1b2441318SGreg Kroah-Hartman // SPDX-License-Identifier: GPL-2.0 277b14db5SEric W. Biederman /* 377b14db5SEric W. Biederman * /proc/sys support 477b14db5SEric W. Biederman */ 51e0edd3fSAlexey Dobriyan #include <linux/init.h> 677b14db5SEric W. Biederman #include <linux/sysctl.h> 7f1ecf068SLucas De Marchi #include <linux/poll.h> 877b14db5SEric W. Biederman #include <linux/proc_fs.h> 987ebdc00SAndrew Morton #include <linux/printk.h> 1077b14db5SEric W. Biederman #include <linux/security.h> 1140401530SAl Viro #include <linux/sched.h> 125b825c3aSIngo Molnar #include <linux/cred.h> 1334286d66SNick Piggin #include <linux/namei.h> 1440401530SAl Viro #include <linux/mm.h> 154bd6a735SMatthew Wilcox (Oracle) #include <linux/uio.h> 161f87f0b5SEric W. Biederman #include <linux/module.h> 177b146cebSAndrey Ignatov #include <linux/bpf-cgroup.h> 183db978d4SVlastimil Babka #include <linux/mount.h> 1977b14db5SEric W. Biederman #include "internal.h" 2077b14db5SEric W. Biederman 21d72f71ebSAl Viro static const struct dentry_operations proc_sys_dentry_operations; 2277b14db5SEric W. Biederman static const struct file_operations proc_sys_file_operations; 2303a44825SJan Engelhardt static const struct inode_operations proc_sys_inode_operations; 249043476fSAl Viro static const struct file_operations proc_sys_dir_file_operations; 259043476fSAl Viro static const struct inode_operations proc_sys_dir_operations; 2677b14db5SEric W. Biederman 27eec4844fSMatteo Croce /* shared constants to be used in various sysctls */ 28eec4844fSMatteo Croce const int sysctl_vals[] = { 0, 1, INT_MAX }; 29eec4844fSMatteo Croce EXPORT_SYMBOL(sysctl_vals); 30eec4844fSMatteo Croce 31f9bd6733SEric W. Biederman /* Support for permanently empty directories */ 32f9bd6733SEric W. Biederman 33f9bd6733SEric W. Biederman struct ctl_table sysctl_mount_point[] = { 34f9bd6733SEric W. Biederman { } 35f9bd6733SEric W. Biederman }; 36f9bd6733SEric W. Biederman 37f9bd6733SEric W. Biederman static bool is_empty_dir(struct ctl_table_header *head) 38f9bd6733SEric W. Biederman { 39f9bd6733SEric W. Biederman return head->ctl_table[0].child == sysctl_mount_point; 40f9bd6733SEric W. Biederman } 41f9bd6733SEric W. Biederman 42f9bd6733SEric W. Biederman static void set_empty_dir(struct ctl_dir *dir) 43f9bd6733SEric W. Biederman { 44f9bd6733SEric W. Biederman dir->header.ctl_table[0].child = sysctl_mount_point; 45f9bd6733SEric W. Biederman } 46f9bd6733SEric W. Biederman 47f9bd6733SEric W. Biederman static void clear_empty_dir(struct ctl_dir *dir) 48f9bd6733SEric W. Biederman 49f9bd6733SEric W. Biederman { 50f9bd6733SEric W. Biederman dir->header.ctl_table[0].child = NULL; 51f9bd6733SEric W. Biederman } 52f9bd6733SEric W. Biederman 53f1ecf068SLucas De Marchi void proc_sys_poll_notify(struct ctl_table_poll *poll) 54f1ecf068SLucas De Marchi { 55f1ecf068SLucas De Marchi if (!poll) 56f1ecf068SLucas De Marchi return; 57f1ecf068SLucas De Marchi 58f1ecf068SLucas De Marchi atomic_inc(&poll->event); 59f1ecf068SLucas De Marchi wake_up_interruptible(&poll->wait); 60f1ecf068SLucas De Marchi } 61f1ecf068SLucas De Marchi 62a194558eSEric W. Biederman static struct ctl_table root_table[] = { 63a194558eSEric W. Biederman { 64a194558eSEric W. Biederman .procname = "", 657ec66d06SEric W. Biederman .mode = S_IFDIR|S_IRUGO|S_IXUGO, 66a194558eSEric W. Biederman }, 67a194558eSEric W. Biederman { } 68a194558eSEric W. Biederman }; 690e47c99dSEric W. Biederman static struct ctl_table_root sysctl_table_root = { 700e47c99dSEric W. Biederman .default_set.dir.header = { 711f87f0b5SEric W. Biederman {{.count = 1, 72938aaa4fSEric W. Biederman .nreg = 1, 73ac13ac6fSEric W. Biederman .ctl_table = root_table }}, 740e47c99dSEric W. Biederman .ctl_table_arg = root_table, 751f87f0b5SEric W. Biederman .root = &sysctl_table_root, 761f87f0b5SEric W. Biederman .set = &sysctl_table_root.default_set, 777ec66d06SEric W. Biederman }, 781f87f0b5SEric W. Biederman }; 791f87f0b5SEric W. Biederman 801f87f0b5SEric W. Biederman static DEFINE_SPINLOCK(sysctl_lock); 811f87f0b5SEric W. Biederman 827ec66d06SEric W. Biederman static void drop_sysctl_table(struct ctl_table_header *header); 830e47c99dSEric W. Biederman static int sysctl_follow_link(struct ctl_table_header **phead, 8413bcc6a2SEric W. Biederman struct ctl_table **pentry); 850e47c99dSEric W. Biederman static int insert_links(struct ctl_table_header *head); 860e47c99dSEric W. Biederman static void put_links(struct ctl_table_header *header); 877ec66d06SEric W. Biederman 886980128fSEric W. Biederman static void sysctl_print_dir(struct ctl_dir *dir) 896980128fSEric W. Biederman { 906980128fSEric W. Biederman if (dir->header.parent) 916980128fSEric W. Biederman sysctl_print_dir(dir->header.parent); 9287ebdc00SAndrew Morton pr_cont("%s/", dir->header.ctl_table[0].procname); 936980128fSEric W. Biederman } 946980128fSEric W. Biederman 95076c3eedSEric W. Biederman static int namecmp(const char *name1, int len1, const char *name2, int len2) 96076c3eedSEric W. Biederman { 97076c3eedSEric W. Biederman int minlen; 98076c3eedSEric W. Biederman int cmp; 99076c3eedSEric W. Biederman 100076c3eedSEric W. Biederman minlen = len1; 101076c3eedSEric W. Biederman if (minlen > len2) 102076c3eedSEric W. Biederman minlen = len2; 103076c3eedSEric W. Biederman 104076c3eedSEric W. Biederman cmp = memcmp(name1, name2, minlen); 105076c3eedSEric W. Biederman if (cmp == 0) 106076c3eedSEric W. Biederman cmp = len1 - len2; 107076c3eedSEric W. Biederman return cmp; 108076c3eedSEric W. Biederman } 109076c3eedSEric W. Biederman 11060f126d9SEric W. Biederman /* Called under sysctl_lock */ 111076c3eedSEric W. Biederman static struct ctl_table *find_entry(struct ctl_table_header **phead, 1120e47c99dSEric W. Biederman struct ctl_dir *dir, const char *name, int namelen) 113076c3eedSEric W. Biederman { 114076c3eedSEric W. Biederman struct ctl_table_header *head; 115076c3eedSEric W. Biederman struct ctl_table *entry; 116ac13ac6fSEric W. Biederman struct rb_node *node = dir->root.rb_node; 117076c3eedSEric W. Biederman 118ac13ac6fSEric W. Biederman while (node) 119ac13ac6fSEric W. Biederman { 120ac13ac6fSEric W. Biederman struct ctl_node *ctl_node; 121ac13ac6fSEric W. Biederman const char *procname; 122ac13ac6fSEric W. Biederman int cmp; 123ac13ac6fSEric W. Biederman 124ac13ac6fSEric W. Biederman ctl_node = rb_entry(node, struct ctl_node, node); 125ac13ac6fSEric W. Biederman head = ctl_node->header; 126ac13ac6fSEric W. Biederman entry = &head->ctl_table[ctl_node - head->node]; 127ac13ac6fSEric W. Biederman procname = entry->procname; 128ac13ac6fSEric W. Biederman 129ac13ac6fSEric W. Biederman cmp = namecmp(name, namelen, procname, strlen(procname)); 130ac13ac6fSEric W. Biederman if (cmp < 0) 131ac13ac6fSEric W. Biederman node = node->rb_left; 132ac13ac6fSEric W. Biederman else if (cmp > 0) 133ac13ac6fSEric W. Biederman node = node->rb_right; 134ac13ac6fSEric W. Biederman else { 135076c3eedSEric W. Biederman *phead = head; 136076c3eedSEric W. Biederman return entry; 137076c3eedSEric W. Biederman } 138076c3eedSEric W. Biederman } 139076c3eedSEric W. Biederman return NULL; 140076c3eedSEric W. Biederman } 141076c3eedSEric W. Biederman 142ac13ac6fSEric W. Biederman static int insert_entry(struct ctl_table_header *head, struct ctl_table *entry) 143ac13ac6fSEric W. Biederman { 144ac13ac6fSEric W. Biederman struct rb_node *node = &head->node[entry - head->ctl_table].node; 145ac13ac6fSEric W. Biederman struct rb_node **p = &head->parent->root.rb_node; 146ac13ac6fSEric W. Biederman struct rb_node *parent = NULL; 147ac13ac6fSEric W. Biederman const char *name = entry->procname; 148ac13ac6fSEric W. Biederman int namelen = strlen(name); 149ac13ac6fSEric W. Biederman 150ac13ac6fSEric W. Biederman while (*p) { 151ac13ac6fSEric W. Biederman struct ctl_table_header *parent_head; 152ac13ac6fSEric W. Biederman struct ctl_table *parent_entry; 153ac13ac6fSEric W. Biederman struct ctl_node *parent_node; 154ac13ac6fSEric W. Biederman const