automatically */ 19182d92ab3cSAl Viro list_del_init(&old_path.mnt->mnt_expire); 19191da177e4SLinus Torvalds out1: 1920b12cea91SAl Viro unlock_mount(path); 19211da177e4SLinus Torvalds out: 19221da177e4SLinus Torvalds if (!err) 19231a390689SAl Viro path_put(&parent_path); 19242d92ab3cSAl Viro path_put(&old_path); 19251da177e4SLinus Torvalds return err; 19261da177e4SLinus Torvalds } 19271da177e4SLinus Torvalds 19289d412a43SAl Viro static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype) 19299d412a43SAl Viro { 19309d412a43SAl Viro int err; 19319d412a43SAl Viro const char *subtype = strchr(fstype, '.'); 19329d412a43SAl Viro if (subtype) { 19339d412a43SAl Viro subtype++; 19349d412a43SAl Viro err = -EINVAL; 19359d412a43SAl Viro if (!subtype[0]) 19369d412a43SAl Viro goto err; 19379d412a43SAl Viro } else 19389d412a43SAl Viro subtype = ""; 19399d412a43SAl Viro 19409d412a43SAl Viro mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL); 19419d412a43SAl Viro err = -ENOMEM; 19429d412a43SAl Viro if (!mnt->mnt_sb->s_subtype) 19439d412a43SAl Viro goto err; 19449d412a43SAl Viro return mnt; 19459d412a43SAl Viro 19469d412a43SAl Viro err: 19479d412a43SAl Viro mntput(mnt); 19489d412a43SAl Viro return ERR_PTR(err); 19499d412a43SAl Viro } 19509d412a43SAl Viro 195179e801a9SAl Viro static struct vfsmount * 19529d412a43SAl Viro do_kern_mount(const char *fstype, int flags, const char *name, void *data) 19539d412a43SAl Viro { 19549d412a43SAl Viro struct file_system_type *type = get_fs_type(fstype); 19559d412a43SAl Viro struct vfsmount *mnt; 19569d412a43SAl Viro if (!type) 19579d412a43SAl Viro return ERR_PTR(-ENODEV); 19589d412a43SAl Viro mnt = vfs_kern_mount(type, flags, name, data); 19599d412a43SAl Viro if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) && 19609d412a43SAl Viro !mnt->mnt_sb->s_subtype) 19619d412a43SAl Viro mnt = fs_set_subtype(mnt, fstype); 19629d412a43SAl Viro put_filesystem(type); 19639d412a43SAl Viro return mnt; 19649d412a43SAl Viro } 19659d412a43SAl Viro 19669d412a43SAl Viro /* 19679d412a43SAl Viro * add a mount into a namespace's mount tree 19689d412a43SAl Viro */ 19699d412a43SAl Viro static int do_add_mount(struct vfsmount *newmnt, struct path *path, int mnt_flags) 19709d412a43SAl Viro { 19719d412a43SAl Viro int err; 19729d412a43SAl Viro 19739d412a43SAl Viro mnt_flags &= ~(MNT_SHARED | MNT_WRITE_HOLD | MNT_INTERNAL); 19749d412a43SAl Viro 1975b12cea91SAl Viro err = lock_mount(path); 1976b12cea91SAl Viro if (err) 1977b12cea91SAl Viro return err; 19789d412a43SAl Viro 19799d412a43SAl Viro err = -EINVAL; 19809d412a43SAl Viro if (!(mnt_flags & MNT_SHRINKABLE) && !check_mnt(path->mnt)) 19819d412a43SAl Viro goto unlock; 19829d412a43SAl Viro 19839d412a43SAl Viro /* Refuse the same filesystem on the same mount point */ 19849d412a43SAl Viro err = -EBUSY; 19859d412a43SAl Viro if (path->mnt->mnt_sb == newmnt->mnt_sb && 19869d412a43SAl Viro path->mnt->mnt_root == path->dentry) 19879d412a43SAl Viro goto unlock; 19889d412a43SAl Viro 19899d412a43SAl Viro err = -EINVAL; 19909d412a43SAl Viro if (S_ISLNK(newmnt->mnt_root->d_inode->i_mode)) 19919d412a43SAl Viro goto unlock; 19929d412a43SAl Viro 19939d412a43SAl Viro newmnt->mnt_flags = mnt_flags; 19949d412a43SAl Viro err = graft_tree(newmnt, path); 19959d412a43SAl Viro 19969d412a43SAl Viro unlock: 1997b12cea91SAl Viro unlock_mount(path); 19989d412a43SAl Viro return err; 19999d412a43SAl Viro } 2000b1e75df4SAl Viro 20011da177e4SLinus Torvalds /* 20021da177e4SLinus Torvalds * create a new mount for userspace and request it to be added into the 20031da177e4SLinus Torvalds * namespace's tree 20041da177e4SLinus Torvalds */ 20050a0d8a46SAl Viro static int do_new_mount(struct path *path, char *type, int flags, 20061da177e4SLinus Torvalds int mnt_flags, char *name, void *data) 20071da177e4SLinus Torvalds { 20081da177e4SLinus Torvalds struct vfsmount *mnt; 200915f9a3f3SAl Viro int err; 20101da177e4SLinus Torvalds 2011eca6f534SVegard Nossum if (!type) 20121da177e4SLinus Torvalds return -EINVAL; 20131da177e4SLinus Torvalds 20141da177e4SLinus Torvalds /* we need capabilities... */ 20151da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 20161da177e4SLinus Torvalds return -EPERM; 20171da177e4SLinus Torvalds 20181da177e4SLinus Torvalds mnt = do_kern_mount(type, flags, name, data); 20191da177e4SLinus Torvalds if (IS_ERR(mnt)) 20201da177e4SLinus Torvalds return PTR_ERR(mnt); 20211da177e4SLinus Torvalds 202215f9a3f3SAl Viro err = do_add_mount(mnt, path, mnt_flags); 202315f9a3f3SAl Viro if (err) 202415f9a3f3SAl Viro mntput(mnt); 202515f9a3f3SAl Viro return err; 20261da177e4SLinus Torvalds } 20271da177e4SLinus Torvalds 202819a167afSAl Viro int finish_automount(struct vfsmount *m, struct path *path) 202919a167afSAl Viro { 203019a167afSAl Viro int err; 203119a167afSAl Viro /* The new mount record should have at least 2 refs to prevent it being 203219a167afSAl Viro * expired before we get a chance to add it 203319a167afSAl Viro */ 203419a167afSAl Viro BUG_ON(mnt_get_count(m) < 2); 203519a167afSAl Viro 203619a167afSAl Viro if (m->mnt_sb == path->mnt->mnt_sb && 203719a167afSAl Viro m->mnt_root == path->dentry) { 2038b1e75df4SAl Viro err = -ELOOP; 2039b1e75df4SAl Viro goto fail; 204019a167afSAl Viro } 