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 */ 8173cd49ecSMiklos Szeredi static int mnt_alloc_id(struct vfsmount *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); 88f21f6220SAl Viro res = ida_get_new_above(&mnt_id_ida, mnt_id_start, &mnt->mnt_id); 89f21f6220SAl Viro if (!res) 90f21f6220SAl Viro mnt_id_start = 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 9873cd49ecSMiklos Szeredi static void mnt_free_id(struct vfsmount *mnt) 9973cd49ecSMiklos Szeredi { 100f21f6220SAl Viro int id = 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 */ 113719f5d7fSMiklos Szeredi static int mnt_alloc_group_id(struct vfsmount *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, 122f21f6220SAl Viro &mnt->mnt_group_id); 123f21f6220SAl Viro if (!res) 124f21f6220SAl Viro mnt_group_start = 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 */ 132719f5d7fSMiklos Szeredi void mnt_release_group_id(struct vfsmount *mnt) 133719f5d7fSMiklos Szeredi { 134f21f6220SAl Viro int id = 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; 138719f5d7fSMiklos Szeredi mnt->mnt_group_id = 0; 139719f5d7fSMiklos Szeredi } 140719f5d7fSMiklos Szeredi 141b3e19d92SNick Piggin /* 142b3e19d92SNick Piggin * vfsmount lock must be held for read 143b3e19d92SNick Piggin */ 144b3e19d92SNick Piggin static inline void mnt_add_count(struct vfsmount *mnt, int n) 145b3e19d92SNick Piggin { 146b3e19d92SNick Piggin #ifdef CONFIG_SMP 147b3e19d92SNick Piggin this_cpu_add(mnt->mnt_pcp->mnt_count, n); 148b3e19d92SNick Piggin #else 149b3e19d92SNick Piggin preempt_disable(); 150b3e19d92SNick Piggin mnt->mnt_count += n; 151b3e19d92SNick Piggin preempt_enable(); 152b3e19d92SNick Piggin #endif 153b3e19d92SNick Piggin } 154b3e19d92SNick Piggin 155b3e19d92SNick Piggin static inline void mnt_set_count(struct vfsmount *mnt, int n) 156b3e19d92SNick Piggin { 157b3e19d92SNick Piggin #ifdef CONFIG_SMP 158b3e19d92SNick Piggin this_cpu_write(mnt->mnt_pcp->mnt_count, n); 159b3e19d92SNick Piggin #else 160b3e19d92SNick Piggin mnt->mnt_count = n; 161b3e19d92SNick Piggin #endif 162b3e19d92SNick Piggin } 163b3e19d92SNick Piggin 164b3e19d92SNick Piggin /* 165b3e19d92SNick Piggin * vfsmount lock must be held for read 166b3e19d92SNick Piggin */ 167b3e19d92SNick Piggin static inline void mnt_inc_count(struct vfsmount *mnt) 168b3e19d92SNick Piggin { 169b3e19d92SNick Piggin mnt_add_count(mnt, 1); 170b3e19d92SNick Piggin } 171b3e19d92SNick Piggin 172b3e19d92SNick Piggin /* 173b3e19d92SNick Piggin * vfsmount lock must be held for read 174b3e19d92SNick Piggin */ 175b3e19d92SNick Piggin static inline void mnt_dec_count(struct vfsmount *mnt) 176b3e19d92SNick Piggin { 177b3e19d92SNick Piggin mnt_add_count(mnt, -1); 178b3e19d92SNick Piggin } 179b3e19d92SNick Piggin 180b3e19d92SNick Piggin /* 181b3e19d92SNick Piggin * vfsmount lock must be held for write 182b3e19d92SNick Piggin */ 183b3e19d92SNick Piggin unsigned int mnt_get_count(struct vfsmount *mnt) 184b3e19d92SNick Piggin { 185b3e19d92SNick Piggin #ifdef CONFIG_SMP 186f03c6599SAl Viro unsigned int count = 0; 187b3e19d92SNick Piggin int cpu; 188b3e19d92SNick Piggin 189b3e19d92SNick Piggin for_each_possible_cpu(cpu) { 190b3e19d92SNick Piggin count += per_cpu_ptr(mnt->mnt_pcp, cpu)->mnt_count; 191b3e19d92SNick Piggin } 192b3e19d92SNick Piggin 193b3e19d92SNick Piggin return count; 194b3e19d92SNick Piggin #else 195b3e19d92SNick Piggin return mnt->mnt_count; 196b3e19d92SNick Piggin #endif 197b3e19d92SNick Piggin } 198b3e19d92SNick Piggin 1999d412a43SAl Viro static struct vfsmount *alloc_vfsmnt(const char *name) 2001da177e4SLinus Torvalds { 201c3762229SRobert P. J. Day struct vfsmount *mnt = kmem_cache_zalloc(mnt_cache, GFP_KERNEL); 2021da177e4SLinus Torvalds if (mnt) { 20373cd49ecSMiklos Szeredi int err; 20473cd49ecSMiklos Szeredi 20573cd49ecSMiklos Szeredi err = mnt_alloc_id(mnt); 20688b38782SLi Zefan if (err) 20788b38782SLi Zefan goto out_free_cache; 20888b38782SLi Zefan 20988b38782SLi Zefan if (name) { 21088b38782SLi Zefan mnt->mnt_devname = kstrdup(name, GFP_KERNEL); 21188b38782SLi Zefan if (!mnt->mnt_devname) 21288b38782SLi Zefan goto out_free_id; 21373cd49ecSMiklos Szeredi } 21473cd49ecSMiklos Szeredi 215b3e19d92SNick Piggin #ifdef CONFIG_SMP 216b3e19d92SNick Piggin mnt->mnt_pcp = alloc_percpu(struct mnt_pcp); 217b3e19d92SNick Piggin if (!mnt->mnt_pcp) 218b3e19d92SNick Piggin goto out_free_devname; 219b3e19d92SNick Piggin 220f03c6599SAl Viro this_cpu_add(mnt->mnt_pcp->mnt_count, 1); 221b3e19d92SNick Piggin #else 222b3e19d92SNick Piggin mnt->mnt_count = 1; 223b3e19d92SNick Piggin mnt->mnt_writers = 0; 224b3e19d92SNick Piggin #endif 225b3e19d92SNick Piggin 2261da177e4SLinus Torvalds INIT_LIST_HEAD(&mnt->mnt_hash); 2271da177e4SLinus Torvalds INIT_LIST_HEAD(&mnt->mnt_child); 2281da177e4SLinus Torvalds INIT_LIST_HEAD(&mnt->mnt_mounts); 2291da177e4SLinus Torvalds INIT_LIST_HEAD(&mnt->mnt_list); 23055e700b9SMiklos Szeredi INIT_LIST_HEAD(&mnt->mnt_expire); 23103e06e68SRam Pai INIT_LIST_HEAD(&mnt->mnt_share); 232a58b0eb8SRam Pai INIT_LIST_HEAD(&mnt->mnt_slave_list); 233a58b0eb8SRam Pai INIT_LIST_HEAD(&mnt->mnt_slave); 2342504c5d6SAndreas Gruenbacher #ifdef CONFIG_FSNOTIFY 2352504c5d6SAndreas Gruenbacher INIT_HLIST_HEAD(&mnt->mnt_fsnotify_marks); 2362504c5d6SAndreas Gruenbacher #endif 2371da177e4SLinus Torvalds } 2381da177e4SLinus Torvalds return mnt; 23988b38782SLi Zefan 240d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 241d3ef3d73Snpiggin@suse.de out_free_devname: 242d3ef3d73Snpiggin@suse.de kfree(mnt->mnt_devname); 243d3ef3d73Snpiggin@suse.de #endif 24488b38782SLi Zefan out_free_id: 24588b38782SLi Zefan mnt_free_id(mnt); 24688b38782SLi Zefan out_free_cache: 24788b38782SLi Zefan kmem_cache_free(mnt_cache, mnt); 24888b38782SLi Zefan return NULL; 2491da177e4SLinus Torvalds } 2501da177e4SLinus Torvalds 2518366025eSDave Hansen /* 2528366025eSDave Hansen * Most r/o checks on a fs are for operations that take 2538366025eSDave Hansen * discrete amounts of time, like a write() or unlink(). 2548366025eSDave Hansen * We must keep track of when those operations start 2558366025eSDave Hansen * (for permission checks) and when they end, so that 2568366025eSDave Hansen * we can determine when writes are able to occur to 2578366025eSDave Hansen * a filesystem. 2588366025eSDave Hansen */ 2593d733633SDave Hansen /* 2603d733633SDave Hansen * __mnt_is_readonly: check whether a mount is read-only 2613d733633SDave Hansen * @mnt: the mount to check for its write status 2623d733633SDave Hansen * 2633d733633SDave Hansen * This shouldn't be used directly ouside of the VFS. 2643d733633SDave Hansen * It does not guarantee that the filesystem will stay 2653d733633SDave Hansen * r/w, just that it is right *now*. This can not and 2663d733633SDave Hansen * should not be used in place of IS_RDONLY(inode). 2673d733633SDave Hansen * mnt_want/drop_write() will _keep_ the filesystem 2683d733633SDave Hansen * r/w. 2693d733633SDave Hansen */ 2703d733633SDave Hansen int __mnt_is_readonly(struct vfsmount *mnt) 2713d733633SDave Hansen { 2722e4b7fcdSDave Hansen if (mnt->mnt_flags & MNT_READONLY) 2732e4b7fcdSDave Hansen return 1; 2742e4b7fcdSDave Hansen if (mnt->mnt_sb->s_flags & MS_RDONLY) 2752e4b7fcdSDave Hansen return 1; 2762e4b7fcdSDave Hansen return 0; 2773d733633SDave Hansen } 2783d733633SDave Hansen EXPORT_SYMBOL_GPL(__mnt_is_readonly); 2793d733633SDave Hansen 280c6653a83SNick Piggin static inline void mnt_inc_writers(struct vfsmount *mnt) 2813d733633SDave Hansen { 282d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 283b3e19d92SNick Piggin this_cpu_inc(mnt->mnt_pcp->mnt_writers); 284d3ef3d73Snpiggin@suse.de #else 285d3ef3d73Snpiggin@suse.de mnt->mnt_writers++; 286d3ef3d73Snpiggin@suse.de #endif 2873d733633SDave Hansen } 2883d733633SDave Hansen 289c6653a83SNick Piggin static inline void mnt_dec_writers(struct vfsmount *mnt) 2903d733633SDave Hansen { 291d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 292b3e19d92SNick Piggin this_cpu_dec(mnt->mnt_pcp->mnt_writers); 293d3ef3d73Snpiggin@suse.de #else 294d3ef3d73Snpiggin@suse.de mnt->mnt_writers--; 295d3ef3d73Snpiggin@suse.de #endif 296d3ef3d73Snpiggin@suse.de } 297d3ef3d73Snpiggin@suse.de 298c6653a83SNick Piggin static unsigned int mnt_get_writers(struct vfsmount *mnt) 299d3ef3d73Snpiggin@suse.de { 300d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 301d3ef3d73Snpiggin@suse.de unsigned int count = 0; 3023d733633SDave Hansen int cpu; 3033d733633SDave Hansen 3043d733633SDave Hansen for_each_possible_cpu(cpu) { 305b3e19d92SNick Piggin count += per_cpu_ptr(mnt->mnt_pcp, cpu)->mnt_writers; 3063d733633SDave Hansen } 3073d733633SDave Hansen 308d3ef3d73Snpiggin@suse.de return count; 309d3ef3d73Snpiggin@suse.de #else 310d3ef3d73Snpiggin@suse.de return mnt->mnt_writers; 311d3ef3d73Snpiggin@suse.de #endif 3123d733633SDave Hansen } 3133d733633SDave Hansen 3143d733633SDave Hansen /* 3153d733633SDave Hansen * Most r/o checks on a fs are for operations that take 3163d733633SDave Hansen * discrete amounts of time, like a write() or unlink(). 3173d733633SDave Hansen * We must keep track of when those operations start 3183d733633SDave Hansen * (for permission checks) and when they end, so that 3193d733633SDave Hansen * we can determine when writes are able to occur to 3203d733633SDave Hansen * a filesystem. 3213d733633SDave Hansen */ 3228366025eSDave Hansen /** 3238366025eSDave Hansen * mnt_want_write - get write access to a mount 3248366025eSDave Hansen * @mnt: the mount on which to take a write 3258366025eSDave Hansen * 3268366025eSDave Hansen * This tells the low-level filesystem that a write is 3278366025eSDave Hansen * about to be performed to it, and makes sure that 3288366025eSDave Hansen * writes are allowed before returning success. When 3298366025eSDave Hansen * the write operation is finished, mnt_drop_write() 3308366025eSDave Hansen * must be called. This is effectively a refcount. 3318366025eSDave Hansen */ 3328366025eSDave Hansen int mnt_want_write(struct vfsmount *mnt) 3338366025eSDave Hansen { 3343d733633SDave Hansen int ret = 0; 3353d733633SDave Hansen 336d3ef3d73Snpiggin@suse.de preempt_disable(); 337c6653a83SNick Piggin mnt_inc_writers(mnt); 338d3ef3d73Snpiggin@suse.de /* 339c6653a83SNick Piggin * The store to mnt_inc_writers must be visible before we pass 340d3ef3d73Snpiggin@suse.de * MNT_WRITE_HOLD loop below, so that the slowpath can see our 341d3ef3d73Snpiggin@suse.de * incremented count after it has set MNT_WRITE_HOLD. 342d3ef3d73Snpiggin@suse.de */ 343d3ef3d73Snpiggin@suse.de smp_mb(); 344d3ef3d73Snpiggin@suse.de while (mnt->mnt_flags & MNT_WRITE_HOLD) 345d3ef3d73Snpiggin@suse.de cpu_relax(); 346d3ef3d73Snpiggin@suse.de /* 347d3ef3d73Snpiggin@suse.de * After the slowpath clears MNT_WRITE_HOLD, mnt_is_readonly will 348d3ef3d73Snpiggin@suse.de * be set to match its requirements. So we must not load that until 349d3ef3d73Snpiggin@suse.de * MNT_WRITE_HOLD is cleared. 350d3ef3d73Snpiggin@suse.de */ 351d3ef3d73Snpiggin@suse.de smp_rmb(); 3523d733633SDave Hansen if (__mnt_is_readonly(mnt)) { 353c6653a83SNick Piggin mnt_dec_writers(mnt); 3543d733633SDave Hansen ret = -EROFS; 3553d733633SDave Hansen goto out; 3563d733633SDave Hansen } 3573d733633SDave Hansen out: 358d3ef3d73Snpiggin@suse.de preempt_enable(); 3593d733633SDave Hansen return ret; 3608366025eSDave Hansen } 3618366025eSDave Hansen EXPORT_SYMBOL_GPL(mnt_want_write); 3628366025eSDave Hansen 3638366025eSDave Hansen /** 36496029c4eSnpiggin@suse.de * mnt_clone_write - get write access to a mount 36596029c4eSnpiggin@suse.de * @mnt: the mount on which to take a write 36696029c4eSnpiggin@suse.de * 36796029c4eSnpiggin@suse.de * This is effectively like mnt_want_write, except 36896029c4eSnpiggin@suse.de * it must only be used to take an extra write reference 36996029c4eSnpiggin@suse.de * on a mountpoint that we already know has a write reference 37096029c4eSnpiggin@suse.de * on it. This allows some optimisation. 37196029c4eSnpiggin@suse.de * 37296029c4eSnpiggin@suse.de * After finished, mnt_drop_write must be called as usual to 37396029c4eSnpiggin@suse.de * drop the reference. 37496029c4eSnpiggin@suse.de */ 37596029c4eSnpiggin@suse.de int mnt_clone_write(struct vfsmount *mnt) 37696029c4eSnpiggin@suse.de { 37796029c4eSnpiggin@suse.de /* superblock may be r/o */ 37896029c4eSnpiggin@suse.de if (__mnt_is_readonly(mnt)) 37996029c4eSnpiggin@suse.de return -EROFS; 38096029c4eSnpiggin@suse.de preempt_disable(); 381c6653a83SNick Piggin mnt_inc_writers(mnt); 38296029c4eSnpiggin@suse.de preempt_enable(); 38396029c4eSnpiggin@suse.de return 0; 38496029c4eSnpiggin@suse.de } 38596029c4eSnpiggin@suse.de EXPORT_SYMBOL_GPL(mnt_clone_write); 38696029c4eSnpiggin@suse.de 38796029c4eSnpiggin@suse.de /** 38896029c4eSnpiggin@suse.de * mnt_want_write_file - get write access to a file's mount 38996029c4eSnpiggin@suse.de * @file: the file who's mount on which to take a write 39096029c4eSnpiggin@suse.de * 39196029c4eSnpiggin@suse.de * This is like mnt_want_write, but it takes a file and can 39296029c4eSnpiggin@suse.de * do some optimisations if the file is open for write already 39396029c4eSnpiggin@suse.de */ 39496029c4eSnpiggin@suse.de int mnt_want_write_file(struct file *file) 39596029c4eSnpiggin@suse.de { 3962d8dd38aSOGAWA Hirofumi struct inode *inode = file->f_dentry->d_inode; 3972d8dd38aSOGAWA Hirofumi if (!(file->f_mode & FMODE_WRITE) || special_file(inode->i_mode)) 39896029c4eSnpiggin@suse.de return mnt_want_write(file->f_path.mnt); 39996029c4eSnpiggin@suse.de else 40096029c4eSnpiggin@suse.de return mnt_clone_write(file->f_path.mnt); 40196029c4eSnpiggin@suse.de } 40296029c4eSnpiggin@suse.de EXPORT_SYMBOL_GPL(mnt_want_write_file); 40396029c4eSnpiggin@suse.de 40496029c4eSnpiggin@suse.de /** 4058366025eSDave Hansen * mnt_drop_write - give up write access to a mount 4068366025eSDave Hansen * @mnt: the mount on which to give up write access 4078366025eSDave Hansen * 4088366025eSDave Hansen * Tells the low-level filesystem that we are done 4098366025eSDave Hansen * performing writes to it. Must be matched with 4108366025eSDave Hansen * mnt_want_write() call above. 4118366025eSDave Hansen */ 4128366025eSDave Hansen void mnt_drop_write(struct vfsmount *mnt) 4138366025eSDave Hansen { 414d3ef3d73Snpiggin@suse.de preempt_disable(); 415c6653a83SNick Piggin mnt_dec_writers(mnt); 416d3ef3d73Snpiggin@suse.de preempt_enable(); 4178366025eSDave Hansen } 4188366025eSDave Hansen EXPORT_SYMBOL_GPL(mnt_drop_write); 4198366025eSDave Hansen 4202e4b7fcdSDave Hansen static int mnt_make_readonly(struct vfsmount *mnt) 4218366025eSDave Hansen { 4223d733633SDave Hansen int ret = 0; 4233d733633SDave Hansen 42499b7db7bSNick Piggin br_write_lock(vfsmount_lock); 425d3ef3d73Snpiggin@suse.de mnt->mnt_flags |= MNT_WRITE_HOLD; 426d3ef3d73Snpiggin@suse.de /* 427d3ef3d73Snpiggin@suse.de * After storing MNT_WRITE_HOLD, we'll read the counters. This store 428d3ef3d73Snpiggin@suse.de * should be visible before we do. 429d3ef3d73Snpiggin@suse.de */ 430d3ef3d73Snpiggin@suse.de smp_mb(); 431d3ef3d73Snpiggin@suse.de 432d3ef3d73Snpiggin@suse.de /* 433d3ef3d73Snpiggin@suse.de * With writers on hold, if this value is zero, then there are 434d3ef3d73Snpiggin@suse.de * definitely no active writers (although held writers may subsequently 435d3ef3d73Snpiggin@suse.de * increment the count, they'll have to wait, and decrement it after 436d3ef3d73Snpiggin@suse.de * seeing MNT_READONLY). 437d3ef3d73Snpiggin@suse.de * 438d3ef3d73Snpiggin@suse.de * It is OK to have counter incremented on one CPU and decremented on 439d3ef3d73Snpiggin@suse.de * another: the sum will add up correctly. The danger would be when we 440d3ef3d73Snpiggin@suse.de * sum up each counter, if we read a counter before it is incremented, 441d3ef3d73Snpiggin@suse.de * but then read another CPU's count which it has been subsequently 442d3ef3d73Snpiggin@suse.de * decremented from -- we would see more decrements than we should. 