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 */ 14483adc753SAl Viro static inline void mnt_add_count(struct mount *mnt, int n) 145b3e19d92SNick Piggin { 146b3e19d92SNick Piggin #ifdef CONFIG_SMP 14768e8a9feSAl Viro this_cpu_add(mnt->mnt_pcp->mnt_count, n); 148b3e19d92SNick Piggin #else 149b3e19d92SNick Piggin preempt_disable(); 15068e8a9feSAl Viro 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 */ 15883adc753SAl Viro unsigned int mnt_get_count(struct mount *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) { 16568e8a9feSAl Viro count += per_cpu_ptr(mnt->mnt_pcp, cpu)->mnt_count; 166b3e19d92SNick Piggin } 167b3e19d92SNick Piggin 168b3e19d92SNick Piggin return count; 169b3e19d92SNick Piggin #else 17068e8a9feSAl Viro 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 19268e8a9feSAl Viro p->mnt_pcp = alloc_percpu(struct mnt_pcp); 19368e8a9feSAl Viro if (!p->mnt_pcp) 194b3e19d92SNick Piggin goto out_free_devname; 195b3e19d92SNick Piggin 19668e8a9feSAl Viro this_cpu_add(p->mnt_pcp->mnt_count, 1); 197b3e19d92SNick Piggin #else 19868e8a9feSAl Viro p->mnt_count = 1; 19968e8a9feSAl Viro p->mnt_writers = 0; 200b3e19d92SNick Piggin #endif 201b3e19d92SNick Piggin 2021b8e5564SAl Viro INIT_LIST_HEAD(&p->mnt_hash); 2036b41d536SAl Viro INIT_LIST_HEAD(&p->mnt_child); 2046b41d536SAl Viro INIT_LIST_HEAD(&p->mnt_mounts); 2051da177e4SLinus Torvalds INIT_LIST_HEAD(&mnt->mnt_list); 2066776db3dSAl Viro INIT_LIST_HEAD(&p->mnt_expire); 2076776db3dSAl Viro INIT_LIST_HEAD(&p->mnt_share); 2086776db3dSAl Viro INIT_LIST_HEAD(&p->mnt_slave_list); 2096776db3dSAl Viro INIT_LIST_HEAD(&p->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 25683adc753SAl Viro static inline void mnt_inc_writers(struct mount *mnt) 2573d733633SDave Hansen { 258d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 25968e8a9feSAl Viro this_cpu_inc(mnt->mnt_pcp->mnt_writers); 260d3ef3d73Snpiggin@suse.de #else 26168e8a9feSAl Viro mnt->mnt_writers++; 262d3ef3d73Snpiggin@suse.de #endif 2633d733633SDave Hansen } 2643d733633SDave Hansen 26583adc753SAl Viro static inline void mnt_dec_writers(struct mount *mnt) 2663d733633SDave Hansen { 267d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 26868e8a9feSAl Viro this_cpu_dec(mnt->mnt_pcp->mnt_writers); 269d3ef3d73Snpiggin@suse.de #else 27068e8a9feSAl Viro mnt->mnt_writers--; 271d3ef3d73Snpiggin@suse.de #endif 272d3ef3d73Snpiggin@suse.de } 273d3ef3d73Snpiggin@suse.de 27483adc753SAl Viro static unsigned int mnt_get_writers(struct mount *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) { 28168e8a9feSAl Viro 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 30083adc753SAl Viro * @m: 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 */ 30883adc753SAl Viro int mnt_want_write(struct vfsmount *m) 3098366025eSDave Hansen { 31083adc753SAl Viro struct mount *mnt = real_mount(m); 3113d733633SDave Hansen int ret = 0; 3123d733633SDave Hansen 313d3ef3d73Snpiggin@suse.de preempt_disable(); 314c6653a83SNick Piggin mnt_inc_writers(mnt); 315d3ef3d73Snpiggin@suse.de /* 316c6653a83SNick Piggin * The store to mnt_inc_writers must be visible before we pass 317d3ef3d73Snpiggin@suse.de * MNT_WRITE_HOLD loop below, so that the slowpath can see our 318d3ef3d73Snpiggin@suse.de * incremented count after it has set MNT_WRITE_HOLD. 319d3ef3d73Snpiggin@suse.de */ 320d3ef3d73Snpiggin@suse.de smp_mb(); 32183adc753SAl Viro while (mnt->mnt.mnt_flags & MNT_WRITE_HOLD) 322d3ef3d73Snpiggin@suse.de cpu_relax(); 323d3ef3d73Snpiggin@suse.de /* 324d3ef3d73Snpiggin@suse.de * After the slowpath clears MNT_WRITE_HOLD, mnt_is_readonly will 325d3ef3d73Snpiggin@suse.de * be set to match its requirements. So we must not load that until 326d3ef3d73Snpiggin@suse.de * MNT_WRITE_HOLD is cleared. 327d3ef3d73Snpiggin@suse.de */ 328d3ef3d73Snpiggin@suse.de smp_rmb(); 32983adc753SAl Viro if (__mnt_is_readonly(m)) { 330c6653a83SNick Piggin mnt_dec_writers(mnt); 3313d733633SDave Hansen ret = -EROFS; 3323d733633SDave Hansen goto out; 3333d733633SDave Hansen } 3343d733633SDave Hansen out: 335d3ef3d73Snpiggin@suse.de preempt_enable(); 3363d733633SDave Hansen return ret; 3378366025eSDave Hansen } 3388366025eSDave Hansen EXPORT_SYMBOL_GPL(mnt_want_write); 3398366025eSDave Hansen 3408366025eSDave Hansen /** 34196029c4eSnpiggin@suse.de * mnt_clone_write - get write access to a mount 34296029c4eSnpiggin@suse.de * @mnt: the mount on which to take a write 34396029c4eSnpiggin@suse.de * 34496029c4eSnpiggin@suse.de * This is effectively like mnt_want_write, except 34596029c4eSnpiggin@suse.de * it must only be used to take an extra write reference 34696029c4eSnpiggin@suse.de * on a mountpoint that we already know has a write reference 34796029c4eSnpiggin@suse.de * on it. This allows some optimisation. 34896029c4eSnpiggin@suse.de * 34996029c4eSnpiggin@suse.de * After finished, mnt_drop_write must be called as usual to 35096029c4eSnpiggin@suse.de * drop the reference. 35196029c4eSnpiggin@suse.de */ 35296029c4eSnpiggin@suse.de int mnt_clone_write(struct vfsmount *mnt) 35396029c4eSnpiggin@suse.de { 35496029c4eSnpiggin@suse.de /* superblock may be r/o */ 35596029c4eSnpiggin@suse.de if (__mnt_is_readonly(mnt)) 35696029c4eSnpiggin@suse.de return -EROFS; 35796029c4eSnpiggin@suse.de preempt_disable(); 35883adc753SAl Viro mnt_inc_writers(real_mount(mnt)); 35996029c4eSnpiggin@suse.de preempt_enable(); 36096029c4eSnpiggin@suse.de return 0; 36196029c4eSnpiggin@suse.de } 36296029c4eSnpiggin@suse.de EXPORT_SYMBOL_GPL(mnt_clone_write); 36396029c4eSnpiggin@suse.de 36496029c4eSnpiggin@suse.de /** 36596029c4eSnpiggin@suse.de * mnt_want_write_file - get write access to a file's mount 36696029c4eSnpiggin@suse.de * @file: the file who's mount on which to take a write 36796029c4eSnpiggin@suse.de * 36896029c4eSnpiggin@suse.de * This is like mnt_want_write, but it takes a file and can 36996029c4eSnpiggin@suse.de * do some optimisations if the file is open for write already 37096029c4eSnpiggin@suse.de */ 37196029c4eSnpiggin@suse.de int mnt_want_write_file(struct file *file) 37296029c4eSnpiggin@suse.de { 3732d8dd38aSOGAWA Hirofumi struct inode *inode = file->f_dentry->d_inode; 3742d8dd38aSOGAWA Hirofumi if (!(file->f_mode & FMODE_WRITE) || special_file(inode->i_mode)) 37596029c4eSnpiggin@suse.de return mnt_want_write(file->f_path.mnt); 37696029c4eSnpiggin@suse.de else 37796029c4eSnpiggin@suse.de return mnt_clone_write(file->f_path.mnt); 37896029c4eSnpiggin@suse.de } 37996029c4eSnpiggin@suse.de EXPORT_SYMBOL_GPL(mnt_want_write_file); 38096029c4eSnpiggin@suse.de 38196029c4eSnpiggin@suse.de /** 3828366025eSDave Hansen * mnt_drop_write - give up write access to a mount 3838366025eSDave Hansen * @mnt: the mount on which to give up write access 3848366025eSDave Hansen * 3858366025eSDave Hansen * Tells the low-level filesystem that we are done 3868366025eSDave Hansen * performing writes to it. Must be matched with 3878366025eSDave Hansen * mnt_want_write() call above. 3888366025eSDave Hansen */ 3898366025eSDave Hansen void mnt_drop_write(struct vfsmount *mnt) 3908366025eSDave Hansen { 391d3ef3d73Snpiggin@suse.de preempt_disable(); 39283adc753SAl Viro mnt_dec_writers(real_mount(mnt)); 393d3ef3d73Snpiggin@suse.de preempt_enable(); 3948366025eSDave Hansen } 3958366025eSDave Hansen EXPORT_SYMBOL_GPL(mnt_drop_write); 3968366025eSDave Hansen 3972a79f17eSAl Viro void mnt_drop_write_file(struct file *file) 3982a79f17eSAl Viro { 3992a79f17eSAl Viro mnt_drop_write(file->f_path.mnt); 4002a79f17eSAl Viro } 4012a79f17eSAl Viro EXPORT_SYMBOL(mnt_drop_write_file); 4022a79f17eSAl Viro 40383adc753SAl Viro static int mnt_make_readonly(struct mount *mnt) 4048366025eSDave Hansen { 4053d733633SDave Hansen int ret = 0; 4063d733633SDave Hansen 40799b7db7bSNick Piggin br_write_lock(vfsmount_lock); 40883adc753SAl Viro mnt->mnt.mnt_flags |= MNT_WRITE_HOLD; 409d3ef3d73Snpiggin@suse.de /* 410d3ef3d73Snpiggin@suse.de * After storing MNT_WRITE_HOLD, we'll read the counters. This store 411d3ef3d73Snpiggin@suse.de * should be visible before we do. 412d3ef3d73Snpiggin@suse.de */ 413d3ef3d73Snpiggin@suse.de smp_mb(); 414d3ef3d73Snpiggin@suse.de 415d3ef3d73Snpiggin@suse.de /* 416d3ef3d73Snpiggin@suse.de * With writers on hold, if this value is zero, then there are 417d3ef3d73Snpiggin@suse.de * definitely no active writers (although held writers may subsequently 418d3ef3d73Snpiggin@suse.de * increment the count, they'll have to wait, and decrement it after 419d3ef3d73Snpiggin@suse.de * seeing MNT_READONLY). 420d3ef3d73Snpiggin@suse.de * 421d3ef3d73Snpiggin@suse.de * It is OK to have counter incremented on one CPU and decremented on 422d3ef3d73Snpiggin@suse.de * another: the sum will add up correctly. The danger would be when we 423d3ef3d73Snpiggin@suse.de * sum up each counter, if we read a counter before it is incremented, 424d3ef3d73Snpiggin@suse.de * but then read another CPU's count which it has been subsequently 425d3ef3d73Snpiggin@suse.de * decremented from -- we would see more decrements than we should. 426d3ef3d73Snpiggin@suse.de * MNT_WRITE_HOLD protects against this scenario, because 427d3ef3d73Snpiggin@suse.de * mnt_want_write first increments count, then smp_mb, then spins on 428d3ef3d73Snpiggin@suse.de * MNT_WRITE_HOLD, so it can't be decremented by another CPU while 429d3ef3d73Snpiggin@suse.de * we're counting up here. 430d3ef3d73Snpiggin@suse.de */ 431c6653a83SNick Piggin if (mnt_get_writers(mnt) > 0) 432d3ef3d73Snpiggin@suse.de ret = -EBUSY; 433d3ef3d73Snpiggin@suse.de else 43483adc753SAl Viro mnt->mnt.mnt_flags |= MNT_READONLY; 435d3ef3d73Snpiggin@suse.de /* 436d3ef3d73Snpiggin@suse.de * MNT_READONLY must become visible before ~MNT_WRITE_HOLD, so writers 437d3ef3d73Snpiggin@suse.de * that become unheld will see MNT_READONLY. 438d3ef3d73Snpiggin@suse.de */ 439d3ef3d73Snpiggin@suse.de smp_wmb(); 44083adc753SAl Viro mnt->mnt.mnt_flags &= ~MNT_WRITE_HOLD; 44199b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 4423d733633SDave Hansen return ret; 4433d733633SDave Hansen } 4448366025eSDave Hansen 44583adc753SAl Viro static void __mnt_unmake_readonly(struct mount *mnt) 4462e4b7fcdSDave Hansen { 44799b7db7bSNick Piggin br_write_lock(vfsmount_lock); 44883adc753SAl Viro mnt->mnt.mnt_flags &= ~MNT_READONLY; 44999b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 4502e4b7fcdSDave Hansen } 4512e4b7fcdSDave Hansen 452b105e270SAl Viro static void free_vfsmnt(struct mount *mnt) 4531da177e4SLinus Torvalds { 454b105e270SAl Viro kfree(mnt->mnt.mnt_devname); 45573cd49ecSMiklos Szeredi mnt_free_id(mnt); 456d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 45768e8a9feSAl Viro free_percpu(mnt->mnt_pcp); 458d3ef3d73Snpiggin@suse.de #endif 459b105e270SAl Viro kmem_cache_free(mnt_cache, mnt); 4601da177e4SLinus Torvalds } 4611da177e4SLinus Torvalds 4621da177e4SLinus Torvalds /* 463a05964f3SRam Pai * find the first or last mount at @dentry on vfsmount @mnt depending on 464a05964f3SRam Pai * @dir. If @dir is set return the first mount else return the last mount. 46599b7db7bSNick Piggin * vfsmount_lock must be held for read or write. 4661da177e4SLinus Torvalds */ 467c7105365SAl Viro struct mount *__lookup_mnt(struct vfsmount *mnt, struct dentry *dentry, 468a05964f3SRam Pai int dir) 4691da177e4SLinus Torvalds { 4701da177e4SLinus Torvalds struct list_head *head = mount_hashtable + hash(mnt, dentry); 4711da177e4SLinus Torvalds struct list_head *tmp = head; 472c7105365SAl Viro struct mount *p, *found = NULL; 4731da177e4SLinus Torvalds 4741da177e4SLinus Torvalds for (;;) { 475a05964f3SRam Pai tmp = dir ? tmp->next : tmp->prev; 4761da177e4SLinus Torvalds p = NULL; 4771da177e4SLinus Torvalds if (tmp == head) 4781da177e4SLinus Torvalds break; 4791b8e5564SAl Viro p = list_entry(tmp, struct mount, mnt_hash); 480a73324daSAl Viro if (&p->mnt_parent->mnt == mnt && p->mnt_mountpoint == dentry) { 481a05964f3SRam Pai found = p; 4821da177e4SLinus Torvalds break; 4831da177e4SLinus Torvalds } 4841da177e4SLinus Torvalds } 4851da177e4SLinus Torvalds return found; 4861da177e4SLinus Torvalds } 4871da177e4SLinus Torvalds 488a05964f3SRam Pai /* 489a05964f3SRam Pai * lookup_mnt increments the ref count before returning 490a05964f3SRam Pai * the vfsmount struct. 491a05964f3SRam Pai */ 4921c755af4SAl Viro struct vfsmount *lookup_mnt(struct path *path) 493a05964f3SRam Pai { 494c7105365SAl Viro struct mount *child_mnt; 49599b7db7bSNick Piggin 49699b7db7bSNick Piggin br_read_lock(vfsmount_lock); 497c7105365SAl Viro child_mnt = __lookup_mnt(path->mnt, path->dentry, 1); 498c7105365SAl Viro if (child_mnt) { 499c7105365SAl Viro mnt_add_count(child_mnt, 1); 50099b7db7bSNick Piggin br_read_unlock(vfsmount_lock); 501c7105365SAl Viro return &child_mnt->mnt; 502c7105365SAl Viro } else { 503c7105365SAl Viro br_read_unlock(vfsmount_lock); 504c7105365SAl Viro return NULL; 505c7105365SAl Viro } 506a05964f3SRam Pai } 507a05964f3SRam Pai 508143c8c91SAl Viro static inline int check_mnt(struct mount *mnt) 5091da177e4SLinus Torvalds { 5106b3286edSKirill Korotaev return mnt->mnt_ns == current->nsproxy->mnt_ns; 5111da177e4SLinus Torvalds } 5121da177e4SLinus Torvalds 51399b7db7bSNick Piggin /* 51499b7db7bSNick Piggin * vfsmount lock must be held for write 51599b7db7bSNick Piggin */ 5166b3286edSKirill Korotaev static void touch_mnt_namespace(struct mnt_namespace *ns) 5175addc5ddSAl Viro { 5185addc5ddSAl Viro if (ns) { 5195addc5ddSAl Viro ns->event = ++event; 5205addc5ddSAl Viro wake_up_interruptible(&ns->poll); 5215addc5ddSAl Viro } 5225addc5ddSAl Viro } 5235addc5ddSAl Viro 52499b7db7bSNick Piggin /* 52599b7db7bSNick Piggin * vfsmount lock must be held for write 52699b7db7bSNick Piggin */ 5276b3286edSKirill Korotaev static void __touch_mnt_namespace(struct mnt_namespace *ns) 5285addc5ddSAl Viro { 5295addc5ddSAl Viro if (ns && ns->event != event) { 5305addc5ddSAl Viro ns->event = event; 5315addc5ddSAl Viro wake_up_interruptible(&ns->poll); 5325addc5ddSAl Viro } 5335addc5ddSAl Viro } 5345addc5ddSAl Viro 53599b7db7bSNick Piggin /* 5365f57cbccSNick Piggin * Clear dentry's mounted state if it has no remaining mounts. 5375f57cbccSNick Piggin * vfsmount_lock must be held for write. 