11da177e4SLinus Torvalds /* 21da177e4SLinus Torvalds * linux/fs/namei.c 31da177e4SLinus Torvalds * 41da177e4SLinus Torvalds * Copyright (C) 1991, 1992 Linus Torvalds 51da177e4SLinus Torvalds */ 61da177e4SLinus Torvalds 71da177e4SLinus Torvalds /* 81da177e4SLinus Torvalds * Some corrections by tytso. 91da177e4SLinus Torvalds */ 101da177e4SLinus Torvalds 111da177e4SLinus Torvalds /* [Feb 1997 T. Schoebel-Theuer] Complete rewrite of the pathname 121da177e4SLinus Torvalds * lookup logic. 131da177e4SLinus Torvalds */ 141da177e4SLinus Torvalds /* [Feb-Apr 2000, AV] Rewrite to the new namespace architecture. 151da177e4SLinus Torvalds */ 161da177e4SLinus Torvalds 171da177e4SLinus Torvalds #include <linux/init.h> 181da177e4SLinus Torvalds #include <linux/module.h> 191da177e4SLinus Torvalds #include <linux/slab.h> 201da177e4SLinus Torvalds #include <linux/fs.h> 211da177e4SLinus Torvalds #include <linux/namei.h> 221da177e4SLinus Torvalds #include <linux/pagemap.h> 230eeca283SRobert Love #include <linux/fsnotify.h> 241da177e4SLinus Torvalds #include <linux/personality.h> 251da177e4SLinus Torvalds #include <linux/security.h> 266146f0d5SMimi Zohar #include <linux/ima.h> 271da177e4SLinus Torvalds #include <linux/syscalls.h> 281da177e4SLinus Torvalds #include <linux/mount.h> 291da177e4SLinus Torvalds #include <linux/audit.h> 3016f7e0feSRandy Dunlap #include <linux/capability.h> 31834f2a4aSTrond Myklebust #include <linux/file.h> 325590ff0dSUlrich Drepper #include <linux/fcntl.h> 3308ce5f16SSerge E. Hallyn #include <linux/device_cgroup.h> 345ad4e53bSAl Viro #include <linux/fs_struct.h> 351da177e4SLinus Torvalds #include <asm/uaccess.h> 361da177e4SLinus Torvalds 37e81e3f4dSEric Paris #include "internal.h" 38e81e3f4dSEric Paris 391da177e4SLinus Torvalds /* [Feb-1997 T. Schoebel-Theuer] 401da177e4SLinus Torvalds * Fundamental changes in the pathname lookup mechanisms (namei) 411da177e4SLinus Torvalds * were necessary because of omirr. The reason is that omirr needs 421da177e4SLinus Torvalds * to know the _real_ pathname, not the user-supplied one, in case 431da177e4SLinus Torvalds * of symlinks (and also when transname replacements occur). 441da177e4SLinus Torvalds * 451da177e4SLinus Torvalds * The new code replaces the old recursive symlink resolution with 461da177e4SLinus Torvalds * an iterative one (in case of non-nested symlink chains). It does 471da177e4SLinus Torvalds * this with calls to <fs>_follow_link(). 481da177e4SLinus Torvalds * As a side effect, dir_namei(), _namei() and follow_link() are now 491da177e4SLinus Torvalds * replaced with a single function lookup_dentry() that can handle all 501da177e4SLinus Torvalds * the special cases of the former code. 511da177e4SLinus Torvalds * 521da177e4SLinus Torvalds * With the new dcache, the pathname is stored at each inode, at least as 531da177e4SLinus Torvalds * long as the refcount of the inode is positive. As a side effect, the 541da177e4SLinus Torvalds * size of the dcache depends on the inode cache and thus is dynamic. 551da177e4SLinus Torvalds * 561da177e4SLinus Torvalds * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink 571da177e4SLinus Torvalds * resolution to correspond with current state of the code. 581da177e4SLinus Torvalds * 591da177e4SLinus Torvalds * Note that the symlink resolution is not *completely* iterative. 601da177e4SLinus Torvalds * There is still a significant amount of tail- and mid- recursion in 611da177e4SLinus Torvalds * the algorithm. Also, note that <fs>_readlink() is not used in 621da177e4SLinus Torvalds * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink() 631da177e4SLinus Torvalds * may return different results than <fs>_follow_link(). Many virtual 641da177e4SLinus Torvalds * filesystems (including /proc) exhibit this behavior. 651da177e4SLinus Torvalds */ 661da177e4SLinus Torvalds 671da177e4SLinus Torvalds /* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation: 681da177e4SLinus Torvalds * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL 691da177e4SLinus Torvalds * and the name already exists in form of a symlink, try to create the new 701da177e4SLinus Torvalds * name indicated by the symlink. The old code always complained that the 711da177e4SLinus Torvalds * name already exists, due to not following the symlink even if its target 721da177e4SLinus Torvalds * is nonexistent. The new semantics affects also mknod() and link() when 731da177e4SLinus Torvalds * the name is a symlink pointing to a non-existant name. 741da177e4SLinus Torvalds * 751da177e4SLinus Torvalds * I don't know which semantics is the right one, since I have no access 761da177e4SLinus Torvalds * to standards. But I found by trial that HP-UX 9.0 has the full "new" 771da177e4SLinus Torvalds * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the 781da177e4SLinus Torvalds * "old" one. Personally, I think the new semantics is much more logical. 791da177e4SLinus Torvalds * Note that "ln old new" where "new" is a symlink pointing to a non-existing 801da177e4SLinus Torvalds * file does succeed in both HP-UX and SunOs, but not in Solaris 811da177e4SLinus Torvalds * and in the old Linux semantics. 821da177e4SLinus Torvalds */ 831da177e4SLinus Torvalds 841da177e4SLinus Torvalds /* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink 851da177e4SLinus Torvalds * semantics. See the comments in "open_namei" and "do_link" below. 861da177e4SLinus Torvalds * 871da177e4SLinus Torvalds * [10-Sep-98 Alan Modra] Another symlink change. 881da177e4SLinus Torvalds */ 891da177e4SLinus Torvalds 901da177e4SLinus Torvalds /* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks: 911da177e4SLinus Torvalds * inside the path - always follow. 921da177e4SLinus Torvalds * in the last component in creation/removal/renaming - never follow. 931da177e4SLinus Torvalds * if LOOKUP_FOLLOW passed - follow. 941da177e4SLinus Torvalds * if the pathname has trailing slashes - follow. 951da177e4SLinus Torvalds * otherwise - don't follow. 961da177e4SLinus Torvalds * (applied in that order). 971da177e4SLinus Torvalds * 981da177e4SLinus Torvalds * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT 991da177e4SLinus Torvalds * restored for 2.4. This is the last surviving part of old 4.2BSD bug. 1001da177e4SLinus Torvalds * During the 2.4 we need to fix the userland stuff depending on it - 1011da177e4SLinus Torvalds * hopefully we will be able to get rid of that wart in 2.5. So far only 1021da177e4SLinus Torvalds * XEmacs seems to be relying on it... 1031da177e4SLinus Torvalds */ 1041da177e4SLinus Torvalds /* 1051da177e4SLinus Torvalds * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland) 106a11f3a05SArjan van de Ven * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives 1071da177e4SLinus Torvalds * any extra contention... 1081da177e4SLinus Torvalds */ 1091da177e4SLinus Torvalds 1101da177e4SLinus Torvalds /* In order to reduce some races, while at the same time doing additional 1111da177e4SLinus Torvalds * checking and hopefully speeding things up, we copy filenames to the 1121da177e4SLinus Torvalds * kernel data space before using them.. 1131da177e4SLinus Torvalds * 1141da177e4SLinus Torvalds * POSIX.1 2.4: an empty pathname is invalid (ENOENT). 1151da177e4SLinus Torvalds * PATH_MAX includes the nul terminator --RR. 1161da177e4SLinus Torvalds */ 117858119e1SArjan van de Ven static int do_getname(const char __user *filename, char *page) 1181da177e4SLinus Torvalds { 1191da177e4SLinus Torvalds int retval; 1201da177e4SLinus Torvalds unsigned long len = PATH_MAX; 1211da177e4SLinus Torvalds 1221da177e4SLinus Torvalds if (!segment_eq(get_fs(), KERNEL_DS)) { 1231da177e4SLinus Torvalds if ((unsigned long) filename >= TASK_SIZE) 1241da177e4SLinus Torvalds return -EFAULT; 1251da177e4SLinus Torvalds if (TASK_SIZE - (unsigned long) filename < PATH_MAX) 1261da177e4SLinus Torvalds len = TASK_SIZE - (unsigned long) filename; 1271da177e4SLinus Torvalds } 1281da177e4SLinus Torvalds 1291da177e4SLinus Torvalds retval = strncpy_from_user(page, filename, len); 1301da177e4SLinus Torvalds if (retval > 0) { 1311da177e4SLinus Torvalds if (retval < len) 1321da177e4SLinus Torvalds return 0; 1331da177e4SLinus Torvalds return -ENAMETOOLONG; 1341da177e4SLinus Torvalds } else if (!retval) 1351da177e4SLinus Torvalds retval = -ENOENT; 1361da177e4SLinus Torvalds return retval; 1371da177e4SLinus Torvalds } 1381da177e4SLinus Torvalds 1391da177e4SLinus Torvalds char * getname(const char __user * filename) 1401da177e4SLinus Torvalds { 1411da177e4SLinus Torvalds char *tmp, *result; 1421da177e4SLinus Torvalds 1431da177e4SLinus Torvalds result = ERR_PTR(-ENOMEM); 1441da177e4SLinus Torvalds tmp = __getname(); 1451da177e4SLinus Torvalds if (tmp) { 1461da177e4SLinus Torvalds int retval = do_getname(filename, tmp); 1471da177e4SLinus Torvalds 1481da177e4SLinus Torvalds result = tmp; 1491da177e4SLinus Torvalds if (retval < 0) { 1501da177e4SLinus Torvalds __putname(tmp); 1511da177e4SLinus Torvalds result = ERR_PTR(retval); 1521da177e4SLinus Torvalds } 1531da177e4SLinus Torvalds } 1541da177e4SLinus Torvalds audit_getname(result); 1551da177e4SLinus Torvalds return result; 1561da177e4SLinus Torvalds } 1571da177e4SLinus Torvalds 1581da177e4SLinus Torvalds #ifdef CONFIG_AUDITSYSCALL 1591da177e4SLinus Torvalds void putname(const char *name) 1601da177e4SLinus Torvalds { 1615ac3a9c2SAl Viro if (unlikely(!audit_dummy_context())) 1621da177e4SLinus Torvalds audit_putname(name); 1631da177e4SLinus Torvalds else 1641da177e4SLinus Torvalds __putname(name); 1651da177e4SLinus Torvalds } 1661da177e4SLinus Torvalds EXPORT_SYMBOL(putname); 1671da177e4SLinus Torvalds #endif 1681da177e4SLinus Torvalds 1695909ccaaSLinus Torvalds /* 1705909ccaaSLinus Torvalds * This does basic POSIX ACL permission checking 1715909ccaaSLinus Torvalds */ 1725909ccaaSLinus Torvalds static int acl_permission_check(struct inode *inode, int mask, 1735909ccaaSLinus Torvalds int (*check_acl)(struct inode *inode, int mask)) 1745909ccaaSLinus Torvalds { 1755909ccaaSLinus Torvalds umode_t mode = inode->i_mode; 1765909ccaaSLinus Torvalds 1775909ccaaSLinus Torvalds mask &= MAY_READ | MAY_WRITE | MAY_EXEC; 1785909ccaaSLinus Torvalds 1795909ccaaSLinus Torvalds if (current_fsuid() == inode->i_uid) 1805909ccaaSLinus Torvalds mode >>= 6; 1815909ccaaSLinus Torvalds else { 1825909ccaaSLinus Torvalds if (IS_POSIXACL(inode) && (mode & S_IRWXG) && check_acl) { 1835909ccaaSLinus Torvalds int error = check_acl(inode, mask); 1845909ccaaSLinus Torvalds if (error != -EAGAIN) 1855909ccaaSLinus Torvalds return error; 1865909ccaaSLinus Torvalds } 1875909ccaaSLinus Torvalds 1885909ccaaSLinus Torvalds if (in_group_p(inode->i_gid)) 1895909ccaaSLinus Torvalds mode >>= 3; 1905909ccaaSLinus Torvalds } 1915909ccaaSLinus Torvalds 1925909ccaaSLinus Torvalds /* 1935909ccaaSLinus Torvalds * If the DACs are ok we don't need any capability check. 1945909ccaaSLinus Torvalds */ 1955909ccaaSLinus Torvalds if ((mask & ~mode) == 0) 1965909ccaaSLinus Torvalds return 0; 1975909ccaaSLinus Torvalds return -EACCES; 1985909ccaaSLinus Torvalds } 1991da177e4SLinus Torvalds 2001da177e4SLinus Torvalds /** 2011da177e4SLinus Torvalds * generic_permission - check for access rights on a Posix-like filesystem 2021da177e4SLinus Torvalds * @inode: inode to check access rights for 2031da177e4SLinus Torvalds * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) 2041da177e4SLinus Torvalds * @check_acl: optional callback to check for Posix ACLs 2051da177e4SLinus Torvalds * 2061da177e4SLinus Torvalds * Used to check for read/write/execute permissions on a file. 2071da177e4SLinus Torvalds * We use "fsuid" for this, letting us set arbitrary permissions 2081da177e4SLinus Torvalds * for filesystem access without changing the "normal" uids which 2091da177e4SLinus Torvalds * are used for other things.. 2101da177e4SLinus Torvalds */ 2111da177e4SLinus Torvalds int generic_permission(struct inode *inode, int mask, 2121da177e4SLinus Torvalds int (*check_acl)(struct inode *inode, int mask)) 2131da177e4SLinus Torvalds { 2145909ccaaSLinus Torvalds int ret; 2151da177e4SLinus Torvalds 2161da177e4SLinus Torvalds /* 2175909ccaaSLinus Torvalds * Do the basic POSIX ACL permission checks. 2181da177e4SLinus Torvalds */ 2195909ccaaSLinus Torvalds ret = acl_permission_check(inode, mask, check_acl); 2205909ccaaSLinus Torvalds if (ret != -EACCES) 2215909ccaaSLinus Torvalds return ret; 2221da177e4SLinus Torvalds 2231da177e4SLinus Torvalds /* 2241da177e4SLinus Torvalds * Read/write DACs are always overridable. 2251da177e4SLinus Torvalds * Executable DACs are overridable if at least one exec bit is set. 2261da177e4SLinus Torvalds */ 227f696a365SMiklos Szeredi if (!(mask & MAY_EXEC) || execute_ok(inode)) 2281da177e4SLinus Torvalds if (capable(CAP_DAC_OVERRIDE)) 2291da177e4SLinus Torvalds return 0; 2301da177e4SLinus Torvalds 2311da177e4SLinus Torvalds /* 2321da177e4SLinus Torvalds * Searching includes executable on directories, else just read. 2331da177e4SLinus Torvalds */ 2347ea66001SSerge E. Hallyn mask &= MAY_READ | MAY_WRITE | MAY_EXEC; 2351da177e4SLinus Torvalds if (mask == MAY_READ || (S_ISDIR(inode->i_mode) && !(mask & MAY_WRITE))) 2361da177e4SLinus Torvalds if (capable(CAP_DAC_READ_SEARCH)) 2371da177e4SLinus Torvalds return 0; 2381da177e4SLinus Torvalds 2391da177e4SLinus Torvalds return -EACCES; 2401da177e4SLinus Torvalds } 2411da177e4SLinus Torvalds 242cb23beb5SChristoph Hellwig /** 243cb23beb5SChristoph Hellwig * inode_permission - check for access rights to a given inode 244cb23beb5SChristoph Hellwig * @inode: inode to check permission on 245cb23beb5SChristoph Hellwig * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) 246cb23beb5SChristoph Hellwig * 247cb23beb5SChristoph Hellwig * Used to check for read/write/execute permissions on an inode. 248cb23beb5SChristoph Hellwig * We use "fsuid" for this, letting us set arbitrary permissions 249cb23beb5SChristoph Hellwig * for filesystem access without changing the "normal" uids which 250cb23beb5SChristoph Hellwig * are used for other things. 251cb23beb5SChristoph Hellwig */ 252f419a2e3SAl Viro int inode_permission(struct inode *inode, int mask) 2531da177e4SLinus Torvalds { 254e6305c43SAl Viro int retval; 2551da177e4SLinus Torvalds 2561da177e4SLinus Torvalds if (mask & MAY_WRITE) { 25722590e41SMiklos Szeredi umode_t mode = inode->i_mode; 2581da177e4SLinus Torvalds 2591da177e4SLinus Torvalds /* 2601da177e4SLinus Torvalds * Nobody gets write access to a read-only fs. 2611da177e4SLinus Torvalds */ 2621da177e4SLinus Torvalds if (IS_RDONLY(inode) && 2631da177e4SLinus Torvalds (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) 2641da177e4SLinus Torvalds return -EROFS; 2651da177e4SLinus Torvalds 2661da177e4SLinus Torvalds /* 2671da177e4SLinus Torvalds * Nobody gets write access to an immutable file. 2681da177e4SLinus Torvalds */ 2691da177e4SLinus Torvalds if (IS_IMMUTABLE(inode)) 2701da177e4SLinus Torvalds return -EACCES; 2711da177e4SLinus Torvalds } 2721da177e4SLinus Torvalds 273acfa4380SAl Viro if (inode->i_op->permission) 274b77b0646SAl Viro retval = inode->i_op->permission(inode, mask); 275f696a365SMiklos Szeredi else 2765909ccaaSLinus Torvalds retval = generic_permission(inode, mask, inode->i_op->check_acl); 277f696a365SMiklos Szeredi 2781da177e4SLinus Torvalds if (retval) 2791da177e4SLinus Torvalds return retval; 2801da177e4SLinus Torvalds 28108ce5f16SSerge E. Hallyn retval = devcgroup_inode_permission(inode, mask); 28208ce5f16SSerge E. Hallyn if (retval) 28308ce5f16SSerge E. Hallyn return retval; 28408ce5f16SSerge E. Hallyn 285d09ca739SEric Paris return security_inode_permission(inode, mask); 2861da177e4SLinus Torvalds } 2871da177e4SLinus Torvalds 288e4543eddSChristoph Hellwig /** 2898c744fb8SChristoph Hellwig * file_permission - check for additional access rights to a given file 2908c744fb8SChristoph Hellwig * @file: file to check access rights for 2918c744fb8SChristoph Hellwig * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) 2928c744fb8SChristoph Hellwig * 2938c744fb8SChristoph Hellwig * Used to check for read/write/execute permissions on an already opened 2948c744fb8SChristoph Hellwig * file. 2958c744fb8SChristoph Hellwig * 2968c744fb8SChristoph Hellwig * Note: 2978c744fb8SChristoph Hellwig * Do not use this function in new code. All access checks should 298cb23beb5SChristoph Hellwig * be done using inode_permission(). 2998c744fb8SChristoph Hellwig */ 3008c744fb8SChristoph Hellwig int file_permission(struct file *file, int mask) 3018c744fb8SChristoph Hellwig { 302f419a2e3SAl Viro return inode_permission(file->f_path.dentry->d_inode, mask); 3038c744fb8SChristoph Hellwig } 3048c744fb8SChristoph Hellwig 3051da177e4SLinus Torvalds /* 3061da177e4SLinus Torvalds * get_write_access() gets write permission for a file. 3071da177e4SLinus Torvalds * put_write_access() releases this write permission. 3081da177e4SLinus Torvalds * This is used for regular files. 3091da177e4SLinus Torvalds * We cannot support write (and maybe mmap read-write shared) accesses and 3101da177e4SLinus Torvalds * MAP_DENYWRITE mmappings simultaneously. The i_writecount field of an inode 3111da177e4SLinus Torvalds * can have the following values: 3121da177e4SLinus Torvalds * 0: no writers, no VM_DENYWRITE mappings 3131da177e4SLinus Torvalds * < 0: (-i_writecount) vm_area_structs with VM_DENYWRITE set exist 3141da177e4SLinus Torvalds * > 0: (i_writecount) users are writing to the file. 3151da177e4SLinus Torvalds * 3161da177e4SLinus Torvalds * Normally we operate on that counter with atomic_{inc,dec} and it's safe 3171da177e4SLinus Torvalds * except for the cases where we don't hold i_writecount yet. Then we need to 3181da177e4SLinus Torvalds * use {get,deny}_write_access() - these functions check the sign and refuse 3191da177e4SLinus Torvalds * to do the change if sign is wrong. Exclusion between them is provided by 3201da177e4SLinus Torvalds * the inode->i_lock spinlock. 