443d3ef3d73Snpiggin@suse.de * MNT_WRITE_HOLD protects against this scenario, because 444d3ef3d73Snpiggin@suse.de * mnt_want_write first increments count, then smp_mb, then spins on 445d3ef3d73Snpiggin@suse.de * MNT_WRITE_HOLD, so it can't be decremented by another CPU while 446d3ef3d73Snpiggin@suse.de * we're counting up here. 447d3ef3d73Snpiggin@suse.de */ 448c6653a83SNick Piggin if (mnt_get_writers(mnt) > 0) 449d3ef3d73Snpiggin@suse.de ret = -EBUSY; 450d3ef3d73Snpiggin@suse.de else 4512e4b7fcdSDave Hansen mnt->mnt_flags |= MNT_READONLY; 452d3ef3d73Snpiggin@suse.de /* 453d3ef3d73Snpiggin@suse.de * MNT_READONLY must become visible before ~MNT_WRITE_HOLD, so writers 454d3ef3d73Snpiggin@suse.de * that become unheld will see MNT_READONLY. 455d3ef3d73Snpiggin@suse.de */ 456d3ef3d73Snpiggin@suse.de smp_wmb(); 457d3ef3d73Snpiggin@suse.de mnt->mnt_flags &= ~MNT_WRITE_HOLD; 45899b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 4593d733633SDave Hansen return ret; 4603d733633SDave Hansen } 4618366025eSDave Hansen 4622e4b7fcdSDave Hansen static void __mnt_unmake_readonly(struct vfsmount *mnt) 4632e4b7fcdSDave Hansen { 46499b7db7bSNick Piggin br_write_lock(vfsmount_lock); 4652e4b7fcdSDave Hansen mnt->mnt_flags &= ~MNT_READONLY; 46699b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 4672e4b7fcdSDave Hansen } 4682e4b7fcdSDave Hansen 4699d412a43SAl Viro static void free_vfsmnt(struct vfsmount *mnt) 4701da177e4SLinus Torvalds { 4711da177e4SLinus Torvalds kfree(mnt->mnt_devname); 47273cd49ecSMiklos Szeredi mnt_free_id(mnt); 473d3ef3d73Snpiggin@suse.de #ifdef CONFIG_SMP 474b3e19d92SNick Piggin free_percpu(mnt->mnt_pcp); 475d3ef3d73Snpiggin@suse.de #endif 4761da177e4SLinus Torvalds kmem_cache_free(mnt_cache, mnt); 4771da177e4SLinus Torvalds } 4781da177e4SLinus Torvalds 4791da177e4SLinus Torvalds /* 480a05964f3SRam Pai * find the first or last mount at @dentry on vfsmount @mnt depending on 481a05964f3SRam Pai * @dir. If @dir is set return the first mount else return the last mount. 48299b7db7bSNick Piggin * vfsmount_lock must be held for read or write. 4831da177e4SLinus Torvalds */ 484a05964f3SRam Pai struct vfsmount *__lookup_mnt(struct vfsmount *mnt, struct dentry *dentry, 485a05964f3SRam Pai int dir) 4861da177e4SLinus Torvalds { 4871da177e4SLinus Torvalds struct list_head *head = mount_hashtable + hash(mnt, dentry); 4881da177e4SLinus Torvalds struct list_head *tmp = head; 4891da177e4SLinus Torvalds struct vfsmount *p, *found = NULL; 4901da177e4SLinus Torvalds 4911da177e4SLinus Torvalds for (;;) { 492a05964f3SRam Pai tmp = dir ? tmp->next : tmp->prev; 4931da177e4SLinus Torvalds p = NULL; 4941da177e4SLinus Torvalds if (tmp == head) 4951da177e4SLinus Torvalds break; 4961da177e4SLinus Torvalds p = list_entry(tmp, struct vfsmount, mnt_hash); 4971da177e4SLinus Torvalds if (p->mnt_parent == mnt && p->mnt_mountpoint == dentry) { 498a05964f3SRam Pai found = p; 4991da177e4SLinus Torvalds break; 5001da177e4SLinus Torvalds } 5011da177e4SLinus Torvalds } 5021da177e4SLinus Torvalds return found; 5031da177e4SLinus Torvalds } 5041da177e4SLinus Torvalds 505a05964f3SRam Pai /* 506a05964f3SRam Pai * lookup_mnt increments the ref count before returning 507a05964f3SRam Pai * the vfsmount struct. 508a05964f3SRam Pai */ 5091c755af4SAl Viro struct vfsmount *lookup_mnt(struct path *path) 510a05964f3SRam Pai { 511a05964f3SRam Pai struct vfsmount *child_mnt; 51299b7db7bSNick Piggin 51399b7db7bSNick Piggin br_read_lock(vfsmount_lock); 5141c755af4SAl Viro if ((child_mnt = __lookup_mnt(path->mnt, path->dentry, 1))) 515a05964f3SRam Pai mntget(child_mnt); 51699b7db7bSNick Piggin br_read_unlock(vfsmount_lock); 517a05964f3SRam Pai return child_mnt; 518a05964f3SRam Pai } 519a05964f3SRam Pai 5201da177e4SLinus Torvalds static inline int check_mnt(struct vfsmount *mnt) 5211da177e4SLinus Torvalds { 5226b3286edSKirill Korotaev return mnt->mnt_ns == current->nsproxy->mnt_ns; 5231da177e4SLinus Torvalds } 5241da177e4SLinus Torvalds 52599b7db7bSNick Piggin /* 52699b7db7bSNick Piggin * vfsmount lock must be held for write 52799b7db7bSNick Piggin */ 5286b3286edSKirill Korotaev static void touch_mnt_namespace(struct mnt_namespace *ns) 5295addc5ddSAl Viro { 5305addc5ddSAl Viro if (ns) { 5315addc5ddSAl Viro ns->event = ++event; 5325addc5ddSAl Viro wake_up_interruptible(&ns->poll); 5335addc5ddSAl Viro } 5345addc5ddSAl Viro } 5355addc5ddSAl Viro 53699b7db7bSNick Piggin /* 53799b7db7bSNick Piggin * vfsmount lock must be held for write 53899b7db7bSNick Piggin */ 5396b3286edSKirill Korotaev static void __touch_mnt_namespace(struct mnt_namespace *ns) 5405addc5ddSAl Viro { 5415addc5ddSAl Viro if (ns && ns->event != event) { 5425addc5ddSAl Viro ns->event = event; 5435addc5ddSAl Viro wake_up_interruptible(&ns->poll); 5445addc5ddSAl Viro } 5455addc5ddSAl Viro } 5465addc5ddSAl Viro 54799b7db7bSNick Piggin /* 5485f57cbccSNick Piggin * Clear dentry's mounted state if it has no remaining mounts. 5495f57cbccSNick Piggin * vfsmount_lock must be held for write. 5505f57cbccSNick Piggin */ 5515f57cbccSNick Piggin static void dentry_reset_mounted(struct vfsmount *mnt, struct dentry *dentry) 5525f57cbccSNick Piggin { 5535f57cbccSNick Piggin unsigned u; 5545f57cbccSNick Piggin 5555f57cbccSNick Piggin for (u = 0; u < HASH_SIZE; u++) { 5565f57cbccSNick Piggin struct vfsmount *p; 5575f57cbccSNick Piggin 5585f57cbccSNick Piggin list_for_each_entry(p, &mount_hashtable[u], mnt_hash) { 5595f57cbccSNick Piggin if (p->mnt_mountpoint == dentry) 5605f57cbccSNick Piggin return; 5615f57cbccSNick Piggin } 5625f57cbccSNick Piggin } 5635f57cbccSNick Piggin spin_lock(&dentry->d_lock); 5645f57cbccSNick Piggin dentry->d_flags &= ~DCACHE_MOUNTED; 5655f57cbccSNick Piggin spin_unlock(&dentry->d_lock); 5665f57cbccSNick Piggin } 5675f57cbccSNick Piggin 5685f57cbccSNick Piggin /* 56999b7db7bSNick Piggin * vfsmount lock must be held for write 57099b7db7bSNick Piggin */ 5711a390689SAl Viro static void detach_mnt(struct vfsmount *mnt, struct path *old_path) 5721da177e4SLinus Torvalds { 5731a390689SAl Viro old_path->dentry = mnt->mnt_mountpoint; 5741a390689SAl Viro old_path->mnt = mnt->mnt_parent; 5751da177e4SLinus Torvalds mnt->mnt_parent = mnt; 5761da177e4SLinus Torvalds mnt->mnt_mountpoint = mnt->mnt_root; 5771da177e4SLinus Torvalds list_del_init(&mnt->mnt_child); 5781da177e4SLinus Torvalds list_del_init(&mnt->mnt_hash); 5795f57cbccSNick Piggin dentry_reset_mounted(old_path->mnt, old_path->dentry); 5801da177e4SLinus Torvalds } 5811da177e4SLinus Torvalds 58299b7db7bSNick Piggin /* 58399b7db7bSNick Piggin * vfsmount lock must be held for write 58499b7db7bSNick Piggin */ 585b90fa9aeSRam Pai void mnt_set_mountpoint(struct vfsmount *mnt, struct dentry *dentry, 586b90fa9aeSRam Pai struct vfsmount *child_mnt) 587b90fa9aeSRam Pai { 588b90fa9aeSRam Pai child_mnt->mnt_parent = mntget(mnt); 589b90fa9aeSRam Pai child_mnt->mnt_mountpoint = dget(dentry); 5905f57cbccSNick Piggin spin_lock(&dentry->d_lock); 5915f57cbccSNick Piggin dentry->d_flags |= DCACHE_MOUNTED; 5925f57cbccSNick Piggin spin_unlock(&dentry->d_lock); 593b90fa9aeSRam Pai } 594b90fa9aeSRam Pai 59599b7db7bSNick Piggin /* 59699b7db7bSNick Piggin * vfsmount lock must be held for write 59799b7db7bSNick Piggin */ 5981a390689SAl Viro static void attach_mnt(struct vfsmount *mnt, struct path *path) 5991da177e4SLinus Torvalds { 6001a390689SAl Viro mnt_set_mountpoint(path->mnt, path->dentry, mnt); 601b90fa9aeSRam Pai list_add_tail(&mnt->mnt_hash, mount_hashtable + 6021a390689SAl Viro hash(path->mnt, path->dentry)); 6031a390689SAl Viro list_add_tail(&mnt->mnt_child, &path->mnt->mnt_mounts); 604b90fa9aeSRam Pai } 605b90fa9aeSRam Pai 6067e3d0eb0SAl Viro static inline void __mnt_make_longterm(struct vfsmount *mnt) 6077e3d0eb0SAl Viro { 6087e3d0eb0SAl Viro #ifdef CONFIG_SMP 6097e3d0eb0SAl Viro atomic_inc(&mnt->mnt_longterm); 6107e3d0eb0SAl Viro #endif 6117e3d0eb0SAl Viro } 6127e3d0eb0SAl Viro 6137e3d0eb0SAl Viro /* needs vfsmount lock for write */ 6147e3d0eb0SAl Viro static inline void __mnt_make_shortterm(struct vfsmount *mnt) 6157e3d0eb0SAl Viro { 6167e3d0eb0SAl Viro #ifdef CONFIG_SMP 6177e3d0eb0SAl Viro atomic_dec(&mnt->mnt_longterm); 6187e3d0eb0SAl Viro #endif 6197e3d0eb0SAl Viro } 6207e3d0eb0SAl Viro 621b90fa9aeSRam Pai /* 62299b7db7bSNick Piggin * vfsmount lock must be held for write 623b90fa9aeSRam Pai */ 624b90fa9aeSRam Pai static void commit_tree(struct vfsmount *mnt) 625b90fa9aeSRam Pai { 626b90fa9aeSRam Pai struct vfsmount *parent = mnt->mnt_parent; 627b90fa9aeSRam Pai struct vfsmount *m; 628b90fa9aeSRam Pai LIST_HEAD(head); 6296b3286edSKirill Korotaev struct mnt_namespace *n = parent->mnt_ns; 630b90fa9aeSRam Pai 631b90fa9aeSRam Pai BUG_ON(parent == mnt); 632b90fa9aeSRam Pai 633b90fa9aeSRam Pai list_add_tail(&head, &mnt->mnt_list); 634f03c6599SAl Viro list_for_each_entry(m, &head, mnt_list) { 6356b3286edSKirill Korotaev m->mnt_ns = n; 6367e3d0eb0SAl Viro __mnt_make_longterm(m); 637f03c6599SAl Viro } 638f03c6599SAl Viro 639b90fa9aeSRam Pai list_splice(&head, n->list.prev); 640b90fa9aeSRam Pai 641b90fa9aeSRam Pai list_add_tail(&mnt->mnt_hash, mount_hashtable + 642b90fa9aeSRam Pai hash(parent, mnt->mnt_mountpoint)); 643b90fa9aeSRam Pai list_add_tail(&mnt->mnt_child, &parent->mnt_mounts); 6446b3286edSKirill Korotaev touch_mnt_namespace(n); 6451da177e4SLinus Torvalds } 6461da177e4SLinus Torvalds 6471da177e4SLinus Torvalds static struct vfsmount *next_mnt(struct vfsmount *p, struct vfsmount *root) 6481da177e4SLinus Torvalds { 6491da177e4SLinus Torvalds struct list_head *next = p->mnt_mounts.next; 6501da177e4SLinus Torvalds if (next == &p->mnt_mounts) { 6511da177e4SLinus Torvalds while (1) { 6521da177e4SLinus Torvalds if (p == root) 6531da177e4SLinus Torvalds return NULL; 6541da177e4SLinus Torvalds next = p->mnt_child.next; 6551da177e4SLinus Torvalds if (next != &p->mnt_parent->mnt_mounts) 6561da177e4SLinus Torvalds break; 6571da177e4SLinus Torvalds p = p->mnt_parent; 6581da177e4SLinus Torvalds } 6591da177e4SLinus Torvalds } 6601da177e4SLinus Torvalds return list_entry(next, struct vfsmount, mnt_child); 6611da177e4SLinus Torvalds } 6621da177e4SLinus Torvalds 6639676f0c6SRam Pai static struct vfsmount *skip_mnt_tree(struct vfsmount *p) 6649676f0c6SRam Pai { 6659676f0c6SRam Pai struct list_head *prev = p->mnt_mounts.prev; 6669676f0c6SRam Pai while (prev != &p->mnt_mounts) { 6679676f0c6SRam Pai p = list_entry(prev, struct vfsmount, mnt_child); 6689676f0c6SRam Pai prev = p->mnt_mounts.prev; 6699676f0c6SRam Pai } 6709676f0c6SRam Pai return p; 6719676f0c6SRam Pai } 6729676f0c6SRam Pai 6739d412a43SAl Viro struct vfsmount * 6749d412a43SAl Viro vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data) 6759d412a43SAl Viro { 6769d412a43SAl Viro struct vfsmount *mnt; 6779d412a43SAl Viro struct dentry *root; 6789d412a43SAl Viro 6799d412a43SAl Viro if (!type) 6809d412a43SAl Viro return ERR_PTR(-ENODEV); 6819d412a43SAl Viro 6829d412a43SAl Viro mnt = alloc_vfsmnt(name); 6839d412a43SAl Viro if (!mnt) 6849d412a43SAl Viro return ERR_PTR(-ENOMEM); 6859d412a43SAl Viro 6869d412a43SAl Viro if (flags & MS_KERNMOUNT) 6879d412a43SAl Viro mnt->mnt_flags = MNT_INTERNAL; 6889d412a43SAl Viro 6899d412a43SAl Viro root = mount_fs(type, flags, name, data); 6909d412a43SAl Viro if (IS_ERR(root)) { 6919d412a43SAl Viro free_vfsmnt(mnt); 6929d412a43SAl Viro return ERR_CAST(root); 6939d412a43SAl Viro } 6949d412a43SAl Viro 6959d412a43SAl Viro mnt->mnt_root = root; 6969d412a43SAl Viro mnt->mnt_sb = root->d_sb; 6979d412a43SAl Viro mnt->mnt_mountpoint = mnt->mnt_root; 6989d412a43SAl Viro mnt->mnt_parent = mnt; 6999d412a43SAl Viro return mnt; 7009d412a43SAl Viro } 7019d412a43SAl Viro EXPORT_SYMBOL_GPL(vfs_kern_mount); 7029d412a43SAl Viro 70336341f64SRam Pai static struct vfsmount *clone_mnt(struct vfsmount *old, struct dentry *root, 70436341f64SRam Pai int flag) 7051da177e4SLinus Torvalds { 7061da177e4SLinus Torvalds struct super_block *sb = old->mnt_sb; 7071da177e4SLinus Torvalds struct vfsmount *mnt = alloc_vfsmnt(old->mnt_devname); 7081da177e4SLinus Torvalds 7091da177e4SLinus Torvalds if (mnt) { 710719f5d7fSMiklos Szeredi if (flag & (CL_SLAVE | CL_PRIVATE)) 711719f5d7fSMiklos Szeredi mnt->mnt_group_id = 0; /* not a peer of original */ 712719f5d7fSMiklos Szeredi else 713719f5d7fSMiklos Szeredi mnt->mnt_group_id = old->mnt_group_id; 714719f5d7fSMiklos Szeredi 715719f5d7fSMiklos Szeredi if ((flag & CL_MAKE_SHARED) && !mnt->mnt_group_id) { 716719f5d7fSMiklos Szeredi int err = mnt_alloc_group_id(mnt); 717719f5d7fSMiklos Szeredi if (err) 718719f5d7fSMiklos Szeredi goto out_free; 719719f5d7fSMiklos Szeredi } 720719f5d7fSMiklos Szeredi 721be1a16a0SMiklos Szeredi mnt->mnt_flags = old->mnt_flags & ~MNT_WRITE_HOLD; 7221da177e4SLinus Torvalds atomic_inc(&sb->s_active); 7231da177e4SLinus Torvalds mnt->mnt_sb = sb; 7241da177e4SLinus Torvalds mnt->mnt_root = dget(root); 7251da177e4SLinus Torvalds mnt->mnt_mountpoint = mnt->mnt_root; 7261da177e4SLinus Torvalds mnt->mnt_parent = mnt; 727b90fa9aeSRam Pai 7285afe0022SRam Pai if (flag & CL_SLAVE) { 7295afe0022SRam Pai list_add(&mnt->mnt_slave, &old->mnt_slave_list); 7305afe0022SRam Pai mnt->mnt_master = old; 7315afe0022SRam Pai CLEAR_MNT_SHARED(mnt); 7328aec0809SAl Viro } else if (!(flag & CL_PRIVATE)) { 733796a6b52SAl Viro if ((flag & CL_MAKE_SHARED) || IS_MNT_SHARED(old)) 734b90fa9aeSRam Pai list_add(&mnt->mnt_share, &old->mnt_share); 7355afe0022SRam Pai if (IS_MNT_SLAVE(old)) 7365afe0022SRam Pai list_add(&mnt->mnt_slave, &old->mnt_slave); 7375afe0022SRam Pai mnt->mnt_master = old->mnt_master; 7385afe0022SRam Pai } 739b90fa9aeSRam Pai if (flag & CL_MAKE_SHARED) 740b90fa9aeSRam Pai set_mnt_shared(mnt); 7411da177e4SLinus Torvalds 7421da177e4SLinus Torvalds /* stick the duplicate mount on the same expiry list 7431da177e4SLinus Torvalds * as the original if that was on one */ 74436341f64SRam Pai if (flag & CL_EXPIRE) { 74555e700b9SMiklos Szeredi if (!list_empty(&old->mnt_expire)) 74655e700b9SMiklos Szeredi list_add(&mnt->mnt_expire, &old->mnt_expire); 7471da177e4SLinus Torvalds } 74836341f64SRam Pai } 7491da177e4SLinus Torvalds return mnt; 750719f5d7fSMiklos Szeredi 751719f5d7fSMiklos Szeredi out_free: 752719f5d7fSMiklos Szeredi free_vfsmnt(mnt); 753719f5d7fSMiklos Szeredi return NULL; 7541da177e4SLinus Torvalds } 7551da177e4SLinus Torvalds 756b3e19d92SNick Piggin static inline void mntfree(struct vfsmount *mnt) 7571da177e4SLinus Torvalds { 7581da177e4SLinus Torvalds struct super_block *sb = mnt->mnt_sb; 759b3e19d92SNick Piggin 7603d733633SDave Hansen /* 7613d733633SDave Hansen * This probably indicates that somebody messed 7623d733633SDave Hansen * up a mnt_want/drop_write() pair. If this 7633d733633SDave Hansen * happens, the filesystem was probably unable 7643d733633SDave Hansen * to make r/w->r/o transitions. 