5385f57cbccSNick Piggin */ 539aa0a4cf0SAl Viro static void dentry_reset_mounted(struct dentry *dentry) 5405f57cbccSNick Piggin { 5415f57cbccSNick Piggin unsigned u; 5425f57cbccSNick Piggin 5435f57cbccSNick Piggin for (u = 0; u < HASH_SIZE; u++) { 544d5e50f74SAl Viro struct mount *p; 5455f57cbccSNick Piggin 5461b8e5564SAl Viro list_for_each_entry(p, &mount_hashtable[u], mnt_hash) { 547a73324daSAl Viro if (p->mnt_mountpoint == dentry) 5485f57cbccSNick Piggin return; 5495f57cbccSNick Piggin } 5505f57cbccSNick Piggin } 5515f57cbccSNick Piggin spin_lock(&dentry->d_lock); 5525f57cbccSNick Piggin dentry->d_flags &= ~DCACHE_MOUNTED; 5535f57cbccSNick Piggin spin_unlock(&dentry->d_lock); 5545f57cbccSNick Piggin } 5555f57cbccSNick Piggin 5565f57cbccSNick Piggin /* 55799b7db7bSNick Piggin * vfsmount lock must be held for write 55899b7db7bSNick Piggin */ 559419148daSAl Viro static void detach_mnt(struct mount *mnt, struct path *old_path) 5601da177e4SLinus Torvalds { 561a73324daSAl Viro old_path->dentry = mnt->mnt_mountpoint; 5620714a533SAl Viro old_path->mnt = &mnt->mnt_parent->mnt; 5630714a533SAl Viro mnt->mnt_parent = mnt; 564a73324daSAl Viro mnt->mnt_mountpoint = mnt->mnt.mnt_root; 5656b41d536SAl Viro list_del_init(&mnt->mnt_child); 5661b8e5564SAl Viro list_del_init(&mnt->mnt_hash); 567aa0a4cf0SAl Viro dentry_reset_mounted(old_path->dentry); 5681da177e4SLinus Torvalds } 5691da177e4SLinus Torvalds 57099b7db7bSNick Piggin /* 57199b7db7bSNick Piggin * vfsmount lock must be held for write 57299b7db7bSNick Piggin */ 57314cf1fa8SAl Viro void mnt_set_mountpoint(struct mount *mnt, struct dentry *dentry, 57444d964d6SAl Viro struct mount *child_mnt) 575b90fa9aeSRam Pai { 57614cf1fa8SAl Viro child_mnt->mnt_parent = real_mount(mntget(&mnt->mnt)); 577a73324daSAl Viro child_mnt->mnt_mountpoint = dget(dentry); 5785f57cbccSNick Piggin spin_lock(&dentry->d_lock); 5795f57cbccSNick Piggin dentry->d_flags |= DCACHE_MOUNTED; 5805f57cbccSNick Piggin spin_unlock(&dentry->d_lock); 581b90fa9aeSRam Pai } 582b90fa9aeSRam Pai 58399b7db7bSNick Piggin /* 58499b7db7bSNick Piggin * vfsmount lock must be held for write 58599b7db7bSNick Piggin */ 586419148daSAl Viro static void attach_mnt(struct mount *mnt, struct path *path) 5871da177e4SLinus Torvalds { 58814cf1fa8SAl Viro mnt_set_mountpoint(real_mount(path->mnt), path->dentry, mnt); 5891b8e5564SAl Viro list_add_tail(&mnt->mnt_hash, mount_hashtable + 5901a390689SAl Viro hash(path->mnt, path->dentry)); 5916b41d536SAl Viro list_add_tail(&mnt->mnt_child, &real_mount(path->mnt)->mnt_mounts); 592b90fa9aeSRam Pai } 593b90fa9aeSRam Pai 59483adc753SAl Viro static inline void __mnt_make_longterm(struct mount *mnt) 5957e3d0eb0SAl Viro { 5967e3d0eb0SAl Viro #ifdef CONFIG_SMP 59768e8a9feSAl Viro atomic_inc(&mnt->mnt_longterm); 5987e3d0eb0SAl Viro #endif 5997e3d0eb0SAl Viro } 6007e3d0eb0SAl Viro 6017e3d0eb0SAl Viro /* needs vfsmount lock for write */ 60283adc753SAl Viro static inline void __mnt_make_shortterm(struct mount *mnt) 6037e3d0eb0SAl Viro { 6047e3d0eb0SAl Viro #ifdef CONFIG_SMP 60568e8a9feSAl Viro atomic_dec(&mnt->mnt_longterm); 6067e3d0eb0SAl Viro #endif 6077e3d0eb0SAl Viro } 6087e3d0eb0SAl Viro 609b90fa9aeSRam Pai /* 61099b7db7bSNick Piggin * vfsmount lock must be held for write 611b90fa9aeSRam Pai */ 6124b2619a5SAl Viro static void commit_tree(struct mount *mnt) 613b90fa9aeSRam Pai { 6140714a533SAl Viro struct mount *parent = mnt->mnt_parent; 61583adc753SAl Viro struct mount *m; 616b90fa9aeSRam Pai LIST_HEAD(head); 617143c8c91SAl Viro struct mnt_namespace *n = parent->mnt_ns; 618b90fa9aeSRam Pai 6190714a533SAl Viro BUG_ON(parent == mnt); 620b90fa9aeSRam Pai 6214b2619a5SAl Viro list_add_tail(&head, &mnt->mnt.mnt_list); 62283adc753SAl Viro list_for_each_entry(m, &head, mnt.mnt_list) { 623143c8c91SAl Viro m->mnt_ns = n; 6247e3d0eb0SAl Viro __mnt_make_longterm(m); 625f03c6599SAl Viro } 626f03c6599SAl Viro 627b90fa9aeSRam Pai list_splice(&head, n->list.prev); 628b90fa9aeSRam Pai 6291b8e5564SAl Viro list_add_tail(&mnt->mnt_hash, mount_hashtable + 630a73324daSAl Viro hash(&parent->mnt, mnt->mnt_mountpoint)); 6316b41d536SAl Viro list_add_tail(&mnt->mnt_child, &parent->mnt_mounts); 6326b3286edSKirill Korotaev touch_mnt_namespace(n); 6331da177e4SLinus Torvalds } 6341da177e4SLinus Torvalds 635315fc83eSAl Viro static struct mount *next_mnt(struct mount *p, struct vfsmount *root) 6361da177e4SLinus Torvalds { 6376b41d536SAl Viro struct list_head *next = p->mnt_mounts.next; 6386b41d536SAl Viro if (next == &p->mnt_mounts) { 6391da177e4SLinus Torvalds while (1) { 640315fc83eSAl Viro if (&p->mnt == root) 6411da177e4SLinus Torvalds return NULL; 6426b41d536SAl Viro next = p->mnt_child.next; 6436b41d536SAl Viro if (next != &p->mnt_parent->mnt_mounts) 6441da177e4SLinus Torvalds break; 6450714a533SAl Viro p = p->mnt_parent; 6461da177e4SLinus Torvalds } 6471da177e4SLinus Torvalds } 6486b41d536SAl Viro return list_entry(next, struct mount, mnt_child); 6491da177e4SLinus Torvalds } 6501da177e4SLinus Torvalds 651315fc83eSAl Viro static struct mount *skip_mnt_tree(struct mount *p) 6529676f0c6SRam Pai { 6536b41d536SAl Viro struct list_head *prev = p->mnt_mounts.prev; 6546b41d536SAl Viro while (prev != &p->mnt_mounts) { 6556b41d536SAl Viro p = list_entry(prev, struct mount, mnt_child); 6566b41d536SAl Viro prev = p->mnt_mounts.prev; 6579676f0c6SRam Pai } 6589676f0c6SRam Pai return p; 6599676f0c6SRam Pai } 6609676f0c6SRam Pai 6619d412a43SAl Viro struct vfsmount * 6629d412a43SAl Viro vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data) 6639d412a43SAl Viro { 664b105e270SAl Viro struct mount *mnt; 6659d412a43SAl Viro struct dentry *root; 6669d412a43SAl Viro 6679d412a43SAl Viro if (!type) 6689d412a43SAl Viro return ERR_PTR(-ENODEV); 6699d412a43SAl Viro 6709d412a43SAl Viro mnt = alloc_vfsmnt(name); 6719d412a43SAl Viro if (!mnt) 6729d412a43SAl Viro return ERR_PTR(-ENOMEM); 6739d412a43SAl Viro 6749d412a43SAl Viro if (flags & MS_KERNMOUNT) 675b105e270SAl Viro mnt->mnt.mnt_flags = MNT_INTERNAL; 6769d412a43SAl Viro 6779d412a43SAl Viro root = mount_fs(type, flags, name, data); 6789d412a43SAl Viro if (IS_ERR(root)) { 6799d412a43SAl Viro free_vfsmnt(mnt); 6809d412a43SAl Viro return ERR_CAST(root); 6819d412a43SAl Viro } 6829d412a43SAl Viro 683b105e270SAl Viro mnt->mnt.mnt_root = root; 684b105e270SAl Viro mnt->mnt.mnt_sb = root->d_sb; 685a73324daSAl Viro mnt->mnt_mountpoint = mnt->mnt.mnt_root; 6860714a533SAl Viro mnt->mnt_parent = mnt; 687b105e270SAl Viro return &mnt->mnt; 6889d412a43SAl Viro } 6899d412a43SAl Viro EXPORT_SYMBOL_GPL(vfs_kern_mount); 6909d412a43SAl Viro 69187129cc0SAl Viro static struct mount *clone_mnt(struct mount *old, struct dentry *root, 69236341f64SRam Pai int flag) 6931da177e4SLinus Torvalds { 69487129cc0SAl Viro struct super_block *sb = old->mnt.mnt_sb; 69587129cc0SAl Viro struct mount *mnt = alloc_vfsmnt(old->mnt.mnt_devname); 6961da177e4SLinus Torvalds 6971da177e4SLinus Torvalds if (mnt) { 698719f5d7fSMiklos Szeredi if (flag & (CL_SLAVE | CL_PRIVATE)) 699b105e270SAl Viro mnt->mnt.mnt_group_id = 0; /* not a peer of original */ 700719f5d7fSMiklos Szeredi else 70187129cc0SAl Viro mnt->mnt.mnt_group_id = old->mnt.mnt_group_id; 702719f5d7fSMiklos Szeredi 703b105e270SAl Viro if ((flag & CL_MAKE_SHARED) && !mnt->mnt.mnt_group_id) { 704b105e270SAl Viro int err = mnt_alloc_group_id(mnt); 705719f5d7fSMiklos Szeredi if (err) 706719f5d7fSMiklos Szeredi goto out_free; 707719f5d7fSMiklos Szeredi } 708719f5d7fSMiklos Szeredi 70987129cc0SAl Viro mnt->mnt.mnt_flags = old->mnt.mnt_flags & ~MNT_WRITE_HOLD; 7101da177e4SLinus Torvalds atomic_inc(&sb->s_active); 711b105e270SAl Viro mnt->mnt.mnt_sb = sb; 712b105e270SAl Viro mnt->mnt.mnt_root = dget(root); 713a73324daSAl Viro mnt->mnt_mountpoint = mnt->mnt.mnt_root; 7140714a533SAl Viro mnt->mnt_parent = mnt; 715b90fa9aeSRam Pai 7165afe0022SRam Pai if (flag & CL_SLAVE) { 7176776db3dSAl Viro list_add(&mnt->mnt_slave, &old->mnt_slave_list); 71832301920SAl Viro mnt->mnt_master = old; 719b105e270SAl Viro CLEAR_MNT_SHARED(&mnt->mnt); 7208aec0809SAl Viro } else if (!(flag & CL_PRIVATE)) { 72187129cc0SAl Viro if ((flag & CL_MAKE_SHARED) || IS_MNT_SHARED(&old->mnt)) 7226776db3dSAl Viro list_add(&mnt->mnt_share, &old->mnt_share); 723d10e8defSAl Viro if (IS_MNT_SLAVE(old)) 7246776db3dSAl Viro list_add(&mnt->mnt_slave, &old->mnt_slave); 725d10e8defSAl Viro mnt->mnt_master = old->mnt_master; 7265afe0022SRam Pai } 727b90fa9aeSRam Pai if (flag & CL_MAKE_SHARED) 7280f0afb1dSAl Viro set_mnt_shared(mnt); 7291da177e4SLinus Torvalds 7301da177e4SLinus Torvalds /* stick the duplicate mount on the same expiry list 7311da177e4SLinus Torvalds * as the original if that was on one */ 73236341f64SRam Pai if (flag & CL_EXPIRE) { 7336776db3dSAl Viro if (!list_empty(&old->mnt_expire)) 7346776db3dSAl Viro list_add(&mnt->mnt_expire, &old->mnt_expire); 7351da177e4SLinus Torvalds } 73636341f64SRam Pai } 737cb338d06SAl Viro return mnt; 738719f5d7fSMiklos Szeredi 739719f5d7fSMiklos Szeredi out_free: 740719f5d7fSMiklos Szeredi free_vfsmnt(mnt); 741719f5d7fSMiklos Szeredi return NULL; 7421da177e4SLinus Torvalds } 7431da177e4SLinus Torvalds 74483adc753SAl Viro static inline void mntfree(struct mount *mnt) 7451da177e4SLinus Torvalds { 74683adc753SAl Viro struct vfsmount *m = &mnt->mnt; 74783adc753SAl Viro struct super_block *sb = m->mnt_sb; 748b3e19d92SNick Piggin 7493d733633SDave Hansen /* 7503d733633SDave Hansen * This probably indicates that somebody messed 7513d733633SDave Hansen * up a mnt_want/drop_write() pair. If this 7523d733633SDave Hansen * happens, the filesystem was probably unable 7533d733633SDave Hansen * to make r/w->r/o transitions. 7543d733633SDave Hansen */ 755d3ef3d73Snpiggin@suse.de /* 756b3e19d92SNick Piggin * The locking used to deal with mnt_count decrement provides barriers, 757b3e19d92SNick Piggin * so mnt_get_writers() below is safe. 758d3ef3d73Snpiggin@suse.de */ 759c6653a83SNick Piggin WARN_ON(mnt_get_writers(mnt)); 76083adc753SAl Viro fsnotify_vfsmount_delete(m); 76183adc753SAl Viro dput(m->mnt_root); 76283adc753SAl Viro free_vfsmnt(mnt); 7631da177e4SLinus Torvalds deactivate_super(sb); 7641da177e4SLinus Torvalds } 7651da177e4SLinus Torvalds 766900148dcSAl Viro static void mntput_no_expire(struct mount *mnt) 7677b7b1aceSAl Viro { 768b3e19d92SNick Piggin put_again: 769f03c6599SAl Viro #ifdef CONFIG_SMP 770b3e19d92SNick Piggin br_read_lock(vfsmount_lock); 77168e8a9feSAl Viro if (likely(atomic_read(&mnt->mnt_longterm))) { 772aa9c0e07SAl Viro mnt_add_count(mnt, -1); 773b3e19d92SNick Piggin br_read_unlock(vfsmount_lock); 77499b7db7bSNick Piggin return; 775b3e19d92SNick Piggin } 776b3e19d92SNick Piggin br_read_unlock(vfsmount_lock); 777b3e19d92SNick Piggin 77899b7db7bSNick Piggin br_write_lock(vfsmount_lock); 779aa9c0e07SAl Viro mnt_add_count(mnt, -1); 780b3e19d92SNick Piggin if (mnt_get_count(mnt)) { 78199b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 78299b7db7bSNick Piggin return; 78399b7db7bSNick Piggin } 784b3e19d92SNick Piggin #else 785aa9c0e07SAl Viro mnt_add_count(mnt, -1); 786b3e19d92SNick Piggin if (likely(mnt_get_count(mnt))) 787b3e19d92SNick Piggin return; 788b3e19d92SNick Piggin br_write_lock(vfsmount_lock); 789f03c6599SAl Viro #endif 79083adc753SAl Viro if (unlikely(mnt->mnt.mnt_pinned)) { 79183adc753SAl Viro mnt_add_count(mnt, mnt->mnt.mnt_pinned + 1); 79283adc753SAl Viro mnt->mnt.mnt_pinned = 0; 793b3e19d92SNick Piggin br_write_unlock(vfsmount_lock); 794900148dcSAl Viro acct_auto_close_mnt(&mnt->mnt); 795b3e19d92SNick Piggin goto put_again; 796b3e19d92SNick Piggin } 797b3e19d92SNick Piggin br_write_unlock(vfsmount_lock); 798b3e19d92SNick Piggin mntfree(mnt); 799b3e19d92SNick Piggin } 800b3e19d92SNick Piggin 801b3e19d92SNick Piggin void mntput(struct vfsmount *mnt) 802b3e19d92SNick Piggin { 803b3e19d92SNick Piggin if (mnt) { 804b3e19d92SNick Piggin /* avoid cacheline pingpong, hope gcc doesn't get "smart" */ 805b3e19d92SNick Piggin if (unlikely(mnt->mnt_expiry_mark)) 806b3e19d92SNick Piggin mnt->mnt_expiry_mark = 0; 807900148dcSAl Viro mntput_no_expire(real_mount(mnt)); 808b3e19d92SNick Piggin } 809b3e19d92SNick Piggin } 810b3e19d92SNick Piggin EXPORT_SYMBOL(mntput); 811b3e19d92SNick Piggin 812b3e19d92SNick Piggin struct vfsmount *mntget(struct vfsmount *mnt) 813b3e19d92SNick Piggin { 814b3e19d92SNick Piggin if (mnt) 81583adc753SAl Viro mnt_add_count(real_mount(mnt), 1); 816b3e19d92SNick Piggin return mnt; 817b3e19d92SNick Piggin } 818b3e19d92SNick Piggin EXPORT_SYMBOL(mntget); 819b3e19d92SNick Piggin 8207b7b1aceSAl Viro void mnt_pin(struct vfsmount *mnt) 8217b7b1aceSAl Viro { 82299b7db7bSNick Piggin br_write_lock(vfsmount_lock); 8237b7b1aceSAl Viro mnt->mnt_pinned++; 82499b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 8257b7b1aceSAl Viro } 8267b7b1aceSAl Viro EXPORT_SYMBOL(mnt_pin); 8277b7b1aceSAl Viro 8287b7b1aceSAl Viro void mnt_unpin(struct vfsmount *mnt) 8297b7b1aceSAl Viro { 83099b7db7bSNick Piggin br_write_lock(vfsmount_lock); 8317b7b1aceSAl Viro if (mnt->mnt_pinned) { 83283adc753SAl Viro mnt_add_count(real_mount(mnt), 1); 8337b7b1aceSAl Viro mnt->mnt_pinned--; 8347b7b1aceSAl Viro } 83599b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 8367b7b1aceSAl Viro } 8377b7b1aceSAl Viro EXPORT_SYMBOL(mnt_unpin); 8381da177e4SLinus Torvalds 839b3b304a2SMiklos Szeredi static inline void mangle(struct seq_file *m, const char *s) 840b3b304a2SMiklos Szeredi { 841b3b304a2SMiklos Szeredi seq_escape(m, s, " \t\n\\"); 842b3b304a2SMiklos Szeredi } 843b3b304a2SMiklos Szeredi 844b3b304a2SMiklos Szeredi /* 845b3b304a2SMiklos Szeredi * Simple .show_options callback for filesystems which don't want to 846b3b304a2SMiklos Szeredi * implement more complex mount option showing. 847b3b304a2SMiklos Szeredi * 848b3b304a2SMiklos Szeredi * See also save_mount_options(). 849b3b304a2SMiklos Szeredi */ 850b3b304a2SMiklos Szeredi int generic_show_options(struct seq_file *m, struct vfsmount *mnt) 851b3b304a2SMiklos Szeredi { 8522a32cebdSAl Viro const char *options; 8532a32cebdSAl Viro 8542a32cebdSAl Viro rcu_read_lock(); 8552a32cebdSAl Viro options = rcu_dereference(mnt->mnt_sb->s_options); 856b3b304a2SMiklos Szeredi 857b3b304a2SMiklos Szeredi if (options != NULL && options[0]) { 858b3b304a2SMiklos Szeredi seq_putc(m, ','); 859b3b304a2SMiklos Szeredi mangle(m, options); 860b3b304a2SMiklos Szeredi } 8612a32cebdSAl Viro rcu_read_unlock(); 862b3b304a2SMiklos Szeredi 863b3b304a2SMiklos Szeredi return 0; 864b3b304a2SMiklos Szeredi } 865b3b304a2SMiklos Szeredi EXPORT_SYMBOL(generic_show_options); 866b3b304a2SMiklos Szeredi 867b3b304a2SMiklos Szeredi /* 868b3b304a2SMiklos Szeredi * If filesystem uses generic_show_options(), this function should be 869b3b304a2SMiklos Szeredi * called from the fill_super() callback. 870b3b304a2SMiklos Szeredi * 871b3b304a2SMiklos Szeredi * The .remount_fs callback usually needs to be handled in a special 872b3b304a2SMiklos Szeredi * way, to make sure, that previous options are not overwritten if the 873b3b304a2SMiklos Szeredi * remount fails. 