char *parent_name; 155ac13ac6fSEric W. Biederman int cmp; 156ac13ac6fSEric W. Biederman 157ac13ac6fSEric W. Biederman parent = *p; 158ac13ac6fSEric W. Biederman parent_node = rb_entry(parent, struct ctl_node, node); 159ac13ac6fSEric W. Biederman parent_head = parent_node->header; 160ac13ac6fSEric W. Biederman parent_entry = &parent_head->ctl_table[parent_node - parent_head->node]; 161ac13ac6fSEric W. Biederman parent_name = parent_entry->procname; 162ac13ac6fSEric W. Biederman 163ac13ac6fSEric W. Biederman cmp = namecmp(name, namelen, parent_name, strlen(parent_name)); 164ac13ac6fSEric W. Biederman if (cmp < 0) 165ac13ac6fSEric W. Biederman p = &(*p)->rb_left; 166ac13ac6fSEric W. Biederman else if (cmp > 0) 167ac13ac6fSEric W. Biederman p = &(*p)->rb_right; 168ac13ac6fSEric W. Biederman else { 16987ebdc00SAndrew Morton pr_err("sysctl duplicate entry: "); 170ac13ac6fSEric W. Biederman sysctl_print_dir(head->parent); 17187ebdc00SAndrew Morton pr_cont("/%s\n", entry->procname); 172ac13ac6fSEric W. Biederman return -EEXIST; 173ac13ac6fSEric W. Biederman } 174ac13ac6fSEric W. Biederman } 175ac13ac6fSEric W. Biederman 176ac13ac6fSEric W. Biederman rb_link_node(node, parent, p); 177ea5272f5SMichel Lespinasse rb_insert_color(node, &head->parent->root); 178ac13ac6fSEric W. Biederman return 0; 179ac13ac6fSEric W. Biederman } 180ac13ac6fSEric W. Biederman 181ac13ac6fSEric W. Biederman static void erase_entry(struct ctl_table_header *head, struct ctl_table *entry) 182ac13ac6fSEric W. Biederman { 183ac13ac6fSEric W. Biederman struct rb_node *node = &head->node[entry - head->ctl_table].node; 184ac13ac6fSEric W. Biederman 185ac13ac6fSEric W. Biederman rb_erase(node, &head->parent->root); 186ac13ac6fSEric W. Biederman } 187ac13ac6fSEric W. Biederman 188e0d04529SEric W. Biederman static void init_header(struct ctl_table_header *head, 189e0d04529SEric W. Biederman struct ctl_table_root *root, struct ctl_table_set *set, 190ac13ac6fSEric W. Biederman struct ctl_node *node, struct ctl_table *table) 191e0d04529SEric W. Biederman { 1927ec66d06SEric W. Biederman head->ctl_table = table; 193e0d04529SEric W. Biederman head->ctl_table_arg = table; 194e0d04529SEric W. Biederman head->used = 0; 195e0d04529SEric W. Biederman head->count = 1; 196e0d04529SEric W. Biederman head->nreg = 1; 197e0d04529SEric W. Biederman head->unregistering = NULL; 198e0d04529SEric W. Biederman head->root = root; 199e0d04529SEric W. Biederman head->set = set; 200e0d04529SEric W. Biederman head->parent = NULL; 201ac13ac6fSEric W. Biederman head->node = node; 2022fd1d2c4SEric W. Biederman INIT_HLIST_HEAD(&head->inodes); 203ac13ac6fSEric W. Biederman if (node) { 204ac13ac6fSEric W. Biederman struct ctl_table *entry; 2054c199a93SMichel Lespinasse for (entry = table; entry->procname; entry++, node++) 206ac13ac6fSEric W. Biederman node->header = head; 207ac13ac6fSEric W. Biederman } 208ac13ac6fSEric W. Biederman } 209e0d04529SEric W. Biederman 2108425d6aaSEric W. Biederman static void erase_header(struct ctl_table_header *head) 2118425d6aaSEric W. Biederman { 212ac13ac6fSEric W. Biederman struct ctl_table *entry; 213ac13ac6fSEric W. Biederman for (entry = head->ctl_table; entry->procname; entry++) 214ac13ac6fSEric W. Biederman erase_entry(head, entry); 2158425d6aaSEric W. Biederman } 2168425d6aaSEric W. Biederman 2170e47c99dSEric W. Biederman static int insert_header(struct ctl_dir *dir, struct ctl_table_header *header) 2188425d6aaSEric W. Biederman { 219ac13ac6fSEric W. Biederman struct ctl_table *entry; 2200e47c99dSEric W. Biederman int err; 2210e47c99dSEric W. Biederman 222f9bd6733SEric W. Biederman /* Is this a permanently empty directory? */ 223f9bd6733SEric W. Biederman if (is_empty_dir(&dir->header)) 224f9bd6733SEric W. Biederman return -EROFS; 225f9bd6733SEric W. Biederman 226f9bd6733SEric W. Biederman /* Am I creating a permanently empty directory? */ 227f9bd6733SEric W. Biederman if (header->ctl_table == sysctl_mount_point) { 228f9bd6733SEric W. Biederman if (!RB_EMPTY_ROOT(&dir->root)) 229f9bd6733SEric W. Biederman return -EINVAL; 230f9bd6733SEric W. Biederman set_empty_dir(dir); 231f9bd6733SEric W. Biederman } 232f9bd6733SEric W. Biederman 2330e47c99dSEric W. Biederman dir->header.nreg++; 2347ec66d06SEric W. Biederman header->parent = dir; 2350e47c99dSEric W. Biederman err = insert_links(header); 2360e47c99dSEric W. Biederman if (err) 2370e47c99dSEric W. Biederman goto fail_links; 238ac13ac6fSEric W. Biederman for (entry = header->ctl_table; entry->procname; entry++) { 239ac13ac6fSEric W. Biederman err = insert_entry(header, entry); 240ac13ac6fSEric W. Biederman if (err) 241ac13ac6fSEric W. Biederman goto fail; 242ac13ac6fSEric W. Biederman } 2430e47c99dSEric W. Biederman return 0; 244ac13ac6fSEric W. Biederman fail: 245ac13ac6fSEric W. Biederman erase_header(header); 246ac13ac6fSEric W. Biederman put_links(header); 2470e47c99dSEric W. Biederman fail_links: 248f9bd6733SEric W. Biederman if (header->ctl_table == sysctl_mount_point) 249f9bd6733SEric W. Biederman clear_empty_dir(dir); 2500e47c99dSEric W. Biederman header->parent = NULL; 2510e47c99dSEric W. Biederman drop_sysctl_table(&dir->header); 2520e47c99dSEric W. Biederman return err; 2538425d6aaSEric W. Biederman } 2548425d6aaSEric W. Biederman 2551f87f0b5SEric W. Biederman /* called under sysctl_lock */ 2561f87f0b5SEric W. Biederman static int use_table(struct ctl_table_header *p) 2571f87f0b5SEric W. Biederman { 2581f87f0b5SEric W. Biederman if (unlikely(p->unregistering)) 2591f87f0b5SEric W. Biederman return 0; 2601f87f0b5SEric W. Biederman p->used++; 2611f87f0b5SEric W. Biederman return 1; 2621f87f0b5SEric W. Biederman } 2631f87f0b5SEric W. Biederman 2641f87f0b5SEric W. Biederman /* called under sysctl_lock */ 2651f87f0b5SEric W. Biederman static void unuse_table(struct ctl_table_header *p) 2661f87f0b5SEric W. Biederman { 2671f87f0b5SEric W. Biederman if (!--p->used) 2681f87f0b5SEric W. Biederman if (unlikely(p->unregistering)) 2691f87f0b5SEric W. Biederman complete(p->unregistering); 2701f87f0b5SEric W. Biederman } 2711f87f0b5SEric W. Biederman 272f90f3cafSEric W. Biederman static void proc_sys_invalidate_dcache(struct ctl_table_header *head) 273d6cffbbeSKonstantin Khlebnikov { 274f90f3cafSEric W. Biederman proc_invalidate_siblings_dcache(&head->inodes, &sysctl_lock); 275d6cffbbeSKonstantin Khlebnikov } 276d6cffbbeSKonstantin Khlebnikov 2771f87f0b5SEric W. Biederman /* called under sysctl_lock, will reacquire if has to wait */ 2781f87f0b5SEric W. Biederman static void start_unregistering(struct ctl_table_header *p) 2791f87f0b5SEric W. Biederman { 2801f87f0b5SEric W. Biederman /* 2811f87f0b5SEric W. Biederman * if p->used is 0, nobody will ever touch that entry again; 2821f87f0b5SEric W. Biederman * we'll eliminate all paths to it before dropping sysctl_lock 2831f87f0b5SEric W. Biederman */ 2841f87f0b5SEric W. Biederman if (unlikely(p->used)) { 2851f87f0b5SEric W. Biederman struct completion wait; 2861f87f0b5SEric W. Biederman init_completion(&wait); 2871f87f0b5SEric W. Biederman p->unregistering = &wait; 2881f87f0b5SEric W. Biederman spin_unlock(&sysctl_lock); 2891f87f0b5SEric W. Biederman wait_for_completion(&wait); 2901f87f0b5SEric W. Biederman } else { 2911f87f0b5SEric W. Biederman /* anything non-NULL; we'll never dereference it */ 2921f87f0b5SEric W. Biederman p->unregistering = ERR_PTR(-EINVAL); 293ace0c791SEric W. Biederman spin_unlock(&sysctl_lock); 2941f87f0b5SEric W. Biederman } 2951f87f0b5SEric W. Biederman /* 296f90f3cafSEric W. Biederman * Invalidate dentries for unregistered sysctls: namespaced sysctls 297d6cffbbeSKonstantin Khlebnikov * can have duplicate names and contaminate dcache very badly. 