7653d733633SDave Hansen */ 766d3ef3d73Snpiggin@suse.de /* 767b3e19d92SNick Piggin * The locking used to deal with mnt_count decrement provides barriers, 768b3e19d92SNick Piggin * so mnt_get_writers() below is safe. 769d3ef3d73Snpiggin@suse.de */ 770c6653a83SNick Piggin WARN_ON(mnt_get_writers(mnt)); 771ca9c726eSAndreas Gruenbacher fsnotify_vfsmount_delete(mnt); 7721da177e4SLinus Torvalds dput(mnt->mnt_root); 7731da177e4SLinus Torvalds free_vfsmnt(mnt); 7741da177e4SLinus Torvalds deactivate_super(sb); 7751da177e4SLinus Torvalds } 7761da177e4SLinus Torvalds 777f03c6599SAl Viro static void mntput_no_expire(struct vfsmount *mnt) 7787b7b1aceSAl Viro { 779b3e19d92SNick Piggin put_again: 780f03c6599SAl Viro #ifdef CONFIG_SMP 781b3e19d92SNick Piggin br_read_lock(vfsmount_lock); 782f03c6599SAl Viro if (likely(atomic_read(&mnt->mnt_longterm))) { 783b3e19d92SNick Piggin mnt_dec_count(mnt); 784b3e19d92SNick Piggin br_read_unlock(vfsmount_lock); 78599b7db7bSNick Piggin return; 786b3e19d92SNick Piggin } 787b3e19d92SNick Piggin br_read_unlock(vfsmount_lock); 788b3e19d92SNick Piggin 78999b7db7bSNick Piggin br_write_lock(vfsmount_lock); 790b3e19d92SNick Piggin mnt_dec_count(mnt); 791b3e19d92SNick Piggin if (mnt_get_count(mnt)) { 79299b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 79399b7db7bSNick Piggin return; 79499b7db7bSNick Piggin } 795b3e19d92SNick Piggin #else 796b3e19d92SNick Piggin mnt_dec_count(mnt); 797b3e19d92SNick Piggin if (likely(mnt_get_count(mnt))) 798b3e19d92SNick Piggin return; 799b3e19d92SNick Piggin br_write_lock(vfsmount_lock); 800f03c6599SAl Viro #endif 801b3e19d92SNick Piggin if (unlikely(mnt->mnt_pinned)) { 802b3e19d92SNick Piggin mnt_add_count(mnt, mnt->mnt_pinned + 1); 803b3e19d92SNick Piggin mnt->mnt_pinned = 0; 804b3e19d92SNick Piggin br_write_unlock(vfsmount_lock); 805b3e19d92SNick Piggin acct_auto_close_mnt(mnt); 806b3e19d92SNick Piggin goto put_again; 807b3e19d92SNick Piggin } 808b3e19d92SNick Piggin br_write_unlock(vfsmount_lock); 809b3e19d92SNick Piggin mntfree(mnt); 810b3e19d92SNick Piggin } 811b3e19d92SNick Piggin 812b3e19d92SNick Piggin void mntput(struct vfsmount *mnt) 813b3e19d92SNick Piggin { 814b3e19d92SNick Piggin if (mnt) { 815b3e19d92SNick Piggin /* avoid cacheline pingpong, hope gcc doesn't get "smart" */ 816b3e19d92SNick Piggin if (unlikely(mnt->mnt_expiry_mark)) 817b3e19d92SNick Piggin mnt->mnt_expiry_mark = 0; 818f03c6599SAl Viro mntput_no_expire(mnt); 819b3e19d92SNick Piggin } 820b3e19d92SNick Piggin } 821b3e19d92SNick Piggin EXPORT_SYMBOL(mntput); 822b3e19d92SNick Piggin 823b3e19d92SNick Piggin struct vfsmount *mntget(struct vfsmount *mnt) 824b3e19d92SNick Piggin { 825b3e19d92SNick Piggin if (mnt) 826b3e19d92SNick Piggin mnt_inc_count(mnt); 827b3e19d92SNick Piggin return mnt; 828b3e19d92SNick Piggin } 829b3e19d92SNick Piggin EXPORT_SYMBOL(mntget); 830b3e19d92SNick Piggin 8317b7b1aceSAl Viro void mnt_pin(struct vfsmount *mnt) 8327b7b1aceSAl Viro { 83399b7db7bSNick Piggin br_write_lock(vfsmount_lock); 8347b7b1aceSAl Viro mnt->mnt_pinned++; 83599b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 8367b7b1aceSAl Viro } 8377b7b1aceSAl Viro EXPORT_SYMBOL(mnt_pin); 8387b7b1aceSAl Viro 8397b7b1aceSAl Viro void mnt_unpin(struct vfsmount *mnt) 8407b7b1aceSAl Viro { 84199b7db7bSNick Piggin br_write_lock(vfsmount_lock); 8427b7b1aceSAl Viro if (mnt->mnt_pinned) { 843b3e19d92SNick Piggin mnt_inc_count(mnt); 8447b7b1aceSAl Viro mnt->mnt_pinned--; 8457b7b1aceSAl Viro } 84699b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 8477b7b1aceSAl Viro } 8487b7b1aceSAl Viro EXPORT_SYMBOL(mnt_unpin); 8491da177e4SLinus Torvalds 850b3b304a2SMiklos Szeredi static inline void mangle(struct seq_file *m, const char *s) 851b3b304a2SMiklos Szeredi { 852b3b304a2SMiklos Szeredi seq_escape(m, s, " \t\n\\"); 853b3b304a2SMiklos Szeredi } 854b3b304a2SMiklos Szeredi 855b3b304a2SMiklos Szeredi /* 856b3b304a2SMiklos Szeredi * Simple .show_options callback for filesystems which don't want to 857b3b304a2SMiklos Szeredi * implement more complex mount option showing. 858b3b304a2SMiklos Szeredi * 859b3b304a2SMiklos Szeredi * See also save_mount_options(). 860b3b304a2SMiklos Szeredi */ 861b3b304a2SMiklos Szeredi int generic_show_options(struct seq_file *m, struct vfsmount *mnt) 862b3b304a2SMiklos Szeredi { 8632a32cebdSAl Viro const char *options; 8642a32cebdSAl Viro 8652a32cebdSAl Viro rcu_read_lock(); 8662a32cebdSAl Viro options = rcu_dereference(mnt->mnt_sb->s_options); 867b3b304a2SMiklos Szeredi 868b3b304a2SMiklos Szeredi if (options != NULL && options[0]) { 869b3b304a2SMiklos Szeredi seq_putc(m, ','); 870b3b304a2SMiklos Szeredi mangle(m, options); 871b3b304a2SMiklos Szeredi } 8722a32cebdSAl Viro rcu_read_unlock(); 873b3b304a2SMiklos Szeredi 874b3b304a2SMiklos Szeredi return 0; 875b3b304a2SMiklos Szeredi } 876b3b304a2SMiklos Szeredi EXPORT_SYMBOL(generic_show_options); 877b3b304a2SMiklos Szeredi 878b3b304a2SMiklos Szeredi /* 879b3b304a2SMiklos Szeredi * If filesystem uses generic_show_options(), this function should be 880b3b304a2SMiklos Szeredi * called from the fill_super() callback. 881b3b304a2SMiklos Szeredi * 882b3b304a2SMiklos Szeredi * The .remount_fs callback usually needs to be handled in a special 883b3b304a2SMiklos Szeredi * way, to make sure, that previous options are not overwritten if the 884b3b304a2SMiklos Szeredi * remount fails. 885b3b304a2SMiklos Szeredi * 886b3b304a2SMiklos Szeredi * Also note, that if the filesystem's .remount_fs function doesn't 887b3b304a2SMiklos Szeredi * reset all options to their default value, but changes only newly 888b3b304a2SMiklos Szeredi * given options, then the displayed options will not reflect reality 889b3b304a2SMiklos Szeredi * any more. 890b3b304a2SMiklos Szeredi */ 891b3b304a2SMiklos Szeredi void save_mount_options(struct super_block *sb, char *options) 892b3b304a2SMiklos Szeredi { 8932a32cebdSAl Viro BUG_ON(sb->s_options); 8942a32cebdSAl Viro rcu_assign_pointer(sb->s_options, kstrdup(options, GFP_KERNEL)); 895b3b304a2SMiklos Szeredi } 896b3b304a2SMiklos Szeredi EXPORT_SYMBOL(save_mount_options); 897b3b304a2SMiklos Szeredi 8982a32cebdSAl Viro void replace_mount_options(struct super_block *sb, char *options) 8992a32cebdSAl Viro { 9002a32cebdSAl Viro char *old = sb->s_options; 9012a32cebdSAl Viro rcu_assign_pointer(sb->s_options, options); 9022a32cebdSAl Viro if (old) { 9032a32cebdSAl Viro synchronize_rcu(); 9042a32cebdSAl Viro kfree(old); 9052a32cebdSAl Viro } 9062a32cebdSAl Viro } 9072a32cebdSAl Viro EXPORT_SYMBOL(replace_mount_options); 9082a32cebdSAl Viro 909a1a2c409SMiklos Szeredi #ifdef CONFIG_PROC_FS 9101da177e4SLinus Torvalds /* iterator */ 9111da177e4SLinus Torvalds static void *m_start(struct seq_file *m, loff_t *pos) 9121da177e4SLinus Torvalds { 913a1a2c409SMiklos Szeredi struct proc_mounts *p = m->private; 9141da177e4SLinus Torvalds 915390c6843SRam Pai down_read(&namespace_sem); 916a1a2c409SMiklos Szeredi return seq_list_start(&p->ns->list, *pos); 9171da177e4SLinus Torvalds } 9181da177e4SLinus Torvalds 9191da177e4SLinus Torvalds static void *m_next(struct seq_file *m, void *v, loff_t *pos) 9201da177e4SLinus Torvalds { 921a1a2c409SMiklos Szeredi struct proc_mounts *p = m->private; 922b0765fb8SPavel Emelianov 923a1a2c409SMiklos Szeredi return seq_list_next(v, &p->ns->list, pos); 9241da177e4SLinus Torvalds } 9251da177e4SLinus Torvalds 9261da177e4SLinus Torvalds static void m_stop(struct seq_file *m, void *v) 9271da177e4SLinus Torvalds { 928390c6843SRam Pai up_read(&namespace_sem); 9291da177e4SLinus Torvalds } 9301da177e4SLinus Torvalds 9319f5596afSAl Viro int mnt_had_events(struct proc_mounts *p) 9329f5596afSAl Viro { 9339f5596afSAl Viro struct mnt_namespace *ns = p->ns; 9349f5596afSAl Viro int res = 0; 9359f5596afSAl Viro 93699b7db7bSNick Piggin br_read_lock(vfsmount_lock); 937f1514638SKay Sievers if (p->m.poll_event != ns->event) { 938f1514638SKay Sievers p->m.poll_event = ns->event; 9399f5596afSAl Viro res = 1; 9409f5596afSAl Viro } 94199b7db7bSNick Piggin br_read_unlock(vfsmount_lock); 9429f5596afSAl Viro 9439f5596afSAl Viro return res; 9449f5596afSAl Viro } 9459f5596afSAl Viro 9462d4d4864SRam Pai struct proc_fs_info { 9471da177e4SLinus Torvalds int flag; 9482d4d4864SRam Pai const char *str; 9492d4d4864SRam Pai }; 9502d4d4864SRam Pai 9512069f457SEric Paris static int show_sb_opts(struct seq_file *m, struct super_block *sb) 9522d4d4864SRam Pai { 9532d4d4864SRam Pai static const struct proc_fs_info fs_info[] = { 9541da177e4SLinus Torvalds { MS_SYNCHRONOUS, ",sync" }, 9551da177e4SLinus Torvalds { MS_DIRSYNC, ",dirsync" }, 9561da177e4SLinus Torvalds { MS_MANDLOCK, ",mand" }, 9571da177e4SLinus Torvalds { 0, NULL } 9581da177e4SLinus Torvalds }; 9592d4d4864SRam Pai const struct proc_fs_info *fs_infop; 9602d4d4864SRam Pai 9612d4d4864SRam Pai for (fs_infop = fs_info; fs_infop->flag; fs_infop++) { 9622d4d4864SRam Pai if (sb->s_flags & fs_infop->flag) 9632d4d4864SRam Pai seq_puts(m, fs_infop->str); 9642d4d4864SRam Pai } 9652069f457SEric Paris 9662069f457SEric Paris return security_sb_show_options(m, sb); 9672d4d4864SRam Pai } 9682d4d4864SRam Pai 9692d4d4864SRam Pai static void show_mnt_opts(struct seq_file *m, struct vfsmount *mnt) 9702d4d4864SRam Pai { 9712d4d4864SRam Pai static const struct proc_fs_info mnt_info[] = { 9721da177e4SLinus Torvalds { MNT_NOSUID, ",nosuid" }, 9731da177e4SLinus Torvalds { MNT_NODEV, ",nodev" }, 9741da177e4SLinus Torvalds { MNT_NOEXEC, ",noexec" }, 975fc33a7bbSChristoph Hellwig { MNT_NOATIME, ",noatime" }, 976fc33a7bbSChristoph Hellwig { MNT_NODIRATIME, ",nodiratime" }, 97747ae32d6SValerie Henson { MNT_RELATIME, ",relatime" }, 9781da177e4SLinus Torvalds { 0, NULL } 9791da177e4SLinus Torvalds }; 9802d4d4864SRam Pai const struct proc_fs_info *fs_infop; 9812d4d4864SRam Pai 9822d4d4864SRam Pai for (fs_infop = mnt_info; fs_infop->flag; fs_infop++) { 9832d4d4864SRam Pai if (mnt->mnt_flags & fs_infop->flag) 9842d4d4864SRam Pai seq_puts(m, fs_infop->str); 9852d4d4864SRam Pai } 9862d4d4864SRam Pai } 9872d4d4864SRam Pai 9882d4d4864SRam Pai static void show_type(struct seq_file *m, struct super_block *sb) 9892d4d4864SRam Pai { 9902d4d4864SRam Pai mangle(m, sb->s_type->name); 9912d4d4864SRam Pai if (sb->s_subtype && sb->s_subtype[0]) { 9922d4d4864SRam Pai seq_putc(m, '.'); 9932d4d4864SRam Pai mangle(m, sb->s_subtype); 9942d4d4864SRam Pai } 9952d4d4864SRam Pai } 9962d4d4864SRam Pai 9972d4d4864SRam Pai static int show_vfsmnt(struct seq_file *m, void *v) 9982d4d4864SRam Pai { 9992d4d4864SRam Pai struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); 10002d4d4864SRam Pai int err = 0; 1001c32c2f63SJan Blunck struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; 10021da177e4SLinus Torvalds 1003c7f404b4SAl Viro if (mnt->mnt_sb->s_op->show_devname) { 1004c7f404b4SAl Viro err = mnt->mnt_sb->s_op->show_devname(m, mnt); 1005c7f404b4SAl Viro if (err) 1006c7f404b4SAl Viro goto out; 1007c7f404b4SAl Viro } else { 10081da177e4SLinus Torvalds mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none"); 1009c7f404b4SAl Viro } 10101da177e4SLinus Torvalds seq_putc(m, ' '); 1011c32c2f63SJan Blunck seq_path(m, &mnt_path, " \t\n\\"); 10121da177e4SLinus Torvalds seq_putc(m, ' '); 10132d4d4864SRam Pai show_type(m, mnt->mnt_sb); 10142e4b7fcdSDave Hansen seq_puts(m, __mnt_is_readonly(mnt) ? " ro" : " rw"); 10152069f457SEric Paris err = show_sb_opts(m, mnt->mnt_sb); 10162069f457SEric Paris if (err) 10172069f457SEric Paris goto out; 10182d4d4864SRam Pai show_mnt_opts(m, mnt); 10191da177e4SLinus Torvalds if (mnt->mnt_sb->s_op->show_options) 10201da177e4SLinus Torvalds err = mnt->mnt_sb->s_op->show_options(m, mnt); 10211da177e4SLinus Torvalds seq_puts(m, " 0 0\n"); 10222069f457SEric Paris out: 10231da177e4SLinus Torvalds return err; 10241da177e4SLinus Torvalds } 10251da177e4SLinus Torvalds 1026a1a2c409SMiklos Szeredi const struct seq_operations mounts_op = { 10271da177e4SLinus Torvalds .start = m_start, 10281da177e4SLinus Torvalds .next = m_next, 10291da177e4SLinus Torvalds .stop = m_stop, 10301da177e4SLinus Torvalds .show = show_vfsmnt 10311da177e4SLinus Torvalds }; 10321da177e4SLinus Torvalds 10332d4d4864SRam Pai static int show_mountinfo(struct seq_file *m, void *v) 10342d4d4864SRam Pai { 10352d4d4864SRam Pai struct proc_mounts *p = m->private; 10362d4d4864SRam Pai struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); 10372d4d4864SRam Pai struct super_block *sb = mnt->mnt_sb; 10382d4d4864SRam Pai struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; 10392d4d4864SRam Pai struct path root = p->root; 10402d4d4864SRam Pai int err = 0; 10412d4d4864SRam Pai 10422d4d4864SRam Pai seq_printf(m, "%i %i %u:%u ", mnt->mnt_id, mnt->mnt_parent->mnt_id, 10432d4d4864SRam Pai MAJOR(sb->s_dev), MINOR(sb->s_dev)); 1044c7f404b4SAl Viro if (sb->s_op->show_path) 1045c7f404b4SAl Viro err = sb->s_op->show_path(m, mnt); 1046c7f404b4SAl Viro else 10472d4d4864SRam Pai seq_dentry(m, mnt->mnt_root, " \t\n\\"); 1048c7f404b4SAl Viro if (err) 1049c7f404b4SAl Viro goto out; 10502d4d4864SRam Pai seq_putc(m, ' '); 10512d4d4864SRam Pai seq_path_root(m, &mnt_path, &root, " \t\n\\"); 10522d4d4864SRam Pai if (root.mnt != p->root.mnt || root.dentry != p->root.dentry) { 10532d4d4864SRam Pai /* 10542d4d4864SRam Pai * Mountpoint is outside root, discard that one. Ugly, 10552d4d4864SRam Pai * but less so than trying to do that in iterator in a 10562d4d4864SRam Pai * race-free way (due to renames). 10572d4d4864SRam Pai */ 10582d4d4864SRam Pai return SEQ_SKIP; 10592d4d4864SRam Pai } 10602d4d4864SRam Pai seq_puts(m, mnt->mnt_flags & MNT_READONLY ? " ro" : " rw"); 10612d4d4864SRam Pai show_mnt_opts(m, mnt); 10622d4d4864SRam Pai 10632d4d4864SRam Pai /* Tagged fields ("foo:X" or "bar") */ 10642d4d4864SRam Pai if (IS_MNT_SHARED(mnt)) 10652d4d4864SRam Pai seq_printf(m, " shared:%i", mnt->mnt_group_id); 106697e7e0f7SMiklos Szeredi if (IS_MNT_SLAVE(mnt)) { 106797e7e0f7SMiklos Szeredi int master = mnt->mnt_master->mnt_group_id; 106897e7e0f7SMiklos Szeredi int dom = get_dominating_id(mnt, &p->root); 106997e7e0f7SMiklos Szeredi seq_printf(m, " master:%i", master); 107097e7e0f7SMiklos Szeredi if (dom && dom != master) 107197e7e0f7SMiklos Szeredi seq_printf(m, " propagate_from:%i", dom); 107297e7e0f7SMiklos Szeredi } 10732d4d4864SRam Pai if (IS_MNT_UNBINDABLE(mnt)) 10742d4d4864SRam Pai seq_puts(m, " unbindable"); 10752d4d4864SRam Pai 10762d4d4864SRam Pai /* Filesystem specific data */ 10772d4d4864SRam Pai seq_puts(m, " - "); 10782d4d4864SRam Pai show_type(m, sb); 10792d4d4864SRam Pai seq_putc(m, ' '); 1080c7f404b4SAl Viro if (sb->s_op->show_devname) 1081c7f404b4SAl Viro err = sb->s_op->show_devname(m, mnt); 1082c7f404b4SAl Viro else 10832d4d4864SRam Pai mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none"); 1084c7f404b4SAl Viro if (err) 1085c7f404b4SAl Viro goto out; 10862d4d4864SRam Pai seq_puts(m, sb->s_flags & MS_RDONLY ? " ro" : " rw"); 10872069f457SEric Paris err = show_sb_opts(m, sb); 10882069f457SEric Paris if (err) 10892069f457SEric Paris goto out; 10902d4d4864SRam Pai if (sb->s_op->show_options) 10912d4d4864SRam Pai err = sb->s_op->show_options(m, mnt); 10922d4d4864SRam Pai seq_putc(m, '\n'); 10932069f457SEric Paris out: 10942d4d4864SRam Pai return err; 10952d4d4864SRam Pai } 10962d4d4864SRam Pai 10972d4d4864SRam Pai const struct seq_operations mountinfo_op = { 10982d4d4864SRam Pai .start = m_start, 10992d4d4864SRam Pai .next = m_next, 11002d4d4864SRam Pai .stop = m_stop, 11012d4d4864SRam Pai .show = show_mountinfo, 11022d4d4864SRam Pai }; 11032d4d4864SRam Pai 1104b4629fe2SChuck Lever static int show_vfsstat(struct seq_file *m, void *v) 1105b4629fe2SChuck Lever { 1106b0765fb8SPavel Emelianov struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); 1107c32c2f63SJan Blunck struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; 1108b4629fe2SChuck Lever int err = 0; 1109b4629fe2SChuck Lever 1110b4629fe2SChuck Lever /* device */ 1111c7f404b4SAl Viro if (mnt->mnt_sb->s_op->show_devname) { 1112a877ee03SBryan Schumaker seq_puts(m, "device "); 1113c7f404b4SAl Viro err = mnt->mnt_sb->s_op->show_devname(m, mnt); 1114c7f404b4SAl Viro } else { 1115b4629fe2SChuck Lever if (mnt->mnt_devname) { 1116b4629fe2SChuck Lever seq_puts(m, "device "); 1117b4629fe2SChuck Lever mangle(m, mnt->mnt_devname); 1118b4629fe2SChuck Lever } else 1119b4629fe2SChuck Lever seq_puts(m, "no device"); 1120c7f404b4SAl Viro } 1121b4629fe2SChuck Lever 1122b4629fe2SChuck Lever /* mount point */ 1123b4629fe2SChuck Lever seq_puts(m, " mounted on "); 1124c32c2f63SJan Blunck seq_path(m, &mnt_path, " \t\n\\"); 1125b4629fe2SChuck Lever seq_putc(m, ' '); 1126b4629fe2SChuck Lever 1127b4629fe2SChuck Lever /* file system type */ 1128b4629fe2SChuck Lever seq_puts(m, "with fstype "); 11292d4d4864SRam Pai show_type(m, mnt->mnt_sb); 1130b4629fe2SChuck Lever 1131b4629fe2SChuck Lever /* optional statistics */ 1132b4629fe2SChuck Lever if (mnt->mnt_sb->s_op->show_stats) { 