3211da177e4SLinus Torvalds */ 3221da177e4SLinus Torvalds 3231da177e4SLinus Torvalds int get_write_access(struct inode * inode) 3241da177e4SLinus Torvalds { 3251da177e4SLinus Torvalds spin_lock(&inode->i_lock); 3261da177e4SLinus Torvalds if (atomic_read(&inode->i_writecount) < 0) { 3271da177e4SLinus Torvalds spin_unlock(&inode->i_lock); 3281da177e4SLinus Torvalds return -ETXTBSY; 3291da177e4SLinus Torvalds } 3301da177e4SLinus Torvalds atomic_inc(&inode->i_writecount); 3311da177e4SLinus Torvalds spin_unlock(&inode->i_lock); 3321da177e4SLinus Torvalds 3331da177e4SLinus Torvalds return 0; 3341da177e4SLinus Torvalds } 3351da177e4SLinus Torvalds 3361da177e4SLinus Torvalds int deny_write_access(struct file * file) 3371da177e4SLinus Torvalds { 3380f7fc9e4SJosef "Jeff" Sipek struct inode *inode = file->f_path.dentry->d_inode; 3391da177e4SLinus Torvalds 3401da177e4SLinus Torvalds spin_lock(&inode->i_lock); 3411da177e4SLinus Torvalds if (atomic_read(&inode->i_writecount) > 0) { 3421da177e4SLinus Torvalds spin_unlock(&inode->i_lock); 3431da177e4SLinus Torvalds return -ETXTBSY; 3441da177e4SLinus Torvalds } 3451da177e4SLinus Torvalds atomic_dec(&inode->i_writecount); 3461da177e4SLinus Torvalds spin_unlock(&inode->i_lock); 3471da177e4SLinus Torvalds 3481da177e4SLinus Torvalds return 0; 3491da177e4SLinus Torvalds } 3501da177e4SLinus Torvalds 3511d957f9bSJan Blunck /** 3525dd784d0SJan Blunck * path_get - get a reference to a path 3535dd784d0SJan Blunck * @path: path to get the reference to 3545dd784d0SJan Blunck * 3555dd784d0SJan Blunck * Given a path increment the reference count to the dentry and the vfsmount. 3565dd784d0SJan Blunck */ 3575dd784d0SJan Blunck void path_get(struct path *path) 3585dd784d0SJan Blunck { 3595dd784d0SJan Blunck mntget(path->mnt); 3605dd784d0SJan Blunck dget(path->dentry); 3615dd784d0SJan Blunck } 3625dd784d0SJan Blunck EXPORT_SYMBOL(path_get); 3635dd784d0SJan Blunck 3645dd784d0SJan Blunck /** 3651d957f9bSJan Blunck * path_put - put a reference to a path 3661d957f9bSJan Blunck * @path: path to put the reference to 3671d957f9bSJan Blunck * 3681d957f9bSJan Blunck * Given a path decrement the reference count to the dentry and the vfsmount. 3691d957f9bSJan Blunck */ 3701d957f9bSJan Blunck void path_put(struct path *path) 3711da177e4SLinus Torvalds { 3721d957f9bSJan Blunck dput(path->dentry); 3731d957f9bSJan Blunck mntput(path->mnt); 3741da177e4SLinus Torvalds } 3751d957f9bSJan Blunck EXPORT_SYMBOL(path_put); 3761da177e4SLinus Torvalds 377834f2a4aSTrond Myklebust /** 378834f2a4aSTrond Myklebust * release_open_intent - free up open intent resources 379834f2a4aSTrond Myklebust * @nd: pointer to nameidata 380834f2a4aSTrond Myklebust */ 381834f2a4aSTrond Myklebust void release_open_intent(struct nameidata *nd) 382834f2a4aSTrond Myklebust { 3830f7fc9e4SJosef "Jeff" Sipek if (nd->intent.open.file->f_path.dentry == NULL) 384834f2a4aSTrond Myklebust put_filp(nd->intent.open.file); 385834f2a4aSTrond Myklebust else 386834f2a4aSTrond Myklebust fput(nd->intent.open.file); 387834f2a4aSTrond Myklebust } 388834f2a4aSTrond Myklebust 389bcdc5e01SIan Kent static inline struct dentry * 390bcdc5e01SIan Kent do_revalidate(struct dentry *dentry, struct nameidata *nd) 391bcdc5e01SIan Kent { 392bcdc5e01SIan Kent int status = dentry->d_op->d_revalidate(dentry, nd); 393bcdc5e01SIan Kent if (unlikely(status <= 0)) { 394bcdc5e01SIan Kent /* 395bcdc5e01SIan Kent * The dentry failed validation. 396bcdc5e01SIan Kent * If d_revalidate returned 0 attempt to invalidate 397bcdc5e01SIan Kent * the dentry otherwise d_revalidate is asking us 398bcdc5e01SIan Kent * to return a fail status. 399bcdc5e01SIan Kent */ 400bcdc5e01SIan Kent if (!status) { 401bcdc5e01SIan Kent if (!d_invalidate(dentry)) { 402bcdc5e01SIan Kent dput(dentry); 403bcdc5e01SIan Kent dentry = NULL; 404bcdc5e01SIan Kent } 405bcdc5e01SIan Kent } else { 406bcdc5e01SIan Kent dput(dentry); 407bcdc5e01SIan Kent dentry = ERR_PTR(status); 408bcdc5e01SIan Kent } 409bcdc5e01SIan Kent } 410bcdc5e01SIan Kent return dentry; 411bcdc5e01SIan Kent } 412bcdc5e01SIan Kent 4131da177e4SLinus Torvalds /* 41439159de2SJeff Layton * force_reval_path - force revalidation of a dentry 41539159de2SJeff Layton * 41639159de2SJeff Layton * In some situations the path walking code will trust dentries without 41739159de2SJeff Layton * revalidating them. This causes problems for filesystems that depend on 41839159de2SJeff Layton * d_revalidate to handle file opens (e.g. NFSv4). When FS_REVAL_DOT is set 41939159de2SJeff Layton * (which indicates that it's possible for the dentry to go stale), force 42039159de2SJeff Layton * a d_revalidate call before proceeding. 42139159de2SJeff Layton * 42239159de2SJeff Layton * Returns 0 if the revalidation was successful. If the revalidation fails, 42339159de2SJeff Layton * either return the error returned by d_revalidate or -ESTALE if the 42439159de2SJeff Layton * revalidation it just returned 0. If d_revalidate returns 0, we attempt to 42539159de2SJeff Layton * invalidate the dentry. It's up to the caller to handle putting references 42639159de2SJeff Layton * to the path if necessary. 42739159de2SJeff Layton */ 42839159de2SJeff Layton static int 42939159de2SJeff Layton force_reval_path(struct path *path, struct nameidata *nd) 43039159de2SJeff Layton { 43139159de2SJeff Layton int status; 43239159de2SJeff Layton struct dentry *dentry = path->dentry; 43339159de2SJeff Layton 43439159de2SJeff Layton /* 43539159de2SJeff Layton * only check on filesystems where it's possible for the dentry to 43639159de2SJeff Layton * become stale. It's assumed that if this flag is set then the 43739159de2SJeff Layton * d_revalidate op will also be defined. 43839159de2SJeff Layton */ 43939159de2SJeff Layton if (!(dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT)) 44039159de2SJeff Layton return 0; 44139159de2SJeff Layton 44239159de2SJeff Layton status = dentry->d_op->d_revalidate(dentry, nd); 44339159de2SJeff Layton if (status > 0) 44439159de2SJeff Layton return 0; 44539159de2SJeff Layton 44639159de2SJeff Layton if (!status) { 44739159de2SJeff Layton d_invalidate(dentry); 44839159de2SJeff Layton status = -ESTALE; 44939159de2SJeff Layton } 45039159de2SJeff Layton return status; 45139159de2SJeff Layton } 45239159de2SJeff Layton 45339159de2SJeff Layton /* 454b75b5086SAl Viro * Short-cut version of permission(), for calling on directories 455b75b5086SAl Viro * during pathname resolution. Combines parts of permission() 456b75b5086SAl Viro * and generic_permission(), and tests ONLY for MAY_EXEC permission. 4571da177e4SLinus Torvalds * 4581da177e4SLinus Torvalds * If appropriate, check DAC only. If not appropriate, or 459b75b5086SAl Viro * short-cut DAC fails, then call ->permission() to do more 4601da177e4SLinus Torvalds * complete permission check. 4611da177e4SLinus Torvalds */ 462b75b5086SAl Viro static int exec_permission(struct inode *inode) 4631da177e4SLinus Torvalds { 4645909ccaaSLinus Torvalds int ret; 4651da177e4SLinus Torvalds 466cb9179eaSLinus Torvalds if (inode->i_op->permission) { 4675909ccaaSLinus Torvalds ret = inode->i_op->permission(inode, MAY_EXEC); 468cb9179eaSLinus Torvalds if (!ret) 469cb9179eaSLinus Torvalds goto ok; 470cb9179eaSLinus Torvalds return ret; 471cb9179eaSLinus Torvalds } 4725909ccaaSLinus Torvalds ret = acl_permission_check(inode, MAY_EXEC, inode->i_op->check_acl); 4735909ccaaSLinus Torvalds if (!ret) 4741da177e4SLinus Torvalds goto ok; 4751da177e4SLinus Torvalds 476f1ac9f6bSLinus Torvalds if (capable(CAP_DAC_OVERRIDE) || capable(CAP_DAC_READ_SEARCH)) 4771da177e4SLinus Torvalds goto ok; 4781da177e4SLinus Torvalds 4795909ccaaSLinus Torvalds return ret; 4801da177e4SLinus Torvalds ok: 481b77b0646SAl Viro return security_inode_permission(inode, MAY_EXEC); 4821da177e4SLinus Torvalds } 4831da177e4SLinus Torvalds 4842a737871SAl Viro static __always_inline void set_root(struct nameidata *nd) 4852a737871SAl Viro { 486f7ad3c6bSMiklos Szeredi if (!nd->root.mnt) 487f7ad3c6bSMiklos Szeredi get_fs_root(current->fs, &nd->root); 4882a737871SAl Viro } 4892a737871SAl Viro 4906de88d72SAl Viro static int link_path_walk(const char *, struct nameidata *); 4916de88d72SAl Viro 492f1662356SArjan van de Ven static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link) 4931da177e4SLinus Torvalds { 4941da177e4SLinus Torvalds if (IS_ERR(link)) 4951da177e4SLinus Torvalds goto fail; 4961da177e4SLinus Torvalds 4971da177e4SLinus Torvalds if (*link == '/') { 4982a737871SAl Viro set_root(nd); 4991d957f9bSJan Blunck path_put(&nd->path); 5002a737871SAl Viro nd->path = nd->root; 5012a737871SAl Viro path_get(&nd->root); 5021da177e4SLinus Torvalds } 503b4091d5fSChristoph Hellwig 504def4af30SAl Viro return link_path_walk(link, nd); 5051da177e4SLinus Torvalds fail: 5061d957f9bSJan Blunck path_put(&nd->path); 5071da177e4SLinus Torvalds return PTR_ERR(link); 5081da177e4SLinus Torvalds } 5091da177e4SLinus Torvalds 5101d957f9bSJan Blunck static void path_put_conditional(struct path *path, struct nameidata *nd) 511051d3812SIan Kent { 512051d3812SIan Kent dput(path->dentry); 5134ac91378SJan Blunck if (path->mnt != nd->path.mnt) 514051d3812SIan Kent mntput(path->mnt); 515051d3812SIan Kent } 516051d3812SIan Kent 517051d3812SIan Kent static inline void path_to_nameidata(struct path *path, struct nameidata *nd) 518051d3812SIan Kent { 5194ac91378SJan Blunck dput(nd->path.dentry); 5209a229683SHuang Shijie if (nd->path.mnt != path->mnt) { 5214ac91378SJan Blunck mntput(nd->path.mnt); 5224ac91378SJan Blunck nd->path.mnt = path->mnt; 5239a229683SHuang Shijie } 5244ac91378SJan Blunck nd->path.dentry = path->dentry; 525051d3812SIan Kent } 526051d3812SIan Kent 527def4af30SAl Viro static __always_inline int 528def4af30SAl Viro __do_follow_link(struct path *path, struct nameidata *nd, void **p) 5291da177e4SLinus Torvalds { 5301da177e4SLinus Torvalds int error; 531cd4e91d3SAl Viro struct dentry *dentry = path->dentry; 5321da177e4SLinus Torvalds 533d671a1cbSAl Viro touch_atime(path->mnt, dentry); 5341da177e4SLinus Torvalds nd_set_link(nd, NULL); 535cd4e91d3SAl Viro 5364ac91378SJan Blunck if (path->mnt != nd->path.mnt) { 537051d3812SIan Kent path_to_nameidata(path, nd); 538051d3812SIan Kent dget(dentry); 539051d3812SIan Kent } 540cd4e91d3SAl Viro mntget(path->mnt); 54186acdca1SAl Viro nd->last_type = LAST_BIND; 542def4af30SAl Viro *p = dentry->d_inode->i_op->follow_link(dentry, nd); 543def4af30SAl Viro error = PTR_ERR(*p); 544def4af30SAl Viro if (!IS_ERR(*p)) { 5451da177e4SLinus Torvalds char *s = nd_get_link(nd); 546cc314eefSLinus Torvalds error = 0; 5471da177e4SLinus Torvalds if (s) 5481da177e4SLinus Torvalds error = __vfs_follow_link(nd, s); 54939159de2SJeff Layton else if (nd->last_type == LAST_BIND) { 55039159de2SJeff Layton error = force_reval_path(&nd->path, nd); 55139159de2SJeff Layton if (error) 55239159de2SJeff Layton path_put(&nd->path); 55339159de2SJeff Layton } 5541da177e4SLinus Torvalds } 5551da177e4SLinus Torvalds return error; 5561da177e4SLinus Torvalds } 5571da177e4SLinus Torvalds 5581da177e4SLinus Torvalds /* 5591da177e4SLinus Torvalds * This limits recursive symlink follows to 8, while 5601da177e4SLinus Torvalds * limiting consecutive symlinks to 40. 5611da177e4SLinus Torvalds * 5621da177e4SLinus Torvalds * Without that kind of total limit, nasty chains of consecutive 5631da177e4SLinus Torvalds * symlinks can cause almost arbitrarily long lookups. 5641da177e4SLinus Torvalds */ 56590ebe565SAl Viro static inline int do_follow_link(struct path *path, struct nameidata *nd) 5661da177e4SLinus Torvalds { 567def4af30SAl Viro void *cookie; 5681da177e4SLinus Torvalds int err = -ELOOP; 5691da177e4SLinus Torvalds if (current->link_count >= MAX_NESTED_LINKS) 5701da177e4SLinus Torvalds goto loop; 5711da177e4SLinus Torvalds if (current->total_link_count >= 40) 5721da177e4SLinus Torvalds goto loop; 5731da177e4SLinus Torvalds BUG_ON(nd->depth >= MAX_NESTED_LINKS); 5741da177e4SLinus Torvalds cond_resched(); 57590ebe565SAl Viro err = security_inode_follow_link(path->dentry, nd); 5761da177e4SLinus Torvalds if (err) 5771da177e4SLinus Torvalds goto loop; 5781da177e4SLinus Torvalds current->link_count++; 5791da177e4SLinus Torvalds current->total_link_count++; 5801da177e4SLinus Torvalds nd->depth++; 581def4af30SAl Viro err = __do_follow_link(path, nd, &cookie); 582def4af30SAl Viro if (!IS_ERR(cookie) && path->dentry->d_inode->i_op->put_link) 583def4af30SAl Viro path->dentry->d_inode->i_op->put_link(path->dentry, nd, cookie); 584258fa999SAl Viro path_put(path); 5851da177e4SLinus Torvalds current->link_count--; 5861da177e4SLinus Torvalds nd->depth--; 5871da177e4SLinus Torvalds return err; 5881da177e4SLinus Torvalds loop: 5891d957f9bSJan Blunck path_put_conditional(path, nd); 5901d957f9bSJan Blunck path_put(&nd->path); 5911da177e4SLinus Torvalds return err; 5921da177e4SLinus Torvalds } 5931da177e4SLinus Torvalds 594bab77ebfSAl Viro int follow_up(struct path *path) 5951da177e4SLinus Torvalds { 5961da177e4SLinus Torvalds struct vfsmount *parent; 5971da177e4SLinus Torvalds struct dentry *mountpoint; 59899b7db7bSNick Piggin 59999b7db7bSNick Piggin br_read_lock(vfsmount_lock); 600bab77ebfSAl Viro parent = path->mnt->mnt_parent; 601bab77ebfSAl Viro if (parent == path->mnt) { 60299b7db7bSNick Piggin br_read_unlock(vfsmount_lock); 6031da177e4SLinus Torvalds return 0; 6041da177e4SLinus Torvalds } 6051da177e4SLinus Torvalds mntget(parent); 606bab77ebfSAl Viro mountpoint = dget(path->mnt->mnt_mountpoint); 60799b7db7bSNick Piggin br_read_unlock(vfsmount_lock); 608bab77ebfSAl Viro dput(path->dentry); 609bab77ebfSAl Viro path->dentry = mountpoint; 610bab77ebfSAl Viro mntput(path->mnt); 611bab77ebfSAl Viro path->mnt = parent; 6121da177e4SLinus Torvalds return 1; 6131da177e4SLinus Torvalds } 6141da177e4SLinus Torvalds 615b5c84bf6SNick Piggin /* 616b5c84bf6SNick Piggin * serialization is taken care of in namespace.c 6171da177e4SLinus Torvalds */ 618463ffb2eSAl Viro static int __follow_mount(struct path *path) 619463ffb2eSAl Viro { 620463ffb2eSAl Viro int res = 0; 621463ffb2eSAl Viro while (d_mountpoint(path->dentry)) { 6221c755af4SAl Viro struct vfsmount *mounted = lookup_mnt(path); 623463ffb2eSAl Viro if (!mounted) 624463ffb2eSAl Viro break; 625463ffb2eSAl Viro dput(path->dentry); 626463ffb2eSAl Viro if (res) 627463ffb2eSAl Viro mntput(path->mnt); 628463ffb2eSAl Viro path->mnt = mounted; 629463ffb2eSAl Viro path->dentry = dget(mounted->mnt_root); 630463ffb2eSAl Viro res = 1; 631463ffb2eSAl Viro } 632463ffb2eSAl Viro return res; 633463ffb2eSAl Viro } 634463ffb2eSAl Viro 63579ed0226SAl Viro static void follow_mount(struct path *path) 6361da177e4SLinus Torvalds { 63779ed0226SAl Viro while (d_mountpoint(path->dentry)) { 6381c755af4SAl Viro struct vfsmount *mounted = lookup_mnt(path); 6391da177e4SLinus Torvalds if (!mounted) 6401da177e4SLinus Torvalds break; 64179ed0226SAl Viro dput(path->dentry); 64279ed0226SAl Viro mntput(path->mnt); 64379ed0226SAl Viro path->mnt = mounted; 64479ed0226SAl Viro path->dentry = dget(mounted->mnt_root); 6451da177e4SLinus Torvalds } 6461da177e4SLinus Torvalds } 6471da177e4SLinus Torvalds 6489393bd07SAl Viro int follow_down(struct path *path) 6491da177e4SLinus Torvalds { 6501da177e4SLinus Torvalds struct vfsmount *mounted; 6511da177e4SLinus Torvalds 6521c755af4SAl Viro mounted = lookup_mnt(path); 6531da177e4SLinus Torvalds if (mounted) { 6549393bd07SAl Viro dput(path->dentry); 6559393bd07SAl Viro mntput(path->mnt); 6569393bd07SAl Viro path->mnt = mounted; 6579393bd07SAl Viro path->dentry = dget(mounted->mnt_root); 6581da177e4SLinus Torvalds return 1; 6591da177e4SLinus Torvalds } 6601da177e4SLinus Torvalds return 0; 6611da177e4SLinus Torvalds } 6621da177e4SLinus Torvalds 663f1662356SArjan van de Ven static __always_inline void follow_dotdot(struct nameidata *nd) 6641da177e4SLinus Torvalds { 6652a737871SAl Viro set_root(nd); 666e518ddb7SAndreas Mohr 6671da177e4SLinus Torvalds while(1) { 6684ac91378SJan Blunck struct dentry *old = nd->path.dentry; 6691da177e4SLinus Torvalds 6702a737871SAl Viro if (nd->path.dentry == nd->root.dentry && 6712a737871SAl Viro nd->path.mnt == nd->root.mnt) { 6721da177e4SLinus Torvalds break; 6731da177e4SLinus Torvalds } 6744ac91378SJan Blunck if (nd->path.dentry != nd->path.mnt->mnt_root) { 6753088dd70SAl Viro /* rare case of legitimate dget_parent()... */ 6763088dd70SAl Viro nd->path.dentry = dget_parent(nd->path.dentry); 6771da177e4SLinus Torvalds dput(old); 6781da177e4SLinus Torvalds break; 6791da177e4SLinus Torvalds } 6803088dd70SAl Viro if (!follow_up(&nd->path)) 6811da177e4SLinus Torvalds break; 6821da177e4SLinus Torvalds } 68379ed0226SAl Viro follow_mount(&nd->path); 6841da177e4SLinus Torvalds } 6851da177e4SLinus Torvalds 6861da177e4SLinus Torvalds /* 687baa03890SNick Piggin * Allocate a dentry with name and parent, and perform a parent 688baa03890SNick Piggin * directory ->lookup on it. Returns the new dentry, or ERR_PTR 689baa03890SNick Piggin * on error. parent->d_inode->i_mutex must be held. d_lookup must 690baa03890SNick Piggin * have verified that no child exists while under i_mutex. 691baa03890SNick Piggin */ 692baa03890SNick Piggin static struct dentry *d_alloc_and_lookup(struct dentry *parent, 693baa03890SNick Piggin struct qstr *name, struct nameidata *nd) 694baa03890SNick Piggin { 695baa03890SNick Piggin struct inode *inode = parent->d_inode; 696baa03890SNick Piggin struct dentry *dentry; 697baa03890SNick Piggin struct dentry *old; 698baa03890SNick Piggin 699baa03890SNick Piggin /* Don't create child dentry for a dead directory. */ 700baa03890SNick Piggin if (unlikely(IS_DEADDIR(inode))) 701baa03890SNick Piggin return ERR_PTR(-ENOENT); 702baa03890SNick Piggin 703baa03890SNick Piggin dentry = d_alloc(parent, name); 704baa03890SNick Piggin if (unlikely(!dentry)) 705baa03890SNick Piggin return ERR_PTR(-ENOMEM); 706baa03890SNick Piggin 707baa03890SNick Piggin old = inode->i_op->lookup(inode, dentry, nd); 708baa03890SNick Piggin if (unlikely(old)) { 709baa03890SNick Piggin dput(dentry); 710baa03890SNick Piggin dentry = old; 711baa03890SNick Piggin } 712baa03890SNick Piggin return dentry; 713baa03890SNick Piggin } 714baa03890SNick Piggin 715baa03890SNick Piggin /* 7161da177e4SLinus Torvalds * It's more convoluted than I'd like it to be, but... it's still fairly 7171da177e4SLinus Torvalds * small and for now I'd prefer to have fast path as straight as possible. 