204119a167afSAl Viro 204219a167afSAl Viro err = do_add_mount(m, path, path->mnt->mnt_flags | MNT_SHRINKABLE); 2043b1e75df4SAl Viro if (!err) 2044b1e75df4SAl Viro return 0; 2045b1e75df4SAl Viro fail: 2046b1e75df4SAl Viro /* remove m from any expiration list it may be on */ 2047b1e75df4SAl Viro if (!list_empty(&m->mnt_expire)) { 2048b1e75df4SAl Viro down_write(&namespace_sem); 2049b1e75df4SAl Viro br_write_lock(vfsmount_lock); 2050b1e75df4SAl Viro list_del_init(&m->mnt_expire); 2051b1e75df4SAl Viro br_write_unlock(vfsmount_lock); 2052b1e75df4SAl Viro up_write(&namespace_sem); 205319a167afSAl Viro } 2054b1e75df4SAl Viro mntput(m); 2055b1e75df4SAl Viro mntput(m); 205619a167afSAl Viro return err; 205719a167afSAl Viro } 205819a167afSAl Viro 2059ea5b778aSDavid Howells /** 2060ea5b778aSDavid Howells * mnt_set_expiry - Put a mount on an expiration list 2061ea5b778aSDavid Howells * @mnt: The mount to list. 2062ea5b778aSDavid Howells * @expiry_list: The list to add the mount to. 2063ea5b778aSDavid Howells */ 2064ea5b778aSDavid Howells void mnt_set_expiry(struct vfsmount *mnt, struct list_head *expiry_list) 2065ea5b778aSDavid Howells { 2066ea5b778aSDavid Howells down_write(&namespace_sem); 2067ea5b778aSDavid Howells br_write_lock(vfsmount_lock); 2068ea5b778aSDavid Howells 2069ea5b778aSDavid Howells list_add_tail(&mnt->mnt_expire, expiry_list); 2070ea5b778aSDavid Howells 2071ea5b778aSDavid Howells br_write_unlock(vfsmount_lock); 2072ea5b778aSDavid Howells up_write(&namespace_sem); 2073ea5b778aSDavid Howells } 2074ea5b778aSDavid Howells EXPORT_SYMBOL(mnt_set_expiry); 2075ea5b778aSDavid Howells 2076ea5b778aSDavid Howells /* 20771da177e4SLinus Torvalds * process a list of expirable mountpoints with the intent of discarding any 20781da177e4SLinus Torvalds * mountpoints that aren't in use and haven't been touched since last we came 20791da177e4SLinus Torvalds * here 20801da177e4SLinus Torvalds */ 20811da177e4SLinus Torvalds void mark_mounts_for_expiry(struct list_head *mounts) 20821da177e4SLinus Torvalds { 20831da177e4SLinus Torvalds struct vfsmount *mnt, *next; 20841da177e4SLinus Torvalds LIST_HEAD(graveyard); 2085bcc5c7d2SAl Viro LIST_HEAD(umounts); 20861da177e4SLinus Torvalds 20871da177e4SLinus Torvalds if (list_empty(mounts)) 20881da177e4SLinus Torvalds return; 20891da177e4SLinus Torvalds 2090bcc5c7d2SAl Viro down_write(&namespace_sem); 209199b7db7bSNick Piggin br_write_lock(vfsmount_lock); 20921da177e4SLinus Torvalds 20931da177e4SLinus Torvalds /* extract from the expiration list every vfsmount that matches the 20941da177e4SLinus Torvalds * following criteria: 20951da177e4SLinus Torvalds * - only referenced by its parent vfsmount 20961da177e4SLinus Torvalds * - still marked for expiry (marked on the last call here; marks are 20971da177e4SLinus Torvalds * cleared by mntput()) 20981da177e4SLinus Torvalds */ 209955e700b9SMiklos Szeredi list_for_each_entry_safe(mnt, next, mounts, mnt_expire) { 21001da177e4SLinus Torvalds if (!xchg(&mnt->mnt_expiry_mark, 1) || 2101bcc5c7d2SAl Viro propagate_mount_busy(mnt, 1)) 21021da177e4SLinus Torvalds continue; 210355e700b9SMiklos Szeredi list_move(&mnt->mnt_expire, &graveyard); 21041da177e4SLinus Torvalds } 2105bcc5c7d2SAl Viro while (!list_empty(&graveyard)) { 2106bcc5c7d2SAl Viro mnt = list_first_entry(&graveyard, struct vfsmount, mnt_expire); 2107bcc5c7d2SAl Viro touch_mnt_namespace(mnt->mnt_ns); 2108bcc5c7d2SAl Viro umount_tree(mnt, 1, &umounts); 2109bcc5c7d2SAl Viro } 211099b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 2111bcc5c7d2SAl Viro up_write(&namespace_sem); 2112bcc5c7d2SAl Viro 2113bcc5c7d2SAl Viro release_mounts(&umounts); 21141da177e4SLinus Torvalds } 21151da177e4SLinus Torvalds 21161da177e4SLinus Torvalds EXPORT_SYMBOL_GPL(mark_mounts_for_expiry); 21171da177e4SLinus Torvalds 21181da177e4SLinus Torvalds /* 21195528f911STrond Myklebust * Ripoff of 'select_parent()' 21205528f911STrond Myklebust * 21215528f911STrond Myklebust * search the list of submounts for a given mountpoint, and move any 21225528f911STrond Myklebust * shrinkable submounts to the 'graveyard' list. 21235528f911STrond Myklebust */ 21245528f911STrond Myklebust static int select_submounts(struct vfsmount *parent, struct list_head *graveyard) 21255528f911STrond Myklebust { 21265528f911STrond Myklebust struct vfsmount *this_parent = parent; 21275528f911STrond Myklebust struct list_head *next; 21285528f911STrond Myklebust int found = 0; 21295528f911STrond Myklebust 21305528f911STrond Myklebust repeat: 21315528f911STrond Myklebust next = this_parent->mnt_mounts.next; 21325528f911STrond Myklebust resume: 21335528f911STrond Myklebust while (next != &this_parent->mnt_mounts) { 21345528f911STrond Myklebust struct list_head *tmp = next; 21355528f911STrond Myklebust struct vfsmount *mnt = list_entry(tmp, struct vfsmount, mnt_child); 21365528f911STrond Myklebust 21375528f911STrond Myklebust next = tmp->next; 21385528f911STrond Myklebust if (!(mnt->mnt_flags & MNT_SHRINKABLE)) 21395528f911STrond Myklebust continue; 21405528f911STrond Myklebust /* 21415528f911STrond Myklebust * Descend a level if the d_mounts list is non-empty. 