1133b4629fe2SChuck Lever seq_putc(m, ' '); 1134c7f404b4SAl Viro if (!err) 1135b4629fe2SChuck Lever err = mnt->mnt_sb->s_op->show_stats(m, mnt); 1136b4629fe2SChuck Lever } 1137b4629fe2SChuck Lever 1138b4629fe2SChuck Lever seq_putc(m, '\n'); 1139b4629fe2SChuck Lever return err; 1140b4629fe2SChuck Lever } 1141b4629fe2SChuck Lever 1142a1a2c409SMiklos Szeredi const struct seq_operations mountstats_op = { 1143b4629fe2SChuck Lever .start = m_start, 1144b4629fe2SChuck Lever .next = m_next, 1145b4629fe2SChuck Lever .stop = m_stop, 1146b4629fe2SChuck Lever .show = show_vfsstat, 1147b4629fe2SChuck Lever }; 1148a1a2c409SMiklos Szeredi #endif /* CONFIG_PROC_FS */ 1149b4629fe2SChuck Lever 11501da177e4SLinus Torvalds /** 11511da177e4SLinus Torvalds * may_umount_tree - check if a mount tree is busy 11521da177e4SLinus Torvalds * @mnt: root of mount tree 11531da177e4SLinus Torvalds * 11541da177e4SLinus Torvalds * This is called to check if a tree of mounts has any 11551da177e4SLinus Torvalds * open files, pwds, chroots or sub mounts that are 11561da177e4SLinus Torvalds * busy. 11571da177e4SLinus Torvalds */ 11581da177e4SLinus Torvalds int may_umount_tree(struct vfsmount *mnt) 11591da177e4SLinus Torvalds { 116036341f64SRam Pai int actual_refs = 0; 116136341f64SRam Pai int minimum_refs = 0; 116236341f64SRam Pai struct vfsmount *p; 11631da177e4SLinus Torvalds 1164b3e19d92SNick Piggin /* write lock needed for mnt_get_count */ 1165b3e19d92SNick Piggin br_write_lock(vfsmount_lock); 116636341f64SRam Pai for (p = mnt; p; p = next_mnt(p, mnt)) { 1167b3e19d92SNick Piggin actual_refs += mnt_get_count(p); 11681da177e4SLinus Torvalds minimum_refs += 2; 11691da177e4SLinus Torvalds } 1170b3e19d92SNick Piggin br_write_unlock(vfsmount_lock); 11711da177e4SLinus Torvalds 11721da177e4SLinus Torvalds if (actual_refs > minimum_refs) 11731da177e4SLinus Torvalds return 0; 1174e3474a8eSIan Kent 1175e3474a8eSIan Kent return 1; 11761da177e4SLinus Torvalds } 11771da177e4SLinus Torvalds 11781da177e4SLinus Torvalds EXPORT_SYMBOL(may_umount_tree); 11791da177e4SLinus Torvalds 11801da177e4SLinus Torvalds /** 11811da177e4SLinus Torvalds * may_umount - check if a mount point is busy 11821da177e4SLinus Torvalds * @mnt: root of mount 11831da177e4SLinus Torvalds * 11841da177e4SLinus Torvalds * This is called to check if a mount point has any 11851da177e4SLinus Torvalds * open files, pwds, chroots or sub mounts. If the 11861da177e4SLinus Torvalds * mount has sub mounts this will return busy 11871da177e4SLinus Torvalds * regardless of whether the sub mounts are busy. 11881da177e4SLinus Torvalds * 11891da177e4SLinus Torvalds * Doesn't take quota and stuff into account. IOW, in some cases it will 11901da177e4SLinus Torvalds * give false negatives. The main reason why it's here is that we need 11911da177e4SLinus Torvalds * a non-destructive way to look for easily umountable filesystems. 11921da177e4SLinus Torvalds */ 11931da177e4SLinus Torvalds int may_umount(struct vfsmount *mnt) 11941da177e4SLinus Torvalds { 1195e3474a8eSIan Kent int ret = 1; 11968ad08d8aSAl Viro down_read(&namespace_sem); 1197b3e19d92SNick Piggin br_write_lock(vfsmount_lock); 1198a05964f3SRam Pai if (propagate_mount_busy(mnt, 2)) 1199e3474a8eSIan Kent ret = 0; 1200b3e19d92SNick Piggin br_write_unlock(vfsmount_lock); 12018ad08d8aSAl Viro up_read(&namespace_sem); 1202a05964f3SRam Pai return ret; 12031da177e4SLinus Torvalds } 12041da177e4SLinus Torvalds 12051da177e4SLinus Torvalds EXPORT_SYMBOL(may_umount); 12061da177e4SLinus Torvalds 1207b90fa9aeSRam Pai void release_mounts(struct list_head *head) 12081da177e4SLinus Torvalds { 120970fbcdf4SRam Pai struct vfsmount *mnt; 121070fbcdf4SRam Pai while (!list_empty(head)) { 1211b5e61818SPavel Emelianov mnt = list_first_entry(head, struct vfsmount, mnt_hash); 121270fbcdf4SRam Pai list_del_init(&mnt->mnt_hash); 121370fbcdf4SRam Pai if (mnt->mnt_parent != mnt) { 121470fbcdf4SRam Pai struct dentry *dentry; 121570fbcdf4SRam Pai struct vfsmount *m; 121699b7db7bSNick Piggin 121799b7db7bSNick Piggin br_write_lock(vfsmount_lock); 121870fbcdf4SRam Pai dentry = mnt->mnt_mountpoint; 121970fbcdf4SRam Pai m = mnt->mnt_parent; 122070fbcdf4SRam Pai mnt->mnt_mountpoint = mnt->mnt_root; 122170fbcdf4SRam Pai mnt->mnt_parent = mnt; 12227c4b93d8SAl Viro m->mnt_ghosts--; 122399b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 122470fbcdf4SRam Pai dput(dentry); 122570fbcdf4SRam Pai mntput(m); 12261da177e4SLinus Torvalds } 1227f03c6599SAl Viro mntput(mnt); 122870fbcdf4SRam Pai } 122970fbcdf4SRam Pai } 123070fbcdf4SRam Pai 123199b7db7bSNick Piggin /* 123299b7db7bSNick Piggin * vfsmount lock must be held for write 123399b7db7bSNick Piggin * namespace_sem must be held for write 123499b7db7bSNick Piggin */ 1235a05964f3SRam Pai void umount_tree(struct vfsmount *mnt, int propagate, struct list_head *kill) 123670fbcdf4SRam Pai { 12377b8a53fdSAl Viro LIST_HEAD(tmp_list); 123870fbcdf4SRam Pai struct vfsmount *p; 123970fbcdf4SRam Pai 12401bfba4e8SAkinobu Mita for (p = mnt; p; p = next_mnt(p, mnt)) 12417b8a53fdSAl Viro list_move(&p->mnt_hash, &tmp_list); 124270fbcdf4SRam Pai 1243a05964f3SRam Pai if (propagate) 12447b8a53fdSAl Viro propagate_umount(&tmp_list); 1245a05964f3SRam Pai 12467b8a53fdSAl Viro list_for_each_entry(p, &tmp_list, mnt_hash) { 124770fbcdf4SRam Pai list_del_init(&p->mnt_expire); 124870fbcdf4SRam Pai list_del_init(&p->mnt_list); 12496b3286edSKirill Korotaev __touch_mnt_namespace(p->mnt_ns); 12506b3286edSKirill Korotaev p->mnt_ns = NULL; 12517e3d0eb0SAl Viro __mnt_make_shortterm(p); 125270fbcdf4SRam Pai list_del_init(&p->mnt_child); 12537c4b93d8SAl Viro if (p->mnt_parent != p) { 12547c4b93d8SAl Viro p->mnt_parent->mnt_ghosts++; 12555f57cbccSNick Piggin dentry_reset_mounted(p->mnt_parent, p->mnt_mountpoint); 12567c4b93d8SAl Viro } 1257a05964f3SRam Pai change_mnt_propagation(p, MS_PRIVATE); 12581da177e4SLinus Torvalds } 12597b8a53fdSAl Viro list_splice(&tmp_list, kill); 12601da177e4SLinus Torvalds } 12611da177e4SLinus Torvalds 1262c35038beSAl Viro static void shrink_submounts(struct vfsmount *mnt, struct list_head *umounts); 1263c35038beSAl Viro 12641da177e4SLinus Torvalds static int do_umount(struct vfsmount *mnt, int flags) 12651da177e4SLinus Torvalds { 12661da177e4SLinus Torvalds struct super_block *sb = mnt->mnt_sb; 12671da177e4SLinus Torvalds int retval; 126870fbcdf4SRam Pai LIST_HEAD(umount_list); 12691da177e4SLinus Torvalds 12701da177e4SLinus Torvalds retval = security_sb_umount(mnt, flags); 12711da177e4SLinus Torvalds if (retval) 12721da177e4SLinus Torvalds return retval; 12731da177e4SLinus Torvalds 12741da177e4SLinus Torvalds /* 12751da177e4SLinus Torvalds * Allow userspace to request a mountpoint be expired rather than 12761da177e4SLinus Torvalds * unmounting unconditionally. Unmount only happens if: 12771da177e4SLinus Torvalds * (1) the mark is already set (the mark is cleared by mntput()) 12781da177e4SLinus Torvalds * (2) the usage count == 1 [parent vfsmount] + 1 [sys_umount] 12791da177e4SLinus Torvalds */ 12801da177e4SLinus Torvalds if (flags & MNT_EXPIRE) { 12816ac08c39SJan Blunck if (mnt == current->fs->root.mnt || 12821da177e4SLinus Torvalds flags & (MNT_FORCE | MNT_DETACH)) 12831da177e4SLinus Torvalds return -EINVAL; 12841da177e4SLinus Torvalds 1285b3e19d92SNick Piggin /* 1286b3e19d92SNick Piggin * probably don't strictly need the lock here if we examined 1287b3e19d92SNick Piggin * all race cases, but it's a slowpath. 1288b3e19d92SNick Piggin */ 1289b3e19d92SNick Piggin br_write_lock(vfsmount_lock); 1290b3e19d92SNick Piggin if (mnt_get_count(mnt) != 2) { 1291bf9faa2aSJ. R. Okajima br_write_unlock(vfsmount_lock); 12921da177e4SLinus Torvalds return -EBUSY; 1293b3e19d92SNick Piggin } 1294b3e19d92SNick Piggin br_write_unlock(vfsmount_lock); 12951da177e4SLinus Torvalds 12961da177e4SLinus Torvalds if (!xchg(&mnt->mnt_expiry_mark, 1)) 12971da177e4SLinus Torvalds return -EAGAIN; 12981da177e4SLinus Torvalds } 12991da177e4SLinus Torvalds 13001da177e4SLinus Torvalds /* 13011da177e4SLinus Torvalds * If we may have to abort operations to get out of this 13021da177e4SLinus Torvalds * mount, and they will themselves hold resources we must 13031da177e4SLinus Torvalds * allow the fs to do things. In the Unix tradition of 13041da177e4SLinus Torvalds * 'Gee thats tricky lets do it in userspace' the umount_begin 13051da177e4SLinus Torvalds * might fail to complete on the first run through as other tasks 13061da177e4SLinus Torvalds * must return, and the like. Thats for the mount program to worry 13071da177e4SLinus Torvalds * about for the moment. 13081da177e4SLinus Torvalds */ 13091da177e4SLinus Torvalds 131042faad99SAl Viro if (flags & MNT_FORCE && sb->s_op->umount_begin) { 131142faad99SAl Viro sb->s_op->umount_begin(sb); 131242faad99SAl Viro } 13131da177e4SLinus Torvalds 13141da177e4SLinus Torvalds /* 13151da177e4SLinus Torvalds * No sense to grab the lock for this test, but test itself looks 13161da177e4SLinus Torvalds * somewhat bogus. Suggestions for better replacement? 13171da177e4SLinus Torvalds * Ho-hum... In principle, we might treat that as umount + switch 13181da177e4SLinus Torvalds * to rootfs. GC would eventually take care of the old vfsmount. 13191da177e4SLinus Torvalds * Actually it makes sense, especially if rootfs would contain a 13201da177e4SLinus Torvalds * /reboot - static binary that would close all descriptors and 13211da177e4SLinus Torvalds * call reboot(9). Then init(8) could umount root and exec /reboot. 13221da177e4SLinus Torvalds */ 13236ac08c39SJan Blunck if (mnt == current->fs->root.mnt && !(flags & MNT_DETACH)) { 13241da177e4SLinus Torvalds /* 13251da177e4SLinus Torvalds * Special case for "unmounting" root ... 13261da177e4SLinus Torvalds * we just try to remount it readonly. 13271da177e4SLinus Torvalds */ 13281da177e4SLinus Torvalds down_write(&sb->s_umount); 13294aa98cf7SAl Viro if (!(sb->s_flags & MS_RDONLY)) 13301da177e4SLinus Torvalds retval = do_remount_sb(sb, MS_RDONLY, NULL, 0); 13311da177e4SLinus Torvalds up_write(&sb->s_umount); 13321da177e4SLinus Torvalds return retval; 13331da177e4SLinus Torvalds } 13341da177e4SLinus Torvalds 1335390c6843SRam Pai down_write(&namespace_sem); 133699b7db7bSNick Piggin br_write_lock(vfsmount_lock); 13375addc5ddSAl Viro event++; 13381da177e4SLinus Torvalds 1339c35038beSAl Viro if (!(flags & MNT_DETACH)) 1340c35038beSAl Viro shrink_submounts(mnt, &umount_list); 1341c35038beSAl Viro 13421da177e4SLinus Torvalds retval = -EBUSY; 1343a05964f3SRam Pai if (flags & MNT_DETACH || !propagate_mount_busy(mnt, 2)) { 13441da177e4SLinus Torvalds if (!list_empty(&mnt->mnt_list)) 1345a05964f3SRam Pai umount_tree(mnt, 1, &umount_list); 13461da177e4SLinus Torvalds retval = 0; 13471da177e4SLinus Torvalds } 134899b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 1349390c6843SRam Pai up_write(&namespace_sem); 135070fbcdf4SRam Pai release_mounts(&umount_list); 13511da177e4SLinus Torvalds return retval; 13521da177e4SLinus Torvalds } 13531da177e4SLinus Torvalds 13541da177e4SLinus Torvalds /* 13551da177e4SLinus Torvalds * Now umount can handle mount points as well as block devices. 13561da177e4SLinus Torvalds * This is important for filesystems which use unnamed block devices. 13571da177e4SLinus Torvalds * 13581da177e4SLinus Torvalds * We now support a flag for forced unmount like the other 'big iron' 13591da177e4SLinus Torvalds * unixes. Our API is identical to OSF/1 to avoid making a mess of AMD 13601da177e4SLinus Torvalds */ 13611da177e4SLinus Torvalds 1362bdc480e3SHeiko Carstens SYSCALL_DEFINE2(umount, char __user *, name, int, flags) 13631da177e4SLinus Torvalds { 13642d8f3038SAl Viro struct path path; 13651da177e4SLinus Torvalds int retval; 1366db1f05bbSMiklos Szeredi int lookup_flags = 0; 13671da177e4SLinus Torvalds 1368db1f05bbSMiklos Szeredi if (flags & ~(MNT_FORCE | MNT_DETACH | MNT_EXPIRE | UMOUNT_NOFOLLOW)) 1369db1f05bbSMiklos Szeredi return -EINVAL; 1370db1f05bbSMiklos Szeredi 1371db1f05bbSMiklos Szeredi if (!(flags & UMOUNT_NOFOLLOW)) 1372db1f05bbSMiklos Szeredi lookup_flags |= LOOKUP_FOLLOW; 1373db1f05bbSMiklos Szeredi 1374db1f05bbSMiklos Szeredi retval = user_path_at(AT_FDCWD, name, lookup_flags, &path); 13751da177e4SLinus Torvalds if (retval) 13761da177e4SLinus Torvalds goto out; 13771da177e4SLinus Torvalds retval = -EINVAL; 13782d8f3038SAl Viro if (path.dentry != path.mnt->mnt_root) 13791da177e4SLinus Torvalds goto dput_and_out; 13802d8f3038SAl Viro if (!check_mnt(path.mnt)) 13811da177e4SLinus Torvalds goto dput_and_out; 13821da177e4SLinus Torvalds 13831da177e4SLinus Torvalds retval = -EPERM; 13841da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 13851da177e4SLinus Torvalds goto dput_and_out; 13861da177e4SLinus Torvalds 13872d8f3038SAl Viro retval = do_umount(path.mnt, flags); 13881da177e4SLinus Torvalds dput_and_out: 1389429731b1SJan Blunck /* we mustn't call path_put() as that would clear mnt_expiry_mark */ 13902d8f3038SAl Viro dput(path.dentry); 13912d8f3038SAl Viro mntput_no_expire(path.mnt); 13921da177e4SLinus Torvalds out: 13931da177e4SLinus Torvalds return retval; 13941da177e4SLinus Torvalds } 13951da177e4SLinus Torvalds 13961da177e4SLinus Torvalds #ifdef __ARCH_WANT_SYS_OLDUMOUNT 13971da177e4SLinus Torvalds 13981da177e4SLinus Torvalds /* 13991da177e4SLinus Torvalds * The 2.0 compatible umount. No flags. 14001da177e4SLinus Torvalds */ 1401bdc480e3SHeiko Carstens SYSCALL_DEFINE1(oldumount, char __user *, name) 14021da177e4SLinus Torvalds { 14031da177e4SLinus Torvalds return sys_umount(name, 0); 14041da177e4SLinus Torvalds } 14051da177e4SLinus Torvalds 14061da177e4SLinus Torvalds #endif 14071da177e4SLinus Torvalds 14082d92ab3cSAl Viro static int mount_is_safe(struct path *path) 14091da177e4SLinus Torvalds { 14101da177e4SLinus Torvalds if (capable(CAP_SYS_ADMIN)) 14111da177e4SLinus Torvalds return 0; 14121da177e4SLinus Torvalds return -EPERM; 14131da177e4SLinus Torvalds #ifdef notyet 14142d92ab3cSAl Viro if (S_ISLNK(path->dentry->d_inode->i_mode)) 14151da177e4SLinus Torvalds return -EPERM; 14162d92ab3cSAl Viro if (path->dentry->d_inode->i_mode & S_ISVTX) { 1417da9592edSDavid Howells if (current_uid() != path->dentry->d_inode->i_uid) 14181da177e4SLinus Torvalds return -EPERM; 14191da177e4SLinus Torvalds } 14202d92ab3cSAl Viro if (inode_permission(path->dentry->d_inode, MAY_WRITE)) 14211da177e4SLinus Torvalds return -EPERM; 14221da177e4SLinus Torvalds return 0; 14231da177e4SLinus Torvalds #endif 14241da177e4SLinus Torvalds } 14251da177e4SLinus Torvalds 1426b90fa9aeSRam Pai struct vfsmount *copy_tree(struct vfsmount *mnt, struct dentry *dentry, 142736341f64SRam Pai int flag) 14281da177e4SLinus Torvalds { 14291da177e4SLinus Torvalds struct vfsmount *res, *p, *q, *r, *s; 14301a390689SAl Viro struct path path; 14311da177e4SLinus Torvalds 14329676f0c6SRam Pai if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(mnt)) 14339676f0c6SRam Pai return NULL; 14349676f0c6SRam Pai 143536341f64SRam Pai res = q = clone_mnt(mnt, dentry, flag); 14361da177e4SLinus Torvalds if (!q) 14371da177e4SLinus Torvalds goto Enomem; 14381da177e4SLinus Torvalds q->mnt_mountpoint = mnt->mnt_mountpoint; 14391da177e4SLinus Torvalds 14401da177e4SLinus Torvalds p = mnt; 1441fdadd65fSDomen Puncer list_for_each_entry(r, &mnt->mnt_mounts, mnt_child) { 14427ec02ef1SJan Blunck if (!is_subdir(r->mnt_mountpoint, dentry)) 14431da177e4SLinus Torvalds continue; 14441da177e4SLinus Torvalds 14451da177e4SLinus Torvalds for (s = r; s; s = next_mnt(s, r)) { 14469676f0c6SRam Pai if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(s)) { 14479676f0c6SRam Pai s = skip_mnt_tree(s); 14489676f0c6SRam Pai continue; 14499676f0c6SRam Pai } 14501da177e4SLinus Torvalds while (p != s->mnt_parent) { 14511da177e4SLinus Torvalds p = p->mnt_parent; 14521da177e4SLinus Torvalds q = q->mnt_parent; 14531da177e4SLinus Torvalds } 14541da177e4SLinus Torvalds p = s; 