298d6cffbbeSKonstantin Khlebnikov */ 299f90f3cafSEric W. Biederman proc_sys_invalidate_dcache(p); 300d6cffbbeSKonstantin Khlebnikov /* 3011f87f0b5SEric W. Biederman * do not remove from the list until nobody holds it; walking the 3021f87f0b5SEric W. Biederman * list in do_sysctl() relies on that. 3031f87f0b5SEric W. Biederman */ 304ace0c791SEric W. Biederman spin_lock(&sysctl_lock); 3058425d6aaSEric W. Biederman erase_header(p); 3061f87f0b5SEric W. Biederman } 3071f87f0b5SEric W. Biederman 3081f87f0b5SEric W. Biederman static struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head) 3091f87f0b5SEric W. Biederman { 310ab4a1f24SPrasad Joshi BUG_ON(!head); 3111f87f0b5SEric W. Biederman spin_lock(&sysctl_lock); 3121f87f0b5SEric W. Biederman if (!use_table(head)) 3131f87f0b5SEric W. Biederman head = ERR_PTR(-ENOENT); 3141f87f0b5SEric W. Biederman spin_unlock(&sysctl_lock); 3151f87f0b5SEric W. Biederman return head; 3161f87f0b5SEric W. Biederman } 3171f87f0b5SEric W. Biederman 3181f87f0b5SEric W. Biederman static void sysctl_head_finish(struct ctl_table_header *head) 3191f87f0b5SEric W. Biederman { 3201f87f0b5SEric W. Biederman if (!head) 3211f87f0b5SEric W. Biederman return; 3221f87f0b5SEric W. Biederman spin_lock(&sysctl_lock); 3231f87f0b5SEric W. Biederman unuse_table(head); 3241f87f0b5SEric W. Biederman spin_unlock(&sysctl_lock); 3251f87f0b5SEric W. Biederman } 3261f87f0b5SEric W. Biederman 3271f87f0b5SEric W. Biederman static struct ctl_table_set * 32813bcc6a2SEric W. Biederman lookup_header_set(struct ctl_table_root *root) 3291f87f0b5SEric W. Biederman { 3301f87f0b5SEric W. Biederman struct ctl_table_set *set = &root->default_set; 3311f87f0b5SEric W. Biederman if (root->lookup) 33213bcc6a2SEric W. Biederman set = root->lookup(root); 3331f87f0b5SEric W. Biederman return set; 3341f87f0b5SEric W. Biederman } 3351f87f0b5SEric W. Biederman 336076c3eedSEric W. Biederman static struct ctl_table *lookup_entry(struct ctl_table_header **phead, 3377ec66d06SEric W. Biederman struct ctl_dir *dir, 338076c3eedSEric W. Biederman const char *name, int namelen) 339076c3eedSEric W. Biederman { 340076c3eedSEric W. Biederman struct ctl_table_header *head; 341076c3eedSEric W. Biederman struct ctl_table *entry; 342076c3eedSEric W. Biederman 343076c3eedSEric W. Biederman spin_lock(&sysctl_lock); 3440e47c99dSEric W. Biederman entry = find_entry(&head, dir, name, namelen); 345076c3eedSEric W. Biederman if (entry && use_table(head)) 346076c3eedSEric W. Biederman *phead = head; 347076c3eedSEric W. Biederman else 348076c3eedSEric W. Biederman entry = NULL; 349076c3eedSEric W. Biederman spin_unlock(&sysctl_lock); 350076c3eedSEric W. Biederman return entry; 351076c3eedSEric W. Biederman } 352076c3eedSEric W. Biederman 353ac13ac6fSEric W. Biederman static struct ctl_node *first_usable_entry(struct rb_node *node) 3541f87f0b5SEric W. Biederman { 355ac13ac6fSEric W. Biederman struct ctl_node *ctl_node; 3561f87f0b5SEric W. Biederman 357ac13ac6fSEric W. Biederman for (;node; node = rb_next(node)) { 358ac13ac6fSEric W. Biederman ctl_node = rb_entry(node, struct ctl_node, node); 359ac13ac6fSEric W. Biederman if (use_table(ctl_node->header)) 360ac13ac6fSEric W. Biederman return ctl_node; 3611f87f0b5SEric W. Biederman } 3621f87f0b5SEric W. Biederman return NULL; 3631f87f0b5SEric W. Biederman } 3641f87f0b5SEric W. Biederman 3657ec66d06SEric W. Biederman static void first_entry(struct ctl_dir *dir, 3666a75ce16SEric W. Biederman struct ctl_table_header **phead, struct ctl_table **pentry) 3671f87f0b5SEric W. Biederman { 368ac13ac6fSEric W. Biederman struct ctl_table_header *head = NULL; 3697ec66d06SEric W. Biederman struct ctl_table *entry = NULL; 370ac13ac6fSEric W. Biederman struct ctl_node *ctl_node; 3716a75ce16SEric W. Biederman 3726a75ce16SEric W. Biederman spin_lock(&sysctl_lock); 373ac13ac6fSEric W. Biederman ctl_node = first_usable_entry(rb_first(&dir->root)); 3746a75ce16SEric W. Biederman spin_unlock(&sysctl_lock); 375ac13ac6fSEric W. Biederman if (ctl_node) { 376ac13ac6fSEric W. Biederman head = ctl_node->header; 377ac13ac6fSEric W. Biederman entry = &head->ctl_table[ctl_node - head->node]; 378ac13ac6fSEric W. Biederman } 3796a75ce16SEric W. Biederman *phead = head; 3806a75ce16SEric W. Biederman *pentry = entry; 3816a75ce16SEric W. Biederman } 3826a75ce16SEric W. Biederman 3837ec66d06SEric W. Biederman static void next_entry(struct ctl_table_header **phead, struct ctl_table **pentry) 3846a75ce16SEric W. Biederman { 3856a75ce16SEric W. Biederman struct ctl_table_header *head = *phead; 3866a75ce16SEric W. Biederman struct ctl_table *entry = *pentry; 387ac13ac6fSEric W. Biederman struct ctl_node *ctl_node = &head->node[entry - head->ctl_table]; 3886a75ce16SEric W. Biederman 3896a75ce16SEric W. Biederman spin_lock(&sysctl_lock); 3906a75ce16SEric W. Biederman unuse_table(head); 391ac13ac6fSEric W. Biederman 392ac13ac6fSEric W. Biederman ctl_node = first_usable_entry(rb_next(&ctl_node->node)); 3936a75ce16SEric W. Biederman spin_unlock(&sysctl_lock); 394ac13ac6fSEric W. Biederman head = NULL; 395ac13ac6fSEric W. Biederman if (ctl_node) { 396ac13ac6fSEric W. Biederman head = ctl_node->header; 397ac13ac6fSEric W. Biederman entry = &head->ctl_table[ctl_node - head->node]; 3986a75ce16SEric W. Biederman } 3996a75ce16SEric W. Biederman *phead = head; 4006a75ce16SEric W. Biederman *pentry = entry; 4011f87f0b5SEric W. Biederman } 4021f87f0b5SEric W. Biederman 4031f87f0b5SEric W. Biederman /* 4041f87f0b5SEric W. Biederman * sysctl_perm does NOT grant the superuser all rights automatically, because 4051f87f0b5SEric W. Biederman * some sysctl variables are readonly even to root. 4061f87f0b5SEric W. Biederman */ 4071f87f0b5SEric W. Biederman 4081f87f0b5SEric W. Biederman static int test_perm(int mode, int op) 4091f87f0b5SEric W. Biederman { 410091bd3eaSEric W. Biederman if (uid_eq(current_euid(), GLOBAL_ROOT_UID)) 4111f87f0b5SEric W. Biederman mode >>= 6; 412091bd3eaSEric W. Biederman else if (in_egroup_p(GLOBAL_ROOT_GID)) 4131f87f0b5SEric W. Biederman mode >>= 3; 4141f87f0b5SEric W. Biederman if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0) 4151f87f0b5SEric W. Biederman return 0; 4161f87f0b5SEric W. Biederman return -EACCES; 4171f87f0b5SEric W. Biederman } 4181f87f0b5SEric W. Biederman 41973f7ef43SEric W. Biederman static int sysctl_perm(struct ctl_table_header *head, struct ctl_table *table, int op) 4201f87f0b5SEric W. Biederman { 42173f7ef43SEric W. Biederman struct ctl_table_root *root = head->root; 4221f87f0b5SEric W. Biederman int mode; 4231f87f0b5SEric W. Biederman 4241f87f0b5SEric W. Biederman if (root->permissions) 42573f7ef43SEric W. Biederman mode = root->permissions(head, table); 4261f87f0b5SEric W. Biederman else 4271f87f0b5SEric W. Biederman mode = table->mode; 4281f87f0b5SEric W. Biederman 4291f87f0b5SEric W. Biederman return test_perm(mode, op); 4301f87f0b5SEric W. Biederman } 4311f87f0b5SEric W. Biederman 4329043476fSAl Viro static struct inode *proc_sys_make_inode(struct super_block *sb, 4339043476fSAl Viro struct ctl_table_header *head, struct ctl_table *table) 43477b14db5SEric W. Biederman { 435e79c6a4fSDmitry Torokhov struct ctl_table_root *root = head->root; 43677b14db5SEric W. Biederman struct inode *inode; 4379043476fSAl Viro struct proc_inode *ei; 43877b14db5SEric W. Biederman 4399043476fSAl Viro inode = new_inode(sb); 44077b14db5SEric W. Biederman if (!inode) 441ea5751ccSIvan Delalande return ERR_PTR(-ENOMEM); 44277b14db5SEric W. Biederman 44385fe4025SChristoph Hellwig inode->i_ino = get_next_ino(); 44485fe4025SChristoph Hellwig 44577b14db5SEric W. Biederman ei = PROC_I(inode); 4469043476fSAl Viro 447d6cffbbeSKonstantin Khlebnikov spin_lock(&sysctl_lock); 448ace0c791SEric W. Biederman if (unlikely(head->unregistering)) { 449ace0c791SEric W. Biederman spin_unlock(&sysctl_lock); 450ace0c791SEric W. Biederman iput(inode); 451ea5751ccSIvan Delalande return ERR_PTR(-ENOENT); 452ace0c791SEric W. Biederman } 453ace0c791SEric W. Biederman ei->sysctl = head; 454ace0c791SEric W. Biederman ei->sysctl_entry = table; 4550afa5ca8SEric W. Biederman hlist_add_head_rcu(&ei->sibling_inodes, &head->inodes); 456d6cffbbeSKonstantin Khlebnikov head->count++; 457d6cffbbeSKonstantin Khlebnikov spin_unlock(&sysctl_lock); 