7181da177e4SLinus Torvalds * It _is_ time-critical. 7191da177e4SLinus Torvalds */ 7201da177e4SLinus Torvalds static int do_lookup(struct nameidata *nd, struct qstr *name, 7211da177e4SLinus Torvalds struct path *path) 7221da177e4SLinus Torvalds { 7234ac91378SJan Blunck struct vfsmount *mnt = nd->path.mnt; 7246e6b1bd1SAl Viro struct dentry *dentry, *parent; 7256e6b1bd1SAl Viro struct inode *dir; 7263cac260aSAl Viro /* 7273cac260aSAl Viro * See if the low-level filesystem might want 7283cac260aSAl Viro * to use its own hash.. 7293cac260aSAl Viro */ 7303cac260aSAl Viro if (nd->path.dentry->d_op && nd->path.dentry->d_op->d_hash) { 731b1e6a015SNick Piggin int err = nd->path.dentry->d_op->d_hash(nd->path.dentry, 732b1e6a015SNick Piggin nd->path.dentry->d_inode, name); 7333cac260aSAl Viro if (err < 0) 7343cac260aSAl Viro return err; 7353cac260aSAl Viro } 7361da177e4SLinus Torvalds 737b04f784eSNick Piggin /* 738b04f784eSNick Piggin * Rename seqlock is not required here because in the off chance 739b04f784eSNick Piggin * of a false negative due to a concurrent rename, we're going to 740b04f784eSNick Piggin * do the non-racy lookup, below. 741b04f784eSNick Piggin */ 7423cac260aSAl Viro dentry = __d_lookup(nd->path.dentry, name); 7431da177e4SLinus Torvalds if (!dentry) 7441da177e4SLinus Torvalds goto need_lookup; 7452e2e88eaSNick Piggin found: 7461da177e4SLinus Torvalds if (dentry->d_op && dentry->d_op->d_revalidate) 7471da177e4SLinus Torvalds goto need_revalidate; 7481da177e4SLinus Torvalds done: 7491da177e4SLinus Torvalds path->mnt = mnt; 7501da177e4SLinus Torvalds path->dentry = dentry; 751634ee701SAl Viro __follow_mount(path); 7521da177e4SLinus Torvalds return 0; 7531da177e4SLinus Torvalds 7541da177e4SLinus Torvalds need_lookup: 7556e6b1bd1SAl Viro parent = nd->path.dentry; 7566e6b1bd1SAl Viro dir = parent->d_inode; 7576e6b1bd1SAl Viro 7586e6b1bd1SAl Viro mutex_lock(&dir->i_mutex); 7596e6b1bd1SAl Viro /* 7606e6b1bd1SAl Viro * First re-do the cached lookup just in case it was created 761b04f784eSNick Piggin * while we waited for the directory semaphore, or the first 762b04f784eSNick Piggin * lookup failed due to an unrelated rename. 7636e6b1bd1SAl Viro * 764b04f784eSNick Piggin * This could use version numbering or similar to avoid unnecessary 765b04f784eSNick Piggin * cache lookups, but then we'd have to do the first lookup in the 766b04f784eSNick Piggin * non-racy way. However in the common case here, everything should 767b04f784eSNick Piggin * be hot in cache, so would it be a big win? 7686e6b1bd1SAl Viro */ 7696e6b1bd1SAl Viro dentry = d_lookup(parent, name); 770baa03890SNick Piggin if (likely(!dentry)) { 771baa03890SNick Piggin dentry = d_alloc_and_lookup(parent, name, nd); 7726e6b1bd1SAl Viro mutex_unlock(&dir->i_mutex); 7736e6b1bd1SAl Viro if (IS_ERR(dentry)) 7746e6b1bd1SAl Viro goto fail; 7756e6b1bd1SAl Viro goto done; 7766e6b1bd1SAl Viro } 7776e6b1bd1SAl Viro /* 7786e6b1bd1SAl Viro * Uhhuh! Nasty case: the cache was re-populated while 7796e6b1bd1SAl Viro * we waited on the semaphore. Need to revalidate. 7806e6b1bd1SAl Viro */ 7816e6b1bd1SAl Viro mutex_unlock(&dir->i_mutex); 7822e2e88eaSNick Piggin goto found; 7831da177e4SLinus Torvalds 7841da177e4SLinus Torvalds need_revalidate: 785bcdc5e01SIan Kent dentry = do_revalidate(dentry, nd); 786bcdc5e01SIan Kent if (!dentry) 7871da177e4SLinus Torvalds goto need_lookup; 788bcdc5e01SIan Kent if (IS_ERR(dentry)) 789bcdc5e01SIan Kent goto fail; 790bcdc5e01SIan Kent goto done; 7911da177e4SLinus Torvalds 7921da177e4SLinus Torvalds fail: 7931da177e4SLinus Torvalds return PTR_ERR(dentry); 7941da177e4SLinus Torvalds } 7951da177e4SLinus Torvalds 7961da177e4SLinus Torvalds /* 797ac278a9cSAl Viro * This is a temporary kludge to deal with "automount" symlinks; proper 798ac278a9cSAl Viro * solution is to trigger them on follow_mount(), so that do_lookup() 799ac278a9cSAl Viro * would DTRT. To be killed before 2.6.34-final. 800ac278a9cSAl Viro */ 801ac278a9cSAl Viro static inline int follow_on_final(struct inode *inode, unsigned lookup_flags) 802ac278a9cSAl Viro { 803ac278a9cSAl Viro return inode && unlikely(inode->i_op->follow_link) && 804ac278a9cSAl Viro ((lookup_flags & LOOKUP_FOLLOW) || S_ISDIR(inode->i_mode)); 805ac278a9cSAl Viro } 806ac278a9cSAl Viro 807ac278a9cSAl Viro /* 8081da177e4SLinus Torvalds * Name resolution. 809ea3834d9SPrasanna Meda * This is the basic name resolution function, turning a pathname into 810ea3834d9SPrasanna Meda * the final dentry. We expect 'base' to be positive and a directory. 8111da177e4SLinus Torvalds * 812ea3834d9SPrasanna Meda * Returns 0 and nd will have valid dentry and mnt on success. 813ea3834d9SPrasanna Meda * Returns error and drops reference to input namei data on failure. 8141da177e4SLinus Torvalds */ 8156de88d72SAl Viro static int link_path_walk(const char *name, struct nameidata *nd) 8161da177e4SLinus Torvalds { 8171da177e4SLinus Torvalds struct path next; 8181da177e4SLinus Torvalds struct inode *inode; 8191da177e4SLinus Torvalds int err; 8201da177e4SLinus Torvalds unsigned int lookup_flags = nd->flags; 8211da177e4SLinus Torvalds 8221da177e4SLinus Torvalds while (*name=='/') 8231da177e4SLinus Torvalds name++; 8241da177e4SLinus Torvalds if (!*name) 8251da177e4SLinus Torvalds goto return_reval; 8261da177e4SLinus Torvalds 8274ac91378SJan Blunck inode = nd->path.dentry->d_inode; 8281da177e4SLinus Torvalds if (nd->depth) 829f55eab82STrond Myklebust lookup_flags = LOOKUP_FOLLOW | (nd->flags & LOOKUP_CONTINUE); 8301da177e4SLinus Torvalds 8311da177e4SLinus Torvalds /* At this point we know we have a real path component. */ 8321da177e4SLinus Torvalds for(;;) { 8331da177e4SLinus Torvalds unsigned long hash; 8341da177e4SLinus Torvalds struct qstr this; 8351da177e4SLinus Torvalds unsigned int c; 8361da177e4SLinus Torvalds 837cdce5d6bSTrond Myklebust nd->flags |= LOOKUP_CONTINUE; 838b75b5086SAl Viro err = exec_permission(inode); 8391da177e4SLinus Torvalds if (err) 8401da177e4SLinus Torvalds break; 8411da177e4SLinus Torvalds 8421da177e4SLinus Torvalds this.name = name; 8431da177e4SLinus Torvalds c = *(const unsigned char *)name; 8441da177e4SLinus Torvalds 8451da177e4SLinus Torvalds hash = init_name_hash(); 8461da177e4SLinus Torvalds do { 8471da177e4SLinus Torvalds name++; 8481da177e4SLinus Torvalds hash = partial_name_hash(c, hash); 8491da177e4SLinus Torvalds c = *(const unsigned char *)name; 8501da177e4SLinus Torvalds } while (c && (c != '/')); 8511da177e4SLinus Torvalds this.len = name - (const char *) this.name; 8521da177e4SLinus Torvalds this.hash = end_name_hash(hash); 8531da177e4SLinus Torvalds 8541da177e4SLinus Torvalds /* remove trailing slashes? */ 8551da177e4SLinus Torvalds if (!c) 8561da177e4SLinus Torvalds goto last_component; 8571da177e4SLinus Torvalds while (*++name == '/'); 8581da177e4SLinus Torvalds if (!*name) 8591da177e4SLinus Torvalds goto last_with_slashes; 8601da177e4SLinus Torvalds 8611da177e4SLinus Torvalds /* 8621da177e4SLinus Torvalds * "." and ".." are special - ".." especially so because it has 8631da177e4SLinus Torvalds * to be able to know about the current root directory and 8641da177e4SLinus Torvalds * parent relationships. 8651da177e4SLinus Torvalds */ 8661da177e4SLinus Torvalds if (this.name[0] == '.') switch (this.len) { 8671da177e4SLinus Torvalds default: 8681da177e4SLinus Torvalds break; 8691da177e4SLinus Torvalds case 2: 8701da177e4SLinus Torvalds if (this.name[1] != '.') 8711da177e4SLinus Torvalds break; 87258c465ebSAl Viro follow_dotdot(nd); 8734ac91378SJan Blunck inode = nd->path.dentry->d_inode; 8741da177e4SLinus Torvalds /* fallthrough */ 8751da177e4SLinus Torvalds case 1: 8761da177e4SLinus Torvalds continue; 8771da177e4SLinus Torvalds } 8781da177e4SLinus Torvalds /* This does the actual lookups.. */ 8791da177e4SLinus Torvalds err = do_lookup(nd, &this, &next); 8801da177e4SLinus Torvalds if (err) 8811da177e4SLinus Torvalds break; 8821da177e4SLinus Torvalds 8831da177e4SLinus Torvalds err = -ENOENT; 8841da177e4SLinus Torvalds inode = next.dentry->d_inode; 8851da177e4SLinus Torvalds if (!inode) 8861da177e4SLinus Torvalds goto out_dput; 8871da177e4SLinus Torvalds 8881da177e4SLinus Torvalds if (inode->i_op->follow_link) { 88990ebe565SAl Viro err = do_follow_link(&next, nd); 8901da177e4SLinus Torvalds if (err) 8911da177e4SLinus Torvalds goto return_err; 8921da177e4SLinus Torvalds err = -ENOENT; 8934ac91378SJan Blunck inode = nd->path.dentry->d_inode; 8941da177e4SLinus Torvalds if (!inode) 8951da177e4SLinus Torvalds break; 89609dd17d3SMiklos Szeredi } else 89709dd17d3SMiklos Szeredi path_to_nameidata(&next, nd); 8981da177e4SLinus Torvalds err = -ENOTDIR; 8991da177e4SLinus Torvalds if (!inode->i_op->lookup) 9001da177e4SLinus Torvalds break; 9011da177e4SLinus Torvalds continue; 9021da177e4SLinus Torvalds /* here ends the main loop */ 9031da177e4SLinus Torvalds 9041da177e4SLinus Torvalds last_with_slashes: 9051da177e4SLinus Torvalds lookup_flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; 9061da177e4SLinus Torvalds last_component: 907f55eab82STrond Myklebust /* Clear LOOKUP_CONTINUE iff it was previously unset */ 908f55eab82STrond Myklebust nd->flags &= lookup_flags | ~LOOKUP_CONTINUE; 9091da177e4SLinus Torvalds if (lookup_flags & LOOKUP_PARENT) 9101da177e4SLinus Torvalds goto lookup_parent; 9111da177e4SLinus Torvalds if (this.name[0] == '.') switch (this.len) { 9121da177e4SLinus Torvalds default: 9131da177e4SLinus Torvalds break; 9141da177e4SLinus Torvalds case 2: 9151da177e4SLinus Torvalds if (this.name[1] != '.') 9161da177e4SLinus Torvalds break; 91758c465ebSAl Viro follow_dotdot(nd); 9184ac91378SJan Blunck inode = nd->path.dentry->d_inode; 9191da177e4SLinus Torvalds /* fallthrough */ 9201da177e4SLinus Torvalds case 1: 9211da177e4SLinus Torvalds goto return_reval; 9221da177e4SLinus Torvalds } 9231da177e4SLinus Torvalds err = do_lookup(nd, &this, &next); 9241da177e4SLinus Torvalds if (err) 9251da177e4SLinus Torvalds break; 9261da177e4SLinus Torvalds inode = next.dentry->d_inode; 927ac278a9cSAl Viro if (follow_on_final(inode, lookup_flags)) { 92890ebe565SAl Viro err = do_follow_link(&next, nd); 9291da177e4SLinus Torvalds if (err) 9301da177e4SLinus Torvalds goto return_err; 9314ac91378SJan Blunck inode = nd->path.dentry->d_inode; 93209dd17d3SMiklos Szeredi } else 93309dd17d3SMiklos Szeredi path_to_nameidata(&next, nd); 9341da177e4SLinus Torvalds err = -ENOENT; 9351da177e4SLinus Torvalds if (!inode) 9361da177e4SLinus Torvalds break; 9371da177e4SLinus Torvalds if (lookup_flags & LOOKUP_DIRECTORY) { 9381da177e4SLinus Torvalds err = -ENOTDIR; 939acfa4380SAl Viro if (!inode->i_op->lookup) 9401da177e4SLinus Torvalds break; 9411da177e4SLinus Torvalds } 9421da177e4SLinus Torvalds goto return_base; 9431da177e4SLinus Torvalds lookup_parent: 9441da177e4SLinus Torvalds nd->last = this; 9451da177e4SLinus Torvalds nd->last_type = LAST_NORM; 9461da177e4SLinus Torvalds if (this.name[0] != '.') 9471da177e4SLinus Torvalds goto return_base; 9481da177e4SLinus Torvalds if (this.len == 1) 9491da177e4SLinus Torvalds nd->last_type = LAST_DOT; 9501da177e4SLinus Torvalds else if (this.len == 2 && this.name[1] == '.') 9511da177e4SLinus Torvalds nd->last_type = LAST_DOTDOT; 9521da177e4SLinus Torvalds else 9531da177e4SLinus Torvalds goto return_base; 9541da177e4SLinus Torvalds return_reval: 9551da177e4SLinus Torvalds /* 9561da177e4SLinus Torvalds * We bypassed the ordinary revalidation routines. 9571da177e4SLinus Torvalds * We may need to check the cached dentry for staleness. 9581da177e4SLinus Torvalds */ 9594ac91378SJan Blunck if (nd->path.dentry && nd->path.dentry->d_sb && 9604ac91378SJan Blunck (nd->path.dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT)) { 9611da177e4SLinus Torvalds err = -ESTALE; 9621da177e4SLinus Torvalds /* Note: we do not d_invalidate() */ 9634ac91378SJan Blunck if (!nd->path.dentry->d_op->d_revalidate( 9644ac91378SJan Blunck nd->path.dentry, nd)) 9651da177e4SLinus Torvalds break; 9661da177e4SLinus Torvalds } 9671da177e4SLinus Torvalds return_base: 9681da177e4SLinus Torvalds return 0; 9691da177e4SLinus Torvalds out_dput: 9701d957f9bSJan Blunck path_put_conditional(&next, nd); 9711da177e4SLinus Torvalds break; 9721da177e4SLinus Torvalds } 9731d957f9bSJan Blunck path_put(&nd->path); 9741da177e4SLinus Torvalds return_err: 9751da177e4SLinus Torvalds return err; 9761da177e4SLinus Torvalds } 9771da177e4SLinus Torvalds 978fc9b52cdSHarvey Harrison static int path_walk(const char *name, struct nameidata *nd) 9791da177e4SLinus Torvalds { 9806de88d72SAl Viro struct path save = nd->path; 9816de88d72SAl Viro int result; 9826de88d72SAl Viro 9831da177e4SLinus Torvalds current->total_link_count = 0; 9846de88d72SAl Viro 9856de88d72SAl Viro /* make sure the stuff we saved doesn't go away */ 9866de88d72SAl Viro path_get(&save); 9876de88d72SAl Viro 9886de88d72SAl Viro result = link_path_walk(name, nd); 9896de88d72SAl Viro if (result == -ESTALE) { 9906de88d72SAl Viro /* nd->path had been dropped */ 9916de88d72SAl Viro current->total_link_count = 0; 9926de88d72SAl Viro nd->path = save; 9936de88d72SAl Viro path_get(&nd->path); 9946de88d72SAl Viro nd->flags |= LOOKUP_REVAL; 9956de88d72SAl Viro result = link_path_walk(name, nd); 9966de88d72SAl Viro } 9976de88d72SAl Viro 9986de88d72SAl Viro path_put(&save); 9996de88d72SAl Viro 10006de88d72SAl Viro return result; 10011da177e4SLinus Torvalds } 10021da177e4SLinus Torvalds 10039b4a9b14SAl Viro static int path_init(int dfd, const char *name, unsigned int flags, struct nameidata *nd) 10041da177e4SLinus Torvalds { 1005ea3834d9SPrasanna Meda int retval = 0; 1006170aa3d0SUlrich Drepper int fput_needed; 1007170aa3d0SUlrich Drepper struct file *file; 10081da177e4SLinus Torvalds 10091da177e4SLinus Torvalds nd->last_type = LAST_ROOT; /* if there are only slashes... */ 10101da177e4SLinus Torvalds nd->flags = flags; 10111da177e4SLinus Torvalds nd->depth = 0; 10122a737871SAl Viro nd->root.mnt = NULL; 10131da177e4SLinus Torvalds 10141da177e4SLinus Torvalds if (*name=='/') { 10152a737871SAl Viro set_root(nd); 10162a737871SAl Viro nd->path = nd->root; 10172a737871SAl Viro path_get(&nd->root); 10185590ff0dSUlrich Drepper } else if (dfd == AT_FDCWD) { 1019f7ad3c6bSMiklos Szeredi get_fs_pwd(current->fs, &nd->path); 10205590ff0dSUlrich Drepper } else { 10215590ff0dSUlrich Drepper struct dentry *dentry; 10225590ff0dSUlrich Drepper 10235590ff0dSUlrich Drepper file = fget_light(dfd, &fput_needed); 10245590ff0dSUlrich Drepper retval = -EBADF; 1025170aa3d0SUlrich Drepper if (!file) 10266d09bb62STrond Myklebust goto out_fail; 10275590ff0dSUlrich Drepper 10280f7fc9e4SJosef "Jeff" Sipek dentry = file->f_path.dentry; 10295590ff0dSUlrich Drepper 10305590ff0dSUlrich Drepper retval = -ENOTDIR; 1031170aa3d0SUlrich Drepper if (!S_ISDIR(dentry->d_inode->i_mode)) 10326d09bb62STrond Myklebust goto fput_fail; 10335590ff0dSUlrich Drepper 10345590ff0dSUlrich Drepper retval = file_permission(file, MAY_EXEC); 1035170aa3d0SUlrich Drepper if (retval) 10366d09bb62STrond Myklebust goto fput_fail; 10375590ff0dSUlrich Drepper 10385dd784d0SJan Blunck nd->path = file->f_path; 10395dd784d0SJan Blunck path_get(&file->f_path); 10405590ff0dSUlrich Drepper 10415590ff0dSUlrich Drepper fput_light(file, fput_needed); 10421da177e4SLinus Torvalds } 10439b4a9b14SAl Viro return 0; 10442dfdd266SJosef 'Jeff' Sipek 10459b4a9b14SAl Viro fput_fail: 10469b4a9b14SAl Viro fput_light(file, fput_needed); 10479b4a9b14SAl Viro out_fail: 10489b4a9b14SAl Viro return retval; 10499b4a9b14SAl Viro } 10509b4a9b14SAl Viro 10519b4a9b14SAl Viro /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ 10529b4a9b14SAl Viro static int do_path_lookup(int dfd, const char *name, 10539b4a9b14SAl Viro unsigned int flags, struct nameidata *nd) 10549b4a9b14SAl Viro { 10559b4a9b14SAl Viro int retval = path_init(dfd, name, flags, nd); 10569b4a9b14SAl Viro if (!retval) 10572dfdd266SJosef 'Jeff' Sipek retval = path_walk(name, nd); 10584ac91378SJan Blunck if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && 10594ac91378SJan Blunck nd->path.dentry->d_inode)) 10604ac91378SJan Blunck audit_inode(name, nd->path.dentry); 10612a737871SAl Viro if (nd->root.mnt) { 10622a737871SAl Viro path_put(&nd->root); 10632a737871SAl Viro nd->root.mnt = NULL; 10642a737871SAl Viro } 1065170aa3d0SUlrich Drepper return retval; 10661da177e4SLinus Torvalds } 10671da177e4SLinus Torvalds 1068fc9b52cdSHarvey Harrison int path_lookup(const char *name, unsigned int flags, 10695590ff0dSUlrich Drepper struct nameidata *nd) 10705590ff0dSUlrich Drepper { 10715590ff0dSUlrich Drepper return do_path_lookup(AT_FDCWD, name, flags, nd); 10725590ff0dSUlrich Drepper } 10735590ff0dSUlrich Drepper 1074d1811465SAl Viro int kern_path(const char *name, unsigned int flags, struct path *path) 1075d1811465SAl Viro { 1076d1811465SAl Viro struct nameidata nd; 1077d1811465SAl Viro int res = do_path_lookup(AT_FDCWD, name, flags, &nd); 1078d1811465SAl Viro if (!res) 1079d1811465SAl Viro *path = nd.path; 1080d1811465SAl Viro return res; 1081d1811465SAl Viro } 1082d1811465SAl Viro 108316f18200SJosef 'Jeff' Sipek /** 108416f18200SJosef 'Jeff' Sipek * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair 108516f18200SJosef 'Jeff' Sipek * @dentry: pointer to dentry of the base directory 108616f18200SJosef 