14551a390689SAl Viro path.mnt = q; 14561a390689SAl Viro path.dentry = p->mnt_mountpoint; 145736341f64SRam Pai q = clone_mnt(p, p->mnt_root, flag); 14581da177e4SLinus Torvalds if (!q) 14591da177e4SLinus Torvalds goto Enomem; 146099b7db7bSNick Piggin br_write_lock(vfsmount_lock); 14611da177e4SLinus Torvalds list_add_tail(&q->mnt_list, &res->mnt_list); 14621a390689SAl Viro attach_mnt(q, &path); 146399b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 14641da177e4SLinus Torvalds } 14651da177e4SLinus Torvalds } 14661da177e4SLinus Torvalds return res; 14671da177e4SLinus Torvalds Enomem: 14681da177e4SLinus Torvalds if (res) { 146970fbcdf4SRam Pai LIST_HEAD(umount_list); 147099b7db7bSNick Piggin br_write_lock(vfsmount_lock); 1471a05964f3SRam Pai umount_tree(res, 0, &umount_list); 147299b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 147370fbcdf4SRam Pai release_mounts(&umount_list); 14741da177e4SLinus Torvalds } 14751da177e4SLinus Torvalds return NULL; 14761da177e4SLinus Torvalds } 14771da177e4SLinus Torvalds 1478589ff870SAl Viro struct vfsmount *collect_mounts(struct path *path) 14798aec0809SAl Viro { 14808aec0809SAl Viro struct vfsmount *tree; 14811a60a280SAl Viro down_write(&namespace_sem); 1482589ff870SAl Viro tree = copy_tree(path->mnt, path->dentry, CL_COPY_ALL | CL_PRIVATE); 14831a60a280SAl Viro up_write(&namespace_sem); 14848aec0809SAl Viro return tree; 14858aec0809SAl Viro } 14868aec0809SAl Viro 14878aec0809SAl Viro void drop_collected_mounts(struct vfsmount *mnt) 14888aec0809SAl Viro { 14898aec0809SAl Viro LIST_HEAD(umount_list); 14901a60a280SAl Viro down_write(&namespace_sem); 149199b7db7bSNick Piggin br_write_lock(vfsmount_lock); 14928aec0809SAl Viro umount_tree(mnt, 0, &umount_list); 149399b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 14941a60a280SAl Viro up_write(&namespace_sem); 14958aec0809SAl Viro release_mounts(&umount_list); 14968aec0809SAl Viro } 14978aec0809SAl Viro 14981f707137SAl Viro int iterate_mounts(int (*f)(struct vfsmount *, void *), void *arg, 14991f707137SAl Viro struct vfsmount *root) 15001f707137SAl Viro { 15011f707137SAl Viro struct vfsmount *mnt; 15021f707137SAl Viro int res = f(root, arg); 15031f707137SAl Viro if (res) 15041f707137SAl Viro return res; 15051f707137SAl Viro list_for_each_entry(mnt, &root->mnt_list, mnt_list) { 15061f707137SAl Viro res = f(mnt, arg); 15071f707137SAl Viro if (res) 15081f707137SAl Viro return res; 15091f707137SAl Viro } 15101f707137SAl Viro return 0; 15111f707137SAl Viro } 15121f707137SAl Viro 1513719f5d7fSMiklos Szeredi static void cleanup_group_ids(struct vfsmount *mnt, struct vfsmount *end) 1514719f5d7fSMiklos Szeredi { 1515719f5d7fSMiklos Szeredi struct vfsmount *p; 1516719f5d7fSMiklos Szeredi 1517719f5d7fSMiklos Szeredi for (p = mnt; p != end; p = next_mnt(p, mnt)) { 1518719f5d7fSMiklos Szeredi if (p->mnt_group_id && !IS_MNT_SHARED(p)) 1519719f5d7fSMiklos Szeredi mnt_release_group_id(p); 1520719f5d7fSMiklos Szeredi } 1521719f5d7fSMiklos Szeredi } 1522719f5d7fSMiklos Szeredi 1523719f5d7fSMiklos Szeredi static int invent_group_ids(struct vfsmount *mnt, bool recurse) 1524719f5d7fSMiklos Szeredi { 1525719f5d7fSMiklos Szeredi struct vfsmount *p; 1526719f5d7fSMiklos Szeredi 1527719f5d7fSMiklos Szeredi for (p = mnt; p; p = recurse ? next_mnt(p, mnt) : NULL) { 1528719f5d7fSMiklos Szeredi if (!p->mnt_group_id && !IS_MNT_SHARED(p)) { 1529719f5d7fSMiklos Szeredi int err = mnt_alloc_group_id(p); 1530719f5d7fSMiklos Szeredi if (err) { 1531719f5d7fSMiklos Szeredi cleanup_group_ids(mnt, p); 1532719f5d7fSMiklos Szeredi return err; 1533719f5d7fSMiklos Szeredi } 1534719f5d7fSMiklos Szeredi } 1535719f5d7fSMiklos Szeredi } 1536719f5d7fSMiklos Szeredi 1537719f5d7fSMiklos Szeredi return 0; 1538719f5d7fSMiklos Szeredi } 1539719f5d7fSMiklos Szeredi 1540b90fa9aeSRam Pai /* 1541b90fa9aeSRam Pai * @source_mnt : mount tree to be attached 1542b90fa9aeSRam Pai * @nd : place the mount tree @source_mnt is attached 154321444403SRam Pai * @parent_nd : if non-null, detach the source_mnt from its parent and 154421444403SRam Pai * store the parent mount and mountpoint dentry. 154521444403SRam Pai * (done when source_mnt is moved) 1546b90fa9aeSRam Pai * 1547b90fa9aeSRam Pai * NOTE: in the table below explains the semantics when a source mount 1548b90fa9aeSRam Pai * of a given type is attached to a destination mount of a given type. 15499676f0c6SRam Pai * --------------------------------------------------------------------------- 1550b90fa9aeSRam Pai * | BIND MOUNT OPERATION | 15519676f0c6SRam Pai * |************************************************************************** 15529676f0c6SRam Pai * | source-->| shared | private | slave | unbindable | 15539676f0c6SRam Pai * | dest | | | | | 15549676f0c6SRam Pai * | | | | | | | 15559676f0c6SRam Pai * | v | | | | | 15569676f0c6SRam Pai * |************************************************************************** 15579676f0c6SRam Pai * | shared | shared (++) | shared (+) | shared(+++)| invalid | 15585afe0022SRam Pai * | | | | | | 15599676f0c6SRam Pai * |non-shared| shared (+) | private | slave (*) | invalid | 15609676f0c6SRam Pai * *************************************************************************** 1561b90fa9aeSRam Pai * A bind operation clones the source mount and mounts the clone on the 1562b90fa9aeSRam Pai * destination mount. 1563b90fa9aeSRam Pai * 1564b90fa9aeSRam Pai * (++) the cloned mount is propagated to all the mounts in the propagation 1565b90fa9aeSRam Pai * tree of the destination mount and the cloned mount is added to 1566b90fa9aeSRam Pai * the peer group of the source mount. 1567b90fa9aeSRam Pai * (+) the cloned mount is created under the destination mount and is marked 1568b90fa9aeSRam Pai * as shared. The cloned mount is added to the peer group of the source 1569b90fa9aeSRam Pai * mount. 15705afe0022SRam Pai * (+++) the mount is propagated to all the mounts in the propagation tree 15715afe0022SRam Pai * of the destination mount and the cloned mount is made slave 15725afe0022SRam Pai * of the same master as that of the source mount. The cloned mount 15735afe0022SRam Pai * is marked as 'shared and slave'. 15745afe0022SRam Pai * (*) the cloned mount is made a slave of the same master as that of the 15755afe0022SRam Pai * source mount. 15765afe0022SRam Pai * 15779676f0c6SRam Pai * --------------------------------------------------------------------------- 157821444403SRam Pai * | MOVE MOUNT OPERATION | 15799676f0c6SRam Pai * |************************************************************************** 15809676f0c6SRam Pai * | source-->| shared | private | slave | unbindable | 15819676f0c6SRam Pai * | dest | | | | | 15829676f0c6SRam Pai * | | | | | | | 15839676f0c6SRam Pai * | v | | | | | 15849676f0c6SRam Pai * |************************************************************************** 15859676f0c6SRam Pai * | shared | shared (+) | shared (+) | shared(+++) | invalid | 15865afe0022SRam Pai * | | | | | | 15879676f0c6SRam Pai * |non-shared| shared (+*) | private | slave (*) | unbindable | 15889676f0c6SRam Pai * *************************************************************************** 15895afe0022SRam Pai * 15905afe0022SRam Pai * (+) the mount is moved to the destination. And is then propagated to 15915afe0022SRam Pai * all the mounts in the propagation tree of the destination mount. 159221444403SRam Pai * (+*) the mount is moved to the destination. 15935afe0022SRam Pai * (+++) the mount is moved to the destination and is then propagated to 15945afe0022SRam Pai * all the mounts belonging to the destination mount's propagation tree. 15955afe0022SRam Pai * the mount is marked as 'shared and slave'. 15965afe0022SRam Pai * (*) the mount continues to be a slave at the new location. 1597b90fa9aeSRam Pai * 1598b90fa9aeSRam Pai * if the source mount is a tree, the operations explained above is 1599b90fa9aeSRam Pai * applied to each mount in the tree. 1600b90fa9aeSRam Pai * Must be called without spinlocks held, since this function can sleep 1601b90fa9aeSRam Pai * in allocations. 1602b90fa9aeSRam Pai */ 1603b90fa9aeSRam Pai static int attach_recursive_mnt(struct vfsmount *source_mnt, 16041a390689SAl Viro struct path *path, struct path *parent_path) 1605b90fa9aeSRam Pai { 1606b90fa9aeSRam Pai LIST_HEAD(tree_list); 16071a390689SAl Viro struct vfsmount *dest_mnt = path->mnt; 16081a390689SAl Viro struct dentry *dest_dentry = path->dentry; 1609b90fa9aeSRam Pai struct vfsmount *child, *p; 1610719f5d7fSMiklos Szeredi int err; 1611b90fa9aeSRam Pai 1612719f5d7fSMiklos Szeredi if (IS_MNT_SHARED(dest_mnt)) { 1613719f5d7fSMiklos Szeredi err = invent_group_ids(source_mnt, true); 1614719f5d7fSMiklos Szeredi if (err) 1615719f5d7fSMiklos Szeredi goto out; 1616719f5d7fSMiklos Szeredi } 1617719f5d7fSMiklos Szeredi err = propagate_mnt(dest_mnt, dest_dentry, source_mnt, &tree_list); 1618719f5d7fSMiklos Szeredi if (err) 1619719f5d7fSMiklos Szeredi goto out_cleanup_ids; 1620b90fa9aeSRam Pai 162199b7db7bSNick Piggin br_write_lock(vfsmount_lock); 1622df1a1ad2SAl Viro 1623b90fa9aeSRam Pai if (IS_MNT_SHARED(dest_mnt)) { 1624b90fa9aeSRam Pai for (p = source_mnt; p; p = next_mnt(p, source_mnt)) 1625b90fa9aeSRam Pai set_mnt_shared(p); 1626b90fa9aeSRam Pai } 16271a390689SAl Viro if (parent_path) { 16281a390689SAl Viro detach_mnt(source_mnt, parent_path); 16291a390689SAl Viro attach_mnt(source_mnt, path); 1630e5d67f07SAl Viro touch_mnt_namespace(parent_path->mnt->mnt_ns); 163121444403SRam Pai } else { 1632b90fa9aeSRam Pai mnt_set_mountpoint(dest_mnt, dest_dentry, source_mnt); 1633b90fa9aeSRam Pai commit_tree(source_mnt); 163421444403SRam Pai } 1635b90fa9aeSRam Pai 1636b90fa9aeSRam Pai list_for_each_entry_safe(child, p, &tree_list, mnt_hash) { 1637b90fa9aeSRam Pai list_del_init(&child->mnt_hash); 1638b90fa9aeSRam Pai commit_tree(child); 1639b90fa9aeSRam Pai } 164099b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 164199b7db7bSNick Piggin 1642b90fa9aeSRam Pai return 0; 1643719f5d7fSMiklos Szeredi 1644719f5d7fSMiklos Szeredi out_cleanup_ids: 1645719f5d7fSMiklos Szeredi if (IS_MNT_SHARED(dest_mnt)) 1646719f5d7fSMiklos Szeredi cleanup_group_ids(source_mnt, NULL); 1647719f5d7fSMiklos Szeredi out: 1648719f5d7fSMiklos Szeredi return err; 1649b90fa9aeSRam Pai } 1650b90fa9aeSRam Pai 1651b12cea91SAl Viro static int lock_mount(struct path *path) 1652b12cea91SAl Viro { 1653b12cea91SAl Viro struct vfsmount *mnt; 1654b12cea91SAl Viro retry: 1655b12cea91SAl Viro mutex_lock(&path->dentry->d_inode->i_mutex); 1656b12cea91SAl Viro if (unlikely(cant_mount(path->dentry))) { 1657b12cea91SAl Viro mutex_unlock(&path->dentry->d_inode->i_mutex); 1658b12cea91SAl Viro return -ENOENT; 1659b12cea91SAl Viro } 1660b12cea91SAl Viro down_write(&namespace_sem); 1661b12cea91SAl Viro mnt = lookup_mnt(path); 1662b12cea91SAl Viro if (likely(!mnt)) 1663b12cea91SAl Viro return 0; 1664b12cea91SAl Viro up_write(&namespace_sem); 1665b12cea91SAl Viro mutex_unlock(&path->dentry->d_inode->i_mutex); 1666b12cea91SAl Viro path_put(path); 1667b12cea91SAl Viro path->mnt = mnt; 1668b12cea91SAl Viro path->dentry = dget(mnt->mnt_root); 1669b12cea91SAl Viro goto retry; 1670b12cea91SAl Viro } 1671b12cea91SAl Viro 1672b12cea91SAl Viro static void unlock_mount(struct path *path) 1673b12cea91SAl Viro { 1674b12cea91SAl Viro up_write(&namespace_sem); 1675b12cea91SAl Viro mutex_unlock(&path->dentry->d_inode->i_mutex); 1676b12cea91SAl Viro } 1677b12cea91SAl Viro 16788c3ee42eSAl Viro static int graft_tree(struct vfsmount *mnt, struct path *path) 16791da177e4SLinus Torvalds { 16801da177e4SLinus Torvalds if (mnt->mnt_sb->s_flags & MS_NOUSER) 16811da177e4SLinus Torvalds return -EINVAL; 16821da177e4SLinus Torvalds 16838c3ee42eSAl Viro if (S_ISDIR(path->dentry->d_inode->i_mode) != 16841da177e4SLinus Torvalds S_ISDIR(mnt->mnt_root->d_inode->i_mode)) 16851da177e4SLinus Torvalds return -ENOTDIR; 16861da177e4SLinus Torvalds 1687b12cea91SAl Viro if (d_unlinked(path->dentry)) 1688b12cea91SAl Viro return -ENOENT; 16891da177e4SLinus Torvalds 1690b12cea91SAl Viro return attach_recursive_mnt(mnt, path, NULL); 16911da177e4SLinus Torvalds } 16921da177e4SLinus Torvalds 16931da177e4SLinus Torvalds /* 16947a2e8a8fSValerie Aurora * Sanity check the flags to change_mnt_propagation. 16957a2e8a8fSValerie Aurora */ 16967a2e8a8fSValerie Aurora 16977a2e8a8fSValerie Aurora static int flags_to_propagation_type(int flags) 16987a2e8a8fSValerie Aurora { 16997c6e984dSRoman Borisov int type = flags & ~(MS_REC | MS_SILENT); 17007a2e8a8fSValerie Aurora 17017a2e8a8fSValerie Aurora /* Fail if any non-propagation flags are set */ 17027a2e8a8fSValerie Aurora if (type & ~(MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE)) 17037a2e8a8fSValerie Aurora return 0; 17047a2e8a8fSValerie Aurora /* Only one propagation flag should be set */ 17057a2e8a8fSValerie Aurora if (!is_power_of_2(type)) 17067a2e8a8fSValerie Aurora return 0; 17077a2e8a8fSValerie Aurora return type; 17087a2e8a8fSValerie Aurora } 17097a2e8a8fSValerie Aurora 17107a2e8a8fSValerie Aurora /* 171107b20889SRam Pai * recursively change the type of the mountpoint. 171207b20889SRam Pai */ 17130a0d8a46SAl Viro static int do_change_type(struct path *path, int flag) 171407b20889SRam Pai { 17152d92ab3cSAl Viro struct vfsmount *m, *mnt = path->mnt; 171607b20889SRam Pai int recurse = flag & MS_REC; 17177a2e8a8fSValerie Aurora int type; 1718719f5d7fSMiklos Szeredi int err = 0; 171907b20889SRam Pai 1720ee6f9582SMiklos Szeredi if (!capable(CAP_SYS_ADMIN)) 1721ee6f9582SMiklos Szeredi return -EPERM; 1722ee6f9582SMiklos Szeredi 17232d92ab3cSAl Viro if (path->dentry != path->mnt->mnt_root) 172407b20889SRam Pai return -EINVAL; 172507b20889SRam Pai 17267a2e8a8fSValerie Aurora type = flags_to_propagation_type(flag); 17277a2e8a8fSValerie Aurora if (!type) 17287a2e8a8fSValerie Aurora return -EINVAL; 17297a2e8a8fSValerie Aurora 173007b20889SRam Pai down_write(&namespace_sem); 1731719f5d7fSMiklos Szeredi if (type == MS_SHARED) { 1732719f5d7fSMiklos Szeredi err = invent_group_ids(mnt, recurse); 1733719f5d7fSMiklos Szeredi if (err) 1734719f5d7fSMiklos Szeredi goto out_unlock; 1735719f5d7fSMiklos Szeredi } 1736719f5d7fSMiklos Szeredi 173799b7db7bSNick Piggin br_write_lock(vfsmount_lock); 173807b20889SRam Pai for (m = mnt; m; m = (recurse ? next_mnt(m, mnt) : NULL)) 173907b20889SRam Pai change_mnt_propagation(m, type); 174099b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 1741719f5d7fSMiklos Szeredi 1742719f5d7fSMiklos Szeredi out_unlock: 174307b20889SRam Pai up_write(&namespace_sem); 1744719f5d7fSMiklos Szeredi return err; 174507b20889SRam Pai } 174607b20889SRam Pai 174707b20889SRam Pai /* 17481da177e4SLinus Torvalds * do loopback mount. 17491da177e4SLinus Torvalds */ 17500a0d8a46SAl Viro static int do_loopback(struct path *path, char *old_name, 17512dafe1c4SEric Sandeen int recurse) 17521da177e4SLinus Torvalds { 1753b12cea91SAl Viro LIST_HEAD(umount_list); 17542d92ab3cSAl Viro struct path old_path; 17551da177e4SLinus Torvalds struct vfsmount *mnt = NULL; 17562d92ab3cSAl Viro int err = mount_is_safe(path); 17571da177e4SLinus Torvalds if (err) 17581da177e4SLinus Torvalds return err; 17591da177e4SLinus Torvalds if (!old_name || !