458d6cffbbeSKonstantin Khlebnikov 459078cd827SDeepa Dinamani inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode); 4609043476fSAl Viro inode->i_mode = table->mode; 4617ec66d06SEric W. Biederman if (!S_ISDIR(table->mode)) { 4629043476fSAl Viro inode->i_mode |= S_IFREG; 46377b14db5SEric W. Biederman inode->i_op = &proc_sys_inode_operations; 46477b14db5SEric W. Biederman inode->i_fop = &proc_sys_file_operations; 4659043476fSAl Viro } else { 4669043476fSAl Viro inode->i_mode |= S_IFDIR; 4679043476fSAl Viro inode->i_op = &proc_sys_dir_operations; 4689043476fSAl Viro inode->i_fop = &proc_sys_dir_file_operations; 469f9bd6733SEric W. Biederman if (is_empty_dir(head)) 470f9bd6733SEric W. Biederman make_empty_dir_inode(inode); 4719043476fSAl Viro } 472e79c6a4fSDmitry Torokhov 473e79c6a4fSDmitry Torokhov if (root->set_ownership) 474e79c6a4fSDmitry Torokhov root->set_ownership(head, table, &inode->i_uid, &inode->i_gid); 4755ec27ec7SRadoslaw Burny else { 4765ec27ec7SRadoslaw Burny inode->i_uid = GLOBAL_ROOT_UID; 4775ec27ec7SRadoslaw Burny inode->i_gid = GLOBAL_ROOT_GID; 4785ec27ec7SRadoslaw Burny } 479e79c6a4fSDmitry Torokhov 48077b14db5SEric W. Biederman return inode; 48177b14db5SEric W. Biederman } 48277b14db5SEric W. Biederman 483d6cffbbeSKonstantin Khlebnikov void proc_sys_evict_inode(struct inode *inode, struct ctl_table_header *head) 484d6cffbbeSKonstantin Khlebnikov { 485d6cffbbeSKonstantin Khlebnikov spin_lock(&sysctl_lock); 4860afa5ca8SEric W. Biederman hlist_del_init_rcu(&PROC_I(inode)->sibling_inodes); 487d6cffbbeSKonstantin Khlebnikov if (!--head->count) 488d6cffbbeSKonstantin Khlebnikov kfree_rcu(head, rcu); 489d6cffbbeSKonstantin Khlebnikov spin_unlock(&sysctl_lock); 490d6cffbbeSKonstantin Khlebnikov } 491d6cffbbeSKonstantin Khlebnikov 49281324364SAdrian Bunk static struct ctl_table_header *grab_header(struct inode *inode) 49377b14db5SEric W. Biederman { 4943cc3e046SEric W. Biederman struct ctl_table_header *head = PROC_I(inode)->sysctl; 4953cc3e046SEric W. Biederman if (!head) 4960e47c99dSEric W. Biederman head = &sysctl_table_root.default_set.dir.header; 4973cc3e046SEric W. Biederman return sysctl_head_grab(head); 49877b14db5SEric W. Biederman } 49977b14db5SEric W. Biederman 50077b14db5SEric W. Biederman static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry, 50100cd8dd3SAl Viro unsigned int flags) 50277b14db5SEric W. Biederman { 5039043476fSAl Viro struct ctl_table_header *head = grab_header(dir); 5049043476fSAl Viro struct ctl_table_header *h = NULL; 505dc12e909SAl Viro const struct qstr *name = &dentry->d_name; 5069043476fSAl Viro struct ctl_table *p; 50777b14db5SEric W. Biederman struct inode *inode; 5089043476fSAl Viro struct dentry *err = ERR_PTR(-ENOENT); 5097ec66d06SEric W. Biederman struct ctl_dir *ctl_dir; 5100e47c99dSEric W. Biederman int ret; 51177b14db5SEric W. Biederman 5129043476fSAl Viro if (IS_ERR(head)) 5139043476fSAl Viro return ERR_CAST(head); 5149043476fSAl Viro 5157ec66d06SEric W. Biederman ctl_dir = container_of(head, struct ctl_dir, header); 5169043476fSAl Viro 5177ec66d06SEric W. Biederman p = lookup_entry(&h, ctl_dir, name->name, name->len); 5189043476fSAl Viro if (!p) 51977b14db5SEric W. Biederman goto out; 52077b14db5SEric W. Biederman 5214e757320SEric W. Biederman if (S_ISLNK(p->mode)) { 52213bcc6a2SEric W. Biederman ret = sysctl_follow_link(&h, &p); 5230e47c99dSEric W. Biederman err = ERR_PTR(ret); 5240e47c99dSEric W. Biederman if (ret) 5250e47c99dSEric W. Biederman goto out; 5264e757320SEric W. Biederman } 5270e47c99dSEric W. Biederman 5289043476fSAl Viro inode = proc_sys_make_inode(dir->i_sb, h ? h : head, p); 529ea5751ccSIvan Delalande if (IS_ERR(inode)) { 530ea5751ccSIvan Delalande err = ERR_CAST(inode); 53177b14db5SEric W. Biederman goto out; 532ea5751ccSIvan Delalande } 53377b14db5SEric W. Biederman 534fb045adbSNick Piggin d_set_d_op(dentry, &proc_sys_dentry_operations); 535888e2b03SAl Viro err = d_splice_alias(inode, dentry); 53677b14db5SEric W. Biederman 53777b14db5SEric W. Biederman out: 5386bf61045SFrancesco Ruggeri if (h) 5396bf61045SFrancesco Ruggeri sysctl_head_finish(h); 54077b14db5SEric W. Biederman sysctl_head_finish(head); 54177b14db5SEric W. Biederman return err; 54277b14db5SEric W. Biederman } 54377b14db5SEric W. Biederman 5444bd6a735SMatthew Wilcox (Oracle) static ssize_t proc_sys_call_handler(struct kiocb *iocb, struct iov_iter *iter, 5454bd6a735SMatthew Wilcox (Oracle) int write) 54677b14db5SEric W. Biederman { 5474bd6a735SMatthew Wilcox (Oracle) struct inode *inode = file_inode(iocb->ki_filp); 5489043476fSAl Viro struct ctl_table_header *head = grab_header(inode); 5499043476fSAl Viro struct ctl_table *table = PROC_I(inode)->sysctl_entry; 5504bd6a735SMatthew Wilcox (Oracle) size_t count = iov_iter_count(iter); 5514bd6a735SMatthew Wilcox (Oracle) char *kbuf; 5522a2da53bSDavid Howells ssize_t error; 55377b14db5SEric W. Biederman 5549043476fSAl Viro if (IS_ERR(head)) 5559043476fSAl Viro return PTR_ERR(head); 55677b14db5SEric W. Biederman 55777b14db5SEric W. Biederman /* 55877b14db5SEric W. Biederman * At this point we know that the sysctl was not unregistered 55977b14db5SEric W. Biederman * and won't be until we finish. 56077b14db5SEric W. Biederman */ 56177b14db5SEric W. Biederman error = -EPERM; 56273f7ef43SEric W. Biederman if (sysctl_perm(head, table, write ? MAY_WRITE : MAY_READ)) 56377b14db5SEric W. Biederman goto out; 56477b14db5SEric W. Biederman 5659043476fSAl Viro /* if that can happen at all, it should be -EINVAL, not -EISDIR */ 5669043476fSAl Viro error = -EINVAL; 5679043476fSAl Viro if (!table->proc_handler) 5689043476fSAl Viro goto out; 5699043476fSAl Viro 570ef9d965bSChristoph Hellwig /* don't even try if the size is too large */ 571d4d80e69SMatthew Wilcox (Oracle) error = -ENOMEM; 572d4d80e69SMatthew Wilcox (Oracle) if (count >= KMALLOC_MAX_SIZE) 573d4d80e69SMatthew Wilcox (Oracle) goto out; 57445089437SJosef Bacik kbuf = kvzalloc(count + 1, GFP_KERNEL); 57532927393SChristoph Hellwig if (!kbuf) 57632927393SChristoph Hellwig goto out; 5774bd6a735SMatthew Wilcox (Oracle) 5784bd6a735SMatthew Wilcox (Oracle) if (write) { 5794bd6a735SMatthew Wilcox (Oracle) error = -EFAULT; 5804bd6a735SMatthew Wilcox (Oracle) if (!copy_from_iter_full(kbuf, count, iter)) 5814bd6a735SMatthew Wilcox (Oracle) goto out_free_buf; 5824bd6a735SMatthew Wilcox (Oracle) kbuf[count] = '\0'; 5834e63acdfSAndrey Ignatov } 5844e63acdfSAndrey Ignatov 58532927393SChristoph Hellwig error = BPF_CGROUP_RUN_PROG_SYSCTL(head, table, write, &kbuf, &count, 5864bd6a735SMatthew Wilcox (Oracle) &iocb->ki_pos); 58732927393SChristoph Hellwig if (error) 58832927393SChristoph Hellwig goto out_free_buf; 58932927393SChristoph Hellwig 59032927393SChristoph Hellwig /* careful: calling conventions are nasty here */ 5914bd6a735SMatthew Wilcox (Oracle) error = table->proc_handler(table, write, kbuf, &count, &iocb->ki_pos); 59232927393SChristoph Hellwig if (error) 59332927393SChristoph Hellwig goto out_free_buf; 59432927393SChristoph Hellwig 59532927393SChristoph Hellwig if (!write) { 59632927393SChristoph Hellwig error = -EFAULT; 5974bd6a735SMatthew Wilcox (Oracle) if (copy_to_iter(kbuf, count, iter) < count) 59832927393SChristoph Hellwig goto out_free_buf; 59932927393SChristoph Hellwig } 60032927393SChristoph Hellwig 6014e63acdfSAndrey Ignatov error = count; 60232927393SChristoph Hellwig out_free_buf: 60345089437SJosef Bacik kvfree(kbuf); 60477b14db5SEric W. Biederman out: 60577b14db5SEric W. Biederman sysctl_head_finish(head); 60677b14db5SEric W. Biederman 60777b14db5SEric W. Biederman return error; 60877b14db5SEric W. Biederman } 60977b14db5SEric W. Biederman 6104bd6a735SMatthew Wilcox (Oracle) static ssize_t proc_sys_read(struct kiocb *iocb, struct iov_iter *iter) 6117708bfb1SPavel Emelyanov { 6124bd6a735SMatthew Wilcox (Oracle) return proc_sys_call_handler(iocb, iter, 0); 6137708bfb1SPavel Emelyanov } 6147708bfb1SPavel Emelyanov 6154bd6a735SMatthew Wilcox (Oracle) static ssize_t proc_sys_write(struct kiocb *iocb, struct iov_iter *iter) 61677b14db5SEric W. Biederman { 6174bd6a735SMatthew Wilcox (Oracle) return proc_sys_call_handler(iocb, iter, 1); 61877b14db5SEric W. Biederman } 61977b14db5SEric W. Biederman 620f1ecf068SLucas De Marchi static int proc_sys_open(struct inode *inode, struct file *filp) 621f1ecf068SLucas De Marchi { 6224e474a00SLucas De Marchi struct ctl_table_header *head = grab_header(inode); 623f1ecf068SLucas De Marchi struct ctl_table *table = PROC_I(inode)->sysctl_entry; 624f1ecf068SLucas De Marchi 6254e474a00SLucas De Marchi /* sysctl was unregistered */ 6264e474a00SLucas De Marchi if (IS_ERR(head)) 6274e474a00SLucas De Marchi return PTR_ERR(head); 6284e474a00SLucas De Marchi 629f1ecf068SLucas De Marchi if (table->poll) 630f1ecf068SLucas De Marchi filp->private_data = proc_sys_poll_event(table->poll); 631f1ecf068SLucas De Marchi 6324e474a00SLucas De Marchi sysctl_head_finish(head); 6334e474a00SLucas De Marchi 634f1ecf068SLucas De Marchi return 0; 635f1ecf068SLucas De Marchi } 636f1ecf068SLucas De Marchi 637076ccb76SAl Viro static __poll_t proc_sys_poll(struct file *filp, poll_table *wait) 638f1ecf068SLucas De Marchi { 639496ad9aaSAl Viro struct inode *inode = file_inode(filp); 6404e474a00SLucas De Marchi struct ctl_table_header *head = grab_header(inode); 641f1ecf068SLucas De Marchi struct ctl_table *table = PROC_I(inode)->sysctl_entry; 642076ccb76SAl Viro __poll_t ret = DEFAULT_POLLMASK; 6434e474a00SLucas De Marchi unsigned long event; 6444e474a00SLucas De Marchi 6454e474a00SLucas De Marchi /* sysctl was unregistered */ 6464e474a00SLucas De Marchi if (IS_ERR(head)) 647a9a08845SLinus Torvalds return EPOLLERR | EPOLLHUP; 648f1ecf068SLucas De Marchi 649f1ecf068SLucas De Marchi if (!table->proc_handler) 650f1ecf068SLucas De Marchi goto out; 651f1ecf068SLucas De Marchi 652f1ecf068SLucas De Marchi if (!table->poll) 653f1ecf068SLucas De Marchi goto out; 654f1ecf068SLucas De Marchi 6554e474a00SLucas De Marchi event = (unsigned long)filp->private_data; 656f1ecf068SLucas De Marchi poll_wait(filp, &table->poll->wait, wait); 657f1ecf068SLucas De Marchi 658f1ecf068SLucas De Marchi if (event != atomic_read(&table->poll->event)) { 659f1ecf068SLucas De Marchi filp->private_data = proc_sys_poll_event(table->poll); 660a9a08845SLinus Torvalds ret = EPOLLIN | EPOLLRDNORM | EPOLLERR | EPOLLPRI; 661f1ecf068SLucas De Marchi } 662f1ecf068SLucas De Marchi 663f1ecf068SLucas De Marchi out: 6644e474a00SLucas De Marchi sysctl_head_finish(head); 6654e474a00SLucas De Marchi 666f1ecf068SLucas De Marchi return ret; 667f1ecf068SLucas De Marchi } 66877b14db5SEric W. Biederman 669f0c3b509SAl Viro static bool proc_sys_fill_cache(struct file *file, 670f0c3b509SAl Viro struct dir_context *ctx, 6719043476fSAl Viro struct ctl_table_header *head, 6729043476fSAl Viro struct ctl_table *table) 67377b14db5SEric W. Biederman { 674f0c3b509SAl Viro struct dentry *child, *dir = file->f_path.dentry; 67577b14db5SEric W. Biederman struct inode *inode; 67677b14db5SEric W. Biederman struct qstr qname; 67777b14db5SEric W. Biederman ino_t ino = 0; 67877b14db5SEric W. Biederman unsigned type = DT_UNKNOWN; 67977b14db5SEric W. Biederman 68077b14db5SEric W. Biederman qname.name = table->procname; 68177b14db5SEric W. Biederman qname.len = strlen(table->procname); 6828387ff25SLinus Torvalds qname.hash = full_name_hash(dir, qname.name, qname.len); 68377b14db5SEric W. Biederman 68477b14db5SEric W. Biederman child = d_lookup(dir, &qname); 68577b14db5SEric W. Biederman if (!child) { 68676aab3abSAl Viro DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); 68776aab3abSAl Viro child = d_alloc_parallel(dir, &qname, &wq); 68876aab3abSAl Viro if (IS_ERR(child)) 68976aab3abSAl Viro return false; 69076aab3abSAl Viro if (d_in_lookup(child)) { 691888e2b03SAl Viro struct dentry *res; 6929043476fSAl Viro inode = proc_sys_make_inode(dir->d_sb, head, table); 693ea5751ccSIvan Delalande if (IS_ERR(inode)) { 69476aab3abSAl Viro d_lookup_done(child); 6959043476fSAl Viro dput(child); 696f0c3b509SAl Viro return false; 69776aab3abSAl Viro } 698fb045adbSNick Piggin d_set_d_op(child, &proc_sys_dentry_operations); 699888e2b03SAl Viro res = d_splice_alias(inode, child); 700888e2b03SAl Viro d_lookup_done(child); 701888e2b03SAl Viro if (unlikely(res)) { 702888e2b03SAl Viro if (IS_ERR(res)) { 703888e2b03SAl Viro dput(child); 704888e2b03SAl Viro return false; 705888e2b03SAl Viro } 706888e2b03SAl Viro dput(child); 707888e2b03SAl Viro child = res; 708888e2b03SAl Viro } 70977b14db5SEric W. Biederman } 71077b14db5SEric W. Biederman } 7112b0143b5SDavid Howells inode = d_inode(child); 71277b14db5SEric W. Biederman ino = inode->i_ino; 71377b14db5SEric W. Biederman type = inode->i_mode >> 12; 71477b14db5SEric W. Biederman dput(child); 715f0c3b509SAl Viro return dir_emit(ctx, qname.name, qname.len, ino, type); 7169043476fSAl Viro } 7179043476fSAl Viro 718f0c3b509SAl Viro static bool proc_sys_link_fill_cache(struct file *file, 719f0c3b509SAl Viro struct dir_context *ctx, 7200e47c99dSEric W. Biederman struct ctl_table_header *head, 7210e47c99dSEric W. Biederman struct ctl_table *table) 7220e47c99dSEric W. Biederman { 723f0c3b509SAl Viro bool ret = true; 724a0b0d1c3SDanilo Krummrich 7250e47c99dSEric W. Biederman head = sysctl_head_grab(head); 726a0b0d1c3SDanilo Krummrich if (IS_ERR(head)) 727a0b0d1c3SDanilo Krummrich return false; 7280e47c99dSEric W. Biederman 7290e47c99dSEric W. Biederman /* It is not an error if we can not follow the link ignore it */ 730835b94e0SDanilo Krummrich if (sysctl_follow_link(&head, &table)) 7310e47c99dSEric W. Biederman goto out; 7320e47c99dSEric W. Biederman 733f0c3b509SAl Viro ret = proc_sys_fill_cache(file, ctx, head, table); 7340e47c99dSEric W. Biederman out: 7350e47c99dSEric W. Biederman sysctl_head_finish(head); 7360e47c99dSEric W. Biederman return ret; 7370e47c99dSEric W. Biederman } 7380e47c99dSEric W. Biederman 739e5eea098SJoe Perches static int scan(struct ctl_table_header *head, struct ctl_table *table, 7409043476fSAl Viro unsigned long *pos, struct file *file, 741f0c3b509SAl Viro struct dir_context *ctx) 7429043476fSAl Viro { 743f0c3b509SAl Viro bool res; 7449043476fSAl Viro 745f0c3b509SAl Viro if ((*pos)++ < ctx->pos) 746f0c3b509SAl Viro return true; 7479043476fSAl Viro 7480e47c99dSEric W. Biederman if (unlikely(S_ISLNK(table->mode))) 749f0c3b509SAl Viro res = proc_sys_link_fill_cache(file, ctx, head, table); 7500e47c99dSEric W. Biederman else 751f0c3b509SAl Viro res = proc_sys_fill_cache(file, ctx, head, table); 7529043476fSAl Viro 753f0c3b509SAl Viro if (res) 754f0c3b509SAl Viro ctx->pos = *pos; 7556a75ce16SEric W. Biederman 7566a75ce16SEric W. Biederman return res; 75777b14db5SEric W. Biederman } 75877b14db5SEric W. Biederman 759f0c3b509SAl Viro static int proc_sys_readdir(struct file *file, struct dir_context *ctx) 76077b14db5SEric W. Biederman { 761f0c3b509SAl Viro struct ctl_table_header *head = grab_header(file_inode(file)); 7629043476fSAl Viro struct ctl_table_header *h = NULL; 7636a75ce16SEric W. Biederman struct ctl_table *entry; 7647ec66d06SEric W. Biederman struct ctl_dir *ctl_dir; 76577b14db5SEric W. Biederman unsigned long pos; 76677b14db5SEric W. Biederman 7679043476fSAl Viro if (IS_ERR(head)) 7689043476fSAl Viro return PTR_ERR(head); 7699043476fSAl Viro 7707ec66d06SEric W. Biederman ctl_dir = container_of(head, struct ctl_dir, header); 77177b14db5SEric W. Biederman 772f0c3b509SAl Viro if (!dir_emit_dots(file, ctx)) 77393362fa4SZhou Chengming goto out; 774f0c3b509SAl Viro 77577b14db5SEric W. Biederman pos = 2; 77677b14db5SEric W. Biederman 7777ec66d06SEric W. Biederman for (first_entry(ctl_dir, &h, &entry); h; next_entry(&h, &entry)) { 778f0c3b509SAl Viro if (!scan(h, entry, &pos, file, ctx)) { 7799043476fSAl Viro sysctl_head_finish(h); 7809043476fSAl Viro break; 78177b14db5SEric W. Biederman } 78277b14db5SEric W. Biederman } 78393362fa4SZhou Chengming out: 78477b14db5SEric W. Biederman