'Jeff' Sipek * @mnt: pointer to vfs mount of the base directory 108716f18200SJosef 'Jeff' Sipek * @name: pointer to file name 108816f18200SJosef 'Jeff' Sipek * @flags: lookup flags 108916f18200SJosef 'Jeff' Sipek * @nd: pointer to nameidata 109016f18200SJosef 'Jeff' Sipek */ 109116f18200SJosef 'Jeff' Sipek int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, 109216f18200SJosef 'Jeff' Sipek const char *name, unsigned int flags, 109316f18200SJosef 'Jeff' Sipek struct nameidata *nd) 109416f18200SJosef 'Jeff' Sipek { 109516f18200SJosef 'Jeff' Sipek int retval; 109616f18200SJosef 'Jeff' Sipek 109716f18200SJosef 'Jeff' Sipek /* same as do_path_lookup */ 109816f18200SJosef 'Jeff' Sipek nd->last_type = LAST_ROOT; 109916f18200SJosef 'Jeff' Sipek nd->flags = flags; 110016f18200SJosef 'Jeff' Sipek nd->depth = 0; 110116f18200SJosef 'Jeff' Sipek 1102c8e7f449SJan Blunck nd->path.dentry = dentry; 1103c8e7f449SJan Blunck nd->path.mnt = mnt; 1104c8e7f449SJan Blunck path_get(&nd->path); 11055b857119SAl Viro nd->root = nd->path; 11065b857119SAl Viro path_get(&nd->root); 110716f18200SJosef 'Jeff' Sipek 110816f18200SJosef 'Jeff' Sipek retval = path_walk(name, nd); 11094ac91378SJan Blunck if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && 11104ac91378SJan Blunck nd->path.dentry->d_inode)) 11114ac91378SJan Blunck audit_inode(name, nd->path.dentry); 111216f18200SJosef 'Jeff' Sipek 11132a737871SAl Viro path_put(&nd->root); 11142a737871SAl Viro nd->root.mnt = NULL; 111516f18200SJosef 'Jeff' Sipek 11162a737871SAl Viro return retval; 111716f18200SJosef 'Jeff' Sipek } 111816f18200SJosef 'Jeff' Sipek 1119eead1911SChristoph Hellwig static struct dentry *__lookup_hash(struct qstr *name, 1120eead1911SChristoph Hellwig struct dentry *base, struct nameidata *nd) 11211da177e4SLinus Torvalds { 112281fca444SChristoph Hellwig struct inode *inode = base->d_inode; 11231da177e4SLinus Torvalds struct dentry *dentry; 11241da177e4SLinus Torvalds int err; 11251da177e4SLinus Torvalds 112681fca444SChristoph Hellwig err = exec_permission(inode); 112781fca444SChristoph Hellwig if (err) 112881fca444SChristoph Hellwig return ERR_PTR(err); 11291da177e4SLinus Torvalds 11301da177e4SLinus Torvalds /* 11311da177e4SLinus Torvalds * See if the low-level filesystem might want 11321da177e4SLinus Torvalds * to use its own hash.. 11331da177e4SLinus Torvalds */ 11341da177e4SLinus Torvalds if (base->d_op && base->d_op->d_hash) { 1135b1e6a015SNick Piggin err = base->d_op->d_hash(base, inode, name); 11361da177e4SLinus Torvalds dentry = ERR_PTR(err); 11371da177e4SLinus Torvalds if (err < 0) 11381da177e4SLinus Torvalds goto out; 11391da177e4SLinus Torvalds } 11401da177e4SLinus Torvalds 1141b04f784eSNick Piggin /* 1142b04f784eSNick Piggin * Don't bother with __d_lookup: callers are for creat as 1143b04f784eSNick Piggin * well as unlink, so a lot of the time it would cost 1144b04f784eSNick Piggin * a double lookup. 11456e6b1bd1SAl Viro */ 11466e6b1bd1SAl Viro dentry = d_lookup(base, name); 11476e6b1bd1SAl Viro 11486e6b1bd1SAl Viro if (dentry && dentry->d_op && dentry->d_op->d_revalidate) 11496e6b1bd1SAl Viro dentry = do_revalidate(dentry, nd); 11506e6b1bd1SAl Viro 11511da177e4SLinus Torvalds if (!dentry) 1152baa03890SNick Piggin dentry = d_alloc_and_lookup(base, name, nd); 11531da177e4SLinus Torvalds out: 11541da177e4SLinus Torvalds return dentry; 11551da177e4SLinus Torvalds } 11561da177e4SLinus Torvalds 1157057f6c01SJames Morris /* 1158057f6c01SJames Morris * Restricted form of lookup. Doesn't follow links, single-component only, 1159057f6c01SJames Morris * needs parent already locked. Doesn't follow mounts. 1160057f6c01SJames Morris * SMP-safe. 1161057f6c01SJames Morris */ 1162a244e169SAdrian Bunk static struct dentry *lookup_hash(struct nameidata *nd) 11631da177e4SLinus Torvalds { 11644ac91378SJan Blunck return __lookup_hash(&nd->last, nd->path.dentry, nd); 11651da177e4SLinus Torvalds } 11661da177e4SLinus Torvalds 1167eead1911SChristoph Hellwig static int __lookup_one_len(const char *name, struct qstr *this, 1168eead1911SChristoph Hellwig struct dentry *base, int len) 11691da177e4SLinus Torvalds { 11701da177e4SLinus Torvalds unsigned long hash; 11711da177e4SLinus Torvalds unsigned int c; 11721da177e4SLinus Torvalds 1173057f6c01SJames Morris this->name = name; 1174057f6c01SJames Morris this->len = len; 11751da177e4SLinus Torvalds if (!len) 1176057f6c01SJames Morris return -EACCES; 11771da177e4SLinus Torvalds 11781da177e4SLinus Torvalds hash = init_name_hash(); 11791da177e4SLinus Torvalds while (len--) { 11801da177e4SLinus Torvalds c = *(const unsigned char *)name++; 11811da177e4SLinus Torvalds if (c == '/' || c == '\0') 1182057f6c01SJames Morris return -EACCES; 11831da177e4SLinus Torvalds hash = partial_name_hash(c, hash); 11841da177e4SLinus Torvalds } 1185057f6c01SJames Morris this->hash = end_name_hash(hash); 1186057f6c01SJames Morris return 0; 1187057f6c01SJames Morris } 11881da177e4SLinus Torvalds 1189eead1911SChristoph Hellwig /** 1190a6b91919SRandy Dunlap * lookup_one_len - filesystem helper to lookup single pathname component 1191eead1911SChristoph Hellwig * @name: pathname component to lookup 1192eead1911SChristoph Hellwig * @base: base directory to lookup from 1193eead1911SChristoph Hellwig * @len: maximum length @len should be interpreted to 1194eead1911SChristoph Hellwig * 1195a6b91919SRandy Dunlap * Note that this routine is purely a helper for filesystem usage and should 1196a6b91919SRandy Dunlap * not be called by generic code. Also note that by using this function the 1197eead1911SChristoph Hellwig * nameidata argument is passed to the filesystem methods and a filesystem 1198eead1911SChristoph Hellwig * using this helper needs to be prepared for that. 1199eead1911SChristoph Hellwig */ 1200057f6c01SJames Morris struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) 1201057f6c01SJames Morris { 1202057f6c01SJames Morris int err; 1203057f6c01SJames Morris struct qstr this; 1204057f6c01SJames Morris 12052f9092e1SDavid Woodhouse WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex)); 12062f9092e1SDavid Woodhouse 1207057f6c01SJames Morris err = __lookup_one_len(name, &this, base, len); 1208057f6c01SJames Morris if (err) 1209057f6c01SJames Morris return ERR_PTR(err); 1210eead1911SChristoph Hellwig 121149705b77SChristoph Hellwig return __lookup_hash(&this, base, NULL); 1212057f6c01SJames Morris } 1213057f6c01SJames Morris 12142d8f3038SAl Viro int user_path_at(int dfd, const char __user *name, unsigned flags, 12152d8f3038SAl Viro struct path *path) 12161da177e4SLinus Torvalds { 12172d8f3038SAl Viro struct nameidata nd; 12181da177e4SLinus Torvalds char *tmp = getname(name); 12191da177e4SLinus Torvalds int err = PTR_ERR(tmp); 12201da177e4SLinus Torvalds if (!IS_ERR(tmp)) { 12212d8f3038SAl Viro 12222d8f3038SAl Viro BUG_ON(flags & LOOKUP_PARENT); 12232d8f3038SAl Viro 12242d8f3038SAl Viro err = do_path_lookup(dfd, tmp, flags, &nd); 12251da177e4SLinus Torvalds putname(tmp); 12262d8f3038SAl Viro if (!err) 12272d8f3038SAl Viro *path = nd.path; 12281da177e4SLinus Torvalds } 12291da177e4SLinus Torvalds return err; 12301da177e4SLinus Torvalds } 12311da177e4SLinus Torvalds 12322ad94ae6SAl Viro static int user_path_parent(int dfd, const char __user *path, 12332ad94ae6SAl Viro struct nameidata *nd, char **name) 12342ad94ae6SAl Viro { 12352ad94ae6SAl Viro char *s = getname(path); 12362ad94ae6SAl Viro int error; 12372ad94ae6SAl Viro 12382ad94ae6SAl Viro if (IS_ERR(s)) 12392ad94ae6SAl Viro return PTR_ERR(s); 12402ad94ae6SAl Viro 12412ad94ae6SAl Viro error = do_path_lookup(dfd, s, LOOKUP_PARENT, nd); 12422ad94ae6SAl Viro if (error) 12432ad94ae6SAl Viro putname(s); 12442ad94ae6SAl Viro else 12452ad94ae6SAl Viro *name = s; 12462ad94ae6SAl Viro 12472ad94ae6SAl Viro return error; 12482ad94ae6SAl Viro } 12492ad94ae6SAl Viro 12501da177e4SLinus Torvalds /* 12511da177e4SLinus Torvalds * It's inline, so penalty for filesystems that don't use sticky bit is 12521da177e4SLinus Torvalds * minimal. 12531da177e4SLinus Torvalds */ 12541da177e4SLinus Torvalds static inline int check_sticky(struct inode *dir, struct inode *inode) 12551da177e4SLinus Torvalds { 1256da9592edSDavid Howells uid_t fsuid = current_fsuid(); 1257da9592edSDavid Howells 12581da177e4SLinus Torvalds if (!(dir->i_mode & S_ISVTX)) 12591da177e4SLinus Torvalds return 0; 1260da9592edSDavid Howells if (inode->i_uid == fsuid) 12611da177e4SLinus Torvalds return 0; 1262da9592edSDavid Howells if (dir->i_uid == fsuid) 12631da177e4SLinus Torvalds return 0; 12641da177e4SLinus Torvalds return !capable(CAP_FOWNER); 12651da177e4SLinus Torvalds } 12661da177e4SLinus Torvalds 12671da177e4SLinus Torvalds /* 12681da177e4SLinus Torvalds * Check whether we can remove a link victim from directory dir, check 12691da177e4SLinus Torvalds * whether the type of victim is right. 12701da177e4SLinus Torvalds * 1. We can't do it if dir is read-only (done in permission()) 12711da177e4SLinus Torvalds * 2. We should have write and exec permissions on dir 12721da177e4SLinus Torvalds * 3. We can't remove anything from append-only dir 12731da177e4SLinus Torvalds * 4. We can't do anything with immutable dir (done in permission()) 12741da177e4SLinus Torvalds * 5. If the sticky bit on dir is set we should either 12751da177e4SLinus Torvalds * a. be owner of dir, or 12761da177e4SLinus Torvalds * b. be owner of victim, or 12771da177e4SLinus Torvalds * c. have CAP_FOWNER capability 12781da177e4SLinus Torvalds * 6. If the victim is append-only or immutable we can't do antyhing with 12791da177e4SLinus Torvalds * links pointing to it. 12801da177e4SLinus Torvalds * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 12811da177e4SLinus Torvalds * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 12821da177e4SLinus Torvalds * 9. We can't remove a root or mountpoint. 12831da177e4SLinus Torvalds * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 12841da177e4SLinus Torvalds * nfs_async_unlink(). 12851da177e4SLinus Torvalds */ 1286858119e1SArjan van de Ven static int may_delete(struct inode *dir,struct dentry *victim,int isdir) 12871da177e4SLinus Torvalds { 12881da177e4SLinus Torvalds int error; 12891da177e4SLinus Torvalds 12901da177e4SLinus Torvalds if (!victim->d_inode) 12911da177e4SLinus Torvalds return -ENOENT; 12921da177e4SLinus Torvalds 12931da177e4SLinus Torvalds BUG_ON(victim->d_parent->d_inode != dir); 1294cccc6bbaSAl Viro audit_inode_child(victim, dir); 12951da177e4SLinus Torvalds 1296f419a2e3SAl Viro error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 12971da177e4SLinus Torvalds if (error) 12981da177e4SLinus Torvalds return error; 12991da177e4SLinus Torvalds if (IS_APPEND(dir)) 13001da177e4SLinus Torvalds return -EPERM; 13011da177e4SLinus Torvalds if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| 1302f9454548SHugh Dickins IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) 13031da177e4SLinus Torvalds return -EPERM; 13041da177e4SLinus Torvalds if (isdir) { 13051da177e4SLinus Torvalds if (!S_ISDIR(victim->d_inode->i_mode)) 13061da177e4SLinus Torvalds return -ENOTDIR; 13071da177e4SLinus Torvalds if (IS_ROOT(victim)) 13081da177e4SLinus Torvalds return -EBUSY; 13091da177e4SLinus Torvalds } else if (S_ISDIR(victim->d_inode->i_mode)) 13101da177e4SLinus Torvalds return -EISDIR; 13111da177e4SLinus Torvalds if (IS_DEADDIR(dir)) 13121da177e4SLinus Torvalds return -ENOENT; 13131da177e4SLinus Torvalds if (victim->d_flags & DCACHE_NFSFS_RENAMED) 13141da177e4SLinus Torvalds return -EBUSY; 13151da177e4SLinus Torvalds return 0; 13161da177e4SLinus Torvalds } 13171da177e4SLinus Torvalds 13181da177e4SLinus Torvalds /* Check whether we can create an object with dentry child in directory 13191da177e4SLinus Torvalds * dir. 13201da177e4SLinus Torvalds * 1. We can't do it if child already exists (open has special treatment for 13211da177e4SLinus Torvalds * this case, but since we are inlined it's OK) 13221da177e4SLinus Torvalds * 2. We can't do it if dir is read-only (done in permission()) 13231da177e4SLinus Torvalds * 3. We should have write and exec permissions on dir 13241da177e4SLinus Torvalds * 4. We can't do it if dir is immutable (done in permission()) 13251da177e4SLinus Torvalds */ 1326a95164d9SMiklos Szeredi static inline int may_create(struct inode *dir, struct dentry *child) 13271da177e4SLinus Torvalds { 13281da177e4SLinus Torvalds if (child->d_inode) 13291da177e4SLinus Torvalds return -EEXIST; 13301da177e4SLinus Torvalds if (IS_DEADDIR(dir)) 13311da177e4SLinus Torvalds return -ENOENT; 1332f419a2e3SAl Viro return inode_permission(dir, MAY_WRITE | MAY_EXEC); 13331da177e4SLinus Torvalds } 13341da177e4SLinus Torvalds 13351da177e4SLinus Torvalds /* 13361da177e4SLinus Torvalds * p1 and p2 should be directories on the same fs. 13371da177e4SLinus Torvalds */ 13381da177e4SLinus Torvalds struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) 13391da177e4SLinus Torvalds { 13401da177e4SLinus Torvalds struct dentry *p; 13411da177e4SLinus Torvalds 13421da177e4SLinus Torvalds if (p1 == p2) { 1343f2eace23SIngo Molnar mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); 13441da177e4SLinus Torvalds return NULL; 13451da177e4SLinus Torvalds } 13461da177e4SLinus Torvalds 1347a11f3a05SArjan van de Ven mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); 13481da177e4SLinus Torvalds 1349e2761a11SOGAWA Hirofumi p = d_ancestor(p2, p1); 1350e2761a11SOGAWA Hirofumi if (p) { 1351f2eace23SIngo Molnar mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); 1352f2eace23SIngo Molnar mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); 13531da177e4SLinus Torvalds return p; 13541da177e4SLinus Torvalds } 13551da177e4SLinus Torvalds 1356e2761a11SOGAWA Hirofumi p = d_ancestor(p1, p2); 1357e2761a11SOGAWA Hirofumi if (p) { 1358f2eace23SIngo Molnar mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); 1359f2eace23SIngo Molnar mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); 13601da177e4SLinus Torvalds return p; 13611da177e4SLinus Torvalds } 13621da177e4SLinus Torvalds 1363f2eace23SIngo Molnar mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); 1364f2eace23SIngo Molnar mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); 13651da177e4SLinus Torvalds return NULL; 13661da177e4SLinus Torvalds } 13671da177e4SLinus Torvalds 13681da177e4SLinus Torvalds void unlock_rename(struct dentry *p1, struct dentry *p2) 13691da177e4SLinus Torvalds { 13701b1dcc1bSJes Sorensen mutex_unlock(&p1->d_inode->i_mutex); 13711da177e4SLinus Torvalds if (p1 != p2) { 13721b1dcc1bSJes Sorensen mutex_unlock(&p2->d_inode->i_mutex); 1373a11f3a05SArjan van de Ven mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); 13741da177e4SLinus Torvalds } 13751da177e4SLinus Torvalds } 13761da177e4SLinus Torvalds 13771da177e4SLinus Torvalds int vfs_create(struct inode *dir, struct dentry *dentry, int mode, 13781da177e4SLinus Torvalds struct nameidata *nd) 13791da177e4SLinus Torvalds { 1380a95164d9SMiklos Szeredi int error = may_create(dir, dentry); 13811da177e4SLinus Torvalds 13821da177e4SLinus Torvalds if (error) 13831da177e4SLinus Torvalds return error; 13841da177e4SLinus Torvalds 1385acfa4380SAl Viro if (!dir->i_op->create) 13861da177e4SLinus Torvalds return -EACCES; /* shouldn't it be ENOSYS? */ 13871da177e4SLinus Torvalds mode &= S_IALLUGO; 13881da177e4SLinus Torvalds mode |= S_IFREG; 13891da177e4SLinus Torvalds error = security_inode_create(dir, dentry, mode); 13901da177e4SLinus Torvalds if (error) 13911da177e4SLinus Torvalds return error; 13921da177e4SLinus Torvalds error = dir->i_op->create(dir, dentry, mode, nd); 1393a74574aaSStephen Smalley if (!error) 1394f38aa942SAmy Griffis fsnotify_create(dir, dentry); 13951da177e4SLinus Torvalds return error; 13961da177e4SLinus Torvalds } 13971da177e4SLinus Torvalds 13983fb64190SChristoph Hellwig int may_open(struct path *path, int acc_mode, int flag) 13991da177e4SLinus Torvalds { 14003fb64190SChristoph Hellwig struct dentry *dentry = path->dentry; 14011da177e4SLinus Torvalds struct inode *inode = dentry->d_inode; 14021da177e4SLinus Torvalds int error; 14031da177e4SLinus Torvalds 14041da177e4SLinus Torvalds if (!inode) 14051da177e4SLinus Torvalds return -ENOENT; 14061da177e4SLinus Torvalds 1407c8fe8f30SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 1408c8fe8f30SChristoph Hellwig case S_IFLNK: 14091da177e4SLinus Torvalds return -ELOOP; 1410c8fe8f30SChristoph Hellwig case S_IFDIR: 1411c8fe8f30SChristoph Hellwig if (acc_mode & MAY_WRITE) 14121da177e4SLinus Torvalds return -EISDIR; 1413c8fe8f30SChristoph Hellwig break; 1414c8fe8f30SChristoph Hellwig case S_IFBLK: 1415c8fe8f30SChristoph Hellwig case S_IFCHR: 14163fb64190SChristoph Hellwig if (path->mnt->mnt_flags & MNT_NODEV) 14171da177e4SLinus Torvalds return -EACCES; 1418c8fe8f30SChristoph Hellwig /*FALLTHRU*/ 1419c8fe8f30SChristoph Hellwig case S_IFIFO: 1420c8fe8f30SChristoph Hellwig case S_IFSOCK: 14211da177e4SLinus Torvalds flag &= ~O_TRUNC; 1422c8fe8f30SChristoph Hellwig break; 14234a3fd211SDave Hansen } 1424b41572e9SDave Hansen 14253fb64190SChristoph Hellwig error = inode_permission(inode, acc_mode); 1426b41572e9SDave Hansen if (error) 1427b41572e9SDave Hansen return error; 14286146f0d5SMimi Zohar 14291da177e4SLinus Torvalds /* 14301da177e4SLinus Torvalds * An append-only file must be opened in append mode for writing. 