874b3b304a2SMiklos Szeredi * 875b3b304a2SMiklos Szeredi * Also note, that if the filesystem's .remount_fs function doesn't 876b3b304a2SMiklos Szeredi * reset all options to their default value, but changes only newly 877b3b304a2SMiklos Szeredi * given options, then the displayed options will not reflect reality 878b3b304a2SMiklos Szeredi * any more. 879b3b304a2SMiklos Szeredi */ 880b3b304a2SMiklos Szeredi void save_mount_options(struct super_block *sb, char *options) 881b3b304a2SMiklos Szeredi { 8822a32cebdSAl Viro BUG_ON(sb->s_options); 8832a32cebdSAl Viro rcu_assign_pointer(sb->s_options, kstrdup(options, GFP_KERNEL)); 884b3b304a2SMiklos Szeredi } 885b3b304a2SMiklos Szeredi EXPORT_SYMBOL(save_mount_options); 886b3b304a2SMiklos Szeredi 8872a32cebdSAl Viro void replace_mount_options(struct super_block *sb, char *options) 8882a32cebdSAl Viro { 8892a32cebdSAl Viro char *old = sb->s_options; 8902a32cebdSAl Viro rcu_assign_pointer(sb->s_options, options); 8912a32cebdSAl Viro if (old) { 8922a32cebdSAl Viro synchronize_rcu(); 8932a32cebdSAl Viro kfree(old); 8942a32cebdSAl Viro } 8952a32cebdSAl Viro } 8962a32cebdSAl Viro EXPORT_SYMBOL(replace_mount_options); 8972a32cebdSAl Viro 898a1a2c409SMiklos Szeredi #ifdef CONFIG_PROC_FS 8991da177e4SLinus Torvalds /* iterator */ 9001da177e4SLinus Torvalds static void *m_start(struct seq_file *m, loff_t *pos) 9011da177e4SLinus Torvalds { 902a1a2c409SMiklos Szeredi struct proc_mounts *p = m->private; 9031da177e4SLinus Torvalds 904390c6843SRam Pai down_read(&namespace_sem); 905a1a2c409SMiklos Szeredi return seq_list_start(&p->ns->list, *pos); 9061da177e4SLinus Torvalds } 9071da177e4SLinus Torvalds 9081da177e4SLinus Torvalds static void *m_next(struct seq_file *m, void *v, loff_t *pos) 9091da177e4SLinus Torvalds { 910a1a2c409SMiklos Szeredi struct proc_mounts *p = m->private; 911b0765fb8SPavel Emelianov 912a1a2c409SMiklos Szeredi return seq_list_next(v, &p->ns->list, pos); 9131da177e4SLinus Torvalds } 9141da177e4SLinus Torvalds 9151da177e4SLinus Torvalds static void m_stop(struct seq_file *m, void *v) 9161da177e4SLinus Torvalds { 917390c6843SRam Pai up_read(&namespace_sem); 9181da177e4SLinus Torvalds } 9191da177e4SLinus Torvalds 9209f5596afSAl Viro int mnt_had_events(struct proc_mounts *p) 9219f5596afSAl Viro { 9229f5596afSAl Viro struct mnt_namespace *ns = p->ns; 9239f5596afSAl Viro int res = 0; 9249f5596afSAl Viro 92599b7db7bSNick Piggin br_read_lock(vfsmount_lock); 926f1514638SKay Sievers if (p->m.poll_event != ns->event) { 927f1514638SKay Sievers p->m.poll_event = ns->event; 9289f5596afSAl Viro res = 1; 9299f5596afSAl Viro } 93099b7db7bSNick Piggin br_read_unlock(vfsmount_lock); 9319f5596afSAl Viro 9329f5596afSAl Viro return res; 9339f5596afSAl Viro } 9349f5596afSAl Viro 9352d4d4864SRam Pai struct proc_fs_info { 9361da177e4SLinus Torvalds int flag; 9372d4d4864SRam Pai const char *str; 9382d4d4864SRam Pai }; 9392d4d4864SRam Pai 9402069f457SEric Paris static int show_sb_opts(struct seq_file *m, struct super_block *sb) 9412d4d4864SRam Pai { 9422d4d4864SRam Pai static const struct proc_fs_info fs_info[] = { 9431da177e4SLinus Torvalds { MS_SYNCHRONOUS, ",sync" }, 9441da177e4SLinus Torvalds { MS_DIRSYNC, ",dirsync" }, 9451da177e4SLinus Torvalds { MS_MANDLOCK, ",mand" }, 9461da177e4SLinus Torvalds { 0, NULL } 9471da177e4SLinus Torvalds }; 9482d4d4864SRam Pai const struct proc_fs_info *fs_infop; 9492d4d4864SRam Pai 9502d4d4864SRam Pai for (fs_infop = fs_info; fs_infop->flag; fs_infop++) { 9512d4d4864SRam Pai if (sb->s_flags & fs_infop->flag) 9522d4d4864SRam Pai seq_puts(m, fs_infop->str); 9532d4d4864SRam Pai } 9542069f457SEric Paris 9552069f457SEric Paris return security_sb_show_options(m, sb); 9562d4d4864SRam Pai } 9572d4d4864SRam Pai 9582d4d4864SRam Pai static void show_mnt_opts(struct seq_file *m, struct vfsmount *mnt) 9592d4d4864SRam Pai { 9602d4d4864SRam Pai static const struct proc_fs_info mnt_info[] = { 9611da177e4SLinus Torvalds { MNT_NOSUID, ",nosuid" }, 9621da177e4SLinus Torvalds { MNT_NODEV, ",nodev" }, 9631da177e4SLinus Torvalds { MNT_NOEXEC, ",noexec" }, 964fc33a7bbSChristoph Hellwig { MNT_NOATIME, ",noatime" }, 965fc33a7bbSChristoph Hellwig { MNT_NODIRATIME, ",nodiratime" }, 96647ae32d6SValerie Henson { MNT_RELATIME, ",relatime" }, 9671da177e4SLinus Torvalds { 0, NULL } 9681da177e4SLinus Torvalds }; 9692d4d4864SRam Pai const struct proc_fs_info *fs_infop; 9702d4d4864SRam Pai 9712d4d4864SRam Pai for (fs_infop = mnt_info; fs_infop->flag; fs_infop++) { 9722d4d4864SRam Pai if (mnt->mnt_flags & fs_infop->flag) 9732d4d4864SRam Pai seq_puts(m, fs_infop->str); 9742d4d4864SRam Pai } 9752d4d4864SRam Pai } 9762d4d4864SRam Pai 9772d4d4864SRam Pai static void show_type(struct seq_file *m, struct super_block *sb) 9782d4d4864SRam Pai { 9792d4d4864SRam Pai mangle(m, sb->s_type->name); 9802d4d4864SRam Pai if (sb->s_subtype && sb->s_subtype[0]) { 9812d4d4864SRam Pai seq_putc(m, '.'); 9822d4d4864SRam Pai mangle(m, sb->s_subtype); 9832d4d4864SRam Pai } 9842d4d4864SRam Pai } 9852d4d4864SRam Pai 9862d4d4864SRam Pai static int show_vfsmnt(struct seq_file *m, void *v) 9872d4d4864SRam Pai { 9882d4d4864SRam Pai struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); 9892d4d4864SRam Pai int err = 0; 990c32c2f63SJan Blunck struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; 9911da177e4SLinus Torvalds 992c7f404b4SAl Viro if (mnt->mnt_sb->s_op->show_devname) { 993c7f404b4SAl Viro err = mnt->mnt_sb->s_op->show_devname(m, mnt); 994c7f404b4SAl Viro if (err) 995c7f404b4SAl Viro goto out; 996c7f404b4SAl Viro } else { 9971da177e4SLinus Torvalds mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none"); 998c7f404b4SAl Viro } 9991da177e4SLinus Torvalds seq_putc(m, ' '); 1000c32c2f63SJan Blunck seq_path(m, &mnt_path, " \t\n\\"); 10011da177e4SLinus Torvalds seq_putc(m, ' '); 10022d4d4864SRam Pai show_type(m, mnt->mnt_sb); 10032e4b7fcdSDave Hansen seq_puts(m, __mnt_is_readonly(mnt) ? " ro" : " rw"); 10042069f457SEric Paris err = show_sb_opts(m, mnt->mnt_sb); 10052069f457SEric Paris if (err) 10062069f457SEric Paris goto out; 10072d4d4864SRam Pai show_mnt_opts(m, mnt); 10081da177e4SLinus Torvalds if (mnt->mnt_sb->s_op->show_options) 10091da177e4SLinus Torvalds err = mnt->mnt_sb->s_op->show_options(m, mnt); 10101da177e4SLinus Torvalds seq_puts(m, " 0 0\n"); 10112069f457SEric Paris out: 10121da177e4SLinus Torvalds return err; 10131da177e4SLinus Torvalds } 10141da177e4SLinus Torvalds 1015a1a2c409SMiklos Szeredi const struct seq_operations mounts_op = { 10161da177e4SLinus Torvalds .start = m_start, 10171da177e4SLinus Torvalds .next = m_next, 10181da177e4SLinus Torvalds .stop = m_stop, 10191da177e4SLinus Torvalds .show = show_vfsmnt 10201da177e4SLinus Torvalds }; 10211da177e4SLinus Torvalds 10222d4d4864SRam Pai static int show_mountinfo(struct seq_file *m, void *v) 10232d4d4864SRam Pai { 10242d4d4864SRam Pai struct proc_mounts *p = m->private; 10252d4d4864SRam Pai struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); 10263376f34fSAl Viro struct mount *r = real_mount(mnt); 10272d4d4864SRam Pai struct super_block *sb = mnt->mnt_sb; 10282d4d4864SRam Pai struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; 10292d4d4864SRam Pai struct path root = p->root; 10302d4d4864SRam Pai int err = 0; 10312d4d4864SRam Pai 10320714a533SAl Viro seq_printf(m, "%i %i %u:%u ", mnt->mnt_id, r->mnt_parent->mnt.mnt_id, 10332d4d4864SRam Pai MAJOR(sb->s_dev), MINOR(sb->s_dev)); 1034c7f404b4SAl Viro if (sb->s_op->show_path) 1035c7f404b4SAl Viro err = sb->s_op->show_path(m, mnt); 1036c7f404b4SAl Viro else 10372d4d4864SRam Pai seq_dentry(m, mnt->mnt_root, " \t\n\\"); 1038c7f404b4SAl Viro if (err) 1039c7f404b4SAl Viro goto out; 10402d4d4864SRam Pai seq_putc(m, ' '); 104102125a82SAl Viro 104202125a82SAl Viro /* mountpoints outside of chroot jail will give SEQ_SKIP on this */ 104302125a82SAl Viro err = seq_path_root(m, &mnt_path, &root, " \t\n\\"); 104402125a82SAl Viro if (err) 104502125a82SAl Viro goto out; 104602125a82SAl Viro 10472d4d4864SRam Pai seq_puts(m, mnt->mnt_flags & MNT_READONLY ? " ro" : " rw"); 10482d4d4864SRam Pai show_mnt_opts(m, mnt); 10492d4d4864SRam Pai 10502d4d4864SRam Pai /* Tagged fields ("foo:X" or "bar") */ 10512d4d4864SRam Pai if (IS_MNT_SHARED(mnt)) 10522d4d4864SRam Pai seq_printf(m, " shared:%i", mnt->mnt_group_id); 1053d10e8defSAl Viro if (IS_MNT_SLAVE(r)) { 105432301920SAl Viro int master = r->mnt_master->mnt.mnt_group_id; 10556fc7871fSAl Viro int dom = get_dominating_id(r, &p->root); 105697e7e0f7SMiklos Szeredi seq_printf(m, " master:%i", master); 105797e7e0f7SMiklos Szeredi if (dom && dom != master) 105897e7e0f7SMiklos Szeredi seq_printf(m, " propagate_from:%i", dom); 105997e7e0f7SMiklos Szeredi } 10602d4d4864SRam Pai if (IS_MNT_UNBINDABLE(mnt)) 10612d4d4864SRam Pai seq_puts(m, " unbindable"); 10622d4d4864SRam Pai 10632d4d4864SRam Pai /* Filesystem specific data */ 10642d4d4864SRam Pai seq_puts(m, " - "); 10652d4d4864SRam Pai show_type(m, sb); 10662d4d4864SRam Pai seq_putc(m, ' '); 1067c7f404b4SAl Viro if (sb->s_op->show_devname) 1068c7f404b4SAl Viro err = sb->s_op->show_devname(m, mnt); 1069c7f404b4SAl Viro else 10702d4d4864SRam Pai mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none"); 1071c7f404b4SAl Viro if (err) 1072c7f404b4SAl Viro goto out; 10732d4d4864SRam Pai seq_puts(m, sb->s_flags & MS_RDONLY ? " ro" : " rw"); 10742069f457SEric Paris err = show_sb_opts(m, sb); 10752069f457SEric Paris if (err) 10762069f457SEric Paris goto out; 10772d4d4864SRam Pai if (sb->s_op->show_options) 10782d4d4864SRam Pai err = sb->s_op->show_options(m, mnt); 10792d4d4864SRam Pai seq_putc(m, '\n'); 10802069f457SEric Paris out: 10812d4d4864SRam Pai return err; 10822d4d4864SRam Pai } 10832d4d4864SRam Pai 10842d4d4864SRam Pai const struct seq_operations mountinfo_op = { 10852d4d4864SRam Pai .start = m_start, 10862d4d4864SRam Pai .next = m_next, 10872d4d4864SRam Pai .stop = m_stop, 10882d4d4864SRam Pai .show = show_mountinfo, 10892d4d4864SRam Pai }; 10902d4d4864SRam Pai 1091b4629fe2SChuck Lever static int show_vfsstat(struct seq_file *m, void *v) 1092b4629fe2SChuck Lever { 1093b0765fb8SPavel Emelianov struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); 1094c32c2f63SJan Blunck struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; 1095b4629fe2SChuck Lever int err = 0; 1096b4629fe2SChuck Lever 1097b4629fe2SChuck Lever /* device */ 1098c7f404b4SAl Viro if (mnt->mnt_sb->s_op->show_devname) { 1099a877ee03SBryan Schumaker seq_puts(m, "device "); 1100c7f404b4SAl Viro err = mnt->mnt_sb->s_op->show_devname(m, mnt); 1101c7f404b4SAl Viro } else { 1102b4629fe2SChuck Lever if (mnt->mnt_devname) { 1103b4629fe2SChuck Lever seq_puts(m, "device "); 1104b4629fe2SChuck Lever mangle(m, mnt->mnt_devname); 1105b4629fe2SChuck Lever } else 1106b4629fe2SChuck Lever seq_puts(m, "no device"); 1107c7f404b4SAl Viro } 1108b4629fe2SChuck Lever 1109b4629fe2SChuck Lever /* mount point */ 1110b4629fe2SChuck Lever seq_puts(m, " mounted on "); 1111c32c2f63SJan Blunck seq_path(m, &mnt_path, " \t\n\\"); 1112b4629fe2SChuck Lever seq_putc(m, ' '); 1113b4629fe2SChuck Lever 1114b4629fe2SChuck Lever /* file system type */ 1115b4629fe2SChuck Lever seq_puts(m, "with fstype "); 11162d4d4864SRam Pai show_type(m, mnt->mnt_sb); 1117b4629fe2SChuck Lever 1118b4629fe2SChuck Lever /* optional statistics */ 1119b4629fe2SChuck Lever if (mnt->mnt_sb->s_op->show_stats) { 1120b4629fe2SChuck Lever seq_putc(m, ' '); 1121c7f404b4SAl Viro if (!err) 1122b4629fe2SChuck Lever err = mnt->mnt_sb->s_op->show_stats(m, mnt); 1123b4629fe2SChuck Lever } 1124b4629fe2SChuck Lever 1125b4629fe2SChuck Lever seq_putc(m, '\n'); 1126b4629fe2SChuck Lever return err; 1127b4629fe2SChuck Lever } 1128b4629fe2SChuck Lever 1129a1a2c409SMiklos Szeredi const struct seq_operations mountstats_op = { 1130b4629fe2SChuck Lever .start = m_start, 1131b4629fe2SChuck Lever .next = m_next, 1132b4629fe2SChuck Lever .stop = m_stop, 1133b4629fe2SChuck Lever .show = show_vfsstat, 1134b4629fe2SChuck Lever }; 1135a1a2c409SMiklos Szeredi #endif /* CONFIG_PROC_FS */ 1136b4629fe2SChuck Lever 11371da177e4SLinus Torvalds /** 11381da177e4SLinus Torvalds * may_umount_tree - check if a mount tree is busy 11391da177e4SLinus Torvalds * @mnt: root of mount tree 11401da177e4SLinus Torvalds * 11411da177e4SLinus Torvalds * This is called to check if a tree of mounts has any 11421da177e4SLinus Torvalds * open files, pwds, chroots or sub mounts that are 11431da177e4SLinus Torvalds * busy. 11441da177e4SLinus Torvalds */ 11451da177e4SLinus Torvalds int may_umount_tree(struct vfsmount *mnt) 11461da177e4SLinus Torvalds { 114736341f64SRam Pai int actual_refs = 0; 114836341f64SRam Pai int minimum_refs = 0; 1149315fc83eSAl Viro struct mount *p; 1150315fc83eSAl Viro BUG_ON(!mnt); 11511da177e4SLinus Torvalds 1152b3e19d92SNick Piggin /* write lock needed for mnt_get_count */ 1153b3e19d92SNick Piggin br_write_lock(vfsmount_lock); 1154315fc83eSAl Viro for (p = real_mount(mnt); p; p = next_mnt(p, mnt)) { 115583adc753SAl Viro actual_refs += mnt_get_count(p); 11561da177e4SLinus Torvalds minimum_refs += 2; 11571da177e4SLinus Torvalds } 1158b3e19d92SNick Piggin br_write_unlock(vfsmount_lock); 11591da177e4SLinus Torvalds 11601da177e4SLinus Torvalds if (actual_refs > minimum_refs) 11611da177e4SLinus Torvalds return 0; 1162e3474a8eSIan Kent 1163e3474a8eSIan Kent return 1; 11641da177e4SLinus Torvalds } 11651da177e4SLinus Torvalds 11661da177e4SLinus Torvalds EXPORT_SYMBOL(may_umount_tree); 11671da177e4SLinus Torvalds 11681da177e4SLinus Torvalds /** 11691da177e4SLinus Torvalds * may_umount - check if a mount point is busy 11701da177e4SLinus Torvalds * @mnt: root of mount 11711da177e4SLinus Torvalds * 11721da177e4SLinus Torvalds * This is called to check if a mount point has any 11731da177e4SLinus Torvalds * open files, pwds, chroots or sub mounts. If the 11741da177e4SLinus Torvalds * mount has sub mounts this will return busy 11751da177e4SLinus Torvalds * regardless of whether the sub mounts are busy. 11761da177e4SLinus Torvalds * 11771da177e4SLinus Torvalds * Doesn't take quota and stuff into account. IOW, in some cases it will 11781da177e4SLinus Torvalds * give false negatives. The main reason why it's here is that we need 11791da177e4SLinus Torvalds * a non-destructive way to look for easily umountable filesystems. 