sysctl_head_finish(head); 785f0c3b509SAl Viro return 0; 78677b14db5SEric W. Biederman } 78777b14db5SEric W. Biederman 788549c7297SChristian Brauner static int proc_sys_permission(struct user_namespace *mnt_userns, 789549c7297SChristian Brauner struct inode *inode, int mask) 79077b14db5SEric W. Biederman { 79177b14db5SEric W. Biederman /* 79277b14db5SEric W. Biederman * sysctl entries that are not writeable, 79377b14db5SEric W. Biederman * are _NOT_ writeable, capabilities or not. 79477b14db5SEric W. Biederman */ 795f696a365SMiklos Szeredi struct ctl_table_header *head; 796f696a365SMiklos Szeredi struct ctl_table *table; 79777b14db5SEric W. Biederman int error; 79877b14db5SEric W. Biederman 799f696a365SMiklos Szeredi /* Executable files are not allowed under /proc/sys/ */ 800f696a365SMiklos Szeredi if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode)) 801f696a365SMiklos Szeredi return -EACCES; 802f696a365SMiklos Szeredi 803f696a365SMiklos Szeredi head = grab_header(inode); 8049043476fSAl Viro if (IS_ERR(head)) 8059043476fSAl Viro return PTR_ERR(head); 80677b14db5SEric W. Biederman 807f696a365SMiklos Szeredi table = PROC_I(inode)->sysctl_entry; 8089043476fSAl Viro if (!table) /* global root - r-xr-xr-x */ 8099043476fSAl Viro error = mask & MAY_WRITE ? -EACCES : 0; 8109043476fSAl Viro else /* Use the permissions on the sysctl table entry */ 81173f7ef43SEric W. Biederman error = sysctl_perm(head, table, mask & ~MAY_NOT_BLOCK); 8129043476fSAl Viro 81377b14db5SEric W. Biederman sysctl_head_finish(head); 81477b14db5SEric W. Biederman return error; 81577b14db5SEric W. Biederman } 81677b14db5SEric W. Biederman 817549c7297SChristian Brauner static int proc_sys_setattr(struct user_namespace *mnt_userns, 818549c7297SChristian Brauner struct dentry *dentry, struct iattr *attr) 81977b14db5SEric W. Biederman { 8202b0143b5SDavid Howells struct inode *inode = d_inode(dentry); 82177b14db5SEric W. Biederman int error; 82277b14db5SEric W. Biederman 82377b14db5SEric W. Biederman if (attr->ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID)) 82477b14db5SEric W. Biederman return -EPERM; 82577b14db5SEric W. Biederman 8262f221d6fSChristian Brauner error = setattr_prepare(&init_user_ns, dentry, attr); 8271025774cSChristoph Hellwig if (error) 82877b14db5SEric W. Biederman return error; 8291025774cSChristoph Hellwig 8302f221d6fSChristian Brauner setattr_copy(&init_user_ns, inode, attr); 8311025774cSChristoph Hellwig mark_inode_dirty(inode); 8321025774cSChristoph Hellwig return 0; 83377b14db5SEric W. Biederman } 83477b14db5SEric W. Biederman 835549c7297SChristian Brauner static int proc_sys_getattr(struct user_namespace *mnt_userns, 836549c7297SChristian Brauner const struct path *path, struct kstat *stat, 837a528d35eSDavid Howells u32 request_mask, unsigned int query_flags) 8389043476fSAl Viro { 839a528d35eSDavid Howells struct inode *inode = d_inode(path->dentry); 8409043476fSAl Viro struct ctl_table_header *head = grab_header(inode); 8419043476fSAl Viro struct ctl_table *table = PROC_I(inode)->sysctl_entry; 8429043476fSAl Viro 8439043476fSAl Viro if (IS_ERR(head)) 8449043476fSAl Viro return PTR_ERR(head); 8459043476fSAl Viro 8460d56a451SChristian Brauner generic_fillattr(&init_user_ns, inode, stat); 8479043476fSAl Viro if (table) 8489043476fSAl Viro stat->mode = (stat->mode & S_IFMT) | table->mode; 8499043476fSAl Viro 8509043476fSAl Viro sysctl_head_finish(head); 8519043476fSAl Viro return 0; 8529043476fSAl Viro } 8539043476fSAl Viro 85477b14db5SEric W. Biederman static const struct file_operations proc_sys_file_operations = { 855f1ecf068SLucas De Marchi .open = proc_sys_open, 856f1ecf068SLucas De Marchi .poll = proc_sys_poll, 8574bd6a735SMatthew Wilcox (Oracle) .read_iter = proc_sys_read, 8584bd6a735SMatthew Wilcox (Oracle) .write_iter = proc_sys_write, 8594bd6a735SMatthew Wilcox (Oracle) .splice_read = generic_file_splice_read, 8604bd6a735SMatthew Wilcox (Oracle) .splice_write = iter_file_splice_write, 8616038f373SArnd Bergmann .llseek = default_llseek, 8629043476fSAl Viro }; 8639043476fSAl Viro 8649043476fSAl Viro static const struct file_operations proc_sys_dir_file_operations = { 865887df078SPavel Emelyanov .read = generic_read_dir, 866f50752eaSAl Viro .iterate_shared = proc_sys_readdir, 8673222a3e5SChristoph Hellwig .llseek = generic_file_llseek, 86877b14db5SEric W. Biederman }; 86977b14db5SEric W. Biederman 87003a44825SJan Engelhardt static const struct inode_operations proc_sys_inode_operations = { 8719043476fSAl Viro .permission = proc_sys_permission, 8729043476fSAl Viro .setattr = proc_sys_setattr, 8739043476fSAl Viro .getattr = proc_sys_getattr, 8749043476fSAl Viro }; 8759043476fSAl Viro 8769043476fSAl Viro static const struct inode_operations proc_sys_dir_operations = { 87777b14db5SEric W. Biederman .lookup = proc_sys_lookup, 87877b14db5SEric W. Biederman .permission = proc_sys_permission, 87977b14db5SEric W. Biederman .setattr = proc_sys_setattr, 8809043476fSAl Viro .getattr = proc_sys_getattr, 88177b14db5SEric W. Biederman }; 88277b14db5SEric W. Biederman 8830b728e19SAl Viro static int proc_sys_revalidate(struct dentry *dentry, unsigned int flags) 88477b14db5SEric W. Biederman { 8850b728e19SAl Viro if (flags & LOOKUP_RCU) 88634286d66SNick Piggin return -ECHILD; 8872b0143b5SDavid Howells return !PROC_I(d_inode(dentry))->sysctl->unregistering; 8889043476fSAl Viro } 8899043476fSAl Viro 890fe15ce44SNick Piggin static int proc_sys_delete(const struct dentry *dentry) 8919043476fSAl Viro { 8922b0143b5SDavid Howells return !!PROC_I(d_inode(dentry))->sysctl->unregistering; 8939043476fSAl Viro } 8949043476fSAl Viro 8951f87f0b5SEric W. Biederman static int sysctl_is_seen(struct ctl_table_header *p) 8961f87f0b5SEric W. Biederman { 8971f87f0b5SEric W. Biederman struct ctl_table_set *set = p->set; 8981f87f0b5SEric W. Biederman int res; 8991f87f0b5SEric W. Biederman spin_lock(&sysctl_lock); 9001f87f0b5SEric W. Biederman if (p->unregistering) 9011f87f0b5SEric W. Biederman res = 0; 9021f87f0b5SEric W. Biederman else if (!set->is_seen) 9031f87f0b5SEric W. Biederman res = 1; 9041f87f0b5SEric W. Biederman else 9051f87f0b5SEric W. Biederman res = set->is_seen(set); 9061f87f0b5SEric W. Biederman spin_unlock(&sysctl_lock); 9071f87f0b5SEric W. Biederman return res; 9081f87f0b5SEric W. Biederman } 9091f87f0b5SEric W. Biederman 9106fa67e70SAl Viro static int proc_sys_compare(const struct dentry *dentry, 911621e155aSNick Piggin unsigned int len, const char *str, const struct qstr *name) 9129043476fSAl Viro { 913dfef6dcdSAl Viro struct ctl_table_header *head; 914da53be12SLinus Torvalds struct inode *inode; 915da53be12SLinus Torvalds 91631e6b01fSNick Piggin /* Although proc doesn't have negative dentries, rcu-walk means 91731e6b01fSNick Piggin * that inode here can be NULL */ 918dfef6dcdSAl Viro /* AV: can it, indeed? */ 9192b0143b5SDavid Howells inode = d_inode_rcu(dentry); 92031e6b01fSNick Piggin if (!inode) 921dfef6dcdSAl Viro return 1; 922621e155aSNick Piggin if (name->len != len) 9239043476fSAl Viro return 1; 924621e155aSNick Piggin if (memcmp(name->name, str, len)) 9259043476fSAl Viro return 1; 926dfef6dcdSAl Viro head = rcu_dereference(PROC_I(inode)->sysctl); 927dfef6dcdSAl Viro return !head || !sysctl_is_seen(head); 92877b14db5SEric W. Biederman } 92977b14db5SEric W. Biederman 930d72f71ebSAl Viro static const struct dentry_operations proc_sys_dentry_operations = { 93177b14db5SEric W. Biederman .d_revalidate = proc_sys_revalidate, 9329043476fSAl Viro .d_delete = proc_sys_delete, 9339043476fSAl Viro .d_compare = proc_sys_compare, 93477b14db5SEric W. Biederman }; 93577b14db5SEric W. Biederman 9360e47c99dSEric W. Biederman static struct ctl_dir *find_subdir(struct ctl_dir *dir, 9377ec66d06SEric W. Biederman const char *name, int namelen) 9381f87f0b5SEric W. Biederman { 9397ec66d06SEric W. Biederman struct ctl_table_header *head; 9407ec66d06SEric W. Biederman struct ctl_table *entry; 9411f87f0b5SEric W. Biederman 