*old_name) 17601da177e4SLinus Torvalds return -EINVAL; 1761815d405cSTrond Myklebust err = kern_path(old_name, LOOKUP_FOLLOW|LOOKUP_AUTOMOUNT, &old_path); 17621da177e4SLinus Torvalds if (err) 17631da177e4SLinus Torvalds return err; 17641da177e4SLinus Torvalds 1765b12cea91SAl Viro err = lock_mount(path); 1766b12cea91SAl Viro if (err) 17679676f0c6SRam Pai goto out; 17689676f0c6SRam Pai 1769b12cea91SAl Viro err = -EINVAL; 1770b12cea91SAl Viro if (IS_MNT_UNBINDABLE(old_path.mnt)) 1771b12cea91SAl Viro goto out2; 1772b12cea91SAl Viro 17732d92ab3cSAl Viro if (!check_mnt(path->mnt) || !check_mnt(old_path.mnt)) 1774b12cea91SAl Viro goto out2; 1775ccd48bc7SAl Viro 17761da177e4SLinus Torvalds err = -ENOMEM; 17771da177e4SLinus Torvalds if (recurse) 17782d92ab3cSAl Viro mnt = copy_tree(old_path.mnt, old_path.dentry, 0); 17791da177e4SLinus Torvalds else 17802d92ab3cSAl Viro mnt = clone_mnt(old_path.mnt, old_path.dentry, 0); 17811da177e4SLinus Torvalds 1782ccd48bc7SAl Viro if (!mnt) 1783b12cea91SAl Viro goto out2; 1784ccd48bc7SAl Viro 17852d92ab3cSAl Viro err = graft_tree(mnt, path); 17861da177e4SLinus Torvalds if (err) { 178799b7db7bSNick Piggin br_write_lock(vfsmount_lock); 1788a05964f3SRam Pai umount_tree(mnt, 0, &umount_list); 178999b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 17905b83d2c5SRam Pai } 1791b12cea91SAl Viro out2: 1792b12cea91SAl Viro unlock_mount(path); 1793b12cea91SAl Viro release_mounts(&umount_list); 1794ccd48bc7SAl Viro out: 17952d92ab3cSAl Viro path_put(&old_path); 17961da177e4SLinus Torvalds return err; 17971da177e4SLinus Torvalds } 17981da177e4SLinus Torvalds 17992e4b7fcdSDave Hansen static int change_mount_flags(struct vfsmount *mnt, int ms_flags) 18002e4b7fcdSDave Hansen { 18012e4b7fcdSDave Hansen int error = 0; 18022e4b7fcdSDave Hansen int readonly_request = 0; 18032e4b7fcdSDave Hansen 18042e4b7fcdSDave Hansen if (ms_flags & MS_RDONLY) 18052e4b7fcdSDave Hansen readonly_request = 1; 18062e4b7fcdSDave Hansen if (readonly_request == __mnt_is_readonly(mnt)) 18072e4b7fcdSDave Hansen return 0; 18082e4b7fcdSDave Hansen 18092e4b7fcdSDave Hansen if (readonly_request) 18102e4b7fcdSDave Hansen error = mnt_make_readonly(mnt); 18112e4b7fcdSDave Hansen else 18122e4b7fcdSDave Hansen __mnt_unmake_readonly(mnt); 18132e4b7fcdSDave Hansen return error; 18142e4b7fcdSDave Hansen } 18152e4b7fcdSDave Hansen 18161da177e4SLinus Torvalds /* 18171da177e4SLinus Torvalds * change filesystem flags. dir should be a physical root of filesystem. 18181da177e4SLinus Torvalds * If you've mounted a non-root directory somewhere and want to do remount 18191da177e4SLinus Torvalds * on it - tough luck. 18201da177e4SLinus Torvalds */ 18210a0d8a46SAl Viro static int do_remount(struct path *path, int flags, int mnt_flags, 18221da177e4SLinus Torvalds void *data) 18231da177e4SLinus Torvalds { 18241da177e4SLinus Torvalds int err; 18252d92ab3cSAl Viro struct super_block *sb = path->mnt->mnt_sb; 18261da177e4SLinus Torvalds 18271da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 18281da177e4SLinus Torvalds return -EPERM; 18291da177e4SLinus Torvalds 18302d92ab3cSAl Viro if (!check_mnt(path->mnt)) 18311da177e4SLinus Torvalds return -EINVAL; 18321da177e4SLinus Torvalds 18332d92ab3cSAl Viro if (path->dentry != path->mnt->mnt_root) 18341da177e4SLinus Torvalds return -EINVAL; 18351da177e4SLinus Torvalds 1836ff36fe2cSEric Paris err = security_sb_remount(sb, data); 1837ff36fe2cSEric Paris if (err) 1838ff36fe2cSEric Paris return err; 1839ff36fe2cSEric Paris 18401da177e4SLinus Torvalds down_write(&sb->s_umount); 18412e4b7fcdSDave Hansen if (flags & MS_BIND) 18422d92ab3cSAl Viro err = change_mount_flags(path->mnt, flags); 18434aa98cf7SAl Viro else 18441da177e4SLinus Torvalds err = do_remount_sb(sb, flags, data, 0); 18457b43a79fSAl Viro if (!err) { 184699b7db7bSNick Piggin br_write_lock(vfsmount_lock); 1847495d6c9cSValerie Aurora mnt_flags |= path->mnt->mnt_flags & MNT_PROPAGATION_MASK; 18482d92ab3cSAl Viro path->mnt->mnt_flags = mnt_flags; 184999b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 18507b43a79fSAl Viro } 18511da177e4SLinus Torvalds up_write(&sb->s_umount); 18520e55a7ccSDan Williams if (!err) { 185399b7db7bSNick Piggin br_write_lock(vfsmount_lock); 18540e55a7ccSDan Williams touch_mnt_namespace(path->mnt->mnt_ns); 185599b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 18560e55a7ccSDan Williams } 18571da177e4SLinus Torvalds return err; 18581da177e4SLinus Torvalds } 18591da177e4SLinus Torvalds 18609676f0c6SRam Pai static inline int tree_contains_unbindable(struct vfsmount *mnt) 18619676f0c6SRam Pai { 18629676f0c6SRam Pai struct vfsmount *p; 18639676f0c6SRam Pai for (p = mnt; p; p = next_mnt(p, mnt)) { 18649676f0c6SRam Pai if (IS_MNT_UNBINDABLE(p)) 18659676f0c6SRam Pai return 1; 18669676f0c6SRam Pai } 18679676f0c6SRam Pai return 0; 18689676f0c6SRam Pai } 18699676f0c6SRam Pai 18700a0d8a46SAl Viro static int do_move_mount(struct path *path, char *old_name) 18711da177e4SLinus Torvalds { 18722d92ab3cSAl Viro struct path old_path, parent_path; 18731da177e4SLinus Torvalds struct vfsmount *p; 18741da177e4SLinus Torvalds int err = 0; 18751da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 18761da177e4SLinus Torvalds return -EPERM; 18771da177e4SLinus Torvalds if (!old_name || !*old_name) 18781da177e4SLinus Torvalds return -EINVAL; 18792d92ab3cSAl Viro err = kern_path(old_name, LOOKUP_FOLLOW, &old_path); 18801da177e4SLinus Torvalds if (err) 18811da177e4SLinus Torvalds return err; 18821da177e4SLinus Torvalds 1883b12cea91SAl Viro err = lock_mount(path); 1884cc53ce53SDavid Howells if (err < 0) 1885cc53ce53SDavid Howells goto out; 1886cc53ce53SDavid Howells 18871da177e4SLinus Torvalds err = -EINVAL; 18882d92ab3cSAl Viro if (!check_mnt(path->mnt) || !check_mnt(old_path.mnt)) 18891da177e4SLinus Torvalds goto out1; 18901da177e4SLinus Torvalds 1891f3da392eSAlexey Dobriyan if (d_unlinked(path->dentry)) 189221444403SRam Pai goto out1; 18931da177e4SLinus Torvalds 18941da177e4SLinus Torvalds err = -EINVAL; 18952d92ab3cSAl Viro if (old_path.dentry != old_path.mnt->mnt_root) 189621444403SRam Pai goto out1; 18971da177e4SLinus Torvalds 18982d92ab3cSAl Viro if (old_path.mnt == old_path.mnt->mnt_parent) 189921444403SRam Pai goto out1; 19001da177e4SLinus Torvalds 19012d92ab3cSAl Viro if (S_ISDIR(path->dentry->d_inode->i_mode) != 19022d92ab3cSAl Viro S_ISDIR(old_path.dentry->d_inode->i_mode)) 190321444403SRam Pai goto out1; 190421444403SRam Pai /* 190521444403SRam Pai * Don't move a mount residing in a shared parent. 190621444403SRam Pai */ 19072d92ab3cSAl Viro if (old_path.mnt->mnt_parent && 19082d92ab3cSAl Viro IS_MNT_SHARED(old_path.mnt->mnt_parent)) 190921444403SRam Pai goto out1; 19109676f0c6SRam Pai /* 19119676f0c6SRam Pai * Don't move a mount tree containing unbindable mounts to a destination 19129676f0c6SRam Pai * mount which is shared. 19139676f0c6SRam Pai */ 19142d92ab3cSAl Viro if (IS_MNT_SHARED(path->mnt) && 19152d92ab3cSAl Viro tree_contains_unbindable(old_path.mnt)) 19169676f0c6SRam Pai goto out1; 19171da177e4SLinus Torvalds err = -ELOOP; 19182d92ab3cSAl Viro for (p = path->mnt; p->mnt_parent != p; p = p->mnt_parent) 19192d92ab3cSAl Viro if (p == old_path.mnt) 192021444403SRam Pai goto out1; 19211da177e4SLinus Torvalds 19222d92ab3cSAl Viro err = attach_recursive_mnt(old_path.mnt, path, &parent_path); 19234ac91378SJan Blunck if (err) 192421444403SRam Pai goto out1; 19251da177e4SLinus Torvalds 19261da177e4SLinus Torvalds /* if the mount is moved, it should no longer be expire 19271da177e4SLinus Torvalds * automatically */ 19282d92ab3cSAl Viro list_del_init(&old_path.mnt->mnt_expire); 19291da177e4SLinus Torvalds out1: 1930b12cea91SAl Viro unlock_mount(path); 19311da177e4SLinus Torvalds out: 19321da177e4SLinus Torvalds if (!err) 19331a390689SAl Viro path_put(&parent_path); 19342d92ab3cSAl Viro path_put(&old_path); 19351da177e4SLinus Torvalds return err; 19361da177e4SLinus Torvalds } 19371da177e4SLinus Torvalds 19389d412a43SAl Viro static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype) 19399d412a43SAl Viro { 19409d412a43SAl Viro int err; 19419d412a43SAl Viro const char *subtype = strchr(fstype, '.'); 19429d412a43SAl Viro if (subtype) { 19439d412a43SAl Viro subtype++; 19449d412a43SAl Viro err = -EINVAL; 19459d412a43SAl Viro if (!subtype[0]) 19469d412a43SAl Viro goto err; 19479d412a43SAl Viro } else 19489d412a43SAl Viro subtype = ""; 19499d412a43SAl Viro 19509d412a43SAl Viro mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL); 19519d412a43SAl Viro err = -ENOMEM; 19529d412a43SAl Viro if (!mnt->mnt_sb->s_subtype) 19539d412a43SAl Viro goto err; 19549d412a43SAl Viro return mnt; 19559d412a43SAl Viro 19569d412a43SAl Viro err: 19579d412a43SAl Viro mntput(mnt); 19589d412a43SAl Viro return ERR_PTR(err); 19599d412a43SAl Viro } 19609d412a43SAl Viro 19619d412a43SAl Viro struct vfsmount * 19629d412a43SAl Viro do_kern_mount(const char *fstype, int flags, const char *name, void *data) 19639d412a43SAl Viro { 19649d412a43SAl Viro struct file_system_type *type = get_fs_type(fstype); 19659d412a43SAl Viro struct vfsmount *mnt; 19669d412a43SAl Viro if (!type) 19679d412a43SAl Viro return ERR_PTR(-ENODEV); 19689d412a43SAl Viro mnt = vfs_kern_mount(type, flags, name, data); 19699d412a43SAl Viro if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) && 19709d412a43SAl Viro !mnt->mnt_sb->s_subtype) 19719d412a43SAl Viro mnt = fs_set_subtype(mnt, fstype); 19729d412a43SAl Viro put_filesystem(type); 19739d412a43SAl Viro return mnt; 19749d412a43SAl Viro } 19759d412a43SAl Viro EXPORT_SYMBOL_GPL(do_kern_mount); 19769d412a43SAl Viro 19779d412a43SAl Viro /* 19789d412a43SAl Viro * add a mount into a namespace's mount tree 19799d412a43SAl Viro */ 19809d412a43SAl Viro static int do_add_mount(struct vfsmount *newmnt, struct path *path, int mnt_flags) 19819d412a43SAl Viro { 19829d412a43SAl Viro int err; 19839d412a43SAl Viro 19849d412a43SAl Viro mnt_flags &= ~(MNT_SHARED | MNT_WRITE_HOLD | MNT_INTERNAL); 19859d412a43SAl Viro 1986b12cea91SAl Viro err = lock_mount(path); 1987b12cea91SAl Viro if (err) 1988b12cea91SAl Viro return err; 19899d412a43SAl Viro 19909d412a43SAl Viro err = -EINVAL; 19919d412a43SAl Viro if (!(mnt_flags & MNT_SHRINKABLE) && !check_mnt(path->mnt)) 19929d412a43SAl Viro goto unlock; 19939d412a43SAl Viro 19949d412a43SAl Viro /* Refuse the same filesystem on the same mount point */ 19959d412a43SAl Viro err = -EBUSY; 19969d412a43SAl Viro if (path->mnt->mnt_sb == newmnt->mnt_sb && 19979d412a43SAl Viro path->mnt->mnt_root == path->dentry) 19989d412a43SAl Viro goto unlock; 19999d412a43SAl Viro 20009d412a43SAl Viro err = -EINVAL; 20019d412a43SAl Viro if (S_ISLNK(newmnt->mnt_root->d_inode->i_mode)) 20029d412a43SAl Viro goto unlock; 20039d412a43SAl Viro 20049d412a43SAl Viro newmnt->mnt_flags = mnt_flags; 20059d412a43SAl Viro err = graft_tree(newmnt, path); 20069d412a43SAl Viro 20079d412a43SAl Viro unlock: 2008b12cea91SAl Viro unlock_mount(path); 20099d412a43SAl Viro return err; 20109d412a43SAl Viro } 2011b1e75df4SAl Viro 20121da177e4SLinus Torvalds /* 20131da177e4SLinus Torvalds * create a new mount for userspace and request it to be added into the 20141da177e4SLinus Torvalds * namespace's tree 20151da177e4SLinus Torvalds */ 20160a0d8a46SAl Viro static int do_new_mount(struct path *path, char *type, int flags, 20171da177e4SLinus Torvalds int mnt_flags, char *name, void *data) 20181da177e4SLinus Torvalds { 20191da177e4SLinus Torvalds struct vfsmount *mnt; 202015f9a3f3SAl Viro int err; 20211da177e4SLinus Torvalds 2022eca6f534SVegard Nossum if (!type) 20231da177e4SLinus Torvalds return -EINVAL; 20241da177e4SLinus Torvalds 20251da177e4SLinus Torvalds /* we need capabilities... */ 20261da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 20271da177e4SLinus Torvalds return -EPERM; 20281da177e4SLinus Torvalds 20291da177e4SLinus Torvalds mnt = do_kern_mount(type, flags, name, data); 20301da177e4SLinus Torvalds if (IS_ERR(mnt)) 20311da177e4SLinus Torvalds return PTR_ERR(mnt); 20321da177e4SLinus Torvalds 203315f9a3f3SAl Viro err = do_add_mount(mnt, path, mnt_flags); 203415f9a3f3SAl Viro if (err) 203515f9a3f3SAl Viro mntput(mnt); 203615f9a3f3SAl Viro return err; 20371da177e4SLinus Torvalds } 20381da177e4SLinus Torvalds 203919a167afSAl Viro int finish_automount(struct vfsmount *m, struct path *path) 204019a167afSAl Viro { 204119a167afSAl Viro int err; 204219a167afSAl Viro /* The new mount record should have at least 2 refs to prevent it being 204319a167afSAl Viro * expired before we get a chance to add it 204419a167afSAl Viro */ 204519a167afSAl Viro BUG_ON(mnt_get_count(m) < 2); 204619a167afSAl Viro 204719a167afSAl Viro if (m->mnt_sb == path->mnt->mnt_sb && 204819a167afSAl Viro m->mnt_root == path->dentry) { 2049b1e75df4SAl Viro err = -ELOOP; 2050b1e75df4SAl Viro goto fail; 205119a167afSAl Viro } 205219a167afSAl Viro 205319a167afSAl Viro err = do_add_mount(m, path, path->mnt->mnt_flags | MNT_SHRINKABLE); 2054b1e75df4SAl Viro if (!err) 2055b1e75df4SAl Viro return 0; 2056b1e75df4SAl Viro fail: 2057b1e75df4SAl Viro /* remove m from any expiration list it may be on */ 2058b1e75df4SAl Viro if (!list_empty(&m->mnt_expire)) { 2059b1e75df4SAl Viro down_write(&namespace_sem); 2060b1e75df4SAl Viro br_write_lock(vfsmount_lock); 2061b1e75df4SAl Viro list_del_init(&m->mnt_expire); 2062b1e75df4SAl Viro br_write_unlock(vfsmount_lock); 2063b1e75df4SAl Viro up_write(&namespace_sem); 206419a167afSAl Viro } 2065b1e75df4SAl Viro mntput(m); 2066b1e75df4SAl Viro mntput(m); 206719a167afSAl Viro return err; 206819a167afSAl Viro } 206919a167afSAl Viro 2070ea5b778aSDavid Howells /** 2071ea5b778aSDavid Howells * mnt_set_expiry - Put a mount on an expiration list 2072ea5b778aSDavid Howells * @mnt: The mount to list. 2073ea5b778aSDavid Howells * @expiry_list: The list to add the mount to. 2074ea5b778aSDavid Howells */ 2075ea5b778aSDavid Howells void mnt_set_expiry(struct vfsmount *mnt, struct list_head *expiry_list) 2076ea5b778aSDavid Howells { 2077ea5b778aSDavid Howells down_write(&namespace_sem); 2078ea5b778aSDavid Howells br_write_lock(vfsmount_lock); 2079ea5b778aSDavid Howells 2080ea5b778aSDavid Howells list_add_tail(&mnt->mnt_expire, expiry_list); 2081ea5b778aSDavid Howells 2082ea5b778aSDavid Howells br_write_unlock(vfsmount_lock); 2083ea5b778aSDavid Howells up_write(&namespace_sem); 2084ea5b778aSDavid Howells } 2085ea5b778aSDavid Howells EXPORT_SYMBOL(mnt_set_expiry); 2086ea5b778aSDavid Howells 2087ea5b778aSDavid Howells /* 20881da177e4SLinus Torvalds * process a list of expirable mountpoints with the intent of discarding any 20891da177e4SLinus Torvalds * mountpoints that aren't in use and haven't been touched since last we came 20901da177e4SLinus Torvalds * here 20911da177e4SLinus Torvalds */ 20921da177e4SLinus Torvalds void mark_mounts_for_expiry(struct list_head *mounts) 20931da177e4SLinus Torvalds { 20941da177e4SLinus Torvalds struct vfsmount *mnt, *next; 20951da177e4SLinus Torvalds LIST_HEAD(graveyard); 2096bcc5c7d2SAl Viro LIST_HEAD(umounts); 20971da177e4SLinus Torvalds 20981da177e4SLinus Torvalds if (list_empty(mounts)) 20991da177e4SLinus Torvalds return; 21001da177e4SLinus Torvalds 2101bcc5c7d2SAl Viro down_write(&namespace_sem); 210299b7db7bSNick Piggin br_write_lock(vfsmount_lock); 21031da177e4SLinus Torvalds 21041da177e4SLinus Torvalds /* extract from the expiration list every vfsmount that matches the 21051da177e4SLinus Torvalds * following criteria: 21061da177e4SLinus Torvalds * - only referenced by its parent vfsmount 21071da177e4SLinus Torvalds * - still marked for expiry (marked on the last call here; marks are 21081da177e4SLinus Torvalds * cleared by mntput()) 21091da177e4SLinus Torvalds */ 211055e700b9SMiklos Szeredi list_for_each_entry_safe(mnt, next, mounts, mnt_expire) { 21111da177e4SLinus Torvalds if (!xchg(&mnt->mnt_expiry_mark, 1) || 2112bcc5c7d2SAl Viro propagate_mount_busy(mnt, 1)) 21131da177e4SLinus Torvalds continue; 211455e700b9SMiklos Szeredi list_move(&mnt->mnt_expire, &graveyard); 21151da177e4SLinus Torvalds } 2116bcc5c7d2SAl Viro while (!list_empty(&graveyard)) { 2117bcc5c7d2SAl Viro mnt = list_first_entry(&graveyard, struct vfsmount, mnt_expire); 2118bcc5c7d2SAl Viro touch_mnt_namespace(mnt->mnt_ns); 2119bcc5c7d2SAl Viro umount_tree(mnt, 1, &umounts); 2120bcc5c7d2SAl Viro } 212199b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 2122bcc5c7d2SAl Viro up_write(&namespace_sem); 2123bcc5c7d2SAl Viro 2124bcc5c7d2SAl Viro release_mounts(&umounts); 21251da177e4SLinus Torvalds } 21261da177e4SLinus Torvalds 21271da177e4SLinus Torvalds EXPORT_SYMBOL_GPL(mark_mounts_for_expiry); 21281da177e4SLinus Torvalds 21291da177e4SLinus Torvalds /* 21305528f911STrond Myklebust * Ripoff of 'select_parent()' 21315528f911STrond Myklebust * 21325528f911STrond Myklebust * search the list of submounts for a given mountpoint, and move any 21335528f911STrond Myklebust * shrinkable submounts to the 'graveyard' list. 21345528f911STrond Myklebust */ 21355528f911STrond Myklebust static int select_submounts(struct vfsmount *parent, struct list_head *graveyard) 21365528f911STrond Myklebust { 21375528f911STrond Myklebust struct vfsmount *this_parent = parent; 21385528f911STrond Myklebust struct list_head *next; 21395528f911STrond Myklebust int found = 0; 21405528f911STrond Myklebust 21415528f911STrond Myklebust repeat: 21425528f911STrond Myklebust next = this_parent->mnt_mounts.next; 21435528f911STrond Myklebust resume: 21445528f911STrond Myklebust while (next != &this_parent->mnt_mounts) { 21455528f911STrond Myklebust struct list_head *tmp = next; 21465528f911STrond Myklebust struct vfsmount *mnt = list_entry(tmp, struct vfsmount, mnt_child); 21475528f911STrond Myklebust 21485528f911STrond Myklebust next = tmp->next; 21495528f911STrond Myklebust if (!