11801da177e4SLinus Torvalds */ 11811da177e4SLinus Torvalds int may_umount(struct vfsmount *mnt) 11821da177e4SLinus Torvalds { 1183e3474a8eSIan Kent int ret = 1; 11848ad08d8aSAl Viro down_read(&namespace_sem); 1185b3e19d92SNick Piggin br_write_lock(vfsmount_lock); 11861ab59738SAl Viro if (propagate_mount_busy(real_mount(mnt), 2)) 1187e3474a8eSIan Kent ret = 0; 1188b3e19d92SNick Piggin br_write_unlock(vfsmount_lock); 11898ad08d8aSAl Viro up_read(&namespace_sem); 1190a05964f3SRam Pai return ret; 11911da177e4SLinus Torvalds } 11921da177e4SLinus Torvalds 11931da177e4SLinus Torvalds EXPORT_SYMBOL(may_umount); 11941da177e4SLinus Torvalds 1195b90fa9aeSRam Pai void release_mounts(struct list_head *head) 11961da177e4SLinus Torvalds { 1197d5e50f74SAl Viro struct mount *mnt; 119870fbcdf4SRam Pai while (!list_empty(head)) { 11991b8e5564SAl Viro mnt = list_first_entry(head, struct mount, mnt_hash); 12001b8e5564SAl Viro list_del_init(&mnt->mnt_hash); 1201676da58dSAl Viro if (mnt_has_parent(mnt)) { 120270fbcdf4SRam Pai struct dentry *dentry; 120370fbcdf4SRam Pai struct vfsmount *m; 120499b7db7bSNick Piggin 120599b7db7bSNick Piggin br_write_lock(vfsmount_lock); 1206a73324daSAl Viro dentry = mnt->mnt_mountpoint; 12070714a533SAl Viro m = &mnt->mnt_parent->mnt; 1208a73324daSAl Viro mnt->mnt_mountpoint = mnt->mnt.mnt_root; 12090714a533SAl Viro mnt->mnt_parent = mnt; 12107c4b93d8SAl Viro m->mnt_ghosts--; 121199b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 121270fbcdf4SRam Pai dput(dentry); 121370fbcdf4SRam Pai mntput(m); 12141da177e4SLinus Torvalds } 1215d5e50f74SAl Viro mntput(&mnt->mnt); 121670fbcdf4SRam Pai } 121770fbcdf4SRam Pai } 121870fbcdf4SRam Pai 121999b7db7bSNick Piggin /* 122099b7db7bSNick Piggin * vfsmount lock must be held for write 122199b7db7bSNick Piggin * namespace_sem must be held for write 122299b7db7bSNick Piggin */ 1223761d5c38SAl Viro void umount_tree(struct mount *mnt, int propagate, struct list_head *kill) 122470fbcdf4SRam Pai { 12257b8a53fdSAl Viro LIST_HEAD(tmp_list); 1226315fc83eSAl Viro struct mount *p; 122770fbcdf4SRam Pai 1228761d5c38SAl Viro for (p = mnt; p; p = next_mnt(p, &mnt->mnt)) 12291b8e5564SAl Viro list_move(&p->mnt_hash, &tmp_list); 123070fbcdf4SRam Pai 1231a05964f3SRam Pai if (propagate) 12327b8a53fdSAl Viro propagate_umount(&tmp_list); 1233a05964f3SRam Pai 12341b8e5564SAl Viro list_for_each_entry(p, &tmp_list, mnt_hash) { 12356776db3dSAl Viro list_del_init(&p->mnt_expire); 1236315fc83eSAl Viro list_del_init(&p->mnt.mnt_list); 1237143c8c91SAl Viro __touch_mnt_namespace(p->mnt_ns); 1238143c8c91SAl Viro p->mnt_ns = NULL; 123983adc753SAl Viro __mnt_make_shortterm(p); 12406b41d536SAl Viro list_del_init(&p->mnt_child); 1241676da58dSAl Viro if (mnt_has_parent(p)) { 12420714a533SAl Viro p->mnt_parent->mnt.mnt_ghosts++; 1243a73324daSAl Viro dentry_reset_mounted(p->mnt_mountpoint); 12447c4b93d8SAl Viro } 12450f0afb1dSAl Viro change_mnt_propagation(p, MS_PRIVATE); 12461da177e4SLinus Torvalds } 12477b8a53fdSAl Viro list_splice(&tmp_list, kill); 12481da177e4SLinus Torvalds } 12491da177e4SLinus Torvalds 1250692afc31SAl Viro static void shrink_submounts(struct mount *mnt, struct list_head *umounts); 1251c35038beSAl Viro 12521ab59738SAl Viro static int do_umount(struct mount *mnt, int flags) 12531da177e4SLinus Torvalds { 12541ab59738SAl Viro struct super_block *sb = mnt->mnt.mnt_sb; 12551da177e4SLinus Torvalds int retval; 125670fbcdf4SRam Pai LIST_HEAD(umount_list); 12571da177e4SLinus Torvalds 12581ab59738SAl Viro retval = security_sb_umount(&mnt->mnt, flags); 12591da177e4SLinus Torvalds if (retval) 12601da177e4SLinus Torvalds return retval; 12611da177e4SLinus Torvalds 12621da177e4SLinus Torvalds /* 12631da177e4SLinus Torvalds * Allow userspace to request a mountpoint be expired rather than 12641da177e4SLinus Torvalds * unmounting unconditionally. Unmount only happens if: 12651da177e4SLinus Torvalds * (1) the mark is already set (the mark is cleared by mntput()) 12661da177e4SLinus Torvalds * (2) the usage count == 1 [parent vfsmount] + 1 [sys_umount] 12671da177e4SLinus Torvalds */ 12681da177e4SLinus Torvalds if (flags & MNT_EXPIRE) { 12691ab59738SAl Viro if (&mnt->mnt == current->fs->root.mnt || 12701da177e4SLinus Torvalds flags & (MNT_FORCE | MNT_DETACH)) 12711da177e4SLinus Torvalds return -EINVAL; 12721da177e4SLinus Torvalds 1273b3e19d92SNick Piggin /* 1274b3e19d92SNick Piggin * probably don't strictly need the lock here if we examined 1275b3e19d92SNick Piggin * all race cases, but it's a slowpath. 1276b3e19d92SNick Piggin */ 1277b3e19d92SNick Piggin br_write_lock(vfsmount_lock); 127883adc753SAl Viro if (mnt_get_count(mnt) != 2) { 1279bf9faa2aSJ. R. Okajima br_write_unlock(vfsmount_lock); 12801da177e4SLinus Torvalds return -EBUSY; 1281b3e19d92SNick Piggin } 1282b3e19d92SNick Piggin br_write_unlock(vfsmount_lock); 12831da177e4SLinus Torvalds 12841ab59738SAl Viro if (!xchg(&mnt->mnt.mnt_expiry_mark, 1)) 12851da177e4SLinus Torvalds return -EAGAIN; 12861da177e4SLinus Torvalds } 12871da177e4SLinus Torvalds 12881da177e4SLinus Torvalds /* 12891da177e4SLinus Torvalds * If we may have to abort operations to get out of this 12901da177e4SLinus Torvalds * mount, and they will themselves hold resources we must 12911da177e4SLinus Torvalds * allow the fs to do things. In the Unix tradition of 12921da177e4SLinus Torvalds * 'Gee thats tricky lets do it in userspace' the umount_begin 12931da177e4SLinus Torvalds * might fail to complete on the first run through as other tasks 12941da177e4SLinus Torvalds * must return, and the like. Thats for the mount program to worry 12951da177e4SLinus Torvalds * about for the moment. 12961da177e4SLinus Torvalds */ 12971da177e4SLinus Torvalds 129842faad99SAl Viro if (flags & MNT_FORCE && sb->s_op->umount_begin) { 129942faad99SAl Viro sb->s_op->umount_begin(sb); 130042faad99SAl Viro } 13011da177e4SLinus Torvalds 13021da177e4SLinus Torvalds /* 13031da177e4SLinus Torvalds * No sense to grab the lock for this test, but test itself looks 13041da177e4SLinus Torvalds * somewhat bogus. Suggestions for better replacement? 13051da177e4SLinus Torvalds * Ho-hum... In principle, we might treat that as umount + switch 13061da177e4SLinus Torvalds * to rootfs. GC would eventually take care of the old vfsmount. 13071da177e4SLinus Torvalds * Actually it makes sense, especially if rootfs would contain a 13081da177e4SLinus Torvalds * /reboot - static binary that would close all descriptors and 13091da177e4SLinus Torvalds * call reboot(9). Then init(8) could umount root and exec /reboot. 13101da177e4SLinus Torvalds */ 13111ab59738SAl Viro if (&mnt->mnt == current->fs->root.mnt && !(flags & MNT_DETACH)) { 13121da177e4SLinus Torvalds /* 13131da177e4SLinus Torvalds * Special case for "unmounting" root ... 13141da177e4SLinus Torvalds * we just try to remount it readonly. 13151da177e4SLinus Torvalds */ 13161da177e4SLinus Torvalds down_write(&sb->s_umount); 13174aa98cf7SAl Viro if (!(sb->s_flags & MS_RDONLY)) 13181da177e4SLinus Torvalds retval = do_remount_sb(sb, MS_RDONLY, NULL, 0); 13191da177e4SLinus Torvalds up_write(&sb->s_umount); 13201da177e4SLinus Torvalds return retval; 13211da177e4SLinus Torvalds } 13221da177e4SLinus Torvalds 1323390c6843SRam Pai down_write(&namespace_sem); 132499b7db7bSNick Piggin br_write_lock(vfsmount_lock); 13255addc5ddSAl Viro event++; 13261da177e4SLinus Torvalds 1327c35038beSAl Viro if (!(flags & MNT_DETACH)) 13281ab59738SAl Viro shrink_submounts(mnt, &umount_list); 1329c35038beSAl Viro 13301da177e4SLinus Torvalds retval = -EBUSY; 1331a05964f3SRam Pai if (flags & MNT_DETACH || !propagate_mount_busy(mnt, 2)) { 13321ab59738SAl Viro if (!list_empty(&mnt->mnt.mnt_list)) 13331ab59738SAl Viro umount_tree(mnt, 1, &umount_list); 13341da177e4SLinus Torvalds retval = 0; 13351da177e4SLinus Torvalds } 133699b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 1337390c6843SRam Pai up_write(&namespace_sem); 133870fbcdf4SRam Pai release_mounts(&umount_list); 13391da177e4SLinus Torvalds return retval; 13401da177e4SLinus Torvalds } 13411da177e4SLinus Torvalds 13421da177e4SLinus Torvalds /* 13431da177e4SLinus Torvalds * Now umount can handle mount points as well as block devices. 13441da177e4SLinus Torvalds * This is important for filesystems which use unnamed block devices. 13451da177e4SLinus Torvalds * 13461da177e4SLinus Torvalds * We now support a flag for forced unmount like the other 'big iron' 13471da177e4SLinus Torvalds * unixes. Our API is identical to OSF/1 to avoid making a mess of AMD 13481da177e4SLinus Torvalds */ 13491da177e4SLinus Torvalds 1350bdc480e3SHeiko Carstens SYSCALL_DEFINE2(umount, char __user *, name, int, flags) 13511da177e4SLinus Torvalds { 13522d8f3038SAl Viro struct path path; 1353900148dcSAl Viro struct mount *mnt; 13541da177e4SLinus Torvalds int retval; 1355db1f05bbSMiklos Szeredi int lookup_flags = 0; 13561da177e4SLinus Torvalds 1357db1f05bbSMiklos Szeredi if (flags & ~(MNT_FORCE | MNT_DETACH | MNT_EXPIRE | UMOUNT_NOFOLLOW)) 1358db1f05bbSMiklos Szeredi return -EINVAL; 1359db1f05bbSMiklos Szeredi 1360db1f05bbSMiklos Szeredi if (!(flags & UMOUNT_NOFOLLOW)) 1361db1f05bbSMiklos Szeredi lookup_flags |= LOOKUP_FOLLOW; 1362db1f05bbSMiklos Szeredi 1363db1f05bbSMiklos Szeredi retval = user_path_at(AT_FDCWD, name, lookup_flags, &path); 13641da177e4SLinus Torvalds if (retval) 13651da177e4SLinus Torvalds goto out; 1366900148dcSAl Viro mnt = real_mount(path.mnt); 13671da177e4SLinus Torvalds retval = -EINVAL; 13682d8f3038SAl Viro if (path.dentry != path.mnt->mnt_root) 13691da177e4SLinus Torvalds goto dput_and_out; 1370143c8c91SAl Viro if (!check_mnt(mnt)) 13711da177e4SLinus Torvalds goto dput_and_out; 13721da177e4SLinus Torvalds 13731da177e4SLinus Torvalds retval = -EPERM; 13741da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 13751da177e4SLinus Torvalds goto dput_and_out; 13761da177e4SLinus Torvalds 1377900148dcSAl Viro retval = do_umount(mnt, flags); 13781da177e4SLinus Torvalds dput_and_out: 1379429731b1SJan Blunck /* we mustn't call path_put() as that would clear mnt_expiry_mark */ 13802d8f3038SAl Viro dput(path.dentry); 1381900148dcSAl Viro mntput_no_expire(mnt); 13821da177e4SLinus Torvalds out: 13831da177e4SLinus Torvalds return retval; 13841da177e4SLinus Torvalds } 13851da177e4SLinus Torvalds 13861da177e4SLinus Torvalds #ifdef __ARCH_WANT_SYS_OLDUMOUNT 13871da177e4SLinus Torvalds 13881da177e4SLinus Torvalds /* 13891da177e4SLinus Torvalds * The 2.0 compatible umount. No flags. 13901da177e4SLinus Torvalds */ 1391bdc480e3SHeiko Carstens SYSCALL_DEFINE1(oldumount, char __user *, name) 13921da177e4SLinus Torvalds { 13931da177e4SLinus Torvalds return sys_umount(name, 0); 13941da177e4SLinus Torvalds } 13951da177e4SLinus Torvalds 13961da177e4SLinus Torvalds #endif 13971da177e4SLinus Torvalds 13982d92ab3cSAl Viro static int mount_is_safe(struct path *path) 13991da177e4SLinus Torvalds { 14001da177e4SLinus Torvalds if (capable(CAP_SYS_ADMIN)) 14011da177e4SLinus Torvalds return 0; 14021da177e4SLinus Torvalds return -EPERM; 14031da177e4SLinus Torvalds #ifdef notyet 14042d92ab3cSAl Viro if (S_ISLNK(path->dentry->d_inode->i_mode)) 14051da177e4SLinus Torvalds return -EPERM; 14062d92ab3cSAl Viro if (path->dentry->d_inode->i_mode & S_ISVTX) { 1407da9592edSDavid Howells if (current_uid() != path->dentry->d_inode->i_uid) 14081da177e4SLinus Torvalds return -EPERM; 14091da177e4SLinus Torvalds } 14102d92ab3cSAl Viro if (inode_permission(path->dentry->d_inode, MAY_WRITE)) 14111da177e4SLinus Torvalds return -EPERM; 14121da177e4SLinus Torvalds return 0; 14131da177e4SLinus Torvalds #endif 14141da177e4SLinus Torvalds } 14151da177e4SLinus Torvalds 141687129cc0SAl Viro struct mount *copy_tree(struct mount *mnt, struct dentry *dentry, 141736341f64SRam Pai int flag) 14181da177e4SLinus Torvalds { 1419a73324daSAl Viro struct mount *res, *p, *q, *r; 14201a390689SAl Viro struct path path; 14211da177e4SLinus Torvalds 142287129cc0SAl Viro if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(&mnt->mnt)) 14239676f0c6SRam Pai return NULL; 14249676f0c6SRam Pai 142536341f64SRam Pai res = q = clone_mnt(mnt, dentry, flag); 14261da177e4SLinus Torvalds if (!q) 14271da177e4SLinus Torvalds goto Enomem; 1428a73324daSAl Viro q->mnt_mountpoint = mnt->mnt_mountpoint; 14291da177e4SLinus Torvalds 14301da177e4SLinus Torvalds p = mnt; 14316b41d536SAl Viro list_for_each_entry(r, &mnt->mnt_mounts, mnt_child) { 1432315fc83eSAl Viro struct mount *s; 14337ec02ef1SJan Blunck if (!is_subdir(r->mnt_mountpoint, dentry)) 14341da177e4SLinus Torvalds continue; 14351da177e4SLinus Torvalds 1436a73324daSAl Viro for (s = r; s; s = next_mnt(s, &r->mnt)) { 1437315fc83eSAl Viro if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(&s->mnt)) { 14389676f0c6SRam Pai s = skip_mnt_tree(s); 14399676f0c6SRam Pai continue; 14409676f0c6SRam Pai } 14410714a533SAl Viro while (p != s->mnt_parent) { 14420714a533SAl Viro p = p->mnt_parent; 14430714a533SAl Viro q = q->mnt_parent; 14441da177e4SLinus Torvalds } 144587129cc0SAl Viro p = s; 1446cb338d06SAl Viro path.mnt = &q->mnt; 1447a73324daSAl Viro path.dentry = p->mnt_mountpoint; 144887129cc0SAl Viro q = clone_mnt(p, p->mnt.mnt_root, flag); 14491da177e4SLinus Torvalds if (!q) 14501da177e4SLinus Torvalds goto Enomem; 145199b7db7bSNick Piggin br_write_lock(vfsmount_lock); 1452cb338d06SAl Viro list_add_tail(&q->mnt.mnt_list, &res->mnt.mnt_list); 1453cb338d06SAl Viro attach_mnt(q, &path); 145499b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 14551da177e4SLinus Torvalds } 14561da177e4SLinus Torvalds } 14571da177e4SLinus Torvalds return res; 14581da177e4SLinus Torvalds Enomem: 14591da177e4SLinus Torvalds if (res) { 146070fbcdf4SRam Pai LIST_HEAD(umount_list); 146199b7db7bSNick Piggin br_write_lock(vfsmount_lock); 1462761d5c38SAl Viro umount_tree(res, 0, &umount_list); 146399b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 146470fbcdf4SRam Pai release_mounts(&umount_list); 14651da177e4SLinus Torvalds } 14661da177e4SLinus Torvalds return NULL; 14671da177e4SLinus Torvalds } 14681da177e4SLinus Torvalds 1469589ff870SAl Viro struct vfsmount *collect_mounts(struct path *path) 14708aec0809SAl Viro { 1471cb338d06SAl Viro struct mount *tree; 14721a60a280SAl Viro down_write(&namespace_sem); 147387129cc0SAl Viro tree = copy_tree(real_mount(path->mnt), path->dentry, 147487129cc0SAl Viro CL_COPY_ALL | CL_PRIVATE); 14751a60a280SAl Viro up_write(&namespace_sem); 1476cb338d06SAl Viro return tree ? &tree->mnt : NULL; 14778aec0809SAl Viro } 14788aec0809SAl Viro 14798aec0809SAl Viro void drop_collected_mounts(struct vfsmount *mnt) 14808aec0809SAl Viro { 14818aec0809SAl Viro LIST_HEAD(umount_list); 14821a60a280SAl Viro down_write(&namespace_sem); 148399b7db7bSNick Piggin br_write_lock(vfsmount_lock); 1484761d5c38SAl Viro umount_tree(real_mount(mnt), 0, &umount_list); 148599b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 14861a60a280SAl Viro up_write(&namespace_sem); 14878aec0809SAl Viro release_mounts(&umount_list); 14888aec0809SAl Viro } 14898aec0809SAl Viro 14901f707137SAl Viro int iterate_mounts(int (*f)(struct vfsmount *, void *), void *arg, 14911f707137SAl Viro struct vfsmount *root) 14921f707137SAl Viro { 14931f707137SAl Viro struct vfsmount *mnt; 14941f707137SAl Viro int res = f(root, arg); 14951f707137SAl Viro if (res) 14961f707137SAl Viro return res; 14971f707137SAl Viro list_for_each_entry(mnt, &root->mnt_list, mnt_list) { 14981f707137SAl Viro res = f(mnt, arg); 14991f707137SAl Viro if (res) 15001f707137SAl Viro return res; 15011f707137SAl Viro } 15021f707137SAl Viro return 0; 15031f707137SAl Viro } 15041f707137SAl Viro 15054b8b21f4SAl Viro static void cleanup_group_ids(struct mount *mnt, struct mount *end) 1506719f5d7fSMiklos Szeredi { 1507315fc83eSAl Viro struct mount *p; 1508719f5d7fSMiklos Szeredi 15094b8b21f4SAl Viro for (p = mnt; p != end; p = next_mnt(p, &mnt->mnt)) { 1510315fc83eSAl Viro if (p->mnt.mnt_group_id && !IS_MNT_SHARED(&p->mnt)) 15114b8b21f4SAl Viro mnt_release_group_id(p); 1512719f5d7fSMiklos Szeredi } 1513719f5d7fSMiklos Szeredi } 1514719f5d7fSMiklos Szeredi 15154b8b21f4SAl Viro static int invent_group_ids(struct mount *mnt, bool recurse) 1516719f5d7fSMiklos Szeredi { 1517315fc83eSAl Viro struct mount *p; 1518719f5d7fSMiklos Szeredi 15194b8b21f4SAl Viro for (p = mnt; p; p = recurse ? next_mnt(p, &mnt->mnt) : NULL) { 1520315fc83eSAl Viro if (!p->mnt.mnt_group_id && !IS_MNT_SHARED(&p->mnt)) { 15214b8b21f4SAl Viro int err = mnt_alloc_group_id(p); 1522719f5d7fSMiklos Szeredi if (err) { 15234b8b21f4SAl Viro cleanup_group_ids(mnt, p); 1524719f5d7fSMiklos Szeredi return err; 1525719f5d7fSMiklos Szeredi } 1526719f5d7fSMiklos Szeredi } 1527719f5d7fSMiklos Szeredi } 1528719f5d7fSMiklos Szeredi 1529719f5d7fSMiklos Szeredi return 0; 1530719f5d7fSMiklos Szeredi } 1531719f5d7fSMiklos Szeredi 1532b90fa9aeSRam Pai /* 1533b90fa9aeSRam Pai * @source_mnt : mount tree to be attached 1534b90fa9aeSRam Pai * @nd : place the mount tree @source_mnt is attached 153521444403SRam Pai * @parent_nd : if non-null, detach the source_mnt from its parent and 153621444403SRam Pai * store the parent mount and mountpoint dentry. 