14311da177e4SLinus Torvalds */ 14321da177e4SLinus Torvalds if (IS_APPEND(inode)) { 14338737c930SAl Viro if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) 14347715b521SAl Viro return -EPERM; 14351da177e4SLinus Torvalds if (flag & O_TRUNC) 14367715b521SAl Viro return -EPERM; 14371da177e4SLinus Torvalds } 14381da177e4SLinus Torvalds 14391da177e4SLinus Torvalds /* O_NOATIME can only be set by the owner or superuser */ 14407715b521SAl Viro if (flag & O_NOATIME && !is_owner_or_cap(inode)) 14417715b521SAl Viro return -EPERM; 14421da177e4SLinus Torvalds 14431da177e4SLinus Torvalds /* 14441da177e4SLinus Torvalds * Ensure there are no outstanding leases on the file. 14451da177e4SLinus Torvalds */ 1446b65a9cfcSAl Viro return break_lease(inode, flag); 14477715b521SAl Viro } 14487715b521SAl Viro 14497715b521SAl Viro static int handle_truncate(struct path *path) 14507715b521SAl Viro { 14517715b521SAl Viro struct inode *inode = path->dentry->d_inode; 14527715b521SAl Viro int error = get_write_access(inode); 14531da177e4SLinus Torvalds if (error) 14547715b521SAl Viro return error; 14551da177e4SLinus Torvalds /* 14561da177e4SLinus Torvalds * Refuse to truncate files with mandatory locks held on them. 14571da177e4SLinus Torvalds */ 14581da177e4SLinus Torvalds error = locks_verify_locked(inode); 1459be6d3e56SKentaro Takeda if (!error) 1460ea0d3ab2STetsuo Handa error = security_path_truncate(path); 14611da177e4SLinus Torvalds if (!error) { 14627715b521SAl Viro error = do_truncate(path->dentry, 0, 1463d139d7ffSMiklos Szeredi ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, 1464d139d7ffSMiklos Szeredi NULL); 14651da177e4SLinus Torvalds } 14661da177e4SLinus Torvalds put_write_access(inode); 1467acd0c935SMimi Zohar return error; 14681da177e4SLinus Torvalds } 14691da177e4SLinus Torvalds 1470d57999e1SDave Hansen /* 1471d57999e1SDave Hansen * Be careful about ever adding any more callers of this 1472d57999e1SDave Hansen * function. Its flags must be in the namei format, not 1473d57999e1SDave Hansen * what get passed to sys_open(). 1474d57999e1SDave Hansen */ 1475d57999e1SDave Hansen static int __open_namei_create(struct nameidata *nd, struct path *path, 14768737c930SAl Viro int open_flag, int mode) 1477aab520e2SDave Hansen { 1478aab520e2SDave Hansen int error; 14794ac91378SJan Blunck struct dentry *dir = nd->path.dentry; 1480aab520e2SDave Hansen 1481aab520e2SDave Hansen if (!IS_POSIXACL(dir->d_inode)) 1482ce3b0f8dSAl Viro mode &= ~current_umask(); 1483be6d3e56SKentaro Takeda error = security_path_mknod(&nd->path, path->dentry, mode, 0); 1484be6d3e56SKentaro Takeda if (error) 1485be6d3e56SKentaro Takeda goto out_unlock; 1486aab520e2SDave Hansen error = vfs_create(dir->d_inode, path->dentry, mode, nd); 1487be6d3e56SKentaro Takeda out_unlock: 1488aab520e2SDave Hansen mutex_unlock(&dir->d_inode->i_mutex); 14894ac91378SJan Blunck dput(nd->path.dentry); 14904ac91378SJan Blunck nd->path.dentry = path->dentry; 1491aab520e2SDave Hansen if (error) 1492aab520e2SDave Hansen return error; 1493aab520e2SDave Hansen /* Don't check for write permission, don't truncate */ 14948737c930SAl Viro return may_open(&nd->path, 0, open_flag & ~O_TRUNC); 1495aab520e2SDave Hansen } 1496aab520e2SDave Hansen 14971da177e4SLinus Torvalds /* 1498d57999e1SDave Hansen * Note that while the flag value (low two bits) for sys_open means: 1499d57999e1SDave Hansen * 00 - read-only 1500d57999e1SDave Hansen * 01 - write-only 1501d57999e1SDave Hansen * 10 - read-write 1502d57999e1SDave Hansen * 11 - special 1503d57999e1SDave Hansen * it is changed into 1504d57999e1SDave Hansen * 00 - no permissions needed 1505d57999e1SDave Hansen * 01 - read-permission 1506d57999e1SDave Hansen * 10 - write-permission 1507d57999e1SDave Hansen * 11 - read-write 1508d57999e1SDave Hansen * for the internal routines (ie open_namei()/follow_link() etc) 1509d57999e1SDave Hansen * This is more logical, and also allows the 00 "no perm needed" 1510d57999e1SDave Hansen * to be used for symlinks (where the permissions are checked 1511d57999e1SDave Hansen * later). 1512d57999e1SDave Hansen * 1513d57999e1SDave Hansen */ 1514d57999e1SDave Hansen static inline int open_to_namei_flags(int flag) 1515d57999e1SDave Hansen { 1516d57999e1SDave Hansen if ((flag+1) & O_ACCMODE) 1517d57999e1SDave Hansen flag++; 1518d57999e1SDave Hansen return flag; 1519d57999e1SDave Hansen } 1520d57999e1SDave Hansen 15217715b521SAl Viro static int open_will_truncate(int flag, struct inode *inode) 15224a3fd211SDave Hansen { 1523d57999e1SDave Hansen /* 15244a3fd211SDave Hansen * We'll never write to the fs underlying 15254a3fd211SDave Hansen * a device file. 15264a3fd211SDave Hansen */ 15274a3fd211SDave Hansen if (special_file(inode->i_mode)) 15284a3fd211SDave Hansen return 0; 15294a3fd211SDave Hansen return (flag & O_TRUNC); 15304a3fd211SDave Hansen } 15314a3fd211SDave Hansen 1532648fa861SAl Viro static struct file *finish_open(struct nameidata *nd, 15339a66179eSAl Viro int open_flag, int acc_mode) 1534648fa861SAl Viro { 1535648fa861SAl Viro struct file *filp; 1536648fa861SAl Viro int will_truncate; 1537648fa861SAl Viro int error; 1538648fa861SAl Viro 15399a66179eSAl Viro will_truncate = open_will_truncate(open_flag, nd->path.dentry->d_inode); 1540648fa861SAl Viro if (will_truncate) { 1541648fa861SAl Viro error = mnt_want_write(nd->path.mnt); 1542648fa861SAl Viro if (error) 1543648fa861SAl Viro goto exit; 1544648fa861SAl Viro } 1545648fa861SAl Viro error = may_open(&nd->path, acc_mode, open_flag); 1546648fa861SAl Viro if (error) { 1547648fa861SAl Viro if (will_truncate) 1548648fa861SAl Viro mnt_drop_write(nd->path.mnt); 1549648fa861SAl Viro goto exit; 1550648fa861SAl Viro } 1551648fa861SAl Viro filp = nameidata_to_filp(nd); 1552648fa861SAl Viro if (!IS_ERR(filp)) { 1553648fa861SAl Viro error = ima_file_check(filp, acc_mode); 1554648fa861SAl Viro if (error) { 1555648fa861SAl Viro fput(filp); 1556648fa861SAl Viro filp = ERR_PTR(error); 1557648fa861SAl Viro } 1558648fa861SAl Viro } 1559648fa861SAl Viro if (!IS_ERR(filp)) { 1560648fa861SAl Viro if (will_truncate) { 1561648fa861SAl Viro error = handle_truncate(&nd->path); 1562648fa861SAl Viro if (error) { 1563648fa861SAl Viro fput(filp); 1564648fa861SAl Viro filp = ERR_PTR(error); 1565648fa861SAl Viro } 1566648fa861SAl Viro } 1567648fa861SAl Viro } 1568648fa861SAl Viro /* 1569648fa861SAl Viro * It is now safe to drop the mnt write 1570648fa861SAl Viro * because the filp has had a write taken 1571648fa861SAl Viro * on its behalf. 1572648fa861SAl Viro */ 1573648fa861SAl Viro if (will_truncate) 1574648fa861SAl Viro mnt_drop_write(nd->path.mnt); 1575d893f1bcSAl Viro path_put(&nd->path); 1576648fa861SAl Viro return filp; 1577648fa861SAl Viro 1578648fa861SAl Viro exit: 1579648fa861SAl Viro if (!IS_ERR(nd->intent.open.file)) 1580648fa861SAl Viro release_open_intent(nd); 1581648fa861SAl Viro path_put(&nd->path); 1582648fa861SAl Viro return ERR_PTR(error); 1583648fa861SAl Viro } 1584648fa861SAl Viro 1585fb1cc555SAl Viro static struct file *do_last(struct nameidata *nd, struct path *path, 15865b369df8SAl Viro int open_flag, int acc_mode, 15873e297b61SAl Viro int mode, const char *pathname) 1588fb1cc555SAl Viro { 1589a1e28038SAl Viro struct dentry *dir = nd->path.dentry; 1590fb1cc555SAl Viro struct file *filp; 15911f36f774SAl Viro int error = -EISDIR; 1592fb1cc555SAl Viro 15931f36f774SAl Viro switch (nd->last_type) { 15941f36f774SAl Viro case LAST_DOTDOT: 15951f36f774SAl Viro follow_dotdot(nd); 15961f36f774SAl Viro dir = nd->path.dentry; 1597176306f5SNeil Brown case LAST_DOT: 15981f36f774SAl Viro if (nd->path.mnt->mnt_sb->s_type->fs_flags & FS_REVAL_DOT) { 15991f36f774SAl Viro if (!dir->d_op->d_revalidate(dir, nd)) { 16001f36f774SAl Viro error = -ESTALE; 1601a2c36b45SAl Viro goto exit; 16021f36f774SAl Viro } 16031f36f774SAl Viro } 16041f36f774SAl Viro /* fallthrough */ 16051f36f774SAl Viro case LAST_ROOT: 16061f36f774SAl Viro if (open_flag & O_CREAT) 16071f36f774SAl Viro goto exit; 16081f36f774SAl Viro /* fallthrough */ 16091f36f774SAl Viro case LAST_BIND: 16101f36f774SAl Viro audit_inode(pathname, dir); 16111f36f774SAl Viro goto ok; 16121f36f774SAl Viro } 1613a2c36b45SAl Viro 16141f36f774SAl Viro /* trailing slashes? */ 16151f36f774SAl Viro if (nd->last.name[nd->last.len]) { 16161f36f774SAl Viro if (open_flag & O_CREAT) 16171f36f774SAl Viro goto exit; 1618002baeecSJan Kara nd->flags |= LOOKUP_DIRECTORY | LOOKUP_FOLLOW; 16191f36f774SAl Viro } 16201f36f774SAl Viro 16211f36f774SAl Viro /* just plain open? */ 16221f36f774SAl Viro if (!(open_flag & O_CREAT)) { 16231f36f774SAl Viro error = do_lookup(nd, &nd->last, path); 16241f36f774SAl Viro if (error) 16251f36f774SAl Viro goto exit; 16261f36f774SAl Viro error = -ENOENT; 16271f36f774SAl Viro if (!path->dentry->d_inode) 16281f36f774SAl Viro goto exit_dput; 16291f36f774SAl Viro if (path->dentry->d_inode->i_op->follow_link) 16301f36f774SAl Viro return NULL; 16311f36f774SAl Viro error = -ENOTDIR; 16323e297b61SAl Viro if (nd->flags & LOOKUP_DIRECTORY) { 16333e297b61SAl Viro if (!path->dentry->d_inode->i_op->lookup) 16341f36f774SAl Viro goto exit_dput; 16353e297b61SAl Viro } 16361f36f774SAl Viro path_to_nameidata(path, nd); 16371f36f774SAl Viro audit_inode(pathname, nd->path.dentry); 16381f36f774SAl Viro goto ok; 16391f36f774SAl Viro } 16401f36f774SAl Viro 16411f36f774SAl Viro /* OK, it's O_CREAT */ 1642a1e28038SAl Viro mutex_lock(&dir->d_inode->i_mutex); 1643a1e28038SAl Viro 1644a1e28038SAl Viro path->dentry = lookup_hash(nd); 1645a1e28038SAl Viro path->mnt = nd->path.mnt; 1646a1e28038SAl Viro 1647fb1cc555SAl Viro error = PTR_ERR(path->dentry); 1648fb1cc555SAl Viro if (IS_ERR(path->dentry)) { 1649fb1cc555SAl Viro mutex_unlock(&dir->d_inode->i_mutex); 1650fb1cc555SAl Viro goto exit; 1651fb1cc555SAl Viro } 1652fb1cc555SAl Viro 1653fb1cc555SAl Viro if (IS_ERR(nd->intent.open.file)) { 1654fb1cc555SAl Viro error = PTR_ERR(nd->intent.open.file); 1655fb1cc555SAl Viro goto exit_mutex_unlock; 1656fb1cc555SAl Viro } 1657fb1cc555SAl Viro 1658fb1cc555SAl Viro /* Negative dentry, just create the file */ 1659fb1cc555SAl Viro if (!path->dentry->d_inode) { 1660fb1cc555SAl Viro /* 1661fb1cc555SAl Viro * This write is needed to ensure that a 1662fb1cc555SAl Viro * ro->rw transition does not occur between 1663fb1cc555SAl Viro * the time when the file is created and when 1664fb1cc555SAl Viro * a permanent write count is taken through 1665fb1cc555SAl Viro * the 'struct file' in nameidata_to_filp(). 1666fb1cc555SAl Viro */ 1667fb1cc555SAl Viro error = mnt_want_write(nd->path.mnt); 1668fb1cc555SAl Viro if (error) 1669fb1cc555SAl Viro goto exit_mutex_unlock; 1670fb1cc555SAl Viro error = __open_namei_create(nd, path, open_flag, mode); 1671fb1cc555SAl Viro if (error) { 1672fb1cc555SAl Viro mnt_drop_write(nd->path.mnt); 1673fb1cc555SAl Viro goto exit; 1674fb1cc555SAl Viro } 1675fb1cc555SAl Viro filp = nameidata_to_filp(nd); 1676fb1cc555SAl Viro mnt_drop_write(nd->path.mnt); 1677d893f1bcSAl Viro path_put(&nd->path); 1678fb1cc555SAl Viro if (!IS_ERR(filp)) { 1679fb1cc555SAl Viro error = ima_file_check(filp, acc_mode); 1680fb1cc555SAl Viro if (error) { 1681fb1cc555SAl Viro fput(filp); 1682fb1cc555SAl Viro filp = ERR_PTR(error); 1683fb1cc555SAl Viro } 1684fb1cc555SAl Viro } 1685fb1cc555SAl Viro return filp; 1686fb1cc555SAl Viro } 1687fb1cc555SAl Viro 1688fb1cc555SAl Viro /* 1689fb1cc555SAl Viro * It already exists. 1690fb1cc555SAl Viro */ 1691fb1cc555SAl Viro mutex_unlock(&dir->d_inode->i_mutex); 1692fb1cc555SAl Viro audit_inode(pathname, path->dentry); 1693fb1cc555SAl Viro 1694fb1cc555SAl Viro error = -EEXIST; 16955b369df8SAl Viro if (open_flag & O_EXCL) 1696fb1cc555SAl Viro goto exit_dput; 1697fb1cc555SAl Viro 1698fb1cc555SAl Viro if (__follow_mount(path)) { 1699fb1cc555SAl Viro error = -ELOOP; 17005b369df8SAl Viro if (open_flag & O_NOFOLLOW) 1701fb1cc555SAl Viro goto exit_dput; 1702fb1cc555SAl Viro } 1703fb1cc555SAl Viro 1704fb1cc555SAl Viro error = -ENOENT; 1705fb1cc555SAl Viro if (!path->dentry->d_inode) 1706fb1cc555SAl Viro goto exit_dput; 17079e67f361SAl Viro 17089e67f361SAl Viro if (path->dentry->d_inode->i_op->follow_link) 1709fb1cc555SAl Viro return NULL; 1710fb1cc555SAl Viro 1711fb1cc555SAl Viro path_to_nameidata(path, nd); 1712fb1cc555SAl Viro error = -EISDIR; 1713fb1cc555SAl Viro if (S_ISDIR(path->dentry->d_inode->i_mode)) 1714fb1cc555SAl Viro goto exit; 171567ee3ad2SAl Viro ok: 17169a66179eSAl Viro filp = finish_open(nd, open_flag, acc_mode); 1717fb1cc555SAl Viro return filp; 1718fb1cc555SAl Viro 1719fb1cc555SAl Viro exit_mutex_unlock: 1720fb1cc555SAl Viro mutex_unlock(&dir->d_inode->i_mutex); 1721fb1cc555SAl Viro exit_dput: 1722fb1cc555SAl Viro path_put_conditional(path, nd); 1723fb1cc555SAl Viro exit: 1724fb1cc555SAl Viro if (!IS_ERR(nd->intent.open.file)) 1725fb1cc555SAl Viro release_open_intent(nd); 1726fb1cc555SAl Viro path_put(&nd->path); 1727fb1cc555SAl Viro return ERR_PTR(error); 1728fb1cc555SAl Viro } 1729fb1cc555SAl Viro 17304a3fd211SDave Hansen /* 17314a3fd211SDave Hansen * Note that the low bits of the passed in "open_flag" 17324a3fd211SDave Hansen * are not the same as in the local variable "flag". See 17334a3fd211SDave Hansen * open_to_namei_flags() for more details. 17341da177e4SLinus Torvalds */ 1735a70e65dfSChristoph Hellwig struct file *do_filp_open(int dfd, const char *pathname, 17366e8341a1SAl Viro int open_flag, int mode, int acc_mode) 17371da177e4SLinus Torvalds { 17384a3fd211SDave Hansen struct file *filp; 1739a70e65dfSChristoph Hellwig struct nameidata nd; 17406e8341a1SAl Viro int error; 17419850c056SAl Viro struct path path; 17421da177e4SLinus Torvalds int count = 0; 1743d57999e1SDave Hansen int flag = open_to_namei_flags(open_flag); 17449850c056SAl Viro int force_reval = 0; 17451f36f774SAl Viro 17461f36f774SAl Viro if (!(open_flag & O_CREAT)) 17471f36f774SAl Viro mode = 0; 17481da177e4SLinus Torvalds 1749b1085ba8SLino Sanfilippo /* Must never be set by userspace */ 1750b1085ba8SLino Sanfilippo open_flag &= ~FMODE_NONOTIFY; 1751b1085ba8SLino Sanfilippo 17526b2f3d1fSChristoph Hellwig /* 17536b2f3d1fSChristoph Hellwig * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only 17546b2f3d1fSChristoph Hellwig * check for O_DSYNC if the need any syncing at all we enforce it's 17556b2f3d1fSChristoph Hellwig * always set instead of having to deal with possibly weird behaviour 17566b2f3d1fSChristoph Hellwig * for malicious applications setting only __O_SYNC. 17576b2f3d1fSChristoph Hellwig */ 17586b2f3d1fSChristoph Hellwig if (open_flag & __O_SYNC) 17596b2f3d1fSChristoph Hellwig open_flag |= O_DSYNC; 17606b2f3d1fSChristoph Hellwig 17616e8341a1SAl Viro if (!acc_mode) 17626d125529SAl Viro acc_mode = MAY_OPEN | ACC_MODE(open_flag); 17631da177e4SLinus Torvalds 1764834f2a4aSTrond Myklebust /* O_TRUNC implies we need access checks for write permissions */ 17654296e2cbSAl Viro if (open_flag & O_TRUNC) 1766834f2a4aSTrond Myklebust acc_mode |= MAY_WRITE; 1767834f2a4aSTrond Myklebust 17681da177e4SLinus Torvalds /* Allow the LSM permission hook to distinguish append 17691da177e4SLinus Torvalds access from general write access. */ 17704296e2cbSAl Viro if (open_flag & O_APPEND) 17711da177e4SLinus Torvalds acc_mode |= MAY_APPEND; 17721da177e4SLinus Torvalds 17731f36f774SAl Viro /* find the parent */ 17749850c056SAl Viro reval: 17759b4a9b14SAl Viro error = path_init(dfd, pathname, LOOKUP_PARENT, &nd); 17761da177e4SLinus Torvalds if (error) 1777a70e65dfSChristoph Hellwig return ERR_PTR(error); 17789850c056SAl Viro if (force_reval) 17799850c056SAl Viro nd.flags |= LOOKUP_REVAL; 17803866248eSAl Viro 17813866248eSAl Viro current->total_link_count = 0; 17823866248eSAl Viro error = link_path_walk(pathname, &nd); 1783654f562cSJ. R. Okajima if (error) { 178410fa8e62SAl Viro filp = ERR_PTR(error); 178510fa8e62SAl Viro goto out; 1786654f562cSJ. R. Okajima } 17871f36f774SAl Viro if (unlikely(!audit_dummy_context()) && (open_flag & O_CREAT)) 17889b4a9b14SAl Viro audit_inode(pathname, nd.path.dentry); 17891da177e4SLinus Torvalds 17901da177e4SLinus Torvalds /* 1791a2c36b45SAl Viro * We have the parent and last component. 17921da177e4SLinus Torvalds */ 17931da177e4SLinus Torvalds 17948737f3a1SAl Viro error = -ENFILE; 17958737f3a1SAl Viro filp = get_empty_filp(); 17968737f3a1SAl Viro if (filp == NULL) 17978737f3a1SAl Viro goto exit_parent; 17988737f3a1SAl Viro nd.intent.open.file = filp; 1799482928d5SAl Viro filp->f_flags = open_flag; 18008737f3a1SAl Viro nd.intent.open.flags = flag; 18018737f3a1SAl Viro nd.intent.open.create_mode = mode; 1802a70e65dfSChristoph Hellwig nd.flags &= ~LOOKUP_PARENT; 18031f36f774SAl Viro nd.flags |= LOOKUP_OPEN; 18041f36f774SAl Viro if (open_flag & O_CREAT) { 18051f36f774SAl Viro nd.flags |= LOOKUP_CREATE; 18064296e2cbSAl Viro if (open_flag & O_EXCL) 18073516586aSAl Viro nd.flags |= LOOKUP_EXCL; 18081f36f774SAl Viro } 18093e297b61SAl Viro if (open_flag & O_DIRECTORY) 18103e297b61SAl Viro nd.flags |= LOOKUP_DIRECTORY; 1811002baeecSJan Kara if (!(open_flag & O_NOFOLLOW)) 1812002baeecSJan Kara nd.flags |= LOOKUP_FOLLOW; 18133e297b61SAl Viro filp = do_last(&nd, &path, open_flag, acc_mode, mode, pathname); 1814806b681cSAl Viro while (unlikely(!filp)) { /* trailing symlink */ 1815def4af30SAl Viro struct path holder; 18161f36f774SAl Viro struct inode *inode = path.dentry->d_inode; 1817def4af30SAl Viro void *cookie; 1818806b681cSAl Viro error = -ELOOP; 18191f36f774SAl Viro /* S_ISDIR part is a temporary automount kludge */ 1820002baeecSJan Kara if (!