21425528f911STrond Myklebust */ 21435528f911STrond Myklebust if (!list_empty(&mnt->mnt_mounts)) { 21445528f911STrond Myklebust this_parent = mnt; 21455528f911STrond Myklebust goto repeat; 21465528f911STrond Myklebust } 21475528f911STrond Myklebust 21485528f911STrond Myklebust if (!propagate_mount_busy(mnt, 1)) { 21495528f911STrond Myklebust list_move_tail(&mnt->mnt_expire, graveyard); 21505528f911STrond Myklebust found++; 21515528f911STrond Myklebust } 21525528f911STrond Myklebust } 21535528f911STrond Myklebust /* 21545528f911STrond Myklebust * All done at this level ... ascend and resume the search 21555528f911STrond Myklebust */ 21565528f911STrond Myklebust if (this_parent != parent) { 21575528f911STrond Myklebust next = this_parent->mnt_child.next; 21585528f911STrond Myklebust this_parent = this_parent->mnt_parent; 21595528f911STrond Myklebust goto resume; 21605528f911STrond Myklebust } 21615528f911STrond Myklebust return found; 21625528f911STrond Myklebust } 21635528f911STrond Myklebust 21645528f911STrond Myklebust /* 21655528f911STrond Myklebust * process a list of expirable mountpoints with the intent of discarding any 21665528f911STrond Myklebust * submounts of a specific parent mountpoint 216799b7db7bSNick Piggin * 216899b7db7bSNick Piggin * vfsmount_lock must be held for write 21695528f911STrond Myklebust */ 2170c35038beSAl Viro static void shrink_submounts(struct vfsmount *mnt, struct list_head *umounts) 21715528f911STrond Myklebust { 21725528f911STrond Myklebust LIST_HEAD(graveyard); 2173c35038beSAl Viro struct vfsmount *m; 21745528f911STrond Myklebust 21755528f911STrond Myklebust /* extract submounts of 'mountpoint' from the expiration list */ 2176c35038beSAl Viro while (select_submounts(mnt, &graveyard)) { 2177bcc5c7d2SAl Viro while (!list_empty(&graveyard)) { 2178c35038beSAl Viro m = list_first_entry(&graveyard, struct vfsmount, 2179bcc5c7d2SAl Viro mnt_expire); 2180afef80b3SEric W. Biederman touch_mnt_namespace(m->mnt_ns); 2181afef80b3SEric W. Biederman umount_tree(m, 1, umounts); 2182bcc5c7d2SAl Viro } 2183bcc5c7d2SAl Viro } 21845528f911STrond Myklebust } 21855528f911STrond Myklebust 21865528f911STrond Myklebust /* 21871da177e4SLinus Torvalds * Some copy_from_user() implementations do not return the exact number of 21881da177e4SLinus Torvalds * bytes remaining to copy on a fault. But copy_mount_options() requires that. 21891da177e4SLinus Torvalds * Note that this function differs from copy_from_user() in that it will oops 21901da177e4SLinus Torvalds * on bad values of `to', rather than returning a short copy. 21911da177e4SLinus Torvalds */ 2192b58fed8bSRam Pai static long exact_copy_from_user(void *to, const void __user * from, 2193b58fed8bSRam Pai unsigned long n) 21941da177e4SLinus Torvalds { 21951da177e4SLinus Torvalds char *t = to; 21961da177e4SLinus Torvalds const char __user *f = from; 21971da177e4SLinus Torvalds char c; 21981da177e4SLinus Torvalds 21991da177e4SLinus Torvalds if (!access_ok(VERIFY_READ, from, n)) 22001da177e4SLinus Torvalds return n; 22011da177e4SLinus Torvalds 22021da177e4SLinus Torvalds while (n) { 22031da177e4SLinus Torvalds if (__get_user(c, f)) { 22041da177e4SLinus Torvalds memset(t, 0, n); 22051da177e4SLinus Torvalds break; 22061da177e4SLinus Torvalds } 22071da177e4SLinus Torvalds *t++ = c; 22081da177e4SLinus Torvalds f++; 22091da177e4SLinus Torvalds n--; 22101da177e4SLinus Torvalds } 22111da177e4SLinus Torvalds return n; 22121da177e4SLinus Torvalds } 22131da177e4SLinus Torvalds 22141da177e4SLinus Torvalds int copy_mount_options(const void __user * data, unsigned long *where) 22151da177e4SLinus Torvalds { 22161da177e4SLinus Torvalds int i; 22171da177e4SLinus Torvalds unsigned long page; 22181da177e4SLinus Torvalds unsigned long size; 22191da177e4SLinus Torvalds 22201da177e4SLinus Torvalds *where = 0; 22211da177e4SLinus Torvalds if (!data) 22221da177e4SLinus Torvalds return 0; 22231da177e4SLinus Torvalds 22241da177e4SLinus Torvalds if (!(page = __get_free_page(GFP_KERNEL))) 22251da177e4SLinus Torvalds return -ENOMEM; 22261da177e4SLinus Torvalds 22271da177e4SLinus Torvalds /* We only care that *some* data at the address the user 22281da177e4SLinus Torvalds * gave us is valid. Just in case, we'll zero 22291da177e4SLinus Torvalds * the remainder of the page. 22301da177e4SLinus Torvalds */ 22311da177e4SLinus Torvalds /* copy_from_user cannot cross TASK_SIZE ! */ 22321da177e4SLinus Torvalds size = TASK_SIZE - (unsigned long)data; 22331da177e4SLinus Torvalds if (size > PAGE_SIZE) 22341da177e4SLinus Torvalds size = PAGE_SIZE; 22351da177e4SLinus Torvalds 22361da177e4SLinus Torvalds i = size - exact_copy_from_user((void *)page, data, size); 22371da177e4SLinus Torvalds if (!i) { 22381da177e4SLinus Torvalds free_page(page); 22391da177e4SLinus Torvalds return -EFAULT; 22401da177e4SLinus Torvalds } 22411da177e4SLinus Torvalds if (i != PAGE_SIZE) 22421da177e4SLinus Torvalds memset((char *)page + i, 0, PAGE_SIZE - i); 22431da177e4SLinus Torvalds *where = page; 22441da177e4SLinus Torvalds return 0; 22451da177e4SLinus Torvalds } 22461da177e4SLinus Torvalds 2247eca6f534SVegard Nossum int copy_mount_string(const void __user *data, char **where) 2248eca6f534SVegard Nossum { 2249eca6f534SVegard Nossum char *tmp; 2250eca6f534SVegard Nossum 2251eca6f534SVegard Nossum if (!data) { 2252eca6f534SVegard Nossum *where = NULL; 2253eca6f534SVegard Nossum return 0; 2254eca6f534SVegard Nossum } 2255eca6f534SVegard Nossum 2256eca6f534SVegard Nossum tmp = strndup_user(data, PAGE_SIZE); 2257eca6f534SVegard Nossum if (IS_ERR(tmp)) 2258eca6f534SVegard Nossum return PTR_ERR(tmp); 2259eca6f534SVegard Nossum 2260eca6f534SVegard Nossum *where = tmp; 2261eca6f534SVegard Nossum return 0; 2262eca6f534SVegard Nossum } 2263eca6f534SVegard Nossum 22641da177e4SLinus Torvalds /* 22651da177e4SLinus Torvalds * Flags is a 32-bit value that allows up to 31 non-fs dependent flags to 22661da177e4SLinus Torvalds * be given to the mount() call (ie: read-only, no-dev, no-suid etc). 