153721444403SRam Pai * (done when source_mnt is moved) 1538b90fa9aeSRam Pai * 1539b90fa9aeSRam Pai * NOTE: in the table below explains the semantics when a source mount 1540b90fa9aeSRam Pai * of a given type is attached to a destination mount of a given type. 15419676f0c6SRam Pai * --------------------------------------------------------------------------- 1542b90fa9aeSRam Pai * | BIND MOUNT OPERATION | 15439676f0c6SRam Pai * |************************************************************************** 15449676f0c6SRam Pai * | source-->| shared | private | slave | unbindable | 15459676f0c6SRam Pai * | dest | | | | | 15469676f0c6SRam Pai * | | | | | | | 15479676f0c6SRam Pai * | v | | | | | 15489676f0c6SRam Pai * |************************************************************************** 15499676f0c6SRam Pai * | shared | shared (++) | shared (+) | shared(+++)| invalid | 15505afe0022SRam Pai * | | | | | | 15519676f0c6SRam Pai * |non-shared| shared (+) | private | slave (*) | invalid | 15529676f0c6SRam Pai * *************************************************************************** 1553b90fa9aeSRam Pai * A bind operation clones the source mount and mounts the clone on the 1554b90fa9aeSRam Pai * destination mount. 1555b90fa9aeSRam Pai * 1556b90fa9aeSRam Pai * (++) the cloned mount is propagated to all the mounts in the propagation 1557b90fa9aeSRam Pai * tree of the destination mount and the cloned mount is added to 1558b90fa9aeSRam Pai * the peer group of the source mount. 1559b90fa9aeSRam Pai * (+) the cloned mount is created under the destination mount and is marked 1560b90fa9aeSRam Pai * as shared. The cloned mount is added to the peer group of the source 1561b90fa9aeSRam Pai * mount. 15625afe0022SRam Pai * (+++) the mount is propagated to all the mounts in the propagation tree 15635afe0022SRam Pai * of the destination mount and the cloned mount is made slave 15645afe0022SRam Pai * of the same master as that of the source mount. The cloned mount 15655afe0022SRam Pai * is marked as 'shared and slave'. 15665afe0022SRam Pai * (*) the cloned mount is made a slave of the same master as that of the 15675afe0022SRam Pai * source mount. 15685afe0022SRam Pai * 15699676f0c6SRam Pai * --------------------------------------------------------------------------- 157021444403SRam Pai * | MOVE MOUNT OPERATION | 15719676f0c6SRam Pai * |************************************************************************** 15729676f0c6SRam Pai * | source-->| shared | private | slave | unbindable | 15739676f0c6SRam Pai * | dest | | | | | 15749676f0c6SRam Pai * | | | | | | | 15759676f0c6SRam Pai * | v | | | | | 15769676f0c6SRam Pai * |************************************************************************** 15779676f0c6SRam Pai * | shared | shared (+) | shared (+) | shared(+++) | invalid | 15785afe0022SRam Pai * | | | | | | 15799676f0c6SRam Pai * |non-shared| shared (+*) | private | slave (*) | unbindable | 15809676f0c6SRam Pai * *************************************************************************** 15815afe0022SRam Pai * 15825afe0022SRam Pai * (+) the mount is moved to the destination. And is then propagated to 15835afe0022SRam Pai * all the mounts in the propagation tree of the destination mount. 158421444403SRam Pai * (+*) the mount is moved to the destination. 15855afe0022SRam Pai * (+++) the mount is moved to the destination and is then propagated to 15865afe0022SRam Pai * all the mounts belonging to the destination mount's propagation tree. 15875afe0022SRam Pai * the mount is marked as 'shared and slave'. 15885afe0022SRam Pai * (*) the mount continues to be a slave at the new location. 1589b90fa9aeSRam Pai * 1590b90fa9aeSRam Pai * if the source mount is a tree, the operations explained above is 1591b90fa9aeSRam Pai * applied to each mount in the tree. 1592b90fa9aeSRam Pai * Must be called without spinlocks held, since this function can sleep 1593b90fa9aeSRam Pai * in allocations. 1594b90fa9aeSRam Pai */ 15950fb54e50SAl Viro static int attach_recursive_mnt(struct mount *source_mnt, 15961a390689SAl Viro struct path *path, struct path *parent_path) 1597b90fa9aeSRam Pai { 1598b90fa9aeSRam Pai LIST_HEAD(tree_list); 1599a8d56d8eSAl Viro struct mount *dest_mnt = real_mount(path->mnt); 16001a390689SAl Viro struct dentry *dest_dentry = path->dentry; 1601315fc83eSAl Viro struct mount *child, *p; 1602719f5d7fSMiklos Szeredi int err; 1603b90fa9aeSRam Pai 1604a8d56d8eSAl Viro if (IS_MNT_SHARED(&dest_mnt->mnt)) { 16050fb54e50SAl Viro err = invent_group_ids(source_mnt, true); 1606719f5d7fSMiklos Szeredi if (err) 1607719f5d7fSMiklos Szeredi goto out; 1608719f5d7fSMiklos Szeredi } 1609a8d56d8eSAl Viro err = propagate_mnt(dest_mnt, dest_dentry, source_mnt, &tree_list); 1610719f5d7fSMiklos Szeredi if (err) 1611719f5d7fSMiklos Szeredi goto out_cleanup_ids; 1612b90fa9aeSRam Pai 161399b7db7bSNick Piggin br_write_lock(vfsmount_lock); 1614df1a1ad2SAl Viro 1615a8d56d8eSAl Viro if (IS_MNT_SHARED(&dest_mnt->mnt)) { 16160fb54e50SAl Viro for (p = source_mnt; p; p = next_mnt(p, &source_mnt->mnt)) 16170f0afb1dSAl Viro set_mnt_shared(p); 1618b90fa9aeSRam Pai } 16191a390689SAl Viro if (parent_path) { 16200fb54e50SAl Viro detach_mnt(source_mnt, parent_path); 16210fb54e50SAl Viro attach_mnt(source_mnt, path); 1622143c8c91SAl Viro touch_mnt_namespace(source_mnt->mnt_ns); 162321444403SRam Pai } else { 162414cf1fa8SAl Viro mnt_set_mountpoint(dest_mnt, dest_dentry, source_mnt); 16250fb54e50SAl Viro commit_tree(source_mnt); 162621444403SRam Pai } 1627b90fa9aeSRam Pai 16281b8e5564SAl Viro list_for_each_entry_safe(child, p, &tree_list, mnt_hash) { 16291b8e5564SAl Viro list_del_init(&child->mnt_hash); 16304b2619a5SAl Viro commit_tree(child); 1631b90fa9aeSRam Pai } 163299b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 163399b7db7bSNick Piggin 1634b90fa9aeSRam Pai return 0; 1635719f5d7fSMiklos Szeredi 1636719f5d7fSMiklos Szeredi out_cleanup_ids: 1637a8d56d8eSAl Viro if (IS_MNT_SHARED(&dest_mnt->mnt)) 16380fb54e50SAl Viro cleanup_group_ids(source_mnt, NULL); 1639719f5d7fSMiklos Szeredi out: 1640719f5d7fSMiklos Szeredi return err; 1641b90fa9aeSRam Pai } 1642b90fa9aeSRam Pai 1643b12cea91SAl Viro static int lock_mount(struct path *path) 1644b12cea91SAl Viro { 1645b12cea91SAl Viro struct vfsmount *mnt; 1646b12cea91SAl Viro retry: 1647b12cea91SAl Viro mutex_lock(&path->dentry->d_inode->i_mutex); 1648b12cea91SAl Viro if (unlikely(cant_mount(path->dentry))) { 1649b12cea91SAl Viro mutex_unlock(&path->dentry->d_inode->i_mutex); 1650b12cea91SAl Viro return -ENOENT; 1651b12cea91SAl Viro } 1652b12cea91SAl Viro down_write(&namespace_sem); 1653b12cea91SAl Viro mnt = lookup_mnt(path); 1654b12cea91SAl Viro if (likely(!mnt)) 1655b12cea91SAl Viro return 0; 1656b12cea91SAl Viro up_write(&namespace_sem); 1657b12cea91SAl Viro mutex_unlock(&path->dentry->d_inode->i_mutex); 1658b12cea91SAl Viro path_put(path); 1659b12cea91SAl Viro path->mnt = mnt; 1660b12cea91SAl Viro path->dentry = dget(mnt->mnt_root); 1661b12cea91SAl Viro goto retry; 1662b12cea91SAl Viro } 1663b12cea91SAl Viro 1664b12cea91SAl Viro static void unlock_mount(struct path *path) 1665b12cea91SAl Viro { 1666b12cea91SAl Viro up_write(&namespace_sem); 1667b12cea91SAl Viro mutex_unlock(&path->dentry->d_inode->i_mutex); 1668b12cea91SAl Viro } 1669b12cea91SAl Viro 167095bc5f25SAl Viro static int graft_tree(struct mount *mnt, struct path *path) 16711da177e4SLinus Torvalds { 167295bc5f25SAl Viro if (mnt->mnt.mnt_sb->s_flags & MS_NOUSER) 16731da177e4SLinus Torvalds return -EINVAL; 16741da177e4SLinus Torvalds 16758c3ee42eSAl Viro if (S_ISDIR(path->dentry->d_inode->i_mode) != 167695bc5f25SAl Viro S_ISDIR(mnt->mnt.mnt_root->d_inode->i_mode)) 16771da177e4SLinus Torvalds return -ENOTDIR; 16781da177e4SLinus Torvalds 1679b12cea91SAl Viro if (d_unlinked(path->dentry)) 1680b12cea91SAl Viro return -ENOENT; 16811da177e4SLinus Torvalds 168295bc5f25SAl Viro return attach_recursive_mnt(mnt, path, NULL); 16831da177e4SLinus Torvalds } 16841da177e4SLinus Torvalds 16851da177e4SLinus Torvalds /* 16867a2e8a8fSValerie Aurora * Sanity check the flags to change_mnt_propagation. 16877a2e8a8fSValerie Aurora */ 16887a2e8a8fSValerie Aurora 16897a2e8a8fSValerie Aurora static int flags_to_propagation_type(int flags) 16907a2e8a8fSValerie Aurora { 16917c6e984dSRoman Borisov int type = flags & ~(MS_REC | MS_SILENT); 16927a2e8a8fSValerie Aurora 16937a2e8a8fSValerie Aurora /* Fail if any non-propagation flags are set */ 16947a2e8a8fSValerie Aurora if (type & ~(MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE)) 16957a2e8a8fSValerie Aurora return 0; 16967a2e8a8fSValerie Aurora /* Only one propagation flag should be set */ 16977a2e8a8fSValerie Aurora if (!is_power_of_2(type)) 16987a2e8a8fSValerie Aurora return 0; 16997a2e8a8fSValerie Aurora return type; 17007a2e8a8fSValerie Aurora } 17017a2e8a8fSValerie Aurora 17027a2e8a8fSValerie Aurora /* 170307b20889SRam Pai * recursively change the type of the mountpoint. 170407b20889SRam Pai */ 17050a0d8a46SAl Viro static int do_change_type(struct path *path, int flag) 170607b20889SRam Pai { 1707315fc83eSAl Viro struct mount *m; 17084b8b21f4SAl Viro struct mount *mnt = real_mount(path->mnt); 170907b20889SRam Pai int recurse = flag & MS_REC; 17107a2e8a8fSValerie Aurora int type; 1711719f5d7fSMiklos Szeredi int err = 0; 171207b20889SRam Pai 1713ee6f9582SMiklos Szeredi if (!capable(CAP_SYS_ADMIN)) 1714ee6f9582SMiklos Szeredi return -EPERM; 1715ee6f9582SMiklos Szeredi 17162d92ab3cSAl Viro if (path->dentry != path->mnt->mnt_root) 171707b20889SRam Pai return -EINVAL; 171807b20889SRam Pai 17197a2e8a8fSValerie Aurora type = flags_to_propagation_type(flag); 17207a2e8a8fSValerie Aurora if (!type) 17217a2e8a8fSValerie Aurora return -EINVAL; 17227a2e8a8fSValerie Aurora 172307b20889SRam Pai down_write(&namespace_sem); 1724719f5d7fSMiklos Szeredi if (type == MS_SHARED) { 1725719f5d7fSMiklos Szeredi err = invent_group_ids(mnt, recurse); 1726719f5d7fSMiklos Szeredi if (err) 1727719f5d7fSMiklos Szeredi goto out_unlock; 1728719f5d7fSMiklos Szeredi } 1729719f5d7fSMiklos Szeredi 173099b7db7bSNick Piggin br_write_lock(vfsmount_lock); 17314b8b21f4SAl Viro for (m = mnt; m; m = (recurse ? next_mnt(m, &mnt->mnt) : NULL)) 17320f0afb1dSAl Viro change_mnt_propagation(m, type); 173399b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 1734719f5d7fSMiklos Szeredi 1735719f5d7fSMiklos Szeredi out_unlock: 173607b20889SRam Pai up_write(&namespace_sem); 1737719f5d7fSMiklos Szeredi return err; 173807b20889SRam Pai } 173907b20889SRam Pai 174007b20889SRam Pai /* 17411da177e4SLinus Torvalds * do loopback mount. 17421da177e4SLinus Torvalds */ 17430a0d8a46SAl Viro static int do_loopback(struct path *path, char *old_name, 17442dafe1c4SEric Sandeen int recurse) 17451da177e4SLinus Torvalds { 1746b12cea91SAl Viro LIST_HEAD(umount_list); 17472d92ab3cSAl Viro struct path old_path; 174887129cc0SAl Viro struct mount *mnt = NULL, *old; 17492d92ab3cSAl Viro int err = mount_is_safe(path); 17501da177e4SLinus Torvalds if (err) 17511da177e4SLinus Torvalds return err; 17521da177e4SLinus Torvalds if (!old_name || !*old_name) 17531da177e4SLinus Torvalds return -EINVAL; 1754815d405cSTrond Myklebust err = kern_path(old_name, LOOKUP_FOLLOW|LOOKUP_AUTOMOUNT, &old_path); 17551da177e4SLinus Torvalds if (err) 17561da177e4SLinus Torvalds return err; 17571da177e4SLinus Torvalds 1758b12cea91SAl Viro err = lock_mount(path); 1759b12cea91SAl Viro if (err) 17609676f0c6SRam Pai goto out; 17619676f0c6SRam Pai 176287129cc0SAl Viro old = real_mount(old_path.mnt); 176387129cc0SAl Viro 1764b12cea91SAl Viro err = -EINVAL; 1765b12cea91SAl Viro if (IS_MNT_UNBINDABLE(old_path.mnt)) 1766b12cea91SAl Viro goto out2; 1767b12cea91SAl Viro 1768143c8c91SAl Viro if (!check_mnt(real_mount(path->mnt)) || !check_mnt(old)) 1769b12cea91SAl Viro goto out2; 1770ccd48bc7SAl Viro 17711da177e4SLinus Torvalds err = -ENOMEM; 17721da177e4SLinus Torvalds if (recurse) 177387129cc0SAl Viro mnt = copy_tree(old, old_path.dentry, 0); 17741da177e4SLinus Torvalds else 177587129cc0SAl Viro mnt = clone_mnt(old, old_path.dentry, 0); 17761da177e4SLinus Torvalds 1777ccd48bc7SAl Viro if (!mnt) 1778b12cea91SAl Viro goto out2; 1779ccd48bc7SAl Viro 178095bc5f25SAl Viro err = graft_tree(mnt, path); 17811da177e4SLinus Torvalds if (err) { 178299b7db7bSNick Piggin br_write_lock(vfsmount_lock); 1783761d5c38SAl Viro umount_tree(mnt, 0, &umount_list); 178499b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 17855b83d2c5SRam Pai } 1786b12cea91SAl Viro out2: 1787b12cea91SAl Viro unlock_mount(path); 1788b12cea91SAl Viro release_mounts(&umount_list); 1789ccd48bc7SAl Viro out: 17902d92ab3cSAl Viro path_put(&old_path); 17911da177e4SLinus Torvalds return err; 17921da177e4SLinus Torvalds } 17931da177e4SLinus Torvalds 17942e4b7fcdSDave Hansen static int change_mount_flags(struct vfsmount *mnt, int ms_flags) 17952e4b7fcdSDave Hansen { 17962e4b7fcdSDave Hansen int error = 0; 17972e4b7fcdSDave Hansen int readonly_request = 0; 17982e4b7fcdSDave Hansen 17992e4b7fcdSDave Hansen if (ms_flags & MS_RDONLY) 18002e4b7fcdSDave Hansen readonly_request = 1; 18012e4b7fcdSDave Hansen if (readonly_request == __mnt_is_readonly(mnt)) 18022e4b7fcdSDave Hansen return 0; 18032e4b7fcdSDave Hansen 18042e4b7fcdSDave Hansen if (readonly_request) 180583adc753SAl Viro error = mnt_make_readonly(real_mount(mnt)); 18062e4b7fcdSDave Hansen else 180783adc753SAl Viro __mnt_unmake_readonly(real_mount(mnt)); 18082e4b7fcdSDave Hansen return error; 18092e4b7fcdSDave Hansen } 18102e4b7fcdSDave Hansen 18111da177e4SLinus Torvalds /* 18121da177e4SLinus Torvalds * change filesystem flags. dir should be a physical root of filesystem. 18131da177e4SLinus Torvalds * If you've mounted a non-root directory somewhere and want to do remount 18141da177e4SLinus Torvalds * on it - tough luck. 18151da177e4SLinus Torvalds */ 18160a0d8a46SAl Viro static int do_remount(struct path *path, int flags, int mnt_flags, 18171da177e4SLinus Torvalds void *data) 18181da177e4SLinus Torvalds { 18191da177e4SLinus Torvalds int err; 18202d92ab3cSAl Viro struct super_block *sb = path->mnt->mnt_sb; 1821143c8c91SAl Viro struct mount *mnt = real_mount(path->mnt); 18221da177e4SLinus Torvalds 18231da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 18241da177e4SLinus Torvalds return -EPERM; 18251da177e4SLinus Torvalds 1826143c8c91SAl Viro if (!check_mnt(mnt)) 18271da177e4SLinus Torvalds return -EINVAL; 18281da177e4SLinus Torvalds 18292d92ab3cSAl Viro if (path->dentry != path->mnt->mnt_root) 18301da177e4SLinus Torvalds return -EINVAL; 18311da177e4SLinus Torvalds 1832ff36fe2cSEric Paris err = security_sb_remount(sb, data); 1833ff36fe2cSEric Paris if (err) 1834ff36fe2cSEric Paris return err; 1835ff36fe2cSEric Paris 18361da177e4SLinus Torvalds down_write(&sb->s_umount); 18372e4b7fcdSDave Hansen if (flags & MS_BIND) 18382d92ab3cSAl Viro err = change_mount_flags(path->mnt, flags); 18394aa98cf7SAl Viro else 18401da177e4SLinus Torvalds err = do_remount_sb(sb, flags, data, 0); 18417b43a79fSAl Viro if (!err) { 184299b7db7bSNick Piggin br_write_lock(vfsmount_lock); 1843143c8c91SAl Viro mnt_flags |= mnt->mnt.mnt_flags & MNT_PROPAGATION_MASK; 1844143c8c91SAl Viro mnt->mnt.mnt_flags = mnt_flags; 184599b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 18467b43a79fSAl Viro } 18471da177e4SLinus Torvalds up_write(&sb->s_umount); 18480e55a7ccSDan Williams if (!err) { 184999b7db7bSNick Piggin br_write_lock(vfsmount_lock); 1850143c8c91SAl Viro touch_mnt_namespace(mnt->mnt_ns); 185199b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 18520e55a7ccSDan Williams } 18531da177e4SLinus Torvalds return err; 18541da177e4SLinus Torvalds } 18551da177e4SLinus Torvalds 1856cbbe362cSAl Viro static inline int tree_contains_unbindable(struct mount *mnt) 18579676f0c6SRam Pai { 1858315fc83eSAl Viro struct mount *p; 1859cbbe362cSAl Viro for (p = mnt; p; p = next_mnt(p, &mnt->mnt)) { 1860315fc83eSAl Viro if (IS_MNT_UNBINDABLE(&p->mnt)) 18619676f0c6SRam Pai return 1; 18629676f0c6SRam Pai } 18639676f0c6SRam Pai return 0; 18649676f0c6SRam Pai } 18659676f0c6SRam Pai 18660a0d8a46SAl Viro static int do_move_mount(struct path *path, char *old_name) 18671da177e4SLinus Torvalds { 18682d92ab3cSAl Viro struct path old_path, parent_path; 1869676da58dSAl Viro struct mount *p; 18700fb54e50SAl Viro struct mount *old; 18711da177e4SLinus Torvalds int err = 0; 18721da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 18731da177e4SLinus Torvalds return -EPERM; 18741da177e4SLinus Torvalds if (!old_name || !