9420e47c99dSEric W. Biederman entry = find_entry(&head, dir, name, namelen); 9437ec66d06SEric W. Biederman if (!entry) 9447ec66d06SEric W. Biederman return ERR_PTR(-ENOENT); 94551f72f4aSEric W. Biederman if (!S_ISDIR(entry->mode)) 9467ec66d06SEric W. Biederman return ERR_PTR(-ENOTDIR); 94751f72f4aSEric W. Biederman return container_of(head, struct ctl_dir, header); 9481f87f0b5SEric W. Biederman } 9491f87f0b5SEric W. Biederman 9507ec66d06SEric W. Biederman static struct ctl_dir *new_dir(struct ctl_table_set *set, 9517ec66d06SEric W. Biederman const char *name, int namelen) 9521f87f0b5SEric W. Biederman { 9537ec66d06SEric W. Biederman struct ctl_table *table; 9547ec66d06SEric W. Biederman struct ctl_dir *new; 955ac13ac6fSEric W. Biederman struct ctl_node *node; 9567ec66d06SEric W. Biederman char *new_name; 9571f87f0b5SEric W. Biederman 958ac13ac6fSEric W. Biederman new = kzalloc(sizeof(*new) + sizeof(struct ctl_node) + 959ac13ac6fSEric W. Biederman sizeof(struct ctl_table)*2 + namelen + 1, 960ac13ac6fSEric W. Biederman GFP_KERNEL); 9617ec66d06SEric W. Biederman if (!new) 9627ec66d06SEric W. Biederman return NULL; 9637ec66d06SEric W. Biederman 964ac13ac6fSEric W. Biederman node = (struct ctl_node *)(new + 1); 965ac13ac6fSEric W. Biederman table = (struct ctl_table *)(node + 1); 9667ec66d06SEric W. Biederman new_name = (char *)(table + 2); 9677ec66d06SEric W. Biederman memcpy(new_name, name, namelen); 9687ec66d06SEric W. Biederman new_name[namelen] = '\0'; 9697ec66d06SEric W. Biederman table[0].procname = new_name; 9707ec66d06SEric W. Biederman table[0].mode = S_IFDIR|S_IRUGO|S_IXUGO; 971ac13ac6fSEric W. Biederman init_header(&new->header, set->dir.header.root, set, node, table); 9727ec66d06SEric W. Biederman 9737ec66d06SEric W. Biederman return new; 9741f87f0b5SEric W. Biederman } 9751f87f0b5SEric W. Biederman 97660f126d9SEric W. Biederman /** 97760f126d9SEric W. Biederman * get_subdir - find or create a subdir with the specified name. 97860f126d9SEric W. Biederman * @dir: Directory to create the subdirectory in 97960f126d9SEric W. Biederman * @name: The name of the subdirectory to find or create 98060f126d9SEric W. Biederman * @namelen: The length of name 98160f126d9SEric W. Biederman * 98260f126d9SEric W. Biederman * Takes a directory with an elevated reference count so we know that 98360f126d9SEric W. Biederman * if we drop the lock the directory will not go away. Upon success 98460f126d9SEric W. Biederman * the reference is moved from @dir to the returned subdirectory. 98560f126d9SEric W. Biederman * Upon error an error code is returned and the reference on @dir is 98660f126d9SEric W. Biederman * simply dropped. 98760f126d9SEric W. Biederman */ 9880e47c99dSEric W. Biederman static struct ctl_dir *get_subdir(struct ctl_dir *dir, 9890e47c99dSEric W. Biederman const char *name, int namelen) 9907ec66d06SEric W. Biederman { 9910e47c99dSEric W. Biederman struct ctl_table_set *set = dir->header.set; 9927ec66d06SEric W. Biederman struct ctl_dir *subdir, *new = NULL; 9930eb97f38SEric W. Biederman int err; 9947ec66d06SEric W. Biederman 9957ec66d06SEric W. Biederman spin_lock(&sysctl_lock); 9960e47c99dSEric W. Biederman subdir = find_subdir(dir, name, namelen); 9977ec66d06SEric W. Biederman if (!IS_ERR(subdir)) 9987ec66d06SEric W. Biederman goto found; 9997ec66d06SEric W. Biederman if (PTR_ERR(subdir) != -ENOENT) 10007ec66d06SEric W. Biederman goto failed; 10017ec66d06SEric W. Biederman 10027ec66d06SEric W. Biederman spin_unlock(&sysctl_lock); 10037ec66d06SEric W. Biederman new = new_dir(set, name, namelen); 10047ec66d06SEric W. Biederman spin_lock(&sysctl_lock); 10057ec66d06SEric W. Biederman subdir = ERR_PTR(-ENOMEM); 10067ec66d06SEric W. Biederman if (!new) 10077ec66d06SEric W. Biederman goto failed; 10087ec66d06SEric W. Biederman 100960f126d9SEric W. Biederman /* Was the subdir added while we dropped the lock? */ 10100e47c99dSEric W. Biederman subdir = find_subdir(dir, name, namelen); 10117ec66d06SEric W. Biederman if (!IS_ERR(subdir)) 10127ec66d06SEric W. Biederman goto found; 10137ec66d06SEric W. Biederman if (PTR_ERR(subdir) != -ENOENT) 10147ec66d06SEric W. Biederman goto failed; 10157ec66d06SEric W. Biederman 101660f126d9SEric W. Biederman /* Nope. Use the our freshly made directory entry. */ 10170eb97f38SEric W. Biederman err = insert_header(dir, &new->header); 10180eb97f38SEric W. Biederman subdir = ERR_PTR(err); 10190eb97f38SEric W. Biederman if (err) 10200e47c99dSEric W. Biederman goto failed; 10217ec66d06SEric W. Biederman subdir = new; 10227ec66d06SEric W. Biederman found: 10237ec66d06SEric W. Biederman subdir->header.nreg++; 10247ec66d06SEric W. Biederman failed: 1025a1c83681SViresh Kumar if (IS_ERR(subdir)) { 102687ebdc00SAndrew Morton pr_err("sysctl could not get directory: "); 10276980128fSEric W. Biederman sysctl_print_dir(dir); 102887ebdc00SAndrew Morton pr_cont("/%*.*s %ld\n", 10297ec66d06SEric W. Biederman namelen, namelen, name, PTR_ERR(subdir)); 10301f87f0b5SEric W. Biederman } 10317ec66d06SEric W. Biederman drop_sysctl_table(&dir->header); 10327ec66d06SEric W. Biederman if (new) 10337ec66d06SEric W. Biederman drop_sysctl_table(&new->header); 10347ec66d06SEric W. Biederman spin_unlock(&sysctl_lock); 10357ec66d06SEric W. Biederman return subdir; 10361f87f0b5SEric W. Biederman } 10371f87f0b5SEric W. Biederman 10380e47c99dSEric W. Biederman static struct ctl_dir *xlate_dir(struct ctl_table_set *set, struct ctl_dir *dir) 10390e47c99dSEric W. Biederman { 10400e47c99dSEric W. Biederman struct ctl_dir *parent; 10410e47c99dSEric W. Biederman const char *procname; 10420e47c99dSEric W. Biederman if (!dir->header.parent) 10430e47c99dSEric W. Biederman return &set->dir; 10440e47c99dSEric W. Biederman parent = xlate_dir(set, dir->header.parent); 10450e47c99dSEric W. Biederman if (IS_ERR(parent)) 10460e47c99dSEric W. Biederman return parent; 10470e47c99dSEric W. Biederman procname = dir->header.ctl_table[0].procname; 10480e47c99dSEric W. Biederman return find_subdir(parent, procname, strlen(procname)); 10490e47c99dSEric W. Biederman } 10500e47c99dSEric W. Biederman 10510e47c99dSEric W. Biederman static int sysctl_follow_link(struct ctl_table_header **phead, 105213bcc6a2SEric W. Biederman struct ctl_table **pentry) 10530e47c99dSEric W. Biederman { 10540e47c99dSEric W. Biederman struct ctl_table_header *head; 10550e47c99dSEric W. Biederman struct ctl_table_root *root; 10560e47c99dSEric W. Biederman struct ctl_table_set *set; 10570e47c99dSEric W. Biederman struct ctl_table *entry; 10580e47c99dSEric W. Biederman struct ctl_dir *dir; 10590e47c99dSEric W. Biederman int ret; 10600e47c99dSEric W. Biederman 10610e47c99dSEric W. Biederman ret = 0; 10620e47c99dSEric W. Biederman spin_lock(&sysctl_lock); 10630e47c99dSEric W. Biederman root = (*pentry)->data; 106413bcc6a2SEric W. Biederman set = lookup_header_set(root); 10650e47c99dSEric W. Biederman dir = xlate_dir(set, (*phead)->parent); 10660e47c99dSEric W. Biederman if (IS_ERR(dir)) 10670e47c99dSEric W. Biederman ret = PTR_ERR(dir); 10680e47c99dSEric W. Biederman else { 10690e47c99dSEric W. Biederman const char *procname = (*pentry)->procname; 10700e47c99dSEric W. Biederman head = NULL; 10710e47c99dSEric W. Biederman entry = find_entry(&head, dir, procname, strlen(procname)); 10720e47c99dSEric W. Biederman ret = -ENOENT; 10730e47c99dSEric W. Biederman if (entry && use_table(head)) { 10740e47c99dSEric W. Biederman unuse_table(*phead); 10750e47c99dSEric W. Biederman *phead = head; 10760e47c99dSEric W. Biederman *pentry = entry; 10770e47c99dSEric W. Biederman ret = 0; 10780e47c99dSEric W. Biederman } 10790e47c99dSEric W. Biederman } 10800e47c99dSEric W. Biederman 10810e47c99dSEric W. Biederman spin_unlock(&sysctl_lock); 10820e47c99dSEric W. Biederman return ret; 10830e47c99dSEric W. Biederman } 10840e47c99dSEric W. Biederman 10857c60c48fSEric W. Biederman static int sysctl_err(const char *path, struct ctl_table *table, char *fmt, ...) 