22671da177e4SLinus Torvalds * 22681da177e4SLinus Torvalds * data is a (void *) that can point to any structure up to 22691da177e4SLinus Torvalds * PAGE_SIZE-1 bytes, which can contain arbitrary fs-dependent 22701da177e4SLinus Torvalds * information (or be NULL). 22711da177e4SLinus Torvalds * 22721da177e4SLinus Torvalds * Pre-0.97 versions of mount() didn't have a flags word. 22731da177e4SLinus Torvalds * When the flags word was introduced its top half was required 22741da177e4SLinus Torvalds * to have the magic value 0xC0ED, and this remained so until 2.4.0-test9. 22751da177e4SLinus Torvalds * Therefore, if this magic number is present, it carries no information 22761da177e4SLinus Torvalds * and must be discarded. 22771da177e4SLinus Torvalds */ 22781da177e4SLinus Torvalds long do_mount(char *dev_name, char *dir_name, char *type_page, 22791da177e4SLinus Torvalds unsigned long flags, void *data_page) 22801da177e4SLinus Torvalds { 22812d92ab3cSAl Viro struct path path; 22821da177e4SLinus Torvalds int retval = 0; 22831da177e4SLinus Torvalds int mnt_flags = 0; 22841da177e4SLinus Torvalds 22851da177e4SLinus Torvalds /* Discard magic */ 22861da177e4SLinus Torvalds if ((flags & MS_MGC_MSK) == MS_MGC_VAL) 22871da177e4SLinus Torvalds flags &= ~MS_MGC_MSK; 22881da177e4SLinus Torvalds 22891da177e4SLinus Torvalds /* Basic sanity checks */ 22901da177e4SLinus Torvalds 22911da177e4SLinus Torvalds if (!dir_name || !*dir_name || !memchr(dir_name, 0, PAGE_SIZE)) 22921da177e4SLinus Torvalds return -EINVAL; 22931da177e4SLinus Torvalds 22941da177e4SLinus Torvalds if (data_page) 22951da177e4SLinus Torvalds ((char *)data_page)[PAGE_SIZE - 1] = 0; 22961da177e4SLinus Torvalds 2297a27ab9f2STetsuo Handa /* ... and get the mountpoint */ 2298a27ab9f2STetsuo Handa retval = kern_path(dir_name, LOOKUP_FOLLOW, &path); 2299a27ab9f2STetsuo Handa if (retval) 2300a27ab9f2STetsuo Handa return retval; 2301a27ab9f2STetsuo Handa 2302a27ab9f2STetsuo Handa retval = security_sb_mount(dev_name, &path, 2303a27ab9f2STetsuo Handa type_page, flags, data_page); 2304a27ab9f2STetsuo Handa if (retval) 2305a27ab9f2STetsuo Handa goto dput_out; 2306a27ab9f2STetsuo Handa 2307613cbe3dSAndi Kleen /* Default to relatime unless overriden */ 2308613cbe3dSAndi Kleen if (!(flags & MS_NOATIME)) 23090a1c01c9SMatthew Garrett mnt_flags |= MNT_RELATIME; 23100a1c01c9SMatthew Garrett 23111da177e4SLinus Torvalds /* Separate the per-mountpoint flags */ 23121da177e4SLinus Torvalds if (flags & MS_NOSUID) 23131da177e4SLinus Torvalds mnt_flags |= MNT_NOSUID; 23141da177e4SLinus Torvalds if (flags & MS_NODEV) 23151da177e4SLinus Torvalds mnt_flags |= MNT_NODEV; 23161da177e4SLinus Torvalds if (flags & MS_NOEXEC) 23171da177e4SLinus Torvalds mnt_flags |= MNT_NOEXEC; 2318fc33a7bbSChristoph Hellwig if (flags & MS_NOATIME) 2319fc33a7bbSChristoph Hellwig mnt_flags |= MNT_NOATIME; 2320fc33a7bbSChristoph Hellwig if (flags & MS_NODIRATIME) 2321fc33a7bbSChristoph Hellwig mnt_flags |= MNT_NODIRATIME; 2322d0adde57SMatthew Garrett if (flags & MS_STRICTATIME) 2323d0adde57SMatthew Garrett mnt_flags &= ~(MNT_RELATIME | MNT_NOATIME); 23242e4b7fcdSDave Hansen if (flags & MS_RDONLY) 23252e4b7fcdSDave Hansen mnt_flags |= MNT_READONLY; 2326fc33a7bbSChristoph Hellwig 23277a4dec53SAl Viro flags &= ~(MS_NOSUID | MS_NOEXEC | MS_NODEV | MS_ACTIVE | MS_BORN | 2328d0adde57SMatthew Garrett MS_NOATIME | MS_NODIRATIME | MS_RELATIME| MS_KERNMOUNT | 2329d0adde57SMatthew Garrett MS_STRICTATIME); 23301da177e4SLinus Torvalds 23311da177e4SLinus Torvalds if (flags & MS_REMOUNT) 23322d92ab3cSAl Viro retval = do_remount(&path, flags & ~MS_REMOUNT, mnt_flags, 23331da177e4SLinus Torvalds data_page); 23341da177e4SLinus Torvalds else if (flags & MS_BIND) 23352d92ab3cSAl Viro retval = do_loopback(&path, dev_name, flags & MS_REC); 23369676f0c6SRam Pai else if (flags & (MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE)) 23372d92ab3cSAl Viro retval = do_change_type(&path, flags); 23381da177e4SLinus Torvalds else if (flags & MS_MOVE) 23392d92ab3cSAl Viro retval = do_move_mount(&path, dev_name); 23401da177e4SLinus Torvalds else 23412d92ab3cSAl Viro retval = do_new_mount(&path, type_page, flags, mnt_flags, 23421da177e4SLinus Torvalds dev_name, data_page); 23431da177e4SLinus Torvalds dput_out: 23442d92ab3cSAl Viro path_put(&path); 23451da177e4SLinus Torvalds return retval; 23461da177e4SLinus Torvalds } 23471da177e4SLinus Torvalds 2348cf8d2c11STrond Myklebust static struct mnt_namespace *alloc_mnt_ns(void) 2349cf8d2c11STrond Myklebust { 2350cf8d2c11STrond Myklebust struct mnt_namespace *new_ns; 