*old_name) 18751da177e4SLinus Torvalds return -EINVAL; 18762d92ab3cSAl Viro err = kern_path(old_name, LOOKUP_FOLLOW, &old_path); 18771da177e4SLinus Torvalds if (err) 18781da177e4SLinus Torvalds return err; 18791da177e4SLinus Torvalds 1880b12cea91SAl Viro err = lock_mount(path); 1881cc53ce53SDavid Howells if (err < 0) 1882cc53ce53SDavid Howells goto out; 1883cc53ce53SDavid Howells 1884143c8c91SAl Viro old = real_mount(old_path.mnt); 1885143c8c91SAl Viro 18861da177e4SLinus Torvalds err = -EINVAL; 1887143c8c91SAl Viro if (!check_mnt(real_mount(path->mnt)) || !check_mnt(old)) 18881da177e4SLinus Torvalds goto out1; 18891da177e4SLinus Torvalds 1890f3da392eSAlexey Dobriyan if (d_unlinked(path->dentry)) 189121444403SRam Pai goto out1; 18921da177e4SLinus Torvalds 18931da177e4SLinus Torvalds err = -EINVAL; 18942d92ab3cSAl Viro if (old_path.dentry != old_path.mnt->mnt_root) 189521444403SRam Pai goto out1; 18961da177e4SLinus Torvalds 1897676da58dSAl Viro if (!mnt_has_parent(old)) 189821444403SRam Pai goto out1; 18991da177e4SLinus Torvalds 19002d92ab3cSAl Viro if (S_ISDIR(path->dentry->d_inode->i_mode) != 19012d92ab3cSAl Viro S_ISDIR(old_path.dentry->d_inode->i_mode)) 190221444403SRam Pai goto out1; 190321444403SRam Pai /* 190421444403SRam Pai * Don't move a mount residing in a shared parent. 190521444403SRam Pai */ 19060714a533SAl Viro if (IS_MNT_SHARED(&old->mnt_parent->mnt)) 190721444403SRam Pai goto out1; 19089676f0c6SRam Pai /* 19099676f0c6SRam Pai * Don't move a mount tree containing unbindable mounts to a destination 19109676f0c6SRam Pai * mount which is shared. 19119676f0c6SRam Pai */ 19122d92ab3cSAl Viro if (IS_MNT_SHARED(path->mnt) && 1913cbbe362cSAl Viro tree_contains_unbindable(old)) 19149676f0c6SRam Pai goto out1; 19151da177e4SLinus Torvalds err = -ELOOP; 19160714a533SAl Viro for (p = real_mount(path->mnt); mnt_has_parent(p); p = p->mnt_parent) 1917676da58dSAl Viro if (p == old) 191821444403SRam Pai goto out1; 19191da177e4SLinus Torvalds 19200fb54e50SAl Viro err = attach_recursive_mnt(old, path, &parent_path); 19214ac91378SJan Blunck if (err) 192221444403SRam Pai goto out1; 19231da177e4SLinus Torvalds 19241da177e4SLinus Torvalds /* if the mount is moved, it should no longer be expire 19251da177e4SLinus Torvalds * automatically */ 19266776db3dSAl Viro list_del_init(&old->mnt_expire); 19271da177e4SLinus Torvalds out1: 1928b12cea91SAl Viro unlock_mount(path); 19291da177e4SLinus Torvalds out: 19301da177e4SLinus Torvalds if (!err) 19311a390689SAl Viro path_put(&parent_path); 19322d92ab3cSAl Viro path_put(&old_path); 19331da177e4SLinus Torvalds return err; 19341da177e4SLinus Torvalds } 19351da177e4SLinus Torvalds 19369d412a43SAl Viro static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype) 19379d412a43SAl Viro { 19389d412a43SAl Viro int err; 19399d412a43SAl Viro const char *subtype = strchr(fstype, '.'); 19409d412a43SAl Viro if (subtype) { 19419d412a43SAl Viro subtype++; 19429d412a43SAl Viro err = -EINVAL; 19439d412a43SAl Viro if (!subtype[0]) 19449d412a43SAl Viro goto err; 19459d412a43SAl Viro } else 19469d412a43SAl Viro subtype = ""; 19479d412a43SAl Viro 19489d412a43SAl Viro mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL); 19499d412a43SAl Viro err = -ENOMEM; 19509d412a43SAl Viro if (!mnt->mnt_sb->s_subtype) 19519d412a43SAl Viro goto err; 19529d412a43SAl Viro return mnt; 19539d412a43SAl Viro 19549d412a43SAl Viro err: 19559d412a43SAl Viro mntput(mnt); 19569d412a43SAl Viro return ERR_PTR(err); 19579d412a43SAl Viro } 19589d412a43SAl Viro 195979e801a9SAl Viro static struct vfsmount * 19609d412a43SAl Viro do_kern_mount(const char *fstype, int flags, const char *name, void *data) 19619d412a43SAl Viro { 19629d412a43SAl Viro struct file_system_type *type = get_fs_type(fstype); 19639d412a43SAl Viro struct vfsmount *mnt; 19649d412a43SAl Viro if (!type) 19659d412a43SAl Viro return ERR_PTR(-ENODEV); 19669d412a43SAl Viro mnt = vfs_kern_mount(type, flags, name, data); 19679d412a43SAl Viro if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) && 19689d412a43SAl Viro !mnt->mnt_sb->s_subtype) 19699d412a43SAl Viro mnt = fs_set_subtype(mnt, fstype); 19709d412a43SAl Viro put_filesystem(type); 19719d412a43SAl Viro return mnt; 19729d412a43SAl Viro } 19739d412a43SAl Viro 19749d412a43SAl Viro /* 19759d412a43SAl Viro * add a mount into a namespace's mount tree 19769d412a43SAl Viro */ 197795bc5f25SAl Viro static int do_add_mount(struct mount *newmnt, struct path *path, int mnt_flags) 19789d412a43SAl Viro { 19799d412a43SAl Viro int err; 19809d412a43SAl Viro 19819d412a43SAl Viro mnt_flags &= ~(MNT_SHARED | MNT_WRITE_HOLD | MNT_INTERNAL); 19829d412a43SAl Viro 1983b12cea91SAl Viro err = lock_mount(path); 1984b12cea91SAl Viro if (err) 1985b12cea91SAl Viro return err; 19869d412a43SAl Viro 19879d412a43SAl Viro err = -EINVAL; 1988143c8c91SAl Viro if (!(mnt_flags & MNT_SHRINKABLE) && !check_mnt(real_mount(path->mnt))) 19899d412a43SAl Viro goto unlock; 19909d412a43SAl Viro 19919d412a43SAl Viro /* Refuse the same filesystem on the same mount point */ 19929d412a43SAl Viro err = -EBUSY; 199395bc5f25SAl Viro if (path->mnt->mnt_sb == newmnt->mnt.mnt_sb && 19949d412a43SAl Viro path->mnt->mnt_root == path->dentry) 19959d412a43SAl Viro goto unlock; 19969d412a43SAl Viro 19979d412a43SAl Viro err = -EINVAL; 199895bc5f25SAl Viro if (S_ISLNK(newmnt->mnt.mnt_root->d_inode->i_mode)) 19999d412a43SAl Viro goto unlock; 20009d412a43SAl Viro 200195bc5f25SAl Viro newmnt->mnt.mnt_flags = mnt_flags; 20029d412a43SAl Viro err = graft_tree(newmnt, path); 20039d412a43SAl Viro 20049d412a43SAl Viro unlock: 2005b12cea91SAl Viro unlock_mount(path); 20069d412a43SAl Viro return err; 20079d412a43SAl Viro } 2008b1e75df4SAl Viro 20091da177e4SLinus Torvalds /* 20101da177e4SLinus Torvalds * create a new mount for userspace and request it to be added into the 20111da177e4SLinus Torvalds * namespace's tree 20121da177e4SLinus Torvalds */ 20130a0d8a46SAl Viro static int do_new_mount(struct path *path, char *type, int flags, 20141da177e4SLinus Torvalds int mnt_flags, char *name, void *data) 20151da177e4SLinus Torvalds { 20161da177e4SLinus Torvalds struct vfsmount *mnt; 201715f9a3f3SAl Viro int err; 20181da177e4SLinus Torvalds 2019eca6f534SVegard Nossum if (!type) 20201da177e4SLinus Torvalds return -EINVAL; 20211da177e4SLinus Torvalds 20221da177e4SLinus Torvalds /* we need capabilities... */ 20231da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 20241da177e4SLinus Torvalds return -EPERM; 20251da177e4SLinus Torvalds 20261da177e4SLinus Torvalds mnt = do_kern_mount(type, flags, name, data); 20271da177e4SLinus Torvalds if (IS_ERR(mnt)) 20281da177e4SLinus Torvalds return PTR_ERR(mnt); 20291da177e4SLinus Torvalds 203095bc5f25SAl Viro err = do_add_mount(real_mount(mnt), path, mnt_flags); 203115f9a3f3SAl Viro if (err) 203215f9a3f3SAl Viro mntput(mnt); 203315f9a3f3SAl Viro return err; 20341da177e4SLinus Torvalds } 20351da177e4SLinus Torvalds 203619a167afSAl Viro int finish_automount(struct vfsmount *m, struct path *path) 203719a167afSAl Viro { 20386776db3dSAl Viro struct mount *mnt = real_mount(m); 203919a167afSAl Viro int err; 204019a167afSAl Viro /* The new mount record should have at least 2 refs to prevent it being 204119a167afSAl Viro * expired before we get a chance to add it 204219a167afSAl Viro */ 20436776db3dSAl Viro BUG_ON(mnt_get_count(mnt) < 2); 204419a167afSAl Viro 204519a167afSAl Viro if (m->mnt_sb == path->mnt->mnt_sb && 204619a167afSAl Viro m->mnt_root == path->dentry) { 2047b1e75df4SAl Viro err = -ELOOP; 2048b1e75df4SAl Viro goto fail; 204919a167afSAl Viro } 205019a167afSAl Viro 205195bc5f25SAl Viro err = do_add_mount(mnt, path, path->mnt->mnt_flags | MNT_SHRINKABLE); 2052b1e75df4SAl Viro if (!err) 2053b1e75df4SAl Viro return 0; 2054b1e75df4SAl Viro fail: 2055b1e75df4SAl Viro /* remove m from any expiration list it may be on */ 20566776db3dSAl Viro if (!list_empty(&mnt->mnt_expire)) { 2057b1e75df4SAl Viro down_write(&namespace_sem); 2058b1e75df4SAl Viro br_write_lock(vfsmount_lock); 20596776db3dSAl Viro list_del_init(&mnt->mnt_expire); 2060b1e75df4SAl Viro br_write_unlock(vfsmount_lock); 2061b1e75df4SAl Viro up_write(&namespace_sem); 206219a167afSAl Viro } 2063b1e75df4SAl Viro mntput(m); 2064b1e75df4SAl Viro mntput(m); 206519a167afSAl Viro return err; 206619a167afSAl Viro } 206719a167afSAl Viro 2068ea5b778aSDavid Howells /** 2069ea5b778aSDavid Howells * mnt_set_expiry - Put a mount on an expiration list 2070ea5b778aSDavid Howells * @mnt: The mount to list. 2071ea5b778aSDavid Howells * @expiry_list: The list to add the mount to. 2072ea5b778aSDavid Howells */ 2073ea5b778aSDavid Howells void mnt_set_expiry(struct vfsmount *mnt, struct list_head *expiry_list) 2074ea5b778aSDavid Howells { 2075ea5b778aSDavid Howells down_write(&namespace_sem); 2076ea5b778aSDavid Howells br_write_lock(vfsmount_lock); 2077ea5b778aSDavid Howells 20786776db3dSAl Viro list_add_tail(&real_mount(mnt)->mnt_expire, expiry_list); 2079ea5b778aSDavid Howells 2080ea5b778aSDavid Howells br_write_unlock(vfsmount_lock); 2081ea5b778aSDavid Howells up_write(&namespace_sem); 2082ea5b778aSDavid Howells } 2083ea5b778aSDavid Howells EXPORT_SYMBOL(mnt_set_expiry); 2084ea5b778aSDavid Howells 2085ea5b778aSDavid Howells /* 20861da177e4SLinus Torvalds * process a list of expirable mountpoints with the intent of discarding any 20871da177e4SLinus Torvalds * mountpoints that aren't in use and haven't been touched since last we came 20881da177e4SLinus Torvalds * here 20891da177e4SLinus Torvalds */ 20901da177e4SLinus Torvalds void mark_mounts_for_expiry(struct list_head *mounts) 20911da177e4SLinus Torvalds { 2092761d5c38SAl Viro struct mount *mnt, *next; 20931da177e4SLinus Torvalds LIST_HEAD(graveyard); 2094bcc5c7d2SAl Viro LIST_HEAD(umounts); 20951da177e4SLinus Torvalds 20961da177e4SLinus Torvalds if (list_empty(mounts)) 20971da177e4SLinus Torvalds return; 20981da177e4SLinus Torvalds 2099bcc5c7d2SAl Viro down_write(&namespace_sem); 210099b7db7bSNick Piggin br_write_lock(vfsmount_lock); 21011da177e4SLinus Torvalds 21021da177e4SLinus Torvalds /* extract from the expiration list every vfsmount that matches the 21031da177e4SLinus Torvalds * following criteria: 21041da177e4SLinus Torvalds * - only referenced by its parent vfsmount 21051da177e4SLinus Torvalds * - still marked for expiry (marked on the last call here; marks are 21061da177e4SLinus Torvalds * cleared by mntput()) 21071da177e4SLinus Torvalds */ 21086776db3dSAl Viro list_for_each_entry_safe(mnt, next, mounts, mnt_expire) { 2109761d5c38SAl Viro if (!xchg(&mnt->mnt.mnt_expiry_mark, 1) || 21101ab59738SAl Viro propagate_mount_busy(mnt, 1)) 21111da177e4SLinus Torvalds continue; 21126776db3dSAl Viro list_move(&mnt->mnt_expire, &graveyard); 21131da177e4SLinus Torvalds } 2114bcc5c7d2SAl Viro while (!list_empty(&graveyard)) { 21156776db3dSAl Viro mnt = list_first_entry(&graveyard, struct mount, mnt_expire); 2116143c8c91SAl Viro touch_mnt_namespace(mnt->mnt_ns); 2117bcc5c7d2SAl Viro umount_tree(mnt, 1, &umounts); 2118bcc5c7d2SAl Viro } 211999b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 2120bcc5c7d2SAl Viro up_write(&namespace_sem); 2121bcc5c7d2SAl Viro 2122bcc5c7d2SAl Viro release_mounts(&umounts); 21231da177e4SLinus Torvalds } 21241da177e4SLinus Torvalds 21251da177e4SLinus Torvalds EXPORT_SYMBOL_GPL(mark_mounts_for_expiry); 21261da177e4SLinus Torvalds 21271da177e4SLinus Torvalds /* 21285528f911STrond Myklebust * Ripoff of 'select_parent()' 21295528f911STrond Myklebust * 21305528f911STrond Myklebust * search the list of submounts for a given mountpoint, and move any 21315528f911STrond Myklebust * shrinkable submounts to the 'graveyard' list. 21325528f911STrond Myklebust */ 2133692afc31SAl Viro static int select_submounts(struct mount *parent, struct list_head *graveyard) 21345528f911STrond Myklebust { 2135692afc31SAl Viro struct mount *this_parent = parent; 21365528f911STrond Myklebust struct list_head *next; 21375528f911STrond Myklebust int found = 0; 21385528f911STrond Myklebust 21395528f911STrond Myklebust repeat: 21406b41d536SAl Viro next = this_parent->mnt_mounts.next; 21415528f911STrond Myklebust resume: 21426b41d536SAl Viro while (next != &this_parent->mnt_mounts) { 21435528f911STrond Myklebust struct list_head *tmp = next; 21446b41d536SAl Viro struct mount *mnt = list_entry(tmp, struct mount, mnt_child); 21455528f911STrond Myklebust 21465528f911STrond Myklebust next = tmp->next; 2147692afc31SAl Viro if (!(mnt->mnt.mnt_flags & MNT_SHRINKABLE)) 21485528f911STrond Myklebust continue; 21495528f911STrond Myklebust /* 21505528f911STrond Myklebust * Descend a level if the d_mounts list is non-empty. 