(mnt->mnt_flags & MNT_SHRINKABLE)) 21505528f911STrond Myklebust continue; 21515528f911STrond Myklebust /* 21525528f911STrond Myklebust * Descend a level if the d_mounts list is non-empty. 21535528f911STrond Myklebust */ 21545528f911STrond Myklebust if (!list_empty(&mnt->mnt_mounts)) { 21555528f911STrond Myklebust this_parent = mnt; 21565528f911STrond Myklebust goto repeat; 21575528f911STrond Myklebust } 21585528f911STrond Myklebust 21595528f911STrond Myklebust if (!propagate_mount_busy(mnt, 1)) { 21605528f911STrond Myklebust list_move_tail(&mnt->mnt_expire, graveyard); 21615528f911STrond Myklebust found++; 21625528f911STrond Myklebust } 21635528f911STrond Myklebust } 21645528f911STrond Myklebust /* 21655528f911STrond Myklebust * All done at this level ... ascend and resume the search 21665528f911STrond Myklebust */ 21675528f911STrond Myklebust if (this_parent != parent) { 21685528f911STrond Myklebust next = this_parent->mnt_child.next; 21695528f911STrond Myklebust this_parent = this_parent->mnt_parent; 21705528f911STrond Myklebust goto resume; 21715528f911STrond Myklebust } 21725528f911STrond Myklebust return found; 21735528f911STrond Myklebust } 21745528f911STrond Myklebust 21755528f911STrond Myklebust /* 21765528f911STrond Myklebust * process a list of expirable mountpoints with the intent of discarding any 21775528f911STrond Myklebust * submounts of a specific parent mountpoint 217899b7db7bSNick Piggin * 217999b7db7bSNick Piggin * vfsmount_lock must be held for write 21805528f911STrond Myklebust */ 2181c35038beSAl Viro static void shrink_submounts(struct vfsmount *mnt, struct list_head *umounts) 21825528f911STrond Myklebust { 21835528f911STrond Myklebust LIST_HEAD(graveyard); 2184c35038beSAl Viro struct vfsmount *m; 21855528f911STrond Myklebust 21865528f911STrond Myklebust /* extract submounts of 'mountpoint' from the expiration list */ 2187c35038beSAl Viro while (select_submounts(mnt, &graveyard)) { 2188bcc5c7d2SAl Viro while (!list_empty(&graveyard)) { 2189c35038beSAl Viro m = list_first_entry(&graveyard, struct vfsmount, 2190bcc5c7d2SAl Viro mnt_expire); 2191afef80b3SEric W. Biederman touch_mnt_namespace(m->mnt_ns); 2192afef80b3SEric W. Biederman umount_tree(m, 1, umounts); 2193bcc5c7d2SAl Viro } 2194bcc5c7d2SAl Viro } 21955528f911STrond Myklebust } 21965528f911STrond Myklebust 21975528f911STrond Myklebust /* 21981da177e4SLinus Torvalds * Some copy_from_user() implementations do not return the exact number of 21991da177e4SLinus Torvalds * bytes remaining to copy on a fault. But copy_mount_options() requires that. 22001da177e4SLinus Torvalds * Note that this function differs from copy_from_user() in that it will oops 22011da177e4SLinus Torvalds * on bad values of `to', rather than returning a short copy. 22021da177e4SLinus Torvalds */ 2203b58fed8bSRam Pai static long exact_copy_from_user(void *to, const void __user * from, 2204b58fed8bSRam Pai unsigned long n) 22051da177e4SLinus Torvalds { 22061da177e4SLinus Torvalds char *t = to; 22071da177e4SLinus Torvalds const char __user *f = from; 22081da177e4SLinus Torvalds char c; 22091da177e4SLinus Torvalds 22101da177e4SLinus Torvalds if (!access_ok(VERIFY_READ, from, n)) 22111da177e4SLinus Torvalds return n; 22121da177e4SLinus Torvalds 22131da177e4SLinus Torvalds while (n) { 22141da177e4SLinus Torvalds if (__get_user(c, f)) { 22151da177e4SLinus Torvalds memset(t, 0, n); 22161da177e4SLinus Torvalds break; 22171da177e4SLinus Torvalds } 22181da177e4SLinus Torvalds *t++ = c; 22191da177e4SLinus Torvalds f++; 22201da177e4SLinus Torvalds n--; 22211da177e4SLinus Torvalds } 22221da177e4SLinus Torvalds return n; 22231da177e4SLinus Torvalds } 22241da177e4SLinus Torvalds 22251da177e4SLinus Torvalds int copy_mount_options(const void __user * data, unsigned long *where) 22261da177e4SLinus Torvalds { 22271da177e4SLinus Torvalds int i; 22281da177e4SLinus Torvalds unsigned long page; 22291da177e4SLinus Torvalds unsigned long size; 22301da177e4SLinus Torvalds 22311da177e4SLinus Torvalds *where = 0; 22321da177e4SLinus Torvalds if (!data) 22331da177e4SLinus Torvalds return 0; 22341da177e4SLinus Torvalds 22351da177e4SLinus Torvalds if (!(page = __get_free_page(GFP_KERNEL))) 22361da177e4SLinus Torvalds return -ENOMEM; 22371da177e4SLinus Torvalds 22381da177e4SLinus Torvalds /* We only care that *some* data at the address the user 22391da177e4SLinus Torvalds * gave us is valid. Just in case, we'll zero 22401da177e4SLinus Torvalds * the remainder of the page. 22411da177e4SLinus Torvalds */ 22421da177e4SLinus Torvalds /* copy_from_user cannot cross TASK_SIZE ! */ 22431da177e4SLinus Torvalds size = TASK_SIZE - (unsigned long)data; 22441da177e4SLinus Torvalds if (size > PAGE_SIZE) 22451da177e4SLinus Torvalds size = PAGE_SIZE; 22461da177e4SLinus Torvalds 22471da177e4SLinus Torvalds i = size - exact_copy_from_user((void *)page, data, size); 22481da177e4SLinus Torvalds if (!i) { 22491da177e4SLinus Torvalds free_page(page); 22501da177e4SLinus Torvalds return -EFAULT; 22511da177e4SLinus Torvalds } 22521da177e4SLinus Torvalds if (i != PAGE_SIZE) 22531da177e4SLinus Torvalds memset((char *)page + i, 0, PAGE_SIZE - i); 22541da177e4SLinus Torvalds *where = page; 22551da177e4SLinus Torvalds return 0; 22561da177e4SLinus Torvalds } 22571da177e4SLinus Torvalds 2258eca6f534SVegard Nossum int copy_mount_string(const void __user *data, char **where) 2259eca6f534SVegard Nossum { 2260eca6f534SVegard Nossum char *tmp; 2261eca6f534SVegard Nossum 2262eca6f534SVegard Nossum if (!data) { 2263eca6f534SVegard Nossum *where = NULL; 2264eca6f534SVegard Nossum return 0; 2265eca6f534SVegard Nossum } 2266eca6f534SVegard Nossum 2267eca6f534SVegard Nossum tmp = strndup_user(data, PAGE_SIZE); 2268eca6f534SVegard Nossum if (IS_ERR(tmp)) 2269eca6f534SVegard Nossum return PTR_ERR(tmp); 2270eca6f534SVegard Nossum 2271eca6f534SVegard Nossum *where = tmp; 2272eca6f534SVegard Nossum return 0; 2273eca6f534SVegard Nossum } 2274eca6f534SVegard Nossum 22751da177e4SLinus Torvalds /* 22761da177e4SLinus Torvalds * Flags is a 32-bit value that allows up to 31 non-fs dependent flags to 22771da177e4SLinus Torvalds * be given to the mount() call (ie: read-only, no-dev, no-suid etc). 22781da177e4SLinus Torvalds * 22791da177e4SLinus Torvalds * data is a (void *) that can point to any structure up to 22801da177e4SLinus Torvalds * PAGE_SIZE-1 bytes, which can contain arbitrary fs-dependent 22811da177e4SLinus Torvalds * information (or be NULL). 22821da177e4SLinus Torvalds * 22831da177e4SLinus Torvalds * Pre-0.97 versions of mount() didn't have a flags word. 22841da177e4SLinus Torvalds * When the flags word was introduced its top half was required 22851da177e4SLinus Torvalds * to have the magic value 0xC0ED, and this remained so until 2.4.0-test9. 22861da177e4SLinus Torvalds * Therefore, if this magic number is present, it carries no information 22871da177e4SLinus Torvalds * and must be discarded. 22881da177e4SLinus Torvalds */ 22891da177e4SLinus Torvalds long do_mount(char *dev_name, char *dir_name, char *type_page, 22901da177e4SLinus Torvalds unsigned long flags, void *data_page) 22911da177e4SLinus Torvalds { 22922d92ab3cSAl Viro struct path path; 22931da177e4SLinus Torvalds int retval = 0; 22941da177e4SLinus Torvalds int mnt_flags = 0; 22951da177e4SLinus Torvalds 22961da177e4SLinus Torvalds /* Discard magic */ 22971da177e4SLinus Torvalds if ((flags & MS_MGC_MSK) == MS_MGC_VAL) 22981da177e4SLinus Torvalds flags &= ~MS_MGC_MSK; 22991da177e4SLinus Torvalds 23001da177e4SLinus Torvalds /* Basic sanity checks */ 23011da177e4SLinus Torvalds 23021da177e4SLinus Torvalds if (!dir_name || !*dir_name || !memchr(dir_name, 0, PAGE_SIZE)) 23031da177e4SLinus Torvalds return -EINVAL; 23041da177e4SLinus Torvalds 23051da177e4SLinus Torvalds if (data_page) 23061da177e4SLinus Torvalds ((char *)data_page)[PAGE_SIZE - 1] = 0; 23071da177e4SLinus Torvalds 2308a27ab9f2STetsuo Handa /* ... and get the mountpoint */ 2309a27ab9f2STetsuo Handa retval = kern_path(dir_name, LOOKUP_FOLLOW, &path); 2310a27ab9f2STetsuo Handa if (retval) 2311a27ab9f2STetsuo Handa return retval; 2312a27ab9f2STetsuo Handa 2313a27ab9f2STetsuo Handa retval = security_sb_mount(dev_name, &path, 2314a27ab9f2STetsuo Handa type_page, flags, data_page); 2315a27ab9f2STetsuo Handa if (retval) 2316a27ab9f2STetsuo Handa goto dput_out; 2317a27ab9f2STetsuo Handa 2318613cbe3dSAndi Kleen /* Default to relatime unless overriden */ 2319613cbe3dSAndi Kleen if (!(flags & MS_NOATIME)) 23200a1c01c9SMatthew Garrett mnt_flags |= MNT_RELATIME; 23210a1c01c9SMatthew Garrett 23221da177e4SLinus Torvalds /* Separate the per-mountpoint flags */ 23231da177e4SLinus Torvalds if (flags & MS_NOSUID) 23241da177e4SLinus Torvalds mnt_flags |= MNT_NOSUID; 23251da177e4SLinus Torvalds if (flags & MS_NODEV) 23261da177e4SLinus Torvalds mnt_flags |= MNT_NODEV; 23271da177e4SLinus Torvalds if (flags & MS_NOEXEC) 23281da177e4SLinus Torvalds mnt_flags |= MNT_NOEXEC; 2329fc33a7bbSChristoph Hellwig if (flags & MS_NOATIME) 2330fc33a7bbSChristoph Hellwig mnt_flags |= MNT_NOATIME; 2331fc33a7bbSChristoph Hellwig if (flags & MS_NODIRATIME) 2332fc33a7bbSChristoph Hellwig mnt_flags |= MNT_NODIRATIME; 2333d0adde57SMatthew Garrett if (flags & MS_STRICTATIME) 2334d0adde57SMatthew Garrett mnt_flags &= ~(MNT_RELATIME | MNT_NOATIME); 23352e4b7fcdSDave Hansen if (flags & MS_RDONLY) 23362e4b7fcdSDave Hansen mnt_flags |= MNT_READONLY; 2337fc33a7bbSChristoph Hellwig 23387a4dec53SAl Viro flags &= ~(MS_NOSUID | MS_NOEXEC | MS_NODEV | MS_ACTIVE | MS_BORN | 2339d0adde57SMatthew Garrett MS_NOATIME | MS_NODIRATIME | MS_RELATIME| MS_KERNMOUNT | 2340d0adde57SMatthew Garrett MS_STRICTATIME); 23411da177e4SLinus Torvalds 23421da177e4SLinus Torvalds if (flags & MS_REMOUNT) 23432d92ab3cSAl Viro retval = do_remount(&path, flags & ~MS_REMOUNT, mnt_flags, 23441da177e4SLinus Torvalds data_page); 23451da177e4SLinus Torvalds else if (flags & MS_BIND) 23462d92ab3cSAl Viro retval = do_loopback(&path, dev_name, flags & MS_REC); 23479676f0c6SRam Pai else if (flags & (MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE)) 23482d92ab3cSAl Viro retval = do_change_type(&path, flags); 23491da177e4SLinus Torvalds else if (flags & MS_MOVE) 23502d92ab3cSAl Viro retval = do_move_mount(&path, dev_name); 23511da177e4SLinus Torvalds else 23522d92ab3cSAl Viro retval = do_new_mount(&path, type_page, flags, mnt_flags, 23531da177e4SLinus Torvalds dev_name, data_page); 23541da177e4SLinus Torvalds dput_out: 23552d92ab3cSAl Viro path_put(&path); 23561da177e4SLinus Torvalds return retval; 23571da177e4SLinus Torvalds } 23581da177e4SLinus Torvalds 2359cf8d2c11STrond Myklebust static struct mnt_namespace *alloc_mnt_ns(void) 2360cf8d2c11STrond Myklebust { 2361cf8d2c11STrond Myklebust struct mnt_namespace *new_ns; 2362cf8d2c11STrond Myklebust 2363cf8d2c11STrond Myklebust new_ns = kmalloc(sizeof(struct mnt_namespace), GFP_KERNEL); 2364cf8d2c11STrond Myklebust if (!new_ns) 2365cf8d2c11STrond Myklebust return ERR_PTR(-ENOMEM); 2366cf8d2c11STrond Myklebust atomic_set(&new_ns->count, 1); 2367cf8d2c11STrond Myklebust new_ns->root = NULL; 2368cf8d2c11STrond Myklebust INIT_LIST_HEAD(&new_ns->list); 2369cf8d2c11STrond Myklebust init_waitqueue_head(&new_ns->poll); 2370cf8d2c11STrond Myklebust new_ns->event = 0; 2371cf8d2c11STrond Myklebust return new_ns; 2372cf8d2c11STrond Myklebust } 2373cf8d2c11STrond Myklebust 2374f03c6599SAl Viro void mnt_make_longterm(struct vfsmount *mnt) 2375f03c6599SAl Viro { 23767e3d0eb0SAl Viro __mnt_make_longterm(mnt); 2377f03c6599SAl Viro } 2378f03c6599SAl Viro 2379f03c6599SAl Viro void mnt_make_shortterm(struct vfsmount *mnt) 2380f03c6599SAl Viro { 23817e3d0eb0SAl Viro #ifdef CONFIG_SMP 2382f03c6599SAl Viro if (atomic_add_unless(&mnt->mnt_longterm, -1, 1)) 2383f03c6599SAl Viro return; 2384f03c6599SAl Viro br_write_lock(vfsmount_lock); 2385f03c6599SAl Viro atomic_dec(&mnt->mnt_longterm); 2386f03c6599SAl Viro br_write_unlock(vfsmount_lock); 23877e3d0eb0SAl Viro #endif 2388f03c6599SAl Viro } 2389f03c6599SAl Viro 2390741a2951SJANAK DESAI /* 2391741a2951SJANAK DESAI * Allocate a new namespace structure and populate it with contents 2392741a2951SJANAK DESAI * copied from the namespace of the passed in task structure. 2393741a2951SJANAK DESAI */ 2394e3222c4eSBadari Pulavarty static struct mnt_namespace *dup_mnt_ns(struct mnt_namespace *mnt_ns, 23956b3286edSKirill Korotaev struct fs_struct *fs) 23961da177e4SLinus Torvalds { 23976b3286edSKirill Korotaev struct mnt_namespace *new_ns; 23987f2da1e7SAl Viro struct vfsmount *rootmnt = NULL, *pwdmnt = NULL; 23991da177e4SLinus Torvalds struct vfsmount *p, *q; 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 */ 24076b3286edSKirill Korotaev new_ns->root = copy_tree(mnt_ns->root, mnt_ns->root->mnt_root, 24089676f0c6SRam Pai CL_COPY_ALL | CL_EXPIRE); 24091da177e4SLinus Torvalds if (!new_ns->root) { 2410390c6843SRam Pai up_write(&namespace_sem); 24111da177e4SLinus Torvalds kfree(new_ns); 24125cc4a034SJulia Lawall return ERR_PTR(-ENOMEM); 24131da177e4SLinus Torvalds } 241499b7db7bSNick Piggin br_write_lock(vfsmount_lock); 24151da177e4SLinus Torvalds list_add_tail(&new_ns->list, &new_ns->root->mnt_list); 241699b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 24171da177e4SLinus Torvalds 24181da177e4SLinus Torvalds /* 24191da177e4SLinus Torvalds * Second pass: switch the tsk->fs->* elements and mark new vfsmounts 24201da177e4SLinus Torvalds * as belonging to new namespace. We have already acquired a private 24211da177e4SLinus Torvalds * fs_struct, so tsk->fs->lock is not needed. 24221da177e4SLinus Torvalds */ 24236b3286edSKirill Korotaev p = mnt_ns->root; 24241da177e4SLinus Torvalds q = new_ns->root; 24251da177e4SLinus Torvalds while (p) { 24266b3286edSKirill Korotaev q->mnt_ns = new_ns; 24277e3d0eb0SAl Viro __mnt_make_longterm(q); 24281da177e4SLinus Torvalds if (fs) { 24296ac08c39SJan Blunck if (p == fs->root.mnt) { 2430f03c6599SAl Viro fs->root.mnt = mntget(q); 24317e3d0eb0SAl Viro __mnt_make_longterm(q); 2432f03c6599SAl Viro mnt_make_shortterm(p); 24331da177e4SLinus Torvalds rootmnt = p; 24341da177e4SLinus Torvalds } 24356ac08c39SJan Blunck if (p == fs->pwd.mnt) { 2436f03c6599SAl Viro fs->pwd.mnt = mntget(q); 24377e3d0eb0SAl Viro __mnt_make_longterm(q); 2438f03c6599SAl Viro mnt_make_shortterm(p); 24391da177e4SLinus Torvalds pwdmnt = p; 24401da177e4SLinus Torvalds } 24411da177e4SLinus Torvalds } 24426b3286edSKirill Korotaev p = next_mnt(p, mnt_ns->root); 24431da177e4SLinus Torvalds q = next_mnt(q, new_ns->root); 24441da177e4SLinus Torvalds } 2445390c6843SRam Pai up_write(&namespace_sem); 24461da177e4SLinus Torvalds 24471da177e4SLinus Torvalds if (rootmnt) 2448f03c6599SAl Viro mntput(rootmnt); 24491da177e4SLinus Torvalds if (pwdmnt) 2450f03c6599SAl Viro mntput(pwdmnt); 24511da177e4SLinus Torvalds 2452741a2951SJANAK DESAI return new_ns; 2453741a2951SJANAK DESAI } 2454741a2951SJANAK DESAI 2455213dd266SEric W. Biederman struct mnt_namespace *copy_mnt_ns(unsigned long flags, struct mnt_namespace *ns, 2456e3222c4eSBadari Pulavarty struct fs_struct *new_fs) 2457741a2951SJANAK DESAI { 24586b3286edSKirill Korotaev struct mnt_namespace *new_ns; 2459741a2951SJANAK DESAI 2460e3222c4eSBadari Pulavarty BUG_ON(!ns); 24616b3286edSKirill Korotaev get_mnt_ns(ns); 2462741a2951SJANAK DESAI 2463741a2951SJANAK DESAI if (!