2351cf8d2c11STrond Myklebust 2352cf8d2c11STrond Myklebust new_ns = kmalloc(sizeof(struct mnt_namespace), GFP_KERNEL); 2353cf8d2c11STrond Myklebust if (!new_ns) 2354cf8d2c11STrond Myklebust return ERR_PTR(-ENOMEM); 2355cf8d2c11STrond Myklebust atomic_set(&new_ns->count, 1); 2356cf8d2c11STrond Myklebust new_ns->root = NULL; 2357cf8d2c11STrond Myklebust INIT_LIST_HEAD(&new_ns->list); 2358cf8d2c11STrond Myklebust init_waitqueue_head(&new_ns->poll); 2359cf8d2c11STrond Myklebust new_ns->event = 0; 2360cf8d2c11STrond Myklebust return new_ns; 2361cf8d2c11STrond Myklebust } 2362cf8d2c11STrond Myklebust 2363f03c6599SAl Viro void mnt_make_longterm(struct vfsmount *mnt) 2364f03c6599SAl Viro { 23657e3d0eb0SAl Viro __mnt_make_longterm(mnt); 2366f03c6599SAl Viro } 2367f03c6599SAl Viro 2368f03c6599SAl Viro void mnt_make_shortterm(struct vfsmount *mnt) 2369f03c6599SAl Viro { 23707e3d0eb0SAl Viro #ifdef CONFIG_SMP 2371f03c6599SAl Viro if (atomic_add_unless(&mnt->mnt_longterm, -1, 1)) 2372f03c6599SAl Viro return; 2373f03c6599SAl Viro br_write_lock(vfsmount_lock); 2374f03c6599SAl Viro atomic_dec(&mnt->mnt_longterm); 2375f03c6599SAl Viro br_write_unlock(vfsmount_lock); 23767e3d0eb0SAl Viro #endif 2377f03c6599SAl Viro } 2378f03c6599SAl Viro 2379741a2951SJANAK DESAI /* 2380741a2951SJANAK DESAI * Allocate a new namespace structure and populate it with contents 2381741a2951SJANAK DESAI * copied from the namespace of the passed in task structure. 2382741a2951SJANAK DESAI */ 2383e3222c4eSBadari Pulavarty static struct mnt_namespace *dup_mnt_ns(struct mnt_namespace *mnt_ns, 23846b3286edSKirill Korotaev struct fs_struct *fs) 23851da177e4SLinus Torvalds { 23866b3286edSKirill Korotaev struct mnt_namespace *new_ns; 23877f2da1e7SAl Viro struct vfsmount *rootmnt = NULL, *pwdmnt = NULL; 2388315fc83eSAl Viro struct mount *p, *q; 2389cb338d06SAl Viro struct mount *new; 23901da177e4SLinus Torvalds 2391cf8d2c11STrond Myklebust new_ns = alloc_mnt_ns(); 2392cf8d2c11STrond Myklebust if (IS_ERR(new_ns)) 2393cf8d2c11STrond Myklebust return new_ns; 23941da177e4SLinus Torvalds 2395390c6843SRam Pai down_write(&namespace_sem); 23961da177e4SLinus Torvalds /* First pass: copy the tree topology */ 2397cb338d06SAl Viro new = copy_tree(mnt_ns->root, mnt_ns->root->mnt_root, 23989676f0c6SRam Pai CL_COPY_ALL | CL_EXPIRE); 2399cb338d06SAl Viro if (!new) { 2400390c6843SRam Pai up_write(&namespace_sem); 24011da177e4SLinus Torvalds kfree(new_ns); 24025cc4a034SJulia Lawall return ERR_PTR(-ENOMEM); 24031da177e4SLinus Torvalds } 2404cb338d06SAl Viro new_ns->root = &new->mnt; 240599b7db7bSNick Piggin br_write_lock(vfsmount_lock); 24061da177e4SLinus Torvalds list_add_tail(&new_ns->list, &new_ns->root->mnt_list); 240799b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 24081da177e4SLinus Torvalds 24091da177e4SLinus Torvalds /* 24101da177e4SLinus Torvalds * Second pass: switch the tsk->fs->* elements and mark new vfsmounts 24111da177e4SLinus Torvalds * as belonging to new namespace. We have already acquired a private 24121da177e4SLinus Torvalds * fs_struct, so tsk->fs->lock is not needed. 24131da177e4SLinus Torvalds */ 2414315fc83eSAl Viro p = real_mount(mnt_ns->root); 2415cb338d06SAl Viro q = new; 24161da177e4SLinus Torvalds while (p) { 2417315fc83eSAl Viro q->mnt.mnt_ns = new_ns; 2418315fc83eSAl Viro __mnt_make_longterm(&q->mnt); 24191da177e4SLinus Torvalds if (fs) { 2420315fc83eSAl Viro if (&p->mnt == fs->root.mnt) { 2421315fc83eSAl Viro fs->root.mnt = mntget(&q->mnt); 2422315fc83eSAl Viro __mnt_make_longterm(&q->mnt); 2423315fc83eSAl Viro mnt_make_shortterm(&p->mnt); 2424315fc83eSAl Viro rootmnt = &p->mnt; 24251da177e4SLinus Torvalds } 2426315fc83eSAl Viro if (&p->mnt == fs->pwd.mnt) { 2427315fc83eSAl Viro fs->pwd.mnt = mntget(&q->mnt); 2428315fc83eSAl Viro __mnt_make_longterm(&q->mnt); 2429315fc83eSAl Viro mnt_make_shortterm(&p->mnt); 2430315fc83eSAl Viro pwdmnt = &p->mnt; 24311da177e4SLinus Torvalds } 24321da177e4SLinus Torvalds } 24336b3286edSKirill Korotaev p = next_mnt(p, mnt_ns->root); 24341da177e4SLinus Torvalds q = next_mnt(q, new_ns->root); 24351da177e4SLinus Torvalds } 2436390c6843SRam Pai up_write(&namespace_sem); 24371da177e4SLinus Torvalds 24381da177e4SLinus Torvalds if (rootmnt) 2439f03c6599SAl Viro mntput(rootmnt); 24401da177e4SLinus Torvalds if (pwdmnt) 2441f03c6599SAl Viro mntput(pwdmnt); 24421da177e4SLinus Torvalds 2443741a2951SJANAK DESAI return new_ns; 2444741a2951SJANAK DESAI } 2445741a2951SJANAK DESAI 2446213dd266SEric W. Biederman struct mnt_namespace *copy_mnt_ns(unsigned long flags, struct mnt_namespace *ns, 2447e3222c4eSBadari Pulavarty struct fs_struct *new_fs) 2448741a2951SJANAK DESAI { 24496b3286edSKirill Korotaev struct mnt_namespace *new_ns; 2450741a2951SJANAK DESAI 2451e3222c4eSBadari Pulavarty BUG_ON(!ns); 24526b3286edSKirill Korotaev get_mnt_ns(ns); 2453741a2951SJANAK DESAI 2454741a2951SJANAK DESAI if (!