21515528f911STrond Myklebust */ 21526b41d536SAl Viro if (!list_empty(&mnt->mnt_mounts)) { 21535528f911STrond Myklebust this_parent = mnt; 21545528f911STrond Myklebust goto repeat; 21555528f911STrond Myklebust } 21565528f911STrond Myklebust 21571ab59738SAl Viro if (!propagate_mount_busy(mnt, 1)) { 21586776db3dSAl Viro list_move_tail(&mnt->mnt_expire, graveyard); 21595528f911STrond Myklebust found++; 21605528f911STrond Myklebust } 21615528f911STrond Myklebust } 21625528f911STrond Myklebust /* 21635528f911STrond Myklebust * All done at this level ... ascend and resume the search 21645528f911STrond Myklebust */ 21655528f911STrond Myklebust if (this_parent != parent) { 21666b41d536SAl Viro next = this_parent->mnt_child.next; 21670714a533SAl Viro this_parent = this_parent->mnt_parent; 21685528f911STrond Myklebust goto resume; 21695528f911STrond Myklebust } 21705528f911STrond Myklebust return found; 21715528f911STrond Myklebust } 21725528f911STrond Myklebust 21735528f911STrond Myklebust /* 21745528f911STrond Myklebust * process a list of expirable mountpoints with the intent of discarding any 21755528f911STrond Myklebust * submounts of a specific parent mountpoint 217699b7db7bSNick Piggin * 217799b7db7bSNick Piggin * vfsmount_lock must be held for write 21785528f911STrond Myklebust */ 2179692afc31SAl Viro static void shrink_submounts(struct mount *mnt, struct list_head *umounts) 21805528f911STrond Myklebust { 21815528f911STrond Myklebust LIST_HEAD(graveyard); 2182761d5c38SAl Viro struct mount *m; 21835528f911STrond Myklebust 21845528f911STrond Myklebust /* extract submounts of 'mountpoint' from the expiration list */ 2185c35038beSAl Viro while (select_submounts(mnt, &graveyard)) { 2186bcc5c7d2SAl Viro while (!list_empty(&graveyard)) { 2187761d5c38SAl Viro m = list_first_entry(&graveyard, struct mount, 21886776db3dSAl Viro mnt_expire); 2189143c8c91SAl Viro touch_mnt_namespace(m->mnt_ns); 2190afef80b3SEric W. Biederman umount_tree(m, 1, umounts); 2191bcc5c7d2SAl Viro } 2192bcc5c7d2SAl Viro } 21935528f911STrond Myklebust } 21945528f911STrond Myklebust 21955528f911STrond Myklebust /* 21961da177e4SLinus Torvalds * Some copy_from_user() implementations do not return the exact number of 21971da177e4SLinus Torvalds * bytes remaining to copy on a fault. But copy_mount_options() requires that. 21981da177e4SLinus Torvalds * Note that this function differs from copy_from_user() in that it will oops 21991da177e4SLinus Torvalds * on bad values of `to', rather than returning a short copy. 22001da177e4SLinus Torvalds */ 2201b58fed8bSRam Pai static long exact_copy_from_user(void *to, const void __user * from, 2202b58fed8bSRam Pai unsigned long n) 22031da177e4SLinus Torvalds { 22041da177e4SLinus Torvalds char *t = to; 22051da177e4SLinus Torvalds const char __user *f = from; 22061da177e4SLinus Torvalds char c; 22071da177e4SLinus Torvalds 22081da177e4SLinus Torvalds if (!access_ok(VERIFY_READ, from, n)) 22091da177e4SLinus Torvalds return n; 22101da177e4SLinus Torvalds 22111da177e4SLinus Torvalds while (n) { 22121da177e4SLinus Torvalds if (__get_user(c, f)) { 22131da177e4SLinus Torvalds memset(t, 0, n); 22141da177e4SLinus Torvalds break; 22151da177e4SLinus Torvalds } 22161da177e4SLinus Torvalds *t++ = c; 22171da177e4SLinus Torvalds f++; 22181da177e4SLinus Torvalds n--; 22191da177e4SLinus Torvalds } 22201da177e4SLinus Torvalds return n; 22211da177e4SLinus Torvalds } 22221da177e4SLinus Torvalds 22231da177e4SLinus Torvalds int copy_mount_options(const void __user * data, unsigned long *where) 22241da177e4SLinus Torvalds { 22251da177e4SLinus Torvalds int i; 22261da177e4SLinus Torvalds unsigned long page; 22271da177e4SLinus Torvalds unsigned long size; 22281da177e4SLinus Torvalds 22291da177e4SLinus Torvalds *where = 0; 22301da177e4SLinus Torvalds if (!data) 22311da177e4SLinus Torvalds return 0; 22321da177e4SLinus Torvalds 22331da177e4SLinus Torvalds if (!(page = __get_free_page(GFP_KERNEL))) 22341da177e4SLinus Torvalds return -ENOMEM; 22351da177e4SLinus Torvalds 22361da177e4SLinus Torvalds /* We only care that *some* data at the address the user 22371da177e4SLinus Torvalds * gave us is valid. Just in case, we'll zero 22381da177e4SLinus Torvalds * the remainder of the page. 22391da177e4SLinus Torvalds */ 22401da177e4SLinus Torvalds /* copy_from_user cannot cross TASK_SIZE ! */ 22411da177e4SLinus Torvalds size = TASK_SIZE - (unsigned long)data; 22421da177e4SLinus Torvalds if (size > PAGE_SIZE) 22431da177e4SLinus Torvalds size = PAGE_SIZE; 22441da177e4SLinus Torvalds 22451da177e4SLinus Torvalds i = size - exact_copy_from_user((void *)page, data, size); 22461da177e4SLinus Torvalds if (!i) { 22471da177e4SLinus Torvalds free_page(page); 22481da177e4SLinus Torvalds return -EFAULT; 22491da177e4SLinus Torvalds } 22501da177e4SLinus Torvalds if (i != PAGE_SIZE) 22511da177e4SLinus Torvalds memset((char *)page + i, 0, PAGE_SIZE - i); 22521da177e4SLinus Torvalds *where = page; 22531da177e4SLinus Torvalds return 0; 22541da177e4SLinus Torvalds } 22551da177e4SLinus Torvalds 2256eca6f534SVegard Nossum int copy_mount_string(const void __user *data, char **where) 2257eca6f534SVegard Nossum { 2258eca6f534SVegard Nossum char *tmp; 2259eca6f534SVegard Nossum 2260eca6f534SVegard Nossum if (!data) { 2261eca6f534SVegard Nossum *where = NULL; 2262eca6f534SVegard Nossum return 0; 2263eca6f534SVegard Nossum } 2264eca6f534SVegard Nossum 2265eca6f534SVegard Nossum tmp = strndup_user(data, PAGE_SIZE); 2266eca6f534SVegard Nossum if (IS_ERR(tmp)) 2267eca6f534SVegard Nossum return PTR_ERR(tmp); 2268eca6f534SVegard Nossum 2269eca6f534SVegard Nossum *where = tmp; 2270eca6f534SVegard Nossum return 0; 2271eca6f534SVegard Nossum } 2272eca6f534SVegard Nossum 22731da177e4SLinus Torvalds /* 22741da177e4SLinus Torvalds * Flags is a 32-bit value that allows up to 31 non-fs dependent flags to 22751da177e4SLinus Torvalds * be given to the mount() call (ie: read-only, no-dev, no-suid etc). 22761da177e4SLinus Torvalds * 22771da177e4SLinus Torvalds * data is a (void *) that can point to any structure up to 22781da177e4SLinus Torvalds * PAGE_SIZE-1 bytes, which can contain arbitrary fs-dependent 22791da177e4SLinus Torvalds * information (or be NULL). 22801da177e4SLinus Torvalds * 22811da177e4SLinus Torvalds * Pre-0.97 versions of mount() didn't have a flags word. 22821da177e4SLinus Torvalds * When the flags word was introduced its top half was required 22831da177e4SLinus Torvalds * to have the magic value 0xC0ED, and this remained so until 2.4.0-test9. 22841da177e4SLinus Torvalds * Therefore, if this magic number is present, it carries no information 22851da177e4SLinus Torvalds * and must be discarded. 22861da177e4SLinus Torvalds */ 22871da177e4SLinus Torvalds long do_mount(char *dev_name, char *dir_name, char *type_page, 22881da177e4SLinus Torvalds unsigned long flags, void *data_page) 22891da177e4SLinus Torvalds { 22902d92ab3cSAl Viro struct path path; 22911da177e4SLinus Torvalds int retval = 0; 22921da177e4SLinus Torvalds int mnt_flags = 0; 22931da177e4SLinus Torvalds 22941da177e4SLinus Torvalds /* Discard magic */ 22951da177e4SLinus Torvalds if ((flags & MS_MGC_MSK) == MS_MGC_VAL) 22961da177e4SLinus Torvalds flags &= ~MS_MGC_MSK; 22971da177e4SLinus Torvalds 22981da177e4SLinus Torvalds /* Basic sanity checks */ 22991da177e4SLinus Torvalds 23001da177e4SLinus Torvalds if (!dir_name || !*dir_name || !memchr(dir_name, 0, PAGE_SIZE)) 23011da177e4SLinus Torvalds return -EINVAL; 23021da177e4SLinus Torvalds 23031da177e4SLinus Torvalds if (data_page) 23041da177e4SLinus Torvalds ((char *)data_page)[PAGE_SIZE - 1] = 0; 23051da177e4SLinus Torvalds 2306a27ab9f2STetsuo Handa /* ... and get the mountpoint */ 2307a27ab9f2STetsuo Handa retval = kern_path(dir_name, LOOKUP_FOLLOW, &path); 2308a27ab9f2STetsuo Handa if (retval) 2309a27ab9f2STetsuo Handa return retval; 2310a27ab9f2STetsuo Handa 2311a27ab9f2STetsuo Handa retval = security_sb_mount(dev_name, &path, 2312a27ab9f2STetsuo Handa type_page, flags, data_page); 2313a27ab9f2STetsuo Handa if (retval) 2314a27ab9f2STetsuo Handa goto dput_out; 2315a27ab9f2STetsuo Handa 2316613cbe3dSAndi Kleen /* Default to relatime unless overriden */ 2317613cbe3dSAndi Kleen if (!(flags & MS_NOATIME)) 23180a1c01c9SMatthew Garrett mnt_flags |= MNT_RELATIME; 23190a1c01c9SMatthew Garrett 23201da177e4SLinus Torvalds /* Separate the per-mountpoint flags */ 23211da177e4SLinus Torvalds if (flags & MS_NOSUID) 23221da177e4SLinus Torvalds mnt_flags |= MNT_NOSUID; 23231da177e4SLinus Torvalds if (flags & MS_NODEV) 23241da177e4SLinus Torvalds mnt_flags |= MNT_NODEV; 23251da177e4SLinus Torvalds if (flags & MS_NOEXEC) 23261da177e4SLinus Torvalds mnt_flags |= MNT_NOEXEC; 2327fc33a7bbSChristoph Hellwig if (flags & MS_NOATIME) 2328fc33a7bbSChristoph Hellwig mnt_flags |= MNT_NOATIME; 2329fc33a7bbSChristoph Hellwig if (flags & MS_NODIRATIME) 2330fc33a7bbSChristoph Hellwig mnt_flags |= MNT_NODIRATIME; 2331d0adde57SMatthew Garrett if (flags & MS_STRICTATIME) 2332d0adde57SMatthew Garrett mnt_flags &= ~(MNT_RELATIME | MNT_NOATIME); 23332e4b7fcdSDave Hansen if (flags & MS_RDONLY) 23342e4b7fcdSDave Hansen mnt_flags |= MNT_READONLY; 2335fc33a7bbSChristoph Hellwig 23367a4dec53SAl Viro flags &= ~(MS_NOSUID | MS_NOEXEC | MS_NODEV | MS_ACTIVE | MS_BORN | 2337d0adde57SMatthew Garrett MS_NOATIME | MS_NODIRATIME | MS_RELATIME| MS_KERNMOUNT | 2338d0adde57SMatthew Garrett MS_STRICTATIME); 23391da177e4SLinus Torvalds 23401da177e4SLinus Torvalds if (flags & MS_REMOUNT) 23412d92ab3cSAl Viro retval = do_remount(&path, flags & ~MS_REMOUNT, mnt_flags, 23421da177e4SLinus Torvalds data_page); 23431da177e4SLinus Torvalds else if (flags & MS_BIND) 23442d92ab3cSAl Viro retval = do_loopback(&path, dev_name, flags & MS_REC); 23459676f0c6SRam Pai else if (flags & (MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE)) 23462d92ab3cSAl Viro retval = do_change_type(&path, flags); 23471da177e4SLinus Torvalds else if (flags & MS_MOVE) 23482d92ab3cSAl Viro retval = do_move_mount(&path, dev_name); 23491da177e4SLinus Torvalds else 23502d92ab3cSAl Viro retval = do_new_mount(&path, type_page, flags, mnt_flags, 23511da177e4SLinus Torvalds dev_name, data_page); 23521da177e4SLinus Torvalds dput_out: 23532d92ab3cSAl Viro path_put(&path); 23541da177e4SLinus Torvalds return retval; 23551da177e4SLinus Torvalds } 23561da177e4SLinus Torvalds 2357cf8d2c11STrond Myklebust static struct mnt_namespace *alloc_mnt_ns(void) 2358cf8d2c11STrond Myklebust { 2359cf8d2c11STrond Myklebust struct mnt_namespace *new_ns; 2360cf8d2c11STrond Myklebust 2361cf8d2c11STrond Myklebust new_ns = kmalloc(sizeof(struct mnt_namespace), GFP_KERNEL); 2362cf8d2c11STrond Myklebust if (!new_ns) 2363cf8d2c11STrond Myklebust return ERR_PTR(-ENOMEM); 2364cf8d2c11STrond Myklebust atomic_set(&new_ns->count, 1); 2365cf8d2c11STrond Myklebust new_ns->root = NULL; 2366cf8d2c11STrond Myklebust INIT_LIST_HEAD(&new_ns->list); 2367cf8d2c11STrond Myklebust init_waitqueue_head(&new_ns->poll); 2368cf8d2c11STrond Myklebust new_ns->event = 0; 2369cf8d2c11STrond Myklebust return new_ns; 2370cf8d2c11STrond Myklebust } 2371cf8d2c11STrond Myklebust 2372f03c6599SAl Viro void mnt_make_longterm(struct vfsmount *mnt) 2373f03c6599SAl Viro { 237483adc753SAl Viro __mnt_make_longterm(real_mount(mnt)); 2375f03c6599SAl Viro } 2376f03c6599SAl Viro 237783adc753SAl Viro void mnt_make_shortterm(struct vfsmount *m) 2378f03c6599SAl Viro { 23797e3d0eb0SAl Viro #ifdef CONFIG_SMP 238083adc753SAl Viro struct mount *mnt = real_mount(m); 238168e8a9feSAl Viro if (atomic_add_unless(&mnt->mnt_longterm, -1, 1)) 2382f03c6599SAl Viro return; 2383f03c6599SAl Viro br_write_lock(vfsmount_lock); 238468e8a9feSAl Viro atomic_dec(&mnt->mnt_longterm); 2385f03c6599SAl Viro br_write_unlock(vfsmount_lock); 23867e3d0eb0SAl Viro #endif 2387f03c6599SAl Viro } 2388f03c6599SAl Viro 2389741a2951SJANAK DESAI /* 2390741a2951SJANAK DESAI * Allocate a new namespace structure and populate it with contents 2391741a2951SJANAK DESAI * copied from the namespace of the passed in task structure. 2392741a2951SJANAK DESAI */ 2393e3222c4eSBadari Pulavarty static struct mnt_namespace *dup_mnt_ns(struct mnt_namespace *mnt_ns, 23946b3286edSKirill Korotaev struct fs_struct *fs) 23951da177e4SLinus Torvalds { 23966b3286edSKirill Korotaev struct mnt_namespace *new_ns; 23977f2da1e7SAl Viro struct vfsmount *rootmnt = NULL, *pwdmnt = NULL; 2398315fc83eSAl Viro struct mount *p, *q; 2399cb338d06SAl Viro struct mount *new; 24001da177e4SLinus Torvalds 2401cf8d2c11STrond Myklebust new_ns = alloc_mnt_ns(); 2402cf8d2c11STrond Myklebust if (IS_ERR(new_ns)) 2403cf8d2c11STrond Myklebust return new_ns; 24041da177e4SLinus Torvalds 2405390c6843SRam Pai down_write(&namespace_sem); 24061da177e4SLinus Torvalds /* First pass: copy the tree topology */ 240787129cc0SAl Viro new = copy_tree(real_mount(mnt_ns->root), mnt_ns->root->mnt_root, 24089676f0c6SRam Pai CL_COPY_ALL | CL_EXPIRE); 2409cb338d06SAl Viro if (!new) { 2410390c6843SRam Pai up_write(&namespace_sem); 24111da177e4SLinus Torvalds kfree(new_ns); 24125cc4a034SJulia Lawall return ERR_PTR(-ENOMEM); 24131da177e4SLinus Torvalds } 2414cb338d06SAl Viro new_ns->root = &new->mnt; 241599b7db7bSNick Piggin br_write_lock(vfsmount_lock); 24161da177e4SLinus Torvalds list_add_tail(&new_ns->list, &new_ns->root->mnt_list); 241799b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 24181da177e4SLinus Torvalds 24191da177e4SLinus Torvalds /* 24201da177e4SLinus Torvalds * Second pass: switch the tsk->fs->* elements and mark new vfsmounts 24211da177e4SLinus Torvalds * as belonging to new namespace. We have already acquired a private 24221da177e4SLinus Torvalds * fs_struct, so tsk->fs->lock is not needed. 24231da177e4SLinus Torvalds */ 2424315fc83eSAl Viro p = real_mount(mnt_ns->root); 2425cb338d06SAl Viro q = new; 24261da177e4SLinus Torvalds while (p) { 2427143c8c91SAl Viro q->mnt_ns = new_ns; 242883adc753SAl Viro __mnt_make_longterm(q); 24291da177e4SLinus Torvalds if (fs) { 2430315fc83eSAl Viro if (&p->mnt == fs->root.mnt) { 2431315fc83eSAl Viro fs->root.mnt = mntget(&q->mnt); 243283adc753SAl Viro __mnt_make_longterm(q); 2433315fc83eSAl Viro mnt_make_shortterm(&p->mnt); 2434315fc83eSAl Viro rootmnt = &p->mnt; 24351da177e4SLinus Torvalds } 2436315fc83eSAl Viro if (&p->mnt == fs->pwd.mnt) { 2437315fc83eSAl Viro fs->pwd.mnt = mntget(&q->mnt); 243883adc753SAl Viro __mnt_make_longterm(q); 2439315fc83eSAl Viro mnt_make_shortterm(&p->mnt); 2440315fc83eSAl Viro pwdmnt = &p->mnt; 24411da177e4SLinus Torvalds } 24421da177e4SLinus Torvalds } 24436b3286edSKirill Korotaev p = next_mnt(p, mnt_ns->root); 24441da177e4SLinus Torvalds q = next_mnt(q, new_ns->root); 24451da177e4SLinus Torvalds } 2446390c6843SRam Pai up_write(&namespace_sem); 24471da177e4SLinus Torvalds 24481da177e4SLinus Torvalds if (rootmnt) 2449f03c6599SAl Viro mntput(rootmnt); 24501da177e4SLinus Torvalds if (pwdmnt) 2451f03c6599SAl Viro mntput(pwdmnt); 24521da177e4SLinus Torvalds 2453741a2951SJANAK DESAI return new_ns; 2454741a2951SJANAK DESAI } 2455741a2951SJANAK DESAI 2456213dd266SEric W. Biederman struct mnt_namespace *copy_mnt_ns(unsigned long flags, struct mnt_namespace *ns, 2457e3222c4eSBadari Pulavarty struct fs_struct *new_fs) 2458741a2951SJANAK DESAI { 24596b3286edSKirill Korotaev struct mnt_namespace *new_ns; 2460741a2951SJANAK DESAI 2461e3222c4eSBadari Pulavarty BUG_ON(!ns); 24626b3286edSKirill Korotaev get_mnt_ns(ns); 2463741a2951SJANAK DESAI 2464741a2951SJANAK DESAI if (!