(nd.flags & LOOKUP_FOLLOW) && !S_ISDIR(inode->i_mode)) 18211f36f774SAl Viro goto exit_dput; 18221f36f774SAl Viro if (count++ == 32) 1823806b681cSAl Viro goto exit_dput; 1824806b681cSAl Viro /* 1825806b681cSAl Viro * This is subtle. Instead of calling do_follow_link() we do 1826806b681cSAl Viro * the thing by hands. The reason is that this way we have zero 1827806b681cSAl Viro * link_count and path_walk() (called from ->follow_link) 1828806b681cSAl Viro * honoring LOOKUP_PARENT. After that we have the parent and 1829806b681cSAl Viro * last component, i.e. we are in the same situation as after 1830806b681cSAl Viro * the first path_walk(). Well, almost - if the last component 1831806b681cSAl Viro * is normal we get its copy stored in nd->last.name and we will 1832806b681cSAl Viro * have to putname() it when we are done. Procfs-like symlinks 1833806b681cSAl Viro * just set LAST_BIND. 1834806b681cSAl Viro */ 1835806b681cSAl Viro nd.flags |= LOOKUP_PARENT; 1836806b681cSAl Viro error = security_inode_follow_link(path.dentry, &nd); 1837806b681cSAl Viro if (error) 1838806b681cSAl Viro goto exit_dput; 1839def4af30SAl Viro error = __do_follow_link(&path, &nd, &cookie); 1840def4af30SAl Viro if (unlikely(error)) { 1841806b681cSAl Viro /* nd.path had been dropped */ 1842def4af30SAl Viro if (!IS_ERR(cookie) && inode->i_op->put_link) 1843def4af30SAl Viro inode->i_op->put_link(path.dentry, &nd, cookie); 1844def4af30SAl Viro path_put(&path); 1845a70e65dfSChristoph Hellwig release_open_intent(&nd); 184610fa8e62SAl Viro filp = ERR_PTR(error); 1847806b681cSAl Viro goto out; 1848806b681cSAl Viro } 1849def4af30SAl Viro holder = path; 1850806b681cSAl Viro nd.flags &= ~LOOKUP_PARENT; 18513e297b61SAl Viro filp = do_last(&nd, &path, open_flag, acc_mode, mode, pathname); 1852def4af30SAl Viro if (inode->i_op->put_link) 1853def4af30SAl Viro inode->i_op->put_link(holder.dentry, &nd, cookie); 1854def4af30SAl Viro path_put(&holder); 1855806b681cSAl Viro } 185610fa8e62SAl Viro out: 18572a737871SAl Viro if (nd.root.mnt) 18582a737871SAl Viro path_put(&nd.root); 185910fa8e62SAl Viro if (filp == ERR_PTR(-ESTALE) && !force_reval) { 186010fa8e62SAl Viro force_reval = 1; 186110fa8e62SAl Viro goto reval; 186210fa8e62SAl Viro } 186310fa8e62SAl Viro return filp; 18641da177e4SLinus Torvalds 1865806b681cSAl Viro exit_dput: 1866806b681cSAl Viro path_put_conditional(&path, &nd); 1867806b681cSAl Viro if (!IS_ERR(nd.intent.open.file)) 1868a70e65dfSChristoph Hellwig release_open_intent(&nd); 1869806b681cSAl Viro exit_parent: 1870806b681cSAl Viro path_put(&nd.path); 187110fa8e62SAl Viro filp = ERR_PTR(error); 187210fa8e62SAl Viro goto out; 1873de459215SKirill Korotaev } 18741da177e4SLinus Torvalds 18751da177e4SLinus Torvalds /** 1876a70e65dfSChristoph Hellwig * filp_open - open file and return file pointer 1877a70e65dfSChristoph Hellwig * 1878a70e65dfSChristoph Hellwig * @filename: path to open 1879a70e65dfSChristoph Hellwig * @flags: open flags as per the open(2) second argument 1880a70e65dfSChristoph Hellwig * @mode: mode for the new file if O_CREAT is set, else ignored 1881a70e65dfSChristoph Hellwig * 1882a70e65dfSChristoph Hellwig * This is the helper to open a file from kernelspace if you really 1883a70e65dfSChristoph Hellwig * have to. But in generally you should not do this, so please move 1884a70e65dfSChristoph Hellwig * along, nothing to see here.. 1885a70e65dfSChristoph Hellwig */ 1886a70e65dfSChristoph Hellwig struct file *filp_open(const char *filename, int flags, int mode) 1887a70e65dfSChristoph Hellwig { 18886e8341a1SAl Viro return do_filp_open(AT_FDCWD, filename, flags, mode, 0); 1889a70e65dfSChristoph Hellwig } 1890a70e65dfSChristoph Hellwig EXPORT_SYMBOL(filp_open); 1891a70e65dfSChristoph Hellwig 1892a70e65dfSChristoph Hellwig /** 18931da177e4SLinus Torvalds * lookup_create - lookup a dentry, creating it if it doesn't exist 18941da177e4SLinus Torvalds * @nd: nameidata info 18951da177e4SLinus Torvalds * @is_dir: directory flag 18961da177e4SLinus Torvalds * 18971da177e4SLinus Torvalds * Simple function to lookup and return a dentry and create it 18981da177e4SLinus Torvalds * if it doesn't exist. Is SMP-safe. 1899c663e5d8SChristoph Hellwig * 19004ac91378SJan Blunck * Returns with nd->path.dentry->d_inode->i_mutex locked. 19011da177e4SLinus Torvalds */ 19021da177e4SLinus Torvalds struct dentry *lookup_create(struct nameidata *nd, int is_dir) 19031da177e4SLinus Torvalds { 1904c663e5d8SChristoph Hellwig struct dentry *dentry = ERR_PTR(-EEXIST); 19051da177e4SLinus Torvalds 19064ac91378SJan Blunck mutex_lock_nested(&nd->path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); 1907c663e5d8SChristoph Hellwig /* 1908c663e5d8SChristoph Hellwig * Yucky last component or no last component at all? 1909c663e5d8SChristoph Hellwig * (foo/., foo/.., /////) 1910c663e5d8SChristoph Hellwig */ 19111da177e4SLinus Torvalds if (nd->last_type != LAST_NORM) 19121da177e4SLinus Torvalds goto fail; 19131da177e4SLinus Torvalds nd->flags &= ~LOOKUP_PARENT; 19143516586aSAl Viro nd->flags |= LOOKUP_CREATE | LOOKUP_EXCL; 1915a634904aSASANO Masahiro nd->intent.open.flags = O_EXCL; 1916c663e5d8SChristoph Hellwig 1917c663e5d8SChristoph Hellwig /* 1918c663e5d8SChristoph Hellwig * Do the final lookup. 1919c663e5d8SChristoph Hellwig */ 192049705b77SChristoph Hellwig dentry = lookup_hash(nd); 19211da177e4SLinus Torvalds if (IS_ERR(dentry)) 19221da177e4SLinus Torvalds goto fail; 1923c663e5d8SChristoph Hellwig 1924e9baf6e5SAl Viro if (dentry->d_inode) 1925e9baf6e5SAl Viro goto eexist; 1926c663e5d8SChristoph Hellwig /* 1927c663e5d8SChristoph Hellwig * Special case - lookup gave negative, but... we had foo/bar/ 1928c663e5d8SChristoph Hellwig * From the vfs_mknod() POV we just have a negative dentry - 1929c663e5d8SChristoph Hellwig * all is fine. Let's be bastards - you had / on the end, you've 1930c663e5d8SChristoph Hellwig * been asking for (non-existent) directory. -ENOENT for you. 1931c663e5d8SChristoph Hellwig */ 1932e9baf6e5SAl Viro if (unlikely(!is_dir && nd->last.name[nd->last.len])) { 19331da177e4SLinus Torvalds dput(dentry); 19341da177e4SLinus Torvalds dentry = ERR_PTR(-ENOENT); 1935e9baf6e5SAl Viro } 1936e9baf6e5SAl Viro return dentry; 1937e9baf6e5SAl Viro eexist: 1938e9baf6e5SAl Viro dput(dentry); 1939e9baf6e5SAl Viro dentry = ERR_PTR(-EEXIST); 19401da177e4SLinus Torvalds fail: 19411da177e4SLinus Torvalds return dentry; 19421da177e4SLinus Torvalds } 1943f81a0bffSChristoph Hellwig EXPORT_SYMBOL_GPL(lookup_create); 19441da177e4SLinus Torvalds 19451da177e4SLinus Torvalds int vfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev) 19461da177e4SLinus Torvalds { 1947a95164d9SMiklos Szeredi int error = may_create(dir, dentry); 19481da177e4SLinus Torvalds 19491da177e4SLinus Torvalds if (error) 19501da177e4SLinus Torvalds return error; 19511da177e4SLinus Torvalds 19521da177e4SLinus Torvalds if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) 19531da177e4SLinus Torvalds return -EPERM; 19541da177e4SLinus Torvalds 1955acfa4380SAl Viro if (!dir->i_op->mknod) 19561da177e4SLinus Torvalds return -EPERM; 19571da177e4SLinus Torvalds 195808ce5f16SSerge E. Hallyn error = devcgroup_inode_mknod(mode, dev); 195908ce5f16SSerge E. Hallyn if (error) 196008ce5f16SSerge E. Hallyn return error; 196108ce5f16SSerge E. Hallyn 19621da177e4SLinus Torvalds error = security_inode_mknod(dir, dentry, mode, dev); 19631da177e4SLinus Torvalds if (error) 19641da177e4SLinus Torvalds return error; 19651da177e4SLinus Torvalds 19661da177e4SLinus Torvalds error = dir->i_op->mknod(dir, dentry, mode, dev); 1967a74574aaSStephen Smalley if (!error) 1968f38aa942SAmy Griffis fsnotify_create(dir, dentry); 19691da177e4SLinus Torvalds return error; 19701da177e4SLinus Torvalds } 19711da177e4SLinus Torvalds 1972463c3197SDave Hansen static int may_mknod(mode_t mode) 1973463c3197SDave Hansen { 1974463c3197SDave Hansen switch (mode & S_IFMT) { 1975463c3197SDave Hansen case S_IFREG: 1976463c3197SDave Hansen case S_IFCHR: 1977463c3197SDave Hansen case S_IFBLK: 1978463c3197SDave Hansen case S_IFIFO: 1979463c3197SDave Hansen case S_IFSOCK: 1980463c3197SDave Hansen case 0: /* zero mode translates to S_IFREG */ 1981463c3197SDave Hansen return 0; 1982463c3197SDave Hansen case S_IFDIR: 1983463c3197SDave Hansen return -EPERM; 1984463c3197SDave Hansen default: 1985463c3197SDave Hansen return -EINVAL; 1986463c3197SDave Hansen } 1987463c3197SDave Hansen } 1988463c3197SDave Hansen 19892e4d0924SHeiko Carstens SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, int, mode, 19902e4d0924SHeiko Carstens unsigned, dev) 19911da177e4SLinus Torvalds { 19922ad94ae6SAl Viro int error; 19931da177e4SLinus Torvalds char *tmp; 19941da177e4SLinus Torvalds struct dentry *dentry; 19951da177e4SLinus Torvalds struct nameidata nd; 19961da177e4SLinus Torvalds 19971da177e4SLinus Torvalds if (S_ISDIR(mode)) 19981da177e4SLinus Torvalds return -EPERM; 19991da177e4SLinus Torvalds 20002ad94ae6SAl Viro error = user_path_parent(dfd, filename, &nd, &tmp); 20011da177e4SLinus Torvalds if (error) 20022ad94ae6SAl Viro return error; 20032ad94ae6SAl Viro 20041da177e4SLinus Torvalds dentry = lookup_create(&nd, 0); 2005463c3197SDave Hansen if (IS_ERR(dentry)) { 20061da177e4SLinus Torvalds error = PTR_ERR(dentry); 2007463c3197SDave Hansen goto out_unlock; 2008463c3197SDave Hansen } 20094ac91378SJan Blunck if (!IS_POSIXACL(nd.path.dentry->d_inode)) 2010ce3b0f8dSAl Viro mode &= ~current_umask(); 2011463c3197SDave Hansen error = may_mknod(mode); 2012463c3197SDave Hansen if (error) 2013463c3197SDave Hansen goto out_dput; 2014463c3197SDave Hansen error = mnt_want_write(nd.path.mnt); 2015463c3197SDave Hansen if (error) 2016463c3197SDave Hansen goto out_dput; 2017be6d3e56SKentaro Takeda error = security_path_mknod(&nd.path, dentry, mode, dev); 2018be6d3e56SKentaro Takeda if (error) 2019be6d3e56SKentaro Takeda goto out_drop_write; 20201da177e4SLinus Torvalds switch (mode & S_IFMT) { 20211da177e4SLinus Torvalds case 0: case S_IFREG: 20224ac91378SJan Blunck error = vfs_create(nd.path.dentry->d_inode,dentry,mode,&nd); 20231da177e4SLinus Torvalds break; 20241da177e4SLinus Torvalds case S_IFCHR: case S_IFBLK: 20254ac91378SJan Blunck error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode, 20261da177e4SLinus Torvalds new_decode_dev(dev)); 20271da177e4SLinus Torvalds break; 20281da177e4SLinus Torvalds case S_IFIFO: case S_IFSOCK: 20294ac91378SJan Blunck error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode,0); 20301da177e4SLinus Torvalds break; 20311da177e4SLinus Torvalds } 2032be6d3e56SKentaro Takeda out_drop_write: 2033463c3197SDave Hansen mnt_drop_write(nd.path.mnt); 2034463c3197SDave Hansen out_dput: 20351da177e4SLinus Torvalds dput(dentry); 2036463c3197SDave Hansen out_unlock: 20374ac91378SJan Blunck mutex_unlock(&nd.path.dentry->d_inode->i_mutex); 20381d957f9bSJan Blunck path_put(&nd.path); 20391da177e4SLinus Torvalds putname(tmp); 20401da177e4SLinus Torvalds 20411da177e4SLinus Torvalds return error; 20421da177e4SLinus Torvalds } 20431da177e4SLinus Torvalds 20443480b257SHeiko Carstens SYSCALL_DEFINE3(mknod, const char __user *, filename, int, mode, unsigned, dev) 20455590ff0dSUlrich Drepper { 20465590ff0dSUlrich Drepper return sys_mknodat(AT_FDCWD, filename, mode, dev); 20475590ff0dSUlrich Drepper } 20485590ff0dSUlrich Drepper 20491da177e4SLinus Torvalds int vfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) 20501da177e4SLinus Torvalds { 2051a95164d9SMiklos Szeredi int error = may_create(dir, dentry); 20521da177e4SLinus Torvalds 20531da177e4SLinus Torvalds if (error) 20541da177e4SLinus Torvalds return error; 20551da177e4SLinus Torvalds 2056acfa4380SAl Viro if (!dir->i_op->mkdir) 20571da177e4SLinus Torvalds return -EPERM; 20581da177e4SLinus Torvalds 20591da177e4SLinus Torvalds mode &= (S_IRWXUGO|S_ISVTX); 20601da177e4SLinus Torvalds error = security_inode_mkdir(dir, dentry, mode); 20611da177e4SLinus Torvalds if (error) 20621da177e4SLinus Torvalds return error; 20631da177e4SLinus Torvalds 20641da177e4SLinus Torvalds error = dir->i_op->mkdir(dir, dentry, mode); 2065a74574aaSStephen Smalley if (!error) 2066f38aa942SAmy Griffis fsnotify_mkdir(dir, dentry); 20671da177e4SLinus Torvalds return error; 20681da177e4SLinus Torvalds } 20691da177e4SLinus Torvalds 20702e4d0924SHeiko Carstens SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, int, mode) 20711da177e4SLinus Torvalds { 20721da177e4SLinus Torvalds int error = 0; 20731da177e4SLinus Torvalds char * tmp; 20746902d925SDave Hansen struct dentry *dentry; 20756902d925SDave Hansen struct nameidata nd; 20761da177e4SLinus Torvalds 20772ad94ae6SAl Viro error = user_path_parent(dfd, pathname, &nd, &tmp); 20782ad94ae6SAl Viro if (error) 20796902d925SDave Hansen goto out_err; 20801da177e4SLinus Torvalds 20811da177e4SLinus Torvalds dentry = lookup_create(&nd, 1); 20821da177e4SLinus Torvalds error = PTR_ERR(dentry); 20836902d925SDave Hansen if (IS_ERR(dentry)) 20846902d925SDave Hansen goto out_unlock; 20856902d925SDave Hansen 20864ac91378SJan Blunck if (!IS_POSIXACL(nd.path.dentry->d_inode)) 2087ce3b0f8dSAl Viro mode &= ~current_umask(); 2088463c3197SDave Hansen error = mnt_want_write(nd.path.mnt); 2089463c3197SDave Hansen if (error) 2090463c3197SDave Hansen goto out_dput; 2091be6d3e56SKentaro Takeda error = security_path_mkdir(&nd.path, dentry, mode); 2092be6d3e56SKentaro Takeda if (error) 2093be6d3e56SKentaro Takeda goto out_drop_write; 20944ac91378SJan Blunck error = vfs_mkdir(nd.path.dentry->d_inode, dentry, mode); 2095be6d3e56SKentaro Takeda out_drop_write: 2096463c3197SDave Hansen mnt_drop_write(nd.path.mnt); 2097463c3197SDave Hansen out_dput: 20981da177e4SLinus Torvalds dput(dentry); 20996902d925SDave Hansen out_unlock: 21004ac91378SJan Blunck mutex_unlock(&nd.path.dentry->d_inode->i_mutex); 21011d957f9bSJan Blunck path_put(&nd.path); 21021da177e4SLinus Torvalds putname(tmp); 21036902d925SDave Hansen out_err: 21041da177e4SLinus Torvalds return error; 21051da177e4SLinus Torvalds } 21061da177e4SLinus Torvalds 21073cdad428SHeiko Carstens SYSCALL_DEFINE2(mkdir, const char __user *, pathname, int, mode) 21085590ff0dSUlrich Drepper { 21095590ff0dSUlrich Drepper return sys_mkdirat(AT_FDCWD, pathname, mode); 21105590ff0dSUlrich Drepper } 21115590ff0dSUlrich Drepper 21121da177e4SLinus Torvalds /* 21131da177e4SLinus Torvalds * We try to drop the dentry early: we should have 21141da177e4SLinus Torvalds * a usage count of 2 if we're the only user of this 21151da177e4SLinus Torvalds * dentry, and if that is true (possibly after pruning 21161da177e4SLinus Torvalds * the dcache), then we drop the dentry now. 21171da177e4SLinus Torvalds * 21181da177e4SLinus Torvalds * A low-level filesystem can, if it choses, legally 21191da177e4SLinus Torvalds * do a 21201da177e4SLinus Torvalds * 21211da177e4SLinus Torvalds * if (!d_unhashed(dentry)) 21221da177e4SLinus Torvalds * return -EBUSY; 21231da177e4SLinus Torvalds * 21241da177e4SLinus Torvalds * if it cannot handle the case of removing a directory 21251da177e4SLinus Torvalds * that is still in use by something else.. 21261da177e4SLinus Torvalds */ 21271da177e4SLinus Torvalds void dentry_unhash(struct dentry *dentry) 21281da177e4SLinus Torvalds { 21291da177e4SLinus Torvalds dget(dentry); 21301da177e4SLinus Torvalds shrink_dcache_parent(dentry); 21311da177e4SLinus Torvalds spin_lock(&dentry->d_lock); 2132b7ab39f6SNick Piggin if (dentry->d_count == 2) 21331da177e4SLinus Torvalds __d_drop(dentry); 21341da177e4SLinus Torvalds spin_unlock(&dentry->d_lock); 21351da177e4SLinus Torvalds } 21361da177e4SLinus Torvalds 21371da177e4SLinus Torvalds int vfs_rmdir(struct inode *dir, struct dentry *dentry) 21381da177e4SLinus Torvalds { 21391da177e4SLinus Torvalds int error = may_delete(dir, dentry, 1); 21401da177e4SLinus Torvalds 21411da177e4SLinus Torvalds if (error) 21421da177e4SLinus Torvalds return error; 21431da177e4SLinus Torvalds 2144acfa4380SAl Viro if (!dir->i_op->rmdir) 21451da177e4SLinus Torvalds return -EPERM; 21461da177e4SLinus Torvalds 21471b1dcc1bSJes Sorensen mutex_lock(&dentry->d_inode->i_mutex); 21481da177e4SLinus Torvalds dentry_unhash(dentry); 21491da177e4SLinus Torvalds if (d_mountpoint(dentry)) 21501da177e4SLinus Torvalds error = -EBUSY; 21511da177e4SLinus Torvalds else { 21521da177e4SLinus Torvalds error = security_inode_rmdir(dir, dentry); 21531da177e4SLinus Torvalds if (!error) { 21541da177e4SLinus Torvalds error = dir->i_op->rmdir(dir, dentry); 2155d83c49f3SAl Viro if (!error) { 21561da177e4SLinus Torvalds dentry->d_inode->i_flags |= S_DEAD; 2157d83c49f3SAl Viro dont_mount(dentry); 2158d83c49f3SAl Viro } 21591da177e4SLinus Torvalds } 21601da177e4SLinus Torvalds } 21611b1dcc1bSJes Sorensen mutex_unlock(&dentry->d_inode->i_mutex); 21621da177e4SLinus Torvalds if (!error) { 21631da177e4SLinus Torvalds d_delete(dentry); 21641da177e4SLinus Torvalds } 21651da177e4SLinus Torvalds dput(dentry); 21661da177e4SLinus Torvalds 21671da177e4SLinus Torvalds return error; 21681da177e4SLinus Torvalds } 21691da177e4SLinus Torvalds 21705590ff0dSUlrich Drepper static long do_rmdir(int dfd, const char __user *pathname) 21711da177e4SLinus Torvalds { 21721da177e4SLinus Torvalds int error = 0; 21731da177e4SLinus Torvalds char * name; 21741da177e4SLinus Torvalds struct dentry *dentry; 21751da177e4SLinus Torvalds struct nameidata nd; 21761da177e4SLinus Torvalds 21772ad94ae6SAl Viro error = user_path_parent(dfd, pathname, &nd, &name); 21781da177e4SLinus Torvalds if (error) 21792ad94ae6SAl Viro return error; 21801da177e4SLinus Torvalds 21811da177e4SLinus Torvalds switch(nd.last_type) { 21821da177e4SLinus Torvalds case LAST_DOTDOT: 21831da177e4SLinus Torvalds error = -ENOTEMPTY; 21841da177e4SLinus Torvalds goto exit1; 21851da177e4SLinus Torvalds case LAST_DOT: 21861da177e4SLinus Torvalds error = -EINVAL; 21871da177e4SLinus Torvalds goto exit1; 21881da177e4SLinus Torvalds case LAST_ROOT: 21891da177e4SLinus Torvalds error = -EBUSY; 21901da177e4SLinus Torvalds goto exit1; 21911da177e4SLinus Torvalds } 21920612d9fbSOGAWA Hirofumi 21930612d9fbSOGAWA Hirofumi nd.flags &= ~LOOKUP_PARENT; 21940612d9fbSOGAWA Hirofumi 21954ac91378SJan Blunck mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); 219649705b77SChristoph Hellwig dentry = lookup_hash(&nd); 21971da177e4SLinus Torvalds error = PTR_ERR(dentry); 21986902d925SDave Hansen if (IS_ERR(dentry)) 21996902d925SDave Hansen goto exit2; 22000622753bSDave Hansen error = mnt_want_write(nd.path.mnt); 22010622753bSDave Hansen if (error) 22020622753bSDave Hansen goto exit3; 2203be6d3e56SKentaro Takeda error = security_path_rmdir(&nd.path, dentry); 2204be6d3e56SKentaro Takeda if (error) 2205be6d3e56SKentaro Takeda goto exit4; 22064ac91378SJan Blunck error = vfs_rmdir(nd.path.dentry->d_inode, dentry); 2207be6d3e56SKentaro Takeda exit4: 22080622753bSDave