(flags & CLONE_NEWNS)) 2455e3222c4eSBadari Pulavarty return ns; 2456741a2951SJANAK DESAI 2457e3222c4eSBadari Pulavarty new_ns = dup_mnt_ns(ns, new_fs); 2458741a2951SJANAK DESAI 24596b3286edSKirill Korotaev put_mnt_ns(ns); 2460e3222c4eSBadari Pulavarty return new_ns; 24611da177e4SLinus Torvalds } 24621da177e4SLinus Torvalds 2463cf8d2c11STrond Myklebust /** 2464cf8d2c11STrond Myklebust * create_mnt_ns - creates a private namespace and adds a root filesystem 2465cf8d2c11STrond Myklebust * @mnt: pointer to the new root filesystem mountpoint 2466cf8d2c11STrond Myklebust */ 24676c449c8dSAl Viro static struct mnt_namespace *create_mnt_ns(struct vfsmount *mnt) 2468cf8d2c11STrond Myklebust { 2469cf8d2c11STrond Myklebust struct mnt_namespace *new_ns; 2470cf8d2c11STrond Myklebust 2471cf8d2c11STrond Myklebust new_ns = alloc_mnt_ns(); 2472cf8d2c11STrond Myklebust if (!IS_ERR(new_ns)) { 2473cf8d2c11STrond Myklebust mnt->mnt_ns = new_ns; 24747e3d0eb0SAl Viro __mnt_make_longterm(mnt); 2475cf8d2c11STrond Myklebust new_ns->root = mnt; 2476cf8d2c11STrond Myklebust list_add(&new_ns->list, &new_ns->root->mnt_list); 2477c1334495SAl Viro } else { 2478c1334495SAl Viro mntput(mnt); 2479cf8d2c11STrond Myklebust } 2480cf8d2c11STrond Myklebust return new_ns; 2481cf8d2c11STrond Myklebust } 2482cf8d2c11STrond Myklebust 2483ea441d11SAl Viro struct dentry *mount_subtree(struct vfsmount *mnt, const char *name) 2484ea441d11SAl Viro { 2485ea441d11SAl Viro struct mnt_namespace *ns; 2486d31da0f0SAl Viro struct super_block *s; 2487ea441d11SAl Viro struct path path; 2488ea441d11SAl Viro int err; 2489ea441d11SAl Viro 2490ea441d11SAl Viro ns = create_mnt_ns(mnt); 2491ea441d11SAl Viro if (IS_ERR(ns)) 2492ea441d11SAl Viro return ERR_CAST(ns); 2493ea441d11SAl Viro 2494ea441d11SAl Viro err = vfs_path_lookup(mnt->mnt_root, mnt, 2495ea441d11SAl Viro name, LOOKUP_FOLLOW|LOOKUP_AUTOMOUNT, &path); 2496ea441d11SAl Viro 2497ea441d11SAl Viro put_mnt_ns(ns); 2498ea441d11SAl Viro 2499ea441d11SAl Viro if (err) 2500ea441d11SAl Viro return ERR_PTR(err); 2501ea441d11SAl Viro 2502ea441d11SAl Viro /* trade a vfsmount reference for active sb one */ 2503d31da0f0SAl Viro s = path.mnt->mnt_sb; 2504d31da0f0SAl Viro atomic_inc(&s->s_active); 2505ea441d11SAl Viro mntput(path.mnt); 2506ea441d11SAl Viro /* lock the sucker */ 2507d31da0f0SAl Viro down_write(&s->s_umount); 2508ea441d11SAl Viro /* ... and return the root of (sub)tree on it */ 2509ea441d11SAl Viro return path.dentry; 2510ea441d11SAl Viro } 2511ea441d11SAl Viro EXPORT_SYMBOL(mount_subtree); 2512ea441d11SAl Viro 2513bdc480e3SHeiko Carstens SYSCALL_DEFINE5(mount, char __user *, dev_name, char __user *, dir_name, 2514bdc480e3SHeiko Carstens char __user *, type, unsigned long, flags, void __user *, data) 25151da177e4SLinus Torvalds { 2516eca6f534SVegard Nossum int ret; 2517eca6f534SVegard Nossum char *kernel_type; 2518eca6f534SVegard Nossum char *kernel_dir; 2519eca6f534SVegard Nossum char *kernel_dev; 25201da177e4SLinus Torvalds unsigned long data_page; 25211da177e4SLinus Torvalds 2522eca6f534SVegard Nossum ret = copy_mount_string(type, &kernel_type); 2523eca6f534SVegard Nossum if (ret < 0) 2524eca6f534SVegard Nossum goto out_type; 25251da177e4SLinus Torvalds 2526eca6f534SVegard Nossum kernel_dir = getname(dir_name); 2527eca6f534SVegard Nossum if (IS_ERR(kernel_dir)) { 2528eca6f534SVegard Nossum ret = PTR_ERR(kernel_dir); 2529eca6f534SVegard Nossum goto out_dir; 2530eca6f534SVegard Nossum } 25311da177e4SLinus Torvalds 2532eca6f534SVegard Nossum ret = copy_mount_string(dev_name, &kernel_dev); 2533eca6f534SVegard Nossum if (ret < 0) 2534eca6f534SVegard Nossum goto out_dev; 25351da177e4SLinus Torvalds 2536eca6f534SVegard Nossum ret = copy_mount_options(data, &data_page); 2537eca6f534SVegard Nossum if (ret < 0) 2538eca6f534SVegard Nossum goto out_data; 25391da177e4SLinus Torvalds 2540eca6f534SVegard Nossum ret = do_mount(kernel_dev, kernel_dir, kernel_type, flags, 2541eca6f534SVegard Nossum (void *) data_page); 2542eca6f534SVegard Nossum 25431da177e4SLinus Torvalds free_page(data_page); 2544eca6f534SVegard Nossum out_data: 2545eca6f534SVegard Nossum kfree(kernel_dev); 2546eca6f534SVegard Nossum out_dev: 2547eca6f534SVegard Nossum putname(kernel_dir); 2548eca6f534SVegard Nossum out_dir: 2549eca6f534SVegard Nossum kfree(kernel_type); 2550eca6f534SVegard Nossum out_type: 2551eca6f534SVegard Nossum return ret; 25521da177e4SLinus Torvalds } 25531da177e4SLinus Torvalds 25541da177e4SLinus Torvalds /* 2555afac7cbaSAl Viro * Return true if path is reachable from root 2556afac7cbaSAl Viro * 2557afac7cbaSAl Viro * namespace_sem or vfsmount_lock is held 2558afac7cbaSAl Viro */ 2559afac7cbaSAl Viro bool is_path_reachable(struct vfsmount *mnt, struct dentry *dentry, 2560afac7cbaSAl Viro const struct path *root) 2561afac7cbaSAl Viro { 2562afac7cbaSAl Viro while (mnt != root->mnt && mnt_has_parent(mnt)) { 2563afac7cbaSAl Viro dentry = mnt->mnt_mountpoint; 2564afac7cbaSAl Viro mnt = mnt->mnt_parent; 2565afac7cbaSAl Viro } 2566afac7cbaSAl Viro return mnt == root->mnt && is_subdir(dentry, root->dentry); 2567afac7cbaSAl Viro } 2568afac7cbaSAl Viro 2569afac7cbaSAl Viro int path_is_under(struct path *path1, struct path *path2) 2570afac7cbaSAl Viro { 2571afac7cbaSAl Viro int res; 2572afac7cbaSAl Viro br_read_lock(vfsmount_lock); 2573afac7cbaSAl Viro res = is_path_reachable(path1->mnt, path1->dentry, path2); 2574afac7cbaSAl Viro br_read_unlock(vfsmount_lock); 2575afac7cbaSAl Viro return res; 2576afac7cbaSAl Viro } 2577afac7cbaSAl Viro EXPORT_SYMBOL(path_is_under); 2578afac7cbaSAl Viro 2579afac7cbaSAl Viro /* 25801da177e4SLinus Torvalds * pivot_root Semantics: 25811da177e4SLinus Torvalds * Moves the root file system of the current process to the directory put_old, 25821da177e4SLinus Torvalds * makes new_root as the new root file system of the current process, and sets 25831da177e4SLinus Torvalds * root/cwd of all processes which had them on the current root to new_root. 25841da177e4SLinus Torvalds * 25851da177e4SLinus Torvalds * Restrictions: 25861da177e4SLinus Torvalds * The new_root and put_old must be directories, and must not be on the 25871da177e4SLinus Torvalds * same file system as the current process root. The put_old must be 25881da177e4SLinus Torvalds * underneath new_root, i.e. adding a non-zero number of /.. to the string 25891da177e4SLinus Torvalds * pointed to by put_old must yield the same directory as new_root. No other 25901da177e4SLinus Torvalds * file system may be mounted on put_old. After all, new_root is a mountpoint. 25911da177e4SLinus Torvalds * 25924a0d11faSNeil Brown * Also, the current root cannot be on the 'rootfs' (initial ramfs) filesystem. 25934a0d11faSNeil Brown * See Documentation/filesystems/ramfs-rootfs-initramfs.txt for alternatives 25944a0d11faSNeil Brown * in this situation. 25954a0d11faSNeil Brown * 25961da177e4SLinus Torvalds * Notes: 25971da177e4SLinus Torvalds * - we don't move root/cwd if they are not at the root (reason: if something 25981da177e4SLinus Torvalds * cared enough to change them, it's probably wrong to force them elsewhere) 25991da177e4SLinus Torvalds * - it's okay to pick a root that isn't the root of a file system, e.g. 26001da177e4SLinus Torvalds * /nfs/my_root where /nfs is the mount point. It must be a mountpoint, 26011da177e4SLinus Torvalds * though, so you may need to say mount --bind /nfs/my_root /nfs/my_root 26021da177e4SLinus Torvalds * first. 26031da177e4SLinus Torvalds */ 26043480b257SHeiko Carstens SYSCALL_DEFINE2(pivot_root, const char __user *, new_root, 26053480b257SHeiko Carstens const char __user *, put_old) 26061da177e4SLinus Torvalds { 26072d8f3038SAl Viro struct path new, old, parent_path, root_parent, root; 2608419148daSAl Viro struct mount *new_mnt, *root_mnt; 26091da177e4SLinus Torvalds int error; 26101da177e4SLinus Torvalds 26111da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 26121da177e4SLinus Torvalds return -EPERM; 26131da177e4SLinus Torvalds 26142d8f3038SAl Viro error = user_path_dir(new_root, &new); 26151da177e4SLinus Torvalds if (error) 26161da177e4SLinus Torvalds goto out0; 26171da177e4SLinus Torvalds 26182d8f3038SAl Viro error = user_path_dir(put_old, &old); 26191da177e4SLinus Torvalds if (error) 26201da177e4SLinus Torvalds goto out1; 26211da177e4SLinus Torvalds 26222d8f3038SAl Viro error = security_sb_pivotroot(&old, &new); 2623b12cea91SAl Viro if (error) 2624b12cea91SAl Viro goto out2; 26251da177e4SLinus Torvalds 2626f7ad3c6bSMiklos Szeredi get_fs_root(current->fs, &root); 2627b12cea91SAl Viro error = lock_mount(&old); 2628b12cea91SAl Viro if (error) 2629b12cea91SAl Viro goto out3; 2630b12cea91SAl Viro 26311da177e4SLinus Torvalds error = -EINVAL; 2632419148daSAl Viro new_mnt = real_mount(new.mnt); 2633419148daSAl Viro root_mnt = real_mount(root.mnt); 26342d8f3038SAl Viro if (IS_MNT_SHARED(old.mnt) || 26352d8f3038SAl Viro IS_MNT_SHARED(new.mnt->mnt_parent) || 26368c3ee42eSAl Viro IS_MNT_SHARED(root.mnt->mnt_parent)) 2637b12cea91SAl Viro goto out4; 263827cb1572SAl Viro if (!check_mnt(root.mnt) || !check_mnt(new.mnt)) 2639b12cea91SAl Viro goto out4; 26401da177e4SLinus Torvalds error = -ENOENT; 2641f3da392eSAlexey Dobriyan if (d_unlinked(new.dentry)) 2642b12cea91SAl Viro goto out4; 2643f3da392eSAlexey Dobriyan if (d_unlinked(old.dentry)) 2644b12cea91SAl Viro goto out4; 26451da177e4SLinus Torvalds error = -EBUSY; 26462d8f3038SAl Viro if (new.mnt == root.mnt || 26472d8f3038SAl Viro old.mnt == root.mnt) 2648b12cea91SAl Viro goto out4; /* loop, on the same file system */ 26491da177e4SLinus Torvalds error = -EINVAL; 26508c3ee42eSAl Viro if (root.mnt->mnt_root != root.dentry) 2651b12cea91SAl Viro goto out4; /* not a mountpoint */ 2652b2dba1afSAl Viro if (!mnt_has_parent(root.mnt)) 2653b12cea91SAl Viro goto out4; /* not attached */ 26542d8f3038SAl Viro if (new.mnt->mnt_root != new.dentry) 2655b12cea91SAl Viro goto out4; /* not a mountpoint */ 2656b2dba1afSAl Viro if (!mnt_has_parent(new.mnt)) 2657b12cea91SAl Viro goto out4; /* not attached */ 26584ac91378SJan Blunck /* make sure we can reach put_old from new_root */ 2659afac7cbaSAl Viro if (!is_path_reachable(old.mnt, old.dentry, &new)) 2660b12cea91SAl Viro goto out4; 266127cb1572SAl Viro br_write_lock(vfsmount_lock); 2662419148daSAl Viro detach_mnt(new_mnt, &parent_path); 2663419148daSAl Viro detach_mnt(root_mnt, &root_parent); 26644ac91378SJan Blunck /* mount old root on put_old */ 2665419148daSAl Viro attach_mnt(root_mnt, &old); 26664ac91378SJan Blunck /* mount new_root on / */ 2667419148daSAl Viro attach_mnt(new_mnt, &root_parent); 26686b3286edSKirill Korotaev touch_mnt_namespace(current->nsproxy->mnt_ns); 266999b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 26702d8f3038SAl Viro chroot_fs_refs(&root, &new); 26711da177e4SLinus Torvalds error = 0; 2672b12cea91SAl Viro out4: 