(flags & CLONE_NEWNS)) 2465e3222c4eSBadari Pulavarty return ns; 2466741a2951SJANAK DESAI 2467e3222c4eSBadari Pulavarty new_ns = dup_mnt_ns(ns, new_fs); 2468741a2951SJANAK DESAI 24696b3286edSKirill Korotaev put_mnt_ns(ns); 2470e3222c4eSBadari Pulavarty return new_ns; 24711da177e4SLinus Torvalds } 24721da177e4SLinus Torvalds 2473cf8d2c11STrond Myklebust /** 2474cf8d2c11STrond Myklebust * create_mnt_ns - creates a private namespace and adds a root filesystem 2475cf8d2c11STrond Myklebust * @mnt: pointer to the new root filesystem mountpoint 2476cf8d2c11STrond Myklebust */ 24776c449c8dSAl Viro static struct mnt_namespace *create_mnt_ns(struct vfsmount *mnt) 2478cf8d2c11STrond Myklebust { 2479cf8d2c11STrond Myklebust struct mnt_namespace *new_ns; 2480cf8d2c11STrond Myklebust 2481cf8d2c11STrond Myklebust new_ns = alloc_mnt_ns(); 2482cf8d2c11STrond Myklebust if (!IS_ERR(new_ns)) { 2483143c8c91SAl Viro real_mount(mnt)->mnt_ns = new_ns; 248483adc753SAl Viro __mnt_make_longterm(real_mount(mnt)); 2485cf8d2c11STrond Myklebust new_ns->root = mnt; 2486cf8d2c11STrond Myklebust list_add(&new_ns->list, &new_ns->root->mnt_list); 2487c1334495SAl Viro } else { 2488c1334495SAl Viro mntput(mnt); 2489cf8d2c11STrond Myklebust } 2490cf8d2c11STrond Myklebust return new_ns; 2491cf8d2c11STrond Myklebust } 2492cf8d2c11STrond Myklebust 2493ea441d11SAl Viro struct dentry *mount_subtree(struct vfsmount *mnt, const char *name) 2494ea441d11SAl Viro { 2495ea441d11SAl Viro struct mnt_namespace *ns; 2496d31da0f0SAl Viro struct super_block *s; 2497ea441d11SAl Viro struct path path; 2498ea441d11SAl Viro int err; 2499ea441d11SAl Viro 2500ea441d11SAl Viro ns = create_mnt_ns(mnt); 2501ea441d11SAl Viro if (IS_ERR(ns)) 2502ea441d11SAl Viro return ERR_CAST(ns); 2503ea441d11SAl Viro 2504ea441d11SAl Viro err = vfs_path_lookup(mnt->mnt_root, mnt, 2505ea441d11SAl Viro name, LOOKUP_FOLLOW|LOOKUP_AUTOMOUNT, &path); 2506ea441d11SAl Viro 2507ea441d11SAl Viro put_mnt_ns(ns); 2508ea441d11SAl Viro 2509ea441d11SAl Viro if (err) 2510ea441d11SAl Viro return ERR_PTR(err); 2511ea441d11SAl Viro 2512ea441d11SAl Viro /* trade a vfsmount reference for active sb one */ 2513d31da0f0SAl Viro s = path.mnt->mnt_sb; 2514d31da0f0SAl Viro atomic_inc(&s->s_active); 2515ea441d11SAl Viro mntput(path.mnt); 2516ea441d11SAl Viro /* lock the sucker */ 2517d31da0f0SAl Viro down_write(&s->s_umount); 2518ea441d11SAl Viro /* ... and return the root of (sub)tree on it */ 2519ea441d11SAl Viro return path.dentry; 2520ea441d11SAl Viro } 2521ea441d11SAl Viro EXPORT_SYMBOL(mount_subtree); 2522ea441d11SAl Viro 2523bdc480e3SHeiko Carstens SYSCALL_DEFINE5(mount, char __user *, dev_name, char __user *, dir_name, 2524bdc480e3SHeiko Carstens char __user *, type, unsigned long, flags, void __user *, data) 25251da177e4SLinus Torvalds { 2526eca6f534SVegard Nossum int ret; 2527eca6f534SVegard Nossum char *kernel_type; 2528eca6f534SVegard Nossum char *kernel_dir; 2529eca6f534SVegard Nossum char *kernel_dev; 25301da177e4SLinus Torvalds unsigned long data_page; 25311da177e4SLinus Torvalds 2532eca6f534SVegard Nossum ret = copy_mount_string(type, &kernel_type); 2533eca6f534SVegard Nossum if (ret < 0) 2534eca6f534SVegard Nossum goto out_type; 25351da177e4SLinus Torvalds 2536eca6f534SVegard Nossum kernel_dir = getname(dir_name); 2537eca6f534SVegard Nossum if (IS_ERR(kernel_dir)) { 2538eca6f534SVegard Nossum ret = PTR_ERR(kernel_dir); 2539eca6f534SVegard Nossum goto out_dir; 2540eca6f534SVegard Nossum } 25411da177e4SLinus Torvalds 2542eca6f534SVegard Nossum ret = copy_mount_string(dev_name, &kernel_dev); 2543eca6f534SVegard Nossum if (ret < 0) 2544eca6f534SVegard Nossum goto out_dev; 25451da177e4SLinus Torvalds 2546eca6f534SVegard Nossum ret = copy_mount_options(data, &data_page); 2547eca6f534SVegard Nossum if (ret < 0) 2548eca6f534SVegard Nossum goto out_data; 25491da177e4SLinus Torvalds 2550eca6f534SVegard Nossum ret = do_mount(kernel_dev, kernel_dir, kernel_type, flags, 2551eca6f534SVegard Nossum (void *) data_page); 2552eca6f534SVegard Nossum 25531da177e4SLinus Torvalds free_page(data_page); 2554eca6f534SVegard Nossum out_data: 2555eca6f534SVegard Nossum kfree(kernel_dev); 2556eca6f534SVegard Nossum out_dev: 2557eca6f534SVegard Nossum putname(kernel_dir); 2558eca6f534SVegard Nossum out_dir: 2559eca6f534SVegard Nossum kfree(kernel_type); 2560eca6f534SVegard Nossum out_type: 2561eca6f534SVegard Nossum return ret; 25621da177e4SLinus Torvalds } 25631da177e4SLinus Torvalds 25641da177e4SLinus Torvalds /* 2565afac7cbaSAl Viro * Return true if path is reachable from root 2566afac7cbaSAl Viro * 2567afac7cbaSAl Viro * namespace_sem or vfsmount_lock is held 2568afac7cbaSAl Viro */ 2569643822b4SAl Viro bool is_path_reachable(struct mount *mnt, struct dentry *dentry, 2570afac7cbaSAl Viro const struct path *root) 2571afac7cbaSAl Viro { 2572643822b4SAl Viro while (&mnt->mnt != root->mnt && mnt_has_parent(mnt)) { 2573a73324daSAl Viro dentry = mnt->mnt_mountpoint; 25740714a533SAl Viro mnt = mnt->mnt_parent; 2575afac7cbaSAl Viro } 2576643822b4SAl Viro return &mnt->mnt == root->mnt && is_subdir(dentry, root->dentry); 2577afac7cbaSAl Viro } 2578afac7cbaSAl Viro 2579afac7cbaSAl Viro int path_is_under(struct path *path1, struct path *path2) 2580afac7cbaSAl Viro { 2581afac7cbaSAl Viro int res; 2582afac7cbaSAl Viro br_read_lock(vfsmount_lock); 2583643822b4SAl Viro res = is_path_reachable(real_mount(path1->mnt), path1->dentry, path2); 2584afac7cbaSAl Viro br_read_unlock(vfsmount_lock); 2585afac7cbaSAl Viro return res; 2586afac7cbaSAl Viro } 2587afac7cbaSAl Viro EXPORT_SYMBOL(path_is_under); 2588afac7cbaSAl Viro 2589afac7cbaSAl Viro /* 25901da177e4SLinus Torvalds * pivot_root Semantics: 25911da177e4SLinus Torvalds * Moves the root file system of the current process to the directory put_old, 25921da177e4SLinus Torvalds * makes new_root as the new root file system of the current process, and sets 25931da177e4SLinus Torvalds * root/cwd of all processes which had them on the current root to new_root. 25941da177e4SLinus Torvalds * 25951da177e4SLinus Torvalds * Restrictions: 25961da177e4SLinus Torvalds * The new_root and put_old must be directories, and must not be on the 25971da177e4SLinus Torvalds * same file system as the current process root. The put_old must be 25981da177e4SLinus Torvalds * underneath new_root, i.e. adding a non-zero number of /.. to the string 25991da177e4SLinus Torvalds * pointed to by put_old must yield the same directory as new_root. No other 26001da177e4SLinus Torvalds * file system may be mounted on put_old. After all, new_root is a mountpoint. 26011da177e4SLinus Torvalds * 26024a0d11faSNeil Brown * Also, the current root cannot be on the 'rootfs' (initial ramfs) filesystem. 26034a0d11faSNeil Brown * See Documentation/filesystems/ramfs-rootfs-initramfs.txt for alternatives 26044a0d11faSNeil Brown * in this situation. 26054a0d11faSNeil Brown * 26061da177e4SLinus Torvalds * Notes: 26071da177e4SLinus Torvalds * - we don't move root/cwd if they are not at the root (reason: if something 26081da177e4SLinus Torvalds * cared enough to change them, it's probably wrong to force them elsewhere) 26091da177e4SLinus Torvalds * - it's okay to pick a root that isn't the root of a file system, e.g. 26101da177e4SLinus Torvalds * /nfs/my_root where /nfs is the mount point. It must be a mountpoint, 26111da177e4SLinus Torvalds * though, so you may need to say mount --bind /nfs/my_root /nfs/my_root 26121da177e4SLinus Torvalds * first. 26131da177e4SLinus Torvalds */ 26143480b257SHeiko Carstens SYSCALL_DEFINE2(pivot_root, const char __user *, new_root, 26153480b257SHeiko Carstens const char __user *, put_old) 26161da177e4SLinus Torvalds { 26172d8f3038SAl Viro struct path new, old, parent_path, root_parent, root; 2618419148daSAl Viro struct mount *new_mnt, *root_mnt; 26191da177e4SLinus Torvalds int error; 26201da177e4SLinus Torvalds 26211da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 26221da177e4SLinus Torvalds return -EPERM; 26231da177e4SLinus Torvalds 26242d8f3038SAl Viro error = user_path_dir(new_root, &new); 26251da177e4SLinus Torvalds if (error) 26261da177e4SLinus Torvalds goto out0; 26271da177e4SLinus Torvalds 26282d8f3038SAl Viro error = user_path_dir(put_old, &old); 26291da177e4SLinus Torvalds if (error) 26301da177e4SLinus Torvalds goto out1; 26311da177e4SLinus Torvalds 26322d8f3038SAl Viro error = security_sb_pivotroot(&old, &new); 2633b12cea91SAl Viro if (error) 2634b12cea91SAl Viro goto out2; 26351da177e4SLinus Torvalds 2636f7ad3c6bSMiklos Szeredi get_fs_root(current->fs, &root); 2637b12cea91SAl Viro error = lock_mount(&old); 2638b12cea91SAl Viro if (error) 2639b12cea91SAl Viro goto out3; 2640b12cea91SAl Viro 26411da177e4SLinus Torvalds error = -EINVAL; 2642419148daSAl Viro new_mnt = real_mount(new.mnt); 2643419148daSAl Viro root_mnt = real_mount(root.mnt); 26442d8f3038SAl Viro if (IS_MNT_SHARED(old.mnt) || 26450714a533SAl Viro IS_MNT_SHARED(&new_mnt->mnt_parent->mnt) || 26460714a533SAl Viro IS_MNT_SHARED(&root_mnt->mnt_parent->mnt)) 2647b12cea91SAl Viro goto out4; 2648143c8c91SAl Viro if (!check_mnt(root_mnt) || !check_mnt(new_mnt)) 2649b12cea91SAl Viro goto out4; 26501da177e4SLinus Torvalds error = -ENOENT; 2651f3da392eSAlexey Dobriyan if (d_unlinked(new.dentry)) 2652b12cea91SAl Viro goto out4; 2653f3da392eSAlexey Dobriyan if (d_unlinked(old.dentry)) 2654b12cea91SAl Viro goto out4; 26551da177e4SLinus Torvalds error = -EBUSY; 26562d8f3038SAl Viro if (new.mnt == root.mnt || 26572d8f3038SAl Viro old.mnt == root.mnt) 2658b12cea91SAl Viro goto out4; /* loop, on the same file system */ 26591da177e4SLinus Torvalds error = -EINVAL; 26608c3ee42eSAl Viro if (root.mnt->mnt_root != root.dentry) 2661b12cea91SAl Viro goto out4; /* not a mountpoint */ 2662676da58dSAl Viro if (!mnt_has_parent(root_mnt)) 2663b12cea91SAl Viro goto out4; /* not attached */ 26642d8f3038SAl Viro if (new.mnt->mnt_root != new.dentry) 2665b12cea91SAl Viro goto out4; /* not a mountpoint */ 2666676da58dSAl Viro if (!mnt_has_parent(new_mnt)) 2667b12cea91SAl Viro goto out4; /* not attached */ 26684ac91378SJan Blunck /* make sure we can reach put_old from new_root */ 2669643822b4SAl Viro if (!is_path_reachable(real_mount(old.mnt), old.dentry, &new)) 2670b12cea91SAl Viro goto out4; 267127cb1572SAl Viro br_write_lock(vfsmount_lock); 2672419148daSAl Viro detach_mnt(new_mnt, &parent_path); 2673419148daSAl Viro detach_mnt(root_mnt, &root_parent); 26744ac91378SJan Blunck /* mount old root on put_old */ 2675419148daSAl Viro attach_mnt(root_mnt, &old); 26764ac91378SJan Blunck /* mount new_root on / */ 2677419148daSAl Viro attach_mnt(new_mnt, &root_parent); 26786b3286edSKirill Korotaev touch_mnt_namespace(current->nsproxy->mnt_ns); 267999b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 26802d8f3038SAl Viro chroot_fs_refs(&root, &new); 26811da177e4SLinus Torvalds error = 0; 2682b12cea91SAl Viro out4: 2683b12cea91SAl Viro unlock_mount(&old); 2684b12cea91SAl Viro if (!error) { 26851a390689SAl Viro path_put(&root_parent); 26861a390689SAl Viro path_put(&parent_path); 2687b12cea91SAl Viro } 2688b12cea91SAl Viro out3: 26898c3ee42eSAl Viro path_put(&root); 2690b12cea91SAl Viro out2: 26912d8f3038SAl Viro path_put(&old); 26921da177e4SLinus Torvalds out1: 26932d8f3038SAl Viro path_put(&new); 26941da177e4SLinus Torvalds out0: 26951da177e4SLinus Torvalds return error; 26961da177e4SLinus Torvalds } 26971da177e4SLinus Torvalds 26981da177e4SLinus Torvalds static void __init init_mount_tree(void) 26991da177e4SLinus Torvalds { 27001da177e4SLinus Torvalds struct vfsmount *mnt; 27016b3286edSKirill Korotaev struct mnt_namespace *ns; 2702ac748a09SJan Blunck struct path root; 27031da177e4SLinus Torvalds 27041da177e4SLinus Torvalds mnt = do_kern_mount("rootfs", 0, "rootfs", NULL); 27051da177e4SLinus Torvalds if (IS_ERR(mnt)) 27061da177e4SLinus Torvalds panic("Can't create rootfs"); 2707b3e19d92SNick Piggin 27083b22edc5STrond Myklebust ns = create_mnt_ns(mnt); 27093b22edc5STrond Myklebust if (IS_ERR(ns)) 27101da177e4SLinus Torvalds panic("Can't allocate initial namespace"); 27111da177e4SLinus Torvalds 27126b3286edSKirill Korotaev init_task.nsproxy->mnt_ns = ns; 27136b3286edSKirill Korotaev get_mnt_ns(ns); 27141da177e4SLinus Torvalds 2715ac748a09SJan Blunck root.mnt = ns->root; 2716ac748a09SJan Blunck root.dentry = ns->root->mnt_root; 2717ac748a09SJan Blunck 2718ac748a09SJan Blunck set_fs_pwd(current->fs, &root); 2719ac748a09SJan Blunck set_fs_root(current->fs, &root); 27201da177e4SLinus Torvalds } 27211da177e4SLinus Torvalds 272274bf17cfSDenis Cheng void __init mnt_init(void) 27231da177e4SLinus Torvalds { 272413f14b4dSEric Dumazet unsigned u; 272515a67dd8SRandy Dunlap int err; 27261da177e4SLinus Torvalds 2727390c6843SRam Pai init_rwsem(&namespace_sem); 2728390c6843SRam Pai 27297d6fec45SAl Viro mnt_cache = kmem_cache_create("mnt_cache", sizeof(struct mount), 273020c2df83SPaul Mundt 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC, NULL); 27311da177e4SLinus Torvalds 2732b58fed8bSRam Pai mount_hashtable = (struct list_head *)__get_free_page(GFP_ATOMIC); 27331da177e4SLinus Torvalds 27341da177e4SLinus Torvalds if (!mount_hashtable) 27351da177e4SLinus Torvalds panic("Failed to allocate mount hash table\n"); 27361da177e4SLinus Torvalds 273780cdc6daSMandeep Singh Baines printk(KERN_INFO "Mount-cache hash table entries: %lu\n", HASH_SIZE); 27381da177e4SLinus Torvalds 273913f14b4dSEric Dumazet for (u = 0; u < HASH_SIZE; u++) 274013f14b4dSEric Dumazet INIT_LIST_HEAD(&mount_hashtable[u]); 27411da177e4SLinus Torvalds 274299b7db7bSNick Piggin br_lock_init(vfsmount_lock); 274399b7db7bSNick Piggin 274415a67dd8SRandy Dunlap err = sysfs_init(); 274515a67dd8SRandy Dunlap if (err) 274615a67dd8SRandy Dunlap printk(KERN_WARNING "%s: sysfs_init error: %d\n", 27478e24eea7SHarvey Harrison __func__, err); 274800d26666SGreg Kroah-Hartman fs_kobj = kobject_create_and_add("fs", NULL); 274900d26666SGreg Kroah-Hartman if (!fs_kobj) 27508e24eea7SHarvey Harrison printk(KERN_WARNING "%s: kobj create error\n", __func__); 27511da177e4SLinus Torvalds init_rootfs(); 27521da177e4SLinus Torvalds init_mount_tree(); 27531da177e4SLinus Torvalds } 27541da177e4SLinus Torvalds 2755616511d0STrond Myklebust void put_mnt_ns(struct mnt_namespace *ns) 27561da177e4SLinus Torvalds { 275770fbcdf4SRam Pai LIST_HEAD(umount_list); 2758616511d0STrond Myklebust 2759d498b25aSAl Viro if (!atomic_dec_and_test(&ns->count)) 2760616511d0STrond Myklebust return; 2761390c6843SRam Pai down_write(&namespace_sem); 276299b7db7bSNick Piggin br_write_lock(vfsmount_lock); 2763761d5c38SAl Viro umount_tree(real_mount(ns->root), 0, &umount_list); 276499b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 2765390c6843SRam Pai up_write(&namespace_sem); 276670fbcdf4SRam Pai release_mounts(&umount_list); 27676b3286edSKirill Korotaev kfree(ns); 27681da177e4SLinus Torvalds } 27699d412a43SAl Viro 27709d412a43SAl Viro struct vfsmount *kern_mount_data(struct file_system_type *type, void *data) 27719d412a43SAl Viro { 2772423e0ab0STim Chen struct vfsmount *mnt; 2773423e0ab0STim Chen mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, data); 2774423e0ab0STim Chen if (!IS_ERR(mnt)) { 2775423e0ab0STim Chen /* 2776423e0ab0STim Chen * it is a longterm mount, don't release mnt until 2777423e0ab0STim Chen * we unmount before file sys is unregistered 2778423e0ab0STim Chen */ 2779423e0ab0STim Chen mnt_make_longterm(mnt); 2780423e0ab0STim Chen } 2781423e0ab0STim Chen return mnt; 27829d412a43SAl Viro } 27839d412a43SAl Viro EXPORT_SYMBOL_GPL(kern_mount_data); 2784423e0ab0STim Chen 2785423e0ab0STim Chen void kern_unmount(struct vfsmount *mnt) 2786423e0ab0STim Chen { 2787423e0ab0STim Chen /* release long term mount so mount point can be released */ 2788423e0ab0STim Chen if (!IS_ERR_OR_NULL(mnt)) { 2789423e0ab0STim Chen mnt_make_shortterm(mnt); 2790423e0ab0STim Chen mntput(mnt); 2791423e0ab0STim Chen } 2792423e0ab0STim Chen } 2793423e0ab0STim Chen EXPORT_SYMBOL(kern_unmount); 279402125a82SAl Viro 279502125a82SAl Viro bool our_mnt(struct vfsmount *mnt) 279602125a82SAl Viro { 2797143c8c91SAl Viro return check_mnt(real_mount(mnt)); 279802125a82SAl Viro } 2799