Hansen mnt_drop_write(nd.path.mnt); 22090622753bSDave Hansen exit3: 22101da177e4SLinus Torvalds dput(dentry); 22116902d925SDave Hansen exit2: 22124ac91378SJan Blunck mutex_unlock(&nd.path.dentry->d_inode->i_mutex); 22131da177e4SLinus Torvalds exit1: 22141d957f9bSJan Blunck path_put(&nd.path); 22151da177e4SLinus Torvalds putname(name); 22161da177e4SLinus Torvalds return error; 22171da177e4SLinus Torvalds } 22181da177e4SLinus Torvalds 22193cdad428SHeiko Carstens SYSCALL_DEFINE1(rmdir, const char __user *, pathname) 22205590ff0dSUlrich Drepper { 22215590ff0dSUlrich Drepper return do_rmdir(AT_FDCWD, pathname); 22225590ff0dSUlrich Drepper } 22235590ff0dSUlrich Drepper 22241da177e4SLinus Torvalds int vfs_unlink(struct inode *dir, struct dentry *dentry) 22251da177e4SLinus Torvalds { 22261da177e4SLinus Torvalds int error = may_delete(dir, dentry, 0); 22271da177e4SLinus Torvalds 22281da177e4SLinus Torvalds if (error) 22291da177e4SLinus Torvalds return error; 22301da177e4SLinus Torvalds 2231acfa4380SAl Viro if (!dir->i_op->unlink) 22321da177e4SLinus Torvalds return -EPERM; 22331da177e4SLinus Torvalds 22341b1dcc1bSJes Sorensen mutex_lock(&dentry->d_inode->i_mutex); 22351da177e4SLinus Torvalds if (d_mountpoint(dentry)) 22361da177e4SLinus Torvalds error = -EBUSY; 22371da177e4SLinus Torvalds else { 22381da177e4SLinus Torvalds error = security_inode_unlink(dir, dentry); 2239bec1052eSAl Viro if (!error) { 22401da177e4SLinus Torvalds error = dir->i_op->unlink(dir, dentry); 2241bec1052eSAl Viro if (!error) 2242d83c49f3SAl Viro dont_mount(dentry); 2243bec1052eSAl Viro } 22441da177e4SLinus Torvalds } 22451b1dcc1bSJes Sorensen mutex_unlock(&dentry->d_inode->i_mutex); 22461da177e4SLinus Torvalds 22471da177e4SLinus Torvalds /* We don't d_delete() NFS sillyrenamed files--they still exist. */ 22481da177e4SLinus Torvalds if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { 2249ece95912SJan Kara fsnotify_link_count(dentry->d_inode); 22501da177e4SLinus Torvalds d_delete(dentry); 22511da177e4SLinus Torvalds } 22520eeca283SRobert Love 22531da177e4SLinus Torvalds return error; 22541da177e4SLinus Torvalds } 22551da177e4SLinus Torvalds 22561da177e4SLinus Torvalds /* 22571da177e4SLinus Torvalds * Make sure that the actual truncation of the file will occur outside its 22581b1dcc1bSJes Sorensen * directory's i_mutex. Truncate can take a long time if there is a lot of 22591da177e4SLinus Torvalds * writeout happening, and we don't want to prevent access to the directory 22601da177e4SLinus Torvalds * while waiting on the I/O. 22611da177e4SLinus Torvalds */ 22625590ff0dSUlrich Drepper static long do_unlinkat(int dfd, const char __user *pathname) 22631da177e4SLinus Torvalds { 22642ad94ae6SAl Viro int error; 22651da177e4SLinus Torvalds char *name; 22661da177e4SLinus Torvalds struct dentry *dentry; 22671da177e4SLinus Torvalds struct nameidata nd; 22681da177e4SLinus Torvalds struct inode *inode = NULL; 22691da177e4SLinus Torvalds 22702ad94ae6SAl Viro error = user_path_parent(dfd, pathname, &nd, &name); 22711da177e4SLinus Torvalds if (error) 22722ad94ae6SAl Viro return error; 22732ad94ae6SAl Viro 22741da177e4SLinus Torvalds error = -EISDIR; 22751da177e4SLinus Torvalds if (nd.last_type != LAST_NORM) 22761da177e4SLinus Torvalds goto exit1; 22770612d9fbSOGAWA Hirofumi 22780612d9fbSOGAWA Hirofumi nd.flags &= ~LOOKUP_PARENT; 22790612d9fbSOGAWA Hirofumi 22804ac91378SJan Blunck mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); 228149705b77SChristoph Hellwig dentry = lookup_hash(&nd); 22821da177e4SLinus Torvalds error = PTR_ERR(dentry); 22831da177e4SLinus Torvalds if (!IS_ERR(dentry)) { 22841da177e4SLinus Torvalds /* Why not before? Because we want correct error value */ 22851da177e4SLinus Torvalds if (nd.last.name[nd.last.len]) 22861da177e4SLinus Torvalds goto slashes; 22871da177e4SLinus Torvalds inode = dentry->d_inode; 22881da177e4SLinus Torvalds if (inode) 22897de9c6eeSAl Viro ihold(inode); 22900622753bSDave Hansen error = mnt_want_write(nd.path.mnt); 22910622753bSDave Hansen if (error) 22920622753bSDave Hansen goto exit2; 2293be6d3e56SKentaro Takeda error = security_path_unlink(&nd.path, dentry); 2294be6d3e56SKentaro Takeda if (error) 2295be6d3e56SKentaro Takeda goto exit3; 22964ac91378SJan Blunck error = vfs_unlink(nd.path.dentry->d_inode, dentry); 2297be6d3e56SKentaro Takeda exit3: 22980622753bSDave Hansen mnt_drop_write(nd.path.mnt); 22991da177e4SLinus Torvalds exit2: 23001da177e4SLinus Torvalds dput(dentry); 23011da177e4SLinus Torvalds } 23024ac91378SJan Blunck mutex_unlock(&nd.path.dentry->d_inode->i_mutex); 23031da177e4SLinus Torvalds if (inode) 23041da177e4SLinus Torvalds iput(inode); /* truncate the inode here */ 23051da177e4SLinus Torvalds exit1: 23061d957f9bSJan Blunck path_put(&nd.path); 23071da177e4SLinus Torvalds putname(name); 23081da177e4SLinus Torvalds return error; 23091da177e4SLinus Torvalds 23101da177e4SLinus Torvalds slashes: 23111da177e4SLinus Torvalds error = !dentry->d_inode ? -ENOENT : 23121da177e4SLinus Torvalds S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; 23131da177e4SLinus Torvalds goto exit2; 23141da177e4SLinus Torvalds } 23151da177e4SLinus Torvalds 23162e4d0924SHeiko Carstens SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) 23175590ff0dSUlrich Drepper { 23185590ff0dSUlrich Drepper if ((flag & ~AT_REMOVEDIR) != 0) 23195590ff0dSUlrich Drepper return -EINVAL; 23205590ff0dSUlrich Drepper 23215590ff0dSUlrich Drepper if (flag & AT_REMOVEDIR) 23225590ff0dSUlrich Drepper return do_rmdir(dfd, pathname); 23235590ff0dSUlrich Drepper 23245590ff0dSUlrich Drepper return do_unlinkat(dfd, pathname); 23255590ff0dSUlrich Drepper } 23265590ff0dSUlrich Drepper 23273480b257SHeiko Carstens SYSCALL_DEFINE1(unlink, const char __user *, pathname) 23285590ff0dSUlrich Drepper { 23295590ff0dSUlrich Drepper return do_unlinkat(AT_FDCWD, pathname); 23305590ff0dSUlrich Drepper } 23315590ff0dSUlrich Drepper 2332db2e747bSMiklos Szeredi int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) 23331da177e4SLinus Torvalds { 2334a95164d9SMiklos Szeredi int error = may_create(dir, dentry); 23351da177e4SLinus Torvalds 23361da177e4SLinus Torvalds if (error) 23371da177e4SLinus Torvalds return error; 23381da177e4SLinus Torvalds 2339acfa4380SAl Viro if (!dir->i_op->symlink) 23401da177e4SLinus Torvalds return -EPERM; 23411da177e4SLinus Torvalds 23421da177e4SLinus Torvalds error = security_inode_symlink(dir, dentry, oldname); 23431da177e4SLinus Torvalds if (error) 23441da177e4SLinus Torvalds return error; 23451da177e4SLinus Torvalds 23461da177e4SLinus Torvalds error = dir->i_op->symlink(dir, dentry, oldname); 2347a74574aaSStephen Smalley if (!error) 2348f38aa942SAmy Griffis fsnotify_create(dir, dentry); 23491da177e4SLinus Torvalds return error; 23501da177e4SLinus Torvalds } 23511da177e4SLinus Torvalds 23522e4d0924SHeiko Carstens SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, 23532e4d0924SHeiko Carstens int, newdfd, const char __user *, newname) 23541da177e4SLinus Torvalds { 23552ad94ae6SAl Viro int error; 23561da177e4SLinus Torvalds char *from; 23571da177e4SLinus Torvalds char *to; 23586902d925SDave Hansen struct dentry *dentry; 23596902d925SDave Hansen struct nameidata nd; 23601da177e4SLinus Torvalds 23611da177e4SLinus Torvalds from = getname(oldname); 23621da177e4SLinus Torvalds if (IS_ERR(from)) 23631da177e4SLinus Torvalds return PTR_ERR(from); 23642ad94ae6SAl Viro 23652ad94ae6SAl Viro error = user_path_parent(newdfd, newname, &nd, &to); 23662ad94ae6SAl Viro if (error) 23676902d925SDave Hansen goto out_putname; 23681da177e4SLinus Torvalds 23691da177e4SLinus Torvalds dentry = lookup_create(&nd, 0); 23701da177e4SLinus Torvalds error = PTR_ERR(dentry); 23716902d925SDave Hansen if (IS_ERR(dentry)) 23726902d925SDave Hansen goto out_unlock; 23736902d925SDave Hansen 237475c3f29dSDave Hansen error = mnt_want_write(nd.path.mnt); 237575c3f29dSDave Hansen if (error) 237675c3f29dSDave Hansen goto out_dput; 2377be6d3e56SKentaro Takeda error = security_path_symlink(&nd.path, dentry, from); 2378be6d3e56SKentaro Takeda if (error) 2379be6d3e56SKentaro Takeda goto out_drop_write; 2380db2e747bSMiklos Szeredi error = vfs_symlink(nd.path.dentry->d_inode, dentry, from); 2381be6d3e56SKentaro Takeda out_drop_write: 238275c3f29dSDave Hansen mnt_drop_write(nd.path.mnt); 238375c3f29dSDave Hansen out_dput: 23841da177e4SLinus Torvalds dput(dentry); 23856902d925SDave Hansen out_unlock: 23864ac91378SJan Blunck mutex_unlock(&nd.path.dentry->d_inode->i_mutex); 23871d957f9bSJan Blunck path_put(&nd.path); 23881da177e4SLinus Torvalds putname(to); 23896902d925SDave Hansen out_putname: 23901da177e4SLinus Torvalds putname(from); 23911da177e4SLinus Torvalds return error; 23921da177e4SLinus Torvalds } 23931da177e4SLinus Torvalds 23943480b257SHeiko Carstens SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) 23955590ff0dSUlrich Drepper { 23965590ff0dSUlrich Drepper return sys_symlinkat(oldname, AT_FDCWD, newname); 23975590ff0dSUlrich Drepper } 23985590ff0dSUlrich Drepper 23991da177e4SLinus Torvalds int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) 24001da177e4SLinus Torvalds { 24011da177e4SLinus Torvalds struct inode *inode = old_dentry->d_inode; 24021da177e4SLinus Torvalds int error; 24031da177e4SLinus Torvalds 24041da177e4SLinus Torvalds if (!inode) 24051da177e4SLinus Torvalds return -ENOENT; 24061da177e4SLinus Torvalds 2407a95164d9SMiklos Szeredi error = may_create(dir, new_dentry); 24081da177e4SLinus Torvalds if (error) 24091da177e4SLinus Torvalds return error; 24101da177e4SLinus Torvalds 24111da177e4SLinus Torvalds if (dir->i_sb != inode->i_sb) 24121da177e4SLinus Torvalds return -EXDEV; 24131da177e4SLinus Torvalds 24141da177e4SLinus Torvalds /* 24151da177e4SLinus Torvalds * A link to an append-only or immutable file cannot be created. 24161da177e4SLinus Torvalds */ 24171da177e4SLinus Torvalds if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) 24181da177e4SLinus Torvalds return -EPERM; 2419acfa4380SAl Viro if (!dir->i_op->link) 24201da177e4SLinus Torvalds return -EPERM; 24217e79eedbSTetsuo Handa if (S_ISDIR(inode->i_mode)) 24221da177e4SLinus Torvalds return -EPERM; 24231da177e4SLinus Torvalds 24241da177e4SLinus Torvalds error = security_inode_link(old_dentry, dir, new_dentry); 24251da177e4SLinus Torvalds if (error) 24261da177e4SLinus Torvalds return error; 24271da177e4SLinus Torvalds 24287e79eedbSTetsuo Handa mutex_lock(&inode->i_mutex); 24291da177e4SLinus Torvalds error = dir->i_op->link(old_dentry, dir, new_dentry); 24307e79eedbSTetsuo Handa mutex_unlock(&inode->i_mutex); 2431e31e14ecSStephen Smalley if (!error) 24327e79eedbSTetsuo Handa fsnotify_link(dir, inode, new_dentry); 24331da177e4SLinus Torvalds return error; 24341da177e4SLinus Torvalds } 24351da177e4SLinus Torvalds 24361da177e4SLinus Torvalds /* 24371da177e4SLinus Torvalds * Hardlinks are often used in delicate situations. We avoid 24381da177e4SLinus Torvalds * security-related surprises by not following symlinks on the 24391da177e4SLinus Torvalds * newname. --KAB 24401da177e4SLinus Torvalds * 24411da177e4SLinus Torvalds * We don't follow them on the oldname either to be compatible 24421da177e4SLinus Torvalds * with linux 2.0, and to avoid hard-linking to directories 24431da177e4SLinus Torvalds * and other special files. --ADM 24441da177e4SLinus Torvalds */ 24452e4d0924SHeiko Carstens SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, 24462e4d0924SHeiko Carstens int, newdfd, const char __user *, newname, int, flags) 24471da177e4SLinus Torvalds { 24481da177e4SLinus Torvalds struct dentry *new_dentry; 24492d8f3038SAl Viro struct nameidata nd; 24502d8f3038SAl Viro struct path old_path; 24511da177e4SLinus Torvalds int error; 24521da177e4SLinus Torvalds char *to; 24531da177e4SLinus Torvalds 245445c9b11aSUlrich Drepper if ((flags & ~AT_SYMLINK_FOLLOW) != 0) 2455c04030e1SUlrich Drepper return -EINVAL; 2456c04030e1SUlrich Drepper 24572d8f3038SAl Viro error = user_path_at(olddfd, oldname, 245845c9b11aSUlrich Drepper flags & AT_SYMLINK_FOLLOW ? LOOKUP_FOLLOW : 0, 24592d8f3038SAl Viro &old_path); 24601da177e4SLinus Torvalds if (error) 24612ad94ae6SAl Viro return error; 24622ad94ae6SAl Viro 24632ad94ae6SAl Viro error = user_path_parent(newdfd, newname, &nd, &to); 24641da177e4SLinus Torvalds if (error) 24651da177e4SLinus Torvalds goto out; 24661da177e4SLinus Torvalds error = -EXDEV; 24672d8f3038SAl Viro if (old_path.mnt != nd.path.mnt) 24681da177e4SLinus Torvalds goto out_release; 24691da177e4SLinus Torvalds new_dentry = lookup_create(&nd, 0); 24701da177e4SLinus Torvalds error = PTR_ERR(new_dentry); 24716902d925SDave Hansen if (IS_ERR(new_dentry)) 24726902d925SDave Hansen goto out_unlock; 247375c3f29dSDave Hansen error = mnt_want_write(nd.path.mnt); 247475c3f29dSDave Hansen if (error) 247575c3f29dSDave Hansen goto out_dput; 2476be6d3e56SKentaro Takeda error = security_path_link(old_path.dentry, &nd.path, new_dentry); 2477be6d3e56SKentaro Takeda if (error) 2478be6d3e56SKentaro Takeda goto out_drop_write; 24792d8f3038SAl Viro error = vfs_link(old_path.dentry, nd.path.dentry->d_inode, new_dentry); 2480be6d3e56SKentaro Takeda out_drop_write: 248175c3f29dSDave Hansen mnt_drop_write(nd.path.mnt); 248275c3f29dSDave Hansen out_dput: 24831da177e4SLinus Torvalds dput(new_dentry); 24846902d925SDave Hansen out_unlock: 24854ac91378SJan Blunck mutex_unlock(&nd.path.dentry->d_inode->i_mutex); 24861da177e4SLinus Torvalds out_release: 24871d957f9bSJan Blunck path_put(&nd.path); 24882ad94ae6SAl Viro putname(to); 24891da177e4SLinus Torvalds out: 24902d8f3038SAl Viro path_put(&old_path); 24911da177e4SLinus Torvalds 24921da177e4SLinus Torvalds return error; 24931da177e4SLinus Torvalds } 24941da177e4SLinus Torvalds 24953480b257SHeiko Carstens SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) 24965590ff0dSUlrich Drepper { 2497c04030e1SUlrich Drepper return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); 24985590ff0dSUlrich Drepper } 24995590ff0dSUlrich Drepper 25001da177e4SLinus Torvalds /* 25011da177e4SLinus Torvalds * The worst of all namespace operations - renaming directory. "Perverted" 25021da177e4SLinus Torvalds * doesn't even start to describe it. Somebody in UCB had a heck of a trip... 25031da177e4SLinus Torvalds * Problems: 25041da177e4SLinus Torvalds * a) we can get into loop creation. Check is done in is_subdir(). 25051da177e4SLinus Torvalds * b) race potential - two innocent renames can create a loop together. 25061da177e4SLinus Torvalds * That's where 4.4 screws up. Current fix: serialization on 2507a11f3a05SArjan van de Ven * sb->s_vfs_rename_mutex. We might be more accurate, but that's another 25081da177e4SLinus Torvalds * story. 25091da177e4SLinus Torvalds * c) we have to lock _three_ objects - parents and victim (if it exists). 25101b1dcc1bSJes Sorensen * And that - after we got ->i_mutex on parents (until then we don't know 25111da177e4SLinus Torvalds * whether the target exists). Solution: try to be smart with locking 25121da177e4SLinus Torvalds * order for inodes. We rely on the fact that tree topology may change 2513a11f3a05SArjan van de Ven * only under ->s_vfs_rename_mutex _and_ that parent of the object we 25141da177e4SLinus Torvalds * move will be locked. Thus we can rank directories by the tree 25151da177e4SLinus Torvalds * (ancestors first) and rank all non-directories after them. 25161da177e4SLinus Torvalds * That works since everybody except rename does "lock parent, lookup, 2517a11f3a05SArjan van de Ven * lock child" and rename is under ->s_vfs_rename_mutex. 25181da177e4SLinus Torvalds * HOWEVER, it relies on the assumption that any object with ->lookup() 25191da177e4SLinus Torvalds * has no more than 1 dentry. If "hybrid" objects will ever appear, 25201da177e4SLinus Torvalds * we'd better make sure that there's no link(2) for them. 25211da177e4SLinus Torvalds * d) some filesystems don't support opened-but-unlinked directories, 25221da177e4SLinus Torvalds * either because of layout or because they are not ready to deal with 25231da177e4SLinus Torvalds * all cases correctly. The latter will be fixed (taking this sort of 25241da177e4SLinus Torvalds * stuff into VFS), but the former is not going away. Solution: the same 25251da177e4SLinus Torvalds * trick as in rmdir(). 25261da177e4SLinus Torvalds * e) conversion from fhandle to dentry may come in the wrong moment - when 25271b1dcc1bSJes Sorensen * we are removing the target. Solution: we will have to grab ->i_mutex 25281da177e4SLinus Torvalds * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on 2529c41b20e7SAdam Buchbinder * ->i_mutex on parents, which works but leads to some truly excessive 25301da177e4SLinus Torvalds * locking]. 25311da177e4SLinus Torvalds */ 253275c96f85SAdrian Bunk static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, 25331da177e4SLinus Torvalds struct inode *new_dir, struct dentry *new_dentry) 25341da177e4SLinus Torvalds { 25351da177e4SLinus Torvalds int error = 0; 25361da177e4SLinus Torvalds struct inode *target; 25371da177e4SLinus Torvalds 25381da177e4SLinus Torvalds /* 25391da177e4SLinus Torvalds * If we are going to change the parent - check write permissions, 25401da177e4SLinus Torvalds * we'll need to flip '..'. 