(flags & CLONE_NEWNS)) 2464e3222c4eSBadari Pulavarty return ns; 2465741a2951SJANAK DESAI 2466e3222c4eSBadari Pulavarty new_ns = dup_mnt_ns(ns, new_fs); 2467741a2951SJANAK DESAI 24686b3286edSKirill Korotaev put_mnt_ns(ns); 2469e3222c4eSBadari Pulavarty return new_ns; 24701da177e4SLinus Torvalds } 24711da177e4SLinus Torvalds 2472cf8d2c11STrond Myklebust /** 2473cf8d2c11STrond Myklebust * create_mnt_ns - creates a private namespace and adds a root filesystem 2474cf8d2c11STrond Myklebust * @mnt: pointer to the new root filesystem mountpoint 2475cf8d2c11STrond Myklebust */ 2476a2770d86SLinus Torvalds struct mnt_namespace *create_mnt_ns(struct vfsmount *mnt) 2477cf8d2c11STrond Myklebust { 2478cf8d2c11STrond Myklebust struct mnt_namespace *new_ns; 2479cf8d2c11STrond Myklebust 2480cf8d2c11STrond Myklebust new_ns = alloc_mnt_ns(); 2481cf8d2c11STrond Myklebust if (!IS_ERR(new_ns)) { 2482cf8d2c11STrond Myklebust mnt->mnt_ns = new_ns; 24837e3d0eb0SAl Viro __mnt_make_longterm(mnt); 2484cf8d2c11STrond Myklebust new_ns->root = mnt; 2485cf8d2c11STrond Myklebust list_add(&new_ns->list, &new_ns->root->mnt_list); 2486c1334495SAl Viro } else { 2487c1334495SAl Viro mntput(mnt); 2488cf8d2c11STrond Myklebust } 2489cf8d2c11STrond Myklebust return new_ns; 2490cf8d2c11STrond Myklebust } 2491a2770d86SLinus Torvalds EXPORT_SYMBOL(create_mnt_ns); 2492cf8d2c11STrond Myklebust 2493ea441d11SAl Viro struct dentry *mount_subtree(struct vfsmount *mnt, const char *name) 2494ea441d11SAl Viro { 2495ea441d11SAl Viro struct mnt_namespace *ns; 2496ea441d11SAl Viro struct path path; 2497ea441d11SAl Viro int err; 2498ea441d11SAl Viro 2499ea441d11SAl Viro ns = create_mnt_ns(mnt); 2500ea441d11SAl Viro if (IS_ERR(ns)) 2501ea441d11SAl Viro return ERR_CAST(ns); 2502ea441d11SAl Viro 2503ea441d11SAl Viro err = vfs_path_lookup(mnt->mnt_root, mnt, 2504ea441d11SAl Viro name, LOOKUP_FOLLOW|LOOKUP_AUTOMOUNT, &path); 2505ea441d11SAl Viro 2506ea441d11SAl Viro put_mnt_ns(ns); 2507ea441d11SAl Viro 2508ea441d11SAl Viro if (err) 2509ea441d11SAl Viro return ERR_PTR(err); 2510ea441d11SAl Viro 2511ea441d11SAl Viro /* trade a vfsmount reference for active sb one */ 2512ea441d11SAl Viro atomic_inc(&path.mnt->mnt_sb->s_active); 2513ea441d11SAl Viro mntput(path.mnt); 2514ea441d11SAl Viro /* lock the sucker */ 2515ea441d11SAl Viro down_write(&path.mnt->mnt_sb->s_umount); 2516ea441d11SAl Viro /* ... and return the root of (sub)tree on it */ 2517ea441d11SAl Viro return path.dentry; 2518ea441d11SAl Viro } 2519ea441d11SAl Viro EXPORT_SYMBOL(mount_subtree); 2520ea441d11SAl Viro 2521bdc480e3SHeiko Carstens SYSCALL_DEFINE5(mount, char __user *, dev_name, char __user *, dir_name, 2522bdc480e3SHeiko Carstens char __user *, type, unsigned long, flags, void __user *, data) 25231da177e4SLinus Torvalds { 2524eca6f534SVegard Nossum int ret; 2525eca6f534SVegard Nossum char *kernel_type; 2526eca6f534SVegard Nossum char *kernel_dir; 2527eca6f534SVegard Nossum char *kernel_dev; 25281da177e4SLinus Torvalds unsigned long data_page; 25291da177e4SLinus Torvalds 2530eca6f534SVegard Nossum ret = copy_mount_string(type, &kernel_type); 2531eca6f534SVegard Nossum if (ret < 0) 2532eca6f534SVegard Nossum goto out_type; 25331da177e4SLinus Torvalds 2534eca6f534SVegard Nossum kernel_dir = getname(dir_name); 2535eca6f534SVegard Nossum if (IS_ERR(kernel_dir)) { 2536eca6f534SVegard Nossum ret = PTR_ERR(kernel_dir); 2537eca6f534SVegard Nossum goto out_dir; 2538eca6f534SVegard Nossum } 25391da177e4SLinus Torvalds 2540eca6f534SVegard Nossum ret = copy_mount_string(dev_name, &kernel_dev); 2541eca6f534SVegard Nossum if (ret < 0) 2542eca6f534SVegard Nossum goto out_dev; 25431da177e4SLinus Torvalds 2544eca6f534SVegard Nossum ret = copy_mount_options(data, &data_page); 2545eca6f534SVegard Nossum if (ret < 0) 2546eca6f534SVegard Nossum goto out_data; 25471da177e4SLinus Torvalds 2548eca6f534SVegard Nossum ret = do_mount(kernel_dev, kernel_dir, kernel_type, flags, 2549eca6f534SVegard Nossum (void *) data_page); 2550eca6f534SVegard Nossum 25511da177e4SLinus Torvalds free_page(data_page); 2552eca6f534SVegard Nossum out_data: 2553eca6f534SVegard Nossum kfree(kernel_dev); 2554eca6f534SVegard Nossum out_dev: 2555eca6f534SVegard Nossum putname(kernel_dir); 2556eca6f534SVegard Nossum out_dir: 2557eca6f534SVegard Nossum kfree(kernel_type); 2558eca6f534SVegard Nossum out_type: 2559eca6f534SVegard Nossum return ret; 25601da177e4SLinus Torvalds } 25611da177e4SLinus Torvalds 25621da177e4SLinus Torvalds /* 25631da177e4SLinus Torvalds * pivot_root Semantics: 25641da177e4SLinus Torvalds * Moves the root file system of the current process to the directory put_old, 25651da177e4SLinus Torvalds * makes new_root as the new root file system of the current process, and sets 25661da177e4SLinus Torvalds * root/cwd of all processes which had them on the current root to new_root. 25671da177e4SLinus Torvalds * 25681da177e4SLinus Torvalds * Restrictions: 25691da177e4SLinus Torvalds * The new_root and put_old must be directories, and must not be on the 25701da177e4SLinus Torvalds * same file system as the current process root. The put_old must be 25711da177e4SLinus Torvalds * underneath new_root, i.e. adding a non-zero number of /.. to the string 25721da177e4SLinus Torvalds * pointed to by put_old must yield the same directory as new_root. No other 25731da177e4SLinus Torvalds * file system may be mounted on put_old. After all, new_root is a mountpoint. 25741da177e4SLinus Torvalds * 25754a0d11faSNeil Brown * Also, the current root cannot be on the 'rootfs' (initial ramfs) filesystem. 25764a0d11faSNeil Brown * See Documentation/filesystems/ramfs-rootfs-initramfs.txt for alternatives 25774a0d11faSNeil Brown * in this situation. 25784a0d11faSNeil Brown * 25791da177e4SLinus Torvalds * Notes: 25801da177e4SLinus Torvalds * - we don't move root/cwd if they are not at the root (reason: if something 25811da177e4SLinus Torvalds * cared enough to change them, it's probably wrong to force them elsewhere) 25821da177e4SLinus Torvalds * - it's okay to pick a root that isn't the root of a file system, e.g. 25831da177e4SLinus Torvalds * /nfs/my_root where /nfs is the mount point. It must be a mountpoint, 25841da177e4SLinus Torvalds * though, so you may need to say mount --bind /nfs/my_root /nfs/my_root 25851da177e4SLinus Torvalds * first. 25861da177e4SLinus Torvalds */ 25873480b257SHeiko Carstens SYSCALL_DEFINE2(pivot_root, const char __user *, new_root, 25883480b257SHeiko Carstens const char __user *, put_old) 25891da177e4SLinus Torvalds { 25901da177e4SLinus Torvalds struct vfsmount *tmp; 25912d8f3038SAl Viro struct path new, old, parent_path, root_parent, root; 25921da177e4SLinus Torvalds int error; 25931da177e4SLinus Torvalds 25941da177e4SLinus Torvalds if (!capable(CAP_SYS_ADMIN)) 25951da177e4SLinus Torvalds return -EPERM; 25961da177e4SLinus Torvalds 25972d8f3038SAl Viro error = user_path_dir(new_root, &new); 25981da177e4SLinus Torvalds if (error) 25991da177e4SLinus Torvalds goto out0; 26001da177e4SLinus Torvalds 26012d8f3038SAl Viro error = user_path_dir(put_old, &old); 26021da177e4SLinus Torvalds if (error) 26031da177e4SLinus Torvalds goto out1; 26041da177e4SLinus Torvalds 26052d8f3038SAl Viro error = security_sb_pivotroot(&old, &new); 2606b12cea91SAl Viro if (error) 2607b12cea91SAl Viro goto out2; 26081da177e4SLinus Torvalds 2609f7ad3c6bSMiklos Szeredi get_fs_root(current->fs, &root); 2610b12cea91SAl Viro error = lock_mount(&old); 2611b12cea91SAl Viro if (error) 2612b12cea91SAl Viro goto out3; 2613b12cea91SAl Viro 26141da177e4SLinus Torvalds error = -EINVAL; 26152d8f3038SAl Viro if (IS_MNT_SHARED(old.mnt) || 26162d8f3038SAl Viro IS_MNT_SHARED(new.mnt->mnt_parent) || 26178c3ee42eSAl Viro IS_MNT_SHARED(root.mnt->mnt_parent)) 2618b12cea91SAl Viro goto out4; 261927cb1572SAl Viro if (!check_mnt(root.mnt) || !check_mnt(new.mnt)) 2620b12cea91SAl Viro goto out4; 26211da177e4SLinus Torvalds error = -ENOENT; 2622f3da392eSAlexey Dobriyan if (d_unlinked(new.dentry)) 2623b12cea91SAl Viro goto out4; 2624f3da392eSAlexey Dobriyan if (d_unlinked(old.dentry)) 2625b12cea91SAl Viro goto out4; 26261da177e4SLinus Torvalds error = -EBUSY; 26272d8f3038SAl Viro if (new.mnt == root.mnt || 26282d8f3038SAl Viro old.mnt == root.mnt) 2629b12cea91SAl Viro goto out4; /* loop, on the same file system */ 26301da177e4SLinus Torvalds error = -EINVAL; 26318c3ee42eSAl Viro if (root.mnt->mnt_root != root.dentry) 2632b12cea91SAl Viro goto out4; /* not a mountpoint */ 26338c3ee42eSAl Viro if (root.mnt->mnt_parent == root.mnt) 2634b12cea91SAl Viro goto out4; /* not attached */ 26352d8f3038SAl Viro if (new.mnt->mnt_root != new.dentry) 2636b12cea91SAl Viro goto out4; /* not a mountpoint */ 26372d8f3038SAl Viro if (new.mnt->mnt_parent == new.mnt) 2638b12cea91SAl Viro goto out4; /* not attached */ 26394ac91378SJan Blunck /* make sure we can reach put_old from new_root */ 26402d8f3038SAl Viro tmp = old.mnt; 26412d8f3038SAl Viro if (tmp != new.mnt) { 26421da177e4SLinus Torvalds for (;;) { 26431da177e4SLinus Torvalds if (tmp->mnt_parent == tmp) 2644b12cea91SAl Viro goto out4; /* already mounted on put_old */ 26452d8f3038SAl Viro if (tmp->mnt_parent == new.mnt) 26461da177e4SLinus Torvalds break; 26471da177e4SLinus Torvalds tmp = tmp->mnt_parent; 26481da177e4SLinus Torvalds } 26492d8f3038SAl Viro if (!is_subdir(tmp->mnt_mountpoint, new.dentry)) 2650b12cea91SAl Viro goto out4; 26512d8f3038SAl Viro } else if (!is_subdir(old.dentry, new.dentry)) 2652b12cea91SAl Viro goto out4; 265327cb1572SAl Viro br_write_lock(vfsmount_lock); 26542d8f3038SAl Viro detach_mnt(new.mnt, &parent_path); 26558c3ee42eSAl Viro detach_mnt(root.mnt, &root_parent); 26564ac91378SJan Blunck /* mount old root on put_old */ 26572d8f3038SAl Viro attach_mnt(root.mnt, &old); 26584ac91378SJan Blunck /* mount new_root on / */ 26592d8f3038SAl Viro attach_mnt(new.mnt, &root_parent); 26606b3286edSKirill Korotaev touch_mnt_namespace(current->nsproxy->mnt_ns); 266199b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 26622d8f3038SAl Viro chroot_fs_refs(&root, &new); 26631da177e4SLinus Torvalds error = 0; 2664b12cea91SAl Viro out4: 2665b12cea91SAl Viro unlock_mount(&old); 2666b12cea91SAl Viro if (!error) { 26671a390689SAl Viro path_put(&root_parent); 26681a390689SAl Viro path_put(&parent_path); 2669b12cea91SAl Viro } 2670b12cea91SAl Viro out3: 26718c3ee42eSAl Viro path_put(&root); 2672b12cea91SAl Viro out2: 26732d8f3038SAl Viro path_put(&old); 26741da177e4SLinus Torvalds out1: 26752d8f3038SAl Viro path_put(&new); 26761da177e4SLinus Torvalds out0: 26771da177e4SLinus Torvalds return error; 26781da177e4SLinus Torvalds } 26791da177e4SLinus Torvalds 26801da177e4SLinus Torvalds static void __init init_mount_tree(void) 26811da177e4SLinus Torvalds { 26821da177e4SLinus Torvalds struct vfsmount *mnt; 26836b3286edSKirill Korotaev struct mnt_namespace *ns; 2684ac748a09SJan Blunck struct path root; 26851da177e4SLinus Torvalds 26861da177e4SLinus Torvalds mnt = do_kern_mount("rootfs", 0, "rootfs", NULL); 26871da177e4SLinus Torvalds if (IS_ERR(mnt)) 26881da177e4SLinus Torvalds panic("Can't create rootfs"); 2689b3e19d92SNick Piggin 26903b22edc5STrond Myklebust ns = create_mnt_ns(mnt); 26913b22edc5STrond Myklebust if (IS_ERR(ns)) 26921da177e4SLinus Torvalds panic("Can't allocate initial namespace"); 26931da177e4SLinus Torvalds 26946b3286edSKirill Korotaev init_task.nsproxy->mnt_ns = ns; 26956b3286edSKirill Korotaev get_mnt_ns(ns); 26961da177e4SLinus Torvalds 2697ac748a09SJan Blunck root.mnt = ns->root; 2698ac748a09SJan Blunck root.dentry = ns->root->mnt_root; 2699ac748a09SJan Blunck 2700ac748a09SJan Blunck set_fs_pwd(current->fs, &root); 2701ac748a09SJan Blunck set_fs_root(current->fs, &root); 27021da177e4SLinus Torvalds } 27031da177e4SLinus Torvalds 270474bf17cfSDenis Cheng void __init mnt_init(void) 27051da177e4SLinus Torvalds { 270613f14b4dSEric Dumazet unsigned u; 270715a67dd8SRandy Dunlap int err; 27081da177e4SLinus Torvalds 2709390c6843SRam Pai init_rwsem(&namespace_sem); 2710390c6843SRam Pai 27111da177e4SLinus Torvalds mnt_cache = kmem_cache_create("mnt_cache", sizeof(struct vfsmount), 271220c2df83SPaul Mundt 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC, NULL); 27131da177e4SLinus Torvalds 2714b58fed8bSRam Pai mount_hashtable = (struct list_head *)__get_free_page(GFP_ATOMIC); 27151da177e4SLinus Torvalds 27161da177e4SLinus Torvalds if (!mount_hashtable) 27171da177e4SLinus Torvalds panic("Failed to allocate mount hash table\n"); 27181da177e4SLinus Torvalds 271980cdc6daSMandeep Singh Baines printk(KERN_INFO "Mount-cache hash table entries: %lu\n", HASH_SIZE); 27201da177e4SLinus Torvalds 272113f14b4dSEric Dumazet for (u = 0; u < HASH_SIZE; u++) 272213f14b4dSEric Dumazet INIT_LIST_HEAD(&mount_hashtable[u]); 27231da177e4SLinus Torvalds 272499b7db7bSNick Piggin br_lock_init(vfsmount_lock); 272599b7db7bSNick Piggin 272615a67dd8SRandy Dunlap err = sysfs_init(); 272715a67dd8SRandy Dunlap if (err) 272815a67dd8SRandy Dunlap printk(KERN_WARNING "%s: sysfs_init error: %d\n", 27298e24eea7SHarvey Harrison __func__, err); 273000d26666SGreg Kroah-Hartman fs_kobj = kobject_create_and_add("fs", NULL); 273100d26666SGreg Kroah-Hartman if (!fs_kobj) 27328e24eea7SHarvey Harrison printk(KERN_WARNING "%s: kobj create error\n", __func__); 27331da177e4SLinus Torvalds init_rootfs(); 27341da177e4SLinus Torvalds init_mount_tree(); 27351da177e4SLinus Torvalds } 27361da177e4SLinus Torvalds 2737616511d0STrond Myklebust void put_mnt_ns(struct mnt_namespace *ns) 27381da177e4SLinus Torvalds { 273970fbcdf4SRam Pai LIST_HEAD(umount_list); 2740616511d0STrond Myklebust 2741d498b25aSAl Viro if (!atomic_dec_and_test(&ns->count)) 2742616511d0STrond Myklebust return; 2743390c6843SRam Pai down_write(&namespace_sem); 274499b7db7bSNick Piggin br_write_lock(vfsmount_lock); 2745d498b25aSAl Viro umount_tree(ns->root, 0, &umount_list); 274699b7db7bSNick Piggin br_write_unlock(vfsmount_lock); 2747390c6843SRam Pai up_write(&namespace_sem); 274870fbcdf4SRam Pai release_mounts(&umount_list); 27496b3286edSKirill Korotaev kfree(ns); 27501da177e4SLinus Torvalds } 2751cf8d2c11STrond Myklebust EXPORT_SYMBOL(put_mnt_ns); 27529d412a43SAl Viro 27539d412a43SAl Viro struct vfsmount *kern_mount_data(struct file_system_type *type, void *data) 27549d412a43SAl Viro { 2755423e0ab0STim Chen struct vfsmount *mnt; 2756423e0ab0STim Chen mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, data); 2757423e0ab0STim Chen if (!IS_ERR(mnt)) { 2758423e0ab0STim Chen /* 2759423e0ab0STim Chen * it is a longterm mount, don't release mnt until 2760423e0ab0STim Chen * we unmount before file sys is unregistered 2761423e0ab0STim Chen */ 2762423e0ab0STim Chen mnt_make_longterm(mnt); 2763423e0ab0STim Chen } 2764423e0ab0STim Chen return mnt; 27659d412a43SAl Viro } 27669d412a43SAl Viro EXPORT_SYMBOL_GPL(kern_mount_data); 2767423e0ab0STim Chen 2768423e0ab0STim Chen void kern_unmount(struct vfsmount *mnt) 2769423e0ab0STim Chen { 2770423e0ab0STim Chen /* release long term mount so mount point can be released */ 2771423e0ab0STim Chen if (!IS_ERR_OR_NULL(mnt)) { 2772423e0ab0STim Chen mnt_make_shortterm(mnt); 2773423e0ab0STim Chen mntput(mnt); 2774423e0ab0STim Chen } 2775423e0ab0STim Chen } 2776423e0ab0STim Chen EXPORT_SYMBOL(kern_unmount); 2777