2673b12cea91SAl Viro unlock_mount(&old); 2674b12cea91SAl Viro if (!error) { 26751a390689SAl Viro path_put(&root_parent); 26761a390689SAl Viro path_put(&parent_path); 2677b12cea91SAl Viro } 2678b12cea91SAl Viro out3: 26798c3ee42eSAl Viro path_put(&root); 2680b12cea91SAl Viro out2: 26812d8f3038SAl Viro path_put(&old); 26821da177e4SLinus Torvalds out1: 26832d8f3038SAl Viro path_put(&new); 26841da177e4SLinus Torvalds out0: 26851da177e4SLinus Torvalds return error; 26861da177e4SLinus Torvalds } 26871da177e4SLinus Torvalds 26881da177e4SLinus Torvalds static void __init init_mount_tree(void) 26891da177e4SLinus Torvalds { 26901da177e4SLinus Torvalds struct vfsmount *mnt; 26916b3286edSKirill Korotaev struct mnt_namespace *ns; 2692ac748a09SJan Blunck struct path root; 26931da177e4SLinus Torvalds 26941da177e4SLinus Torvalds mnt = do_kern_mount("rootfs", 0, "rootfs", NULL); 26951da177e4SLinus Torvalds if (IS_ERR(mnt)) 26961da177e4SLinus Torvalds panic("Can't create rootfs"); 2697b3e19d92SNick Piggin 26983b22edc5STrond Myklebust ns = create_mnt_ns(mnt); 26993b22edc5STrond Myklebust if (IS_ERR(ns)) 27001da177e4SLinus Torvalds panic("Can't allocate initial namespace"); 27011da177e4SLinus Torvalds 27026b3286edSKirill Korotaev init_task.nsproxy->mnt_ns = ns; 27036b3286edSKirill Korotaev get_mnt_ns(ns); 27041da177e4SLinus Torvalds 2705ac748a09SJan Blunck root.mnt = ns->root; 2706ac748a09SJan Blunck root.dentry = ns->root->mnt_root; 2707ac748a09SJan Blunck 2708ac748a09SJan Blunck set_fs_pwd(current->fs, &root); 2709ac748a09SJan Blunck set_fs_root(current->fs, &root); 27101da177e4SLinus Torvalds } 27111da177e4SLinus Torvalds 271274bf17cfSDenis Cheng void __init mnt_init(void) 27131da177e4SLinus Torvalds { 271413f14b4dSEric Dumazet unsigned u; 271515a67dd8SRandy Dunlap int err; 27161da177e4SLinus Torvalds 2717390c6843SRam Pai init_rwsem(&namespace_sem); 2718390c6843SRam Pai 27197d6fec45SAl Viro mnt_cache = kmem_cache_create("mnt_cache", sizeof(struct mount), 272020c2df83SPaul Mundt 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC, NULL); 27211da177e4SLinus Torvalds 2722b58fed8bSRam Pai mount_hashtable = (struct list_head *)__get_free_page(GFP_ATOMIC); 27231da177e4SLinus Torvalds 27241da177e4SLinus Torvalds if (!mount_hashtable) 27251da177e4SLinus Torvalds panic("Failed to allocate mount hash table\n"); 27261da177e4SLinus Torvalds 272780cdc6daSMandeep Singh Baines printk(KERN_INFO "Mount-cache hash table entries: %lu\n", HASH_SIZE); 27281da177e4SLinus Torvalds 272913f14b4dSEric Dumazet for (u = 0; u < HASH_SIZE; u++) 273013f14b4dSEric Dumazet INIT_LIST_HEAD(&mount_hashtable[u]); 27311da177e4SLinus Torvalds 273299b7db7bSNick Piggin br_lock_init(vfsmount_lock); 273399b7db7bSNick Piggin 273415a67dd8SRandy Dunlap err = sysfs_init(); 273515a67dd8SRandy Dunlap if (err) 273615a67dd8SRandy Dunlap printk(KERN_WARNING "%s: sysfs_init error: %d\n", 27378e24eea7SHarvey Harrison __func__, err); 273800d26666SGreg Kroah-Hartman fs_kobj = kobject_create_and_add("fs", NULL); 273900d26666SGreg Kroah-Hartman if (!fs_kobj) 27408e24eea7SHarvey Harrison printk(KERN_WARNING "%s: kobj create error\n", __func__); 27411da177e4SLinus Torvalds init_rootfs(); 27421da177e4SLinus Torvalds init_mount_tree(); 27431da177e4SLinus Torvalds } 27441da177e4SLinus Torvalds 2745616511d0STrond Myklebust void put_mnt_ns(struct mnt_namespace *ns) 27461da177e4SLinus Torvalds { 274770fbcdf4SRam Pai LIST_HEAD(umount_list); 2748616511d0STrond Myklebust 2749d498b25aSAl Viro if (!atomic_dec_and_test(&ns->count)) 2750616511d0STrond Myklebust return; 2751390c6843SRam Pai down_write(&namespace_sem); 275299b7db7bSNick Piggin br_write_lock(vfsmount_lock); 2753d498b25aSAl Viro umount_tree(ns->root, 0, &umount_list); 275499b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 2755390c6843SRam Pai up_write(&namespace_sem); 275670fbcdf4SRam Pai release_mounts(&umount_list); 27576b3286edSKirill Korotaev kfree(ns); 27581da177e4SLinus Torvalds } 27599d412a43SAl Viro 27609d412a43SAl Viro struct vfsmount *kern_mount_data(struct file_system_type *type, void *data) 27619d412a43SAl Viro { 2762423e0ab0STim Chen struct vfsmount *mnt; 2763423e0ab0STim Chen mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, data); 2764423e0ab0STim Chen if (!IS_ERR(mnt)) { 2765423e0ab0STim Chen /* 2766423e0ab0STim Chen * it is a longterm mount, don't release mnt until 2767423e0ab0STim Chen * we unmount before file sys is unregistered 2768423e0ab0STim Chen */ 2769423e0ab0STim Chen mnt_make_longterm(mnt); 2770423e0ab0STim Chen } 2771423e0ab0STim Chen return mnt; 27729d412a43SAl Viro } 27739d412a43SAl Viro EXPORT_SYMBOL_GPL(kern_mount_data); 2774423e0ab0STim Chen 2775423e0ab0STim Chen void kern_unmount(struct vfsmount *mnt) 2776423e0ab0STim Chen { 2777423e0ab0STim Chen /* release long term mount so mount point can be released */ 2778423e0ab0STim Chen if (!IS_ERR_OR_NULL(mnt)) { 2779423e0ab0STim Chen mnt_make_shortterm(mnt); 2780423e0ab0STim Chen mntput(mnt); 2781423e0ab0STim Chen } 2782423e0ab0STim Chen } 2783423e0ab0STim Chen EXPORT_SYMBOL(kern_unmount); 278402125a82SAl Viro 278502125a82SAl Viro bool our_mnt(struct vfsmount *mnt) 278602125a82SAl Viro { 278702125a82SAl Viro return check_mnt(mnt); 278802125a82SAl Viro } 2789