25411da177e4SLinus Torvalds */ 25421da177e4SLinus Torvalds if (new_dir != old_dir) { 2543f419a2e3SAl Viro error = inode_permission(old_dentry->d_inode, MAY_WRITE); 25441da177e4SLinus Torvalds if (error) 25451da177e4SLinus Torvalds return error; 25461da177e4SLinus Torvalds } 25471da177e4SLinus Torvalds 25481da177e4SLinus Torvalds error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); 25491da177e4SLinus Torvalds if (error) 25501da177e4SLinus Torvalds return error; 25511da177e4SLinus Torvalds 25521da177e4SLinus Torvalds target = new_dentry->d_inode; 2553d83c49f3SAl Viro if (target) 25541b1dcc1bSJes Sorensen mutex_lock(&target->i_mutex); 25551da177e4SLinus Torvalds if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) 25561da177e4SLinus Torvalds error = -EBUSY; 2557d83c49f3SAl Viro else { 2558d83c49f3SAl Viro if (target) 2559d83c49f3SAl Viro dentry_unhash(new_dentry); 25601da177e4SLinus Torvalds error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); 2561d83c49f3SAl Viro } 25621da177e4SLinus Torvalds if (target) { 2563d83c49f3SAl Viro if (!error) { 25641da177e4SLinus Torvalds target->i_flags |= S_DEAD; 2565d83c49f3SAl Viro dont_mount(new_dentry); 2566d83c49f3SAl Viro } 25671b1dcc1bSJes Sorensen mutex_unlock(&target->i_mutex); 25681da177e4SLinus Torvalds if (d_unhashed(new_dentry)) 25691da177e4SLinus Torvalds d_rehash(new_dentry); 25701da177e4SLinus Torvalds dput(new_dentry); 25711da177e4SLinus Torvalds } 2572e31e14ecSStephen Smalley if (!error) 2573349457ccSMark Fasheh if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) 25741da177e4SLinus Torvalds d_move(old_dentry,new_dentry); 25751da177e4SLinus Torvalds return error; 25761da177e4SLinus Torvalds } 25771da177e4SLinus Torvalds 257875c96f85SAdrian Bunk static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, 25791da177e4SLinus Torvalds struct inode *new_dir, struct dentry *new_dentry) 25801da177e4SLinus Torvalds { 25811da177e4SLinus Torvalds struct inode *target; 25821da177e4SLinus Torvalds int error; 25831da177e4SLinus Torvalds 25841da177e4SLinus Torvalds error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); 25851da177e4SLinus Torvalds if (error) 25861da177e4SLinus Torvalds return error; 25871da177e4SLinus Torvalds 25881da177e4SLinus Torvalds dget(new_dentry); 25891da177e4SLinus Torvalds target = new_dentry->d_inode; 25901da177e4SLinus Torvalds if (target) 25911b1dcc1bSJes Sorensen mutex_lock(&target->i_mutex); 25921da177e4SLinus Torvalds if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) 25931da177e4SLinus Torvalds error = -EBUSY; 25941da177e4SLinus Torvalds else 25951da177e4SLinus Torvalds error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); 25961da177e4SLinus Torvalds if (!error) { 2597bec1052eSAl Viro if (target) 2598d83c49f3SAl Viro dont_mount(new_dentry); 2599349457ccSMark Fasheh if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) 26001da177e4SLinus Torvalds d_move(old_dentry, new_dentry); 26011da177e4SLinus Torvalds } 26021da177e4SLinus Torvalds if (target) 26031b1dcc1bSJes Sorensen mutex_unlock(&target->i_mutex); 26041da177e4SLinus Torvalds dput(new_dentry); 26051da177e4SLinus Torvalds return error; 26061da177e4SLinus Torvalds } 26071da177e4SLinus Torvalds 26081da177e4SLinus Torvalds int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, 26091da177e4SLinus Torvalds struct inode *new_dir, struct dentry *new_dentry) 26101da177e4SLinus Torvalds { 26111da177e4SLinus Torvalds int error; 26121da177e4SLinus Torvalds int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); 261359b0df21SEric Paris const unsigned char *old_name; 26141da177e4SLinus Torvalds 26151da177e4SLinus Torvalds if (old_dentry->d_inode == new_dentry->d_inode) 26161da177e4SLinus Torvalds return 0; 26171da177e4SLinus Torvalds 26181da177e4SLinus Torvalds error = may_delete(old_dir, old_dentry, is_dir); 26191da177e4SLinus Torvalds if (error) 26201da177e4SLinus Torvalds return error; 26211da177e4SLinus Torvalds 26221da177e4SLinus Torvalds if (!new_dentry->d_inode) 2623a95164d9SMiklos Szeredi error = may_create(new_dir, new_dentry); 26241da177e4SLinus Torvalds else 26251da177e4SLinus Torvalds error = may_delete(new_dir, new_dentry, is_dir); 26261da177e4SLinus Torvalds if (error) 26271da177e4SLinus Torvalds return error; 26281da177e4SLinus Torvalds 2629acfa4380SAl Viro if (!old_dir->i_op->rename) 26301da177e4SLinus Torvalds return -EPERM; 26311da177e4SLinus Torvalds 26320eeca283SRobert Love old_name = fsnotify_oldname_init(old_dentry->d_name.name); 26330eeca283SRobert Love 26341da177e4SLinus Torvalds if (is_dir) 26351da177e4SLinus Torvalds error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); 26361da177e4SLinus Torvalds else 26371da177e4SLinus Torvalds error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); 2638123df294SAl Viro if (!error) 2639123df294SAl Viro fsnotify_move(old_dir, new_dir, old_name, is_dir, 26405a190ae6SAl Viro new_dentry->d_inode, old_dentry); 26410eeca283SRobert Love fsnotify_oldname_free(old_name); 26420eeca283SRobert Love 26431da177e4SLinus Torvalds return error; 26441da177e4SLinus Torvalds } 26451da177e4SLinus Torvalds 26462e4d0924SHeiko Carstens SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, 26472e4d0924SHeiko Carstens int, newdfd, const char __user *, newname) 26481da177e4SLinus Torvalds { 26491da177e4SLinus Torvalds struct dentry *old_dir, *new_dir; 26501da177e4SLinus Torvalds struct dentry *old_dentry, *new_dentry; 26511da177e4SLinus Torvalds struct dentry *trap; 26521da177e4SLinus Torvalds struct nameidata oldnd, newnd; 26532ad94ae6SAl Viro char *from; 26542ad94ae6SAl Viro char *to; 26552ad94ae6SAl Viro int error; 26561da177e4SLinus Torvalds 26572ad94ae6SAl Viro error = user_path_parent(olddfd, oldname, &oldnd, &from); 26581da177e4SLinus Torvalds if (error) 26591da177e4SLinus Torvalds goto exit; 26601da177e4SLinus Torvalds 26612ad94ae6SAl Viro error = user_path_parent(newdfd, newname, &newnd, &to); 26621da177e4SLinus Torvalds if (error) 26631da177e4SLinus Torvalds goto exit1; 26641da177e4SLinus Torvalds 26651da177e4SLinus Torvalds error = -EXDEV; 26664ac91378SJan Blunck if (oldnd.path.mnt != newnd.path.mnt) 26671da177e4SLinus Torvalds goto exit2; 26681da177e4SLinus Torvalds 26694ac91378SJan Blunck old_dir = oldnd.path.dentry; 26701da177e4SLinus Torvalds error = -EBUSY; 26711da177e4SLinus Torvalds if (oldnd.last_type != LAST_NORM) 26721da177e4SLinus Torvalds goto exit2; 26731da177e4SLinus Torvalds 26744ac91378SJan Blunck new_dir = newnd.path.dentry; 26751da177e4SLinus Torvalds if (newnd.last_type != LAST_NORM) 26761da177e4SLinus Torvalds goto exit2; 26771da177e4SLinus Torvalds 26780612d9fbSOGAWA Hirofumi oldnd.flags &= ~LOOKUP_PARENT; 26790612d9fbSOGAWA Hirofumi newnd.flags &= ~LOOKUP_PARENT; 26804e9ed2f8SOGAWA Hirofumi newnd.flags |= LOOKUP_RENAME_TARGET; 26810612d9fbSOGAWA Hirofumi 26821da177e4SLinus Torvalds trap = lock_rename(new_dir, old_dir); 26831da177e4SLinus Torvalds 268449705b77SChristoph Hellwig old_dentry = lookup_hash(&oldnd); 26851da177e4SLinus Torvalds error = PTR_ERR(old_dentry); 26861da177e4SLinus Torvalds if (IS_ERR(old_dentry)) 26871da177e4SLinus Torvalds goto exit3; 26881da177e4SLinus Torvalds /* source must exist */ 26891da177e4SLinus Torvalds error = -ENOENT; 26901da177e4SLinus Torvalds if (!old_dentry->d_inode) 26911da177e4SLinus Torvalds goto exit4; 26921da177e4SLinus Torvalds /* unless the source is a directory trailing slashes give -ENOTDIR */ 26931da177e4SLinus Torvalds if (!S_ISDIR(old_dentry->d_inode->i_mode)) { 26941da177e4SLinus Torvalds error = -ENOTDIR; 26951da177e4SLinus Torvalds if (oldnd.last.name[oldnd.last.len]) 26961da177e4SLinus Torvalds goto exit4; 26971da177e4SLinus Torvalds if (newnd.last.name[newnd.last.len]) 26981da177e4SLinus Torvalds goto exit4; 26991da177e4SLinus Torvalds } 27001da177e4SLinus Torvalds /* source should not be ancestor of target */ 27011da177e4SLinus Torvalds error = -EINVAL; 27021da177e4SLinus Torvalds if (old_dentry == trap) 27031da177e4SLinus Torvalds goto exit4; 270449705b77SChristoph Hellwig new_dentry = lookup_hash(&newnd); 27051da177e4SLinus Torvalds error = PTR_ERR(new_dentry); 27061da177e4SLinus Torvalds if (IS_ERR(new_dentry)) 27071da177e4SLinus Torvalds goto exit4; 27081da177e4SLinus Torvalds /* target should not be an ancestor of source */ 27091da177e4SLinus Torvalds error = -ENOTEMPTY; 27101da177e4SLinus Torvalds if (new_dentry == trap) 27111da177e4SLinus Torvalds goto exit5; 27121da177e4SLinus Torvalds 27139079b1ebSDave Hansen error = mnt_want_write(oldnd.path.mnt); 27149079b1ebSDave Hansen if (error) 27159079b1ebSDave Hansen goto exit5; 2716be6d3e56SKentaro Takeda error = security_path_rename(&oldnd.path, old_dentry, 2717be6d3e56SKentaro Takeda &newnd.path, new_dentry); 2718be6d3e56SKentaro Takeda if (error) 2719be6d3e56SKentaro Takeda goto exit6; 27201da177e4SLinus Torvalds error = vfs_rename(old_dir->d_inode, old_dentry, 27211da177e4SLinus Torvalds new_dir->d_inode, new_dentry); 2722be6d3e56SKentaro Takeda exit6: 27239079b1ebSDave Hansen mnt_drop_write(oldnd.path.mnt); 27241da177e4SLinus Torvalds exit5: 27251da177e4SLinus Torvalds dput(new_dentry); 27261da177e4SLinus Torvalds exit4: 27271da177e4SLinus Torvalds dput(old_dentry); 27281da177e4SLinus Torvalds exit3: 27291da177e4SLinus Torvalds unlock_rename(new_dir, old_dir); 27301da177e4SLinus Torvalds exit2: 27311d957f9bSJan Blunck path_put(&newnd.path); 27322ad94ae6SAl Viro putname(to); 27331da177e4SLinus Torvalds exit1: 27341d957f9bSJan Blunck path_put(&oldnd.path); 27351da177e4SLinus Torvalds putname(from); 27362ad94ae6SAl Viro exit: 27371da177e4SLinus Torvalds return error; 27381da177e4SLinus Torvalds } 27391da177e4SLinus Torvalds 2740a26eab24SHeiko Carstens SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) 27415590ff0dSUlrich Drepper { 27425590ff0dSUlrich Drepper return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); 27435590ff0dSUlrich Drepper } 27445590ff0dSUlrich Drepper 27451da177e4SLinus Torvalds int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) 27461da177e4SLinus Torvalds { 27471da177e4SLinus Torvalds int len; 27481da177e4SLinus Torvalds 27491da177e4SLinus Torvalds len = PTR_ERR(link); 27501da177e4SLinus Torvalds if (IS_ERR(link)) 27511da177e4SLinus Torvalds goto out; 27521da177e4SLinus Torvalds 27531da177e4SLinus Torvalds len = strlen(link); 27541da177e4SLinus Torvalds if (len > (unsigned) buflen) 27551da177e4SLinus Torvalds len = buflen; 27561da177e4SLinus Torvalds if (copy_to_user(buffer, link, len)) 27571da177e4SLinus Torvalds len = -EFAULT; 27581da177e4SLinus Torvalds out: 27591da177e4SLinus Torvalds return len; 27601da177e4SLinus Torvalds } 27611da177e4SLinus Torvalds 27621da177e4SLinus Torvalds /* 27631da177e4SLinus Torvalds * A helper for ->readlink(). This should be used *ONLY* for symlinks that 27641da177e4SLinus Torvalds * have ->follow_link() touching nd only in nd_set_link(). Using (or not 27651da177e4SLinus Torvalds * using) it for any given inode is up to filesystem. 27661da177e4SLinus Torvalds */ 27671da177e4SLinus Torvalds int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) 27681da177e4SLinus Torvalds { 27691da177e4SLinus Torvalds struct nameidata nd; 2770cc314eefSLinus Torvalds void *cookie; 2771694a1764SMarcin Slusarz int res; 2772cc314eefSLinus Torvalds 27731da177e4SLinus Torvalds nd.depth = 0; 2774cc314eefSLinus Torvalds cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); 2775694a1764SMarcin Slusarz if (IS_ERR(cookie)) 2776694a1764SMarcin Slusarz return PTR_ERR(cookie); 2777694a1764SMarcin Slusarz 2778694a1764SMarcin Slusarz res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); 27791da177e4SLinus Torvalds if (dentry->d_inode->i_op->put_link) 2780cc314eefSLinus Torvalds dentry->d_inode->i_op->put_link(dentry, &nd, cookie); 2781694a1764SMarcin Slusarz return res; 27821da177e4SLinus Torvalds } 27831da177e4SLinus Torvalds 27841da177e4SLinus Torvalds int vfs_follow_link(struct nameidata *nd, const char *link) 27851da177e4SLinus Torvalds { 27861da177e4SLinus Torvalds return __vfs_follow_link(nd, link); 27871da177e4SLinus Torvalds } 27881da177e4SLinus Torvalds 27891da177e4SLinus Torvalds /* get the link contents into pagecache */ 27901da177e4SLinus Torvalds static char *page_getlink(struct dentry * dentry, struct page **ppage) 27911da177e4SLinus Torvalds { 2792ebd09abbSDuane Griffin char *kaddr; 27931da177e4SLinus Torvalds struct page *page; 27941da177e4SLinus Torvalds struct address_space *mapping = dentry->d_inode->i_mapping; 2795090d2b18SPekka Enberg page = read_mapping_page(mapping, 0, NULL); 27961da177e4SLinus Torvalds if (IS_ERR(page)) 27976fe6900eSNick Piggin return (char*)page; 27981da177e4SLinus Torvalds *ppage = page; 2799ebd09abbSDuane Griffin kaddr = kmap(page); 2800ebd09abbSDuane Griffin nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1); 2801ebd09abbSDuane Griffin return kaddr; 28021da177e4SLinus Torvalds } 28031da177e4SLinus Torvalds 28041da177e4SLinus Torvalds int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) 28051da177e4SLinus Torvalds { 28061da177e4SLinus Torvalds struct page *page = NULL; 28071da177e4SLinus Torvalds char *s = page_getlink(dentry, &page); 28081da177e4SLinus Torvalds int res = vfs_readlink(dentry,buffer,buflen,s); 28091da177e4SLinus Torvalds if (page) { 28101da177e4SLinus Torvalds kunmap(page); 28111da177e4SLinus Torvalds page_cache_release(page); 28121da177e4SLinus Torvalds } 28131da177e4SLinus Torvalds return res; 28141da177e4SLinus Torvalds } 28151da177e4SLinus Torvalds 2816cc314eefSLinus Torvalds void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) 28171da177e4SLinus Torvalds { 2818cc314eefSLinus Torvalds struct page *page = NULL; 28191da177e4SLinus Torvalds nd_set_link(nd, page_getlink(dentry, &page)); 2820cc314eefSLinus Torvalds return page; 28211da177e4SLinus Torvalds } 28221da177e4SLinus Torvalds 2823cc314eefSLinus Torvalds void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) 28241da177e4SLinus Torvalds { 2825cc314eefSLinus Torvalds struct page *page = cookie; 2826cc314eefSLinus Torvalds 2827cc314eefSLinus Torvalds if (page) { 28281da177e4SLinus Torvalds kunmap(page); 28291da177e4SLinus Torvalds page_cache_release(page); 28301da177e4SLinus Torvalds } 28311da177e4SLinus Torvalds } 28321da177e4SLinus Torvalds 283354566b2cSNick Piggin /* 283454566b2cSNick Piggin * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS 283554566b2cSNick Piggin */ 283654566b2cSNick Piggin int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) 28371da177e4SLinus Torvalds { 28381da177e4SLinus Torvalds struct address_space *mapping = inode->i_mapping; 28390adb25d2SKirill Korotaev struct page *page; 2840afddba49SNick Piggin void *fsdata; 2841beb497abSDmitriy Monakhov int err; 28421da177e4SLinus Torvalds char *kaddr; 284354566b2cSNick Piggin unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; 284454566b2cSNick Piggin if (nofs) 284554566b2cSNick Piggin flags |= AOP_FLAG_NOFS; 28461da177e4SLinus Torvalds 28477e53cac4SNeilBrown retry: 2848afddba49SNick Piggin err = pagecache_write_begin(NULL, mapping, 0, len-1, 284954566b2cSNick Piggin flags, &page, &fsdata); 28501da177e4SLinus Torvalds if (err) 2851afddba49SNick Piggin goto fail; 2852afddba49SNick Piggin 28531da177e4SLinus Torvalds kaddr = kmap_atomic(page, KM_USER0); 28541da177e4SLinus Torvalds memcpy(kaddr, symname, len-1); 28551da177e4SLinus Torvalds kunmap_atomic(kaddr, KM_USER0); 2856afddba49SNick Piggin 2857afddba49SNick Piggin err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, 2858afddba49SNick Piggin page, fsdata); 28591da177e4SLinus Torvalds if (err < 0) 28601da177e4SLinus Torvalds goto fail; 2861afddba49SNick Piggin if (err < len-1) 2862afddba49SNick Piggin goto retry; 2863afddba49SNick Piggin 28641da177e4SLinus Torvalds mark_inode_dirty(inode); 28651da177e4SLinus Torvalds return 0; 28661da177e4SLinus Torvalds fail: 28671da177e4SLinus Torvalds return err; 28681da177e4SLinus Torvalds } 28691da177e4SLinus Torvalds 28700adb25d2SKirill Korotaev int page_symlink(struct inode *inode, const char *symname, int len) 28710adb25d2SKirill Korotaev { 28720adb25d2SKirill Korotaev return __page_symlink(inode, symname, len, 287354566b2cSNick Piggin !(mapping_gfp_mask(inode->i_mapping) & __GFP_FS)); 28740adb25d2SKirill Korotaev } 28750adb25d2SKirill Korotaev 287692e1d5beSArjan van de Ven const struct inode_operations page_symlink_inode_operations = { 28771da177e4SLinus Torvalds .readlink = generic_readlink, 28781da177e4SLinus Torvalds .follow_link = page_follow_link_light, 28791da177e4SLinus Torvalds .put_link = page_put_link, 28801da177e4SLinus Torvalds }; 28811da177e4SLinus Torvalds 28822d8f3038SAl Viro EXPORT_SYMBOL(user_path_at); 28831da177e4SLinus Torvalds EXPORT_SYMBOL(follow_down); 28841da177e4SLinus Torvalds EXPORT_SYMBOL(follow_up); 28851da177e4SLinus Torvalds EXPORT_SYMBOL(get_write_access); /* binfmt_aout */ 28861da177e4SLinus Torvalds EXPORT_SYMBOL(getname); 28871da177e4SLinus Torvalds EXPORT_SYMBOL(lock_rename); 28881da177e4SLinus Torvalds EXPORT_SYMBOL(lookup_one_len); 28891da177e4SLinus Torvalds EXPORT_SYMBOL(page_follow_link_light); 28901da177e4SLinus Torvalds EXPORT_SYMBOL(page_put_link); 28911da177e4SLinus Torvalds EXPORT_SYMBOL(page_readlink); 28920adb25d2SKirill Korotaev EXPORT_SYMBOL(__page_symlink); 28931da177e4SLinus Torvalds EXPORT_SYMBOL(page_symlink); 28941da177e4SLinus Torvalds EXPORT_SYMBOL(page_symlink_inode_operations); 28951da177e4SLinus Torvalds EXPORT_SYMBOL(path_lookup); 2896d1811465SAl Viro EXPORT_SYMBOL(kern_path); 289716f18200SJosef 'Jeff' Sipek EXPORT_SYMBOL(vfs_path_lookup); 2898f419a2e3SAl Viro EXPORT_SYMBOL(inode_permission); 28998c744fb8SChristoph Hellwig EXPORT_SYMBOL(file_permission); 29001da177e4SLinus Torvalds EXPORT_SYMBOL(unlock_rename); 29011da177e4SLinus Torvalds EXPORT_SYMBOL(vfs_create); 29021da177e4SLinus Torvalds EXPORT_SYMBOL(vfs_follow_link); 29031da177e4SLinus Torvalds EXPORT_SYMBOL(vfs_link); 29041da177e4SLinus Torvalds EXPORT_SYMBOL(vfs_mkdir); 29051da177e4SLinus Torvalds EXPORT_SYMBOL(vfs_mknod); 29061da177e4SLinus Torvalds EXPORT_SYMBOL(generic_permission); 29071da177e4SLinus Torvalds EXPORT_SYMBOL(vfs_readlink); 29081da177e4SLinus Torvalds EXPORT_SYMBOL(vfs_rename); 29091da177e4SLinus Torvalds EXPORT_SYMBOL(vfs_rmdir); 29101da177e4SLinus Torvalds EXPORT_SYMBOL(vfs_symlink); 29111da177e4SLinus Torvalds EXPORT_SYMBOL(vfs_unlink); 29121da177e4SLinus Torvalds EXPORT_SYMBOL(dentry_unhash); 29131da177e4SLinus Torvalds EXPORT_SYMBOL(generic_readlink); 2914