10867c60c48fSEric W. Biederman { 10877c60c48fSEric W. Biederman struct va_format vaf; 10887c60c48fSEric W. Biederman va_list args; 10897c60c48fSEric W. Biederman 10907c60c48fSEric W. Biederman va_start(args, fmt); 10917c60c48fSEric W. Biederman vaf.fmt = fmt; 10927c60c48fSEric W. Biederman vaf.va = &args; 10937c60c48fSEric W. Biederman 109487ebdc00SAndrew Morton pr_err("sysctl table check failed: %s/%s %pV\n", 10957c60c48fSEric W. Biederman path, table->procname, &vaf); 10967c60c48fSEric W. Biederman 10977c60c48fSEric W. Biederman va_end(args); 10987c60c48fSEric W. Biederman return -EINVAL; 10997c60c48fSEric W. Biederman } 11007c60c48fSEric W. Biederman 11014f2fec00SLuis R. Rodriguez static int sysctl_check_table_array(const char *path, struct ctl_table *table) 11024f2fec00SLuis R. Rodriguez { 11034f2fec00SLuis R. Rodriguez int err = 0; 11044f2fec00SLuis R. Rodriguez 110561d9b56aSLuis R. Rodriguez if ((table->proc_handler == proc_douintvec) || 110661d9b56aSLuis R. Rodriguez (table->proc_handler == proc_douintvec_minmax)) { 11074f2fec00SLuis R. Rodriguez if (table->maxlen != sizeof(unsigned int)) 110864a11f3dSWaiman Long err |= sysctl_err(path, table, "array not allowed"); 11094f2fec00SLuis R. Rodriguez } 11104f2fec00SLuis R. Rodriguez 11114f2fec00SLuis R. Rodriguez return err; 11124f2fec00SLuis R. Rodriguez } 11134f2fec00SLuis R. Rodriguez 11147c60c48fSEric W. Biederman static int sysctl_check_table(const char *path, struct ctl_table *table) 11157c60c48fSEric W. Biederman { 11167c60c48fSEric W. Biederman int err = 0; 11177c60c48fSEric W. Biederman for (; table->procname; table++) { 11187c60c48fSEric W. Biederman if (table->child) 111989c5b53bSLuis R. Rodriguez err |= sysctl_err(path, table, "Not a file"); 11207c60c48fSEric W. Biederman 11211f87f0b5SEric W. Biederman if ((table->proc_handler == proc_dostring) || 11221f87f0b5SEric W. Biederman (table->proc_handler == proc_dointvec) || 11231680a386SLiping Zhang (table->proc_handler == proc_douintvec) || 112461d9b56aSLuis R. Rodriguez (table->proc_handler == proc_douintvec_minmax) || 11251f87f0b5SEric W. Biederman (table->proc_handler == proc_dointvec_minmax) || 11261f87f0b5SEric W. Biederman (table->proc_handler == proc_dointvec_jiffies) || 11271f87f0b5SEric W. Biederman (table->proc_handler == proc_dointvec_userhz_jiffies) || 11281f87f0b5SEric W. Biederman (table->proc_handler == proc_dointvec_ms_jiffies) || 11291f87f0b5SEric W. Biederman (table->proc_handler == proc_doulongvec_minmax) || 11301f87f0b5SEric W. Biederman (table->proc_handler == proc_doulongvec_ms_jiffies_minmax)) { 11311f87f0b5SEric W. Biederman if (!table->data) 113289c5b53bSLuis R. Rodriguez err |= sysctl_err(path, table, "No data"); 11331f87f0b5SEric W. Biederman if (!table->maxlen) 113489c5b53bSLuis R. Rodriguez err |= sysctl_err(path, table, "No maxlen"); 11354f2fec00SLuis R. Rodriguez else 11364f2fec00SLuis R. Rodriguez err |= sysctl_check_table_array(path, table); 11371f87f0b5SEric W. Biederman } 11381f87f0b5SEric W. Biederman if (!table->proc_handler) 113989c5b53bSLuis R. Rodriguez err |= sysctl_err(path, table, "No proc_handler"); 11407c60c48fSEric W. Biederman 11417c60c48fSEric W. Biederman if ((table->mode & (S_IRUGO|S_IWUGO)) != table->mode) 114289c5b53bSLuis R. Rodriguez err |= sysctl_err(path, table, "bogus .mode 0%o", 11437c60c48fSEric W. Biederman table->mode); 11441f87f0b5SEric W. Biederman } 11457c60c48fSEric W. Biederman return err; 11461f87f0b5SEric W. Biederman } 11471f87f0b5SEric W. Biederman 11480e47c99dSEric W. Biederman static struct ctl_table_header *new_links(struct ctl_dir *dir, struct ctl_table *table, 11490e47c99dSEric W. Biederman struct ctl_table_root *link_root) 11500e47c99dSEric W. Biederman { 11510e47c99dSEric W. Biederman struct ctl_table *link_table, *entry, *link; 11520e47c99dSEric W. Biederman struct ctl_table_header *links; 1153ac13ac6fSEric W. Biederman struct ctl_node *node; 11540e47c99dSEric W. Biederman char *link_name; 11550e47c99dSEric W. Biederman int nr_entries, name_bytes; 11560e47c99dSEric W. Biederman 11570e47c99dSEric W. Biederman name_bytes = 0; 11580e47c99dSEric W. Biederman nr_entries = 0; 11590e47c99dSEric W. Biederman for (entry = table; entry->procname; entry++) { 11600e47c99dSEric W. Biederman nr_entries++; 11610e47c99dSEric W. Biederman name_bytes += strlen(entry->procname) + 1; 11620e47c99dSEric W. Biederman } 11630e47c99dSEric W. Biederman 11640e47c99dSEric W. Biederman links = kzalloc(sizeof(struct ctl_table_header) + 1165ac13ac6fSEric W. Biederman sizeof(struct ctl_node)*nr_entries + 11660e47c99dSEric W. Biederman sizeof(struct ctl_table)*(nr_entries + 1) + 11670e47c99dSEric W. Biederman name_bytes, 11680e47c99dSEric W. Biederman GFP_KERNEL); 11690e47c99dSEric W. Biederman 11700e47c99dSEric W. Biederman if (!links) 11710e47c99dSEric W. Biederman return NULL; 11720e47c99dSEric W. Biederman 1173ac13ac6fSEric W. Biederman node = (struct ctl_node *)(links + 1); 1174ac13ac6fSEric W. Biederman link_table = (struct ctl_table *)(node + nr_entries); 11750e47c99dSEric W. Biederman link_name = (char *)&link_table[nr_entries + 1]; 11760e47c99dSEric W. Biederman 11770e47c99dSEric W. Biederman for (link = link_table, entry = table; entry->procname; link++, entry++) { 11780e47c99dSEric W. Biederman int len = strlen(entry->procname) + 1; 11790e47c99dSEric W. Biederman memcpy(link_name, entry->procname, len); 11800e47c99dSEric W. Biederman link->procname = link_name; 11810e47c99dSEric W. Biederman link->mode = S_IFLNK|S_IRWXUGO; 11820e47c99dSEric W. Biederman link->data = link_root; 11830e47c99dSEric W. Biederman link_name += len; 11840e47c99dSEric W. Biederman } 1185ac13ac6fSEric W. Biederman init_header(links, dir->header.root, dir->header.set, node, link_table); 11860e47c99dSEric W. Biederman links->nreg = nr_entries; 11870e47c99dSEric W. Biederman 11880e47c99dSEric W. Biederman return links; 11890e47c99dSEric W. Biederman } 11900e47c99dSEric W. Biederman 11910e47c99dSEric W. Biederman static bool get_links(struct ctl_dir *dir, 11920e47c99dSEric W. Biederman struct ctl_table *table, struct ctl_table_root *link_root) 11930e47c99dSEric W. Biederman { 11940e47c99dSEric W. Biederman struct ctl_table_header *head; 11950e47c99dSEric W. Biederman struct ctl_table *entry, *link; 11960e47c99dSEric W. Biederman 11970e47c99dSEric W. Biederman /* Are there links available for every entry in table? */ 11980e47c99dSEric W. Biederman for (entry = table; entry->procname; entry++) { 11990e47c99dSEric W. Biederman const char *procname = entry->procname; 12000e47c99dSEric W. Biederman link = find_entry(&head, dir, procname, strlen(procname)); 12010e47c99dSEric W. Biederman if (!link) 12020e47c99dSEric W. Biederman return false; 12030e47c99dSEric W. Biederman if (S_ISDIR(link->mode) && S_ISDIR(entry->mode)) 12040e47c99dSEric W. Biederman continue; 12050e47c99dSEric W. Biederman if (S_ISLNK(link->mode) && (link->data == link_root)) 12060e47c99dSEric W. Biederman continue; 12070e47c99dSEric W. Biederman return false; 12080e47c99dSEric W. Biederman } 12090e47c99dSEric W. Biederman 12100e47c99dSEric W. Biederman /* The checks passed. Increase the registration count on the links */ 12110e47c99dSEric W. Biederman for (entry = table; entry->procname; entry++) { 12120e47c99dSEric W. Biederman const char *procname = entry->procname; 12130e47c99dSEric W. Biederman link = find_entry(&head, dir, procname, strlen(procname)); 12140e47c99dSEric W. Biederman head->nreg++; 12150e47c99dSEric W. Biederman } 12160e47c99dSEric W. Biederman return true; 12170e47c99dSEric W. Biederman } 12180e47c99dSEric W. Biederman 12190e47c99dSEric W. Biederman static int insert_links(struct ctl_table_header *head) 12200e47c99dSEric W. Biederman { 12210e47c99dSEric W. Biederman struct ctl_table_set *root_set = &sysctl_table_root.default_set; 12220e47c99dSEric W. Biederman struct ctl_dir *core_parent = NULL; 12230e47c99dSEric W. Biederman struct ctl_table_header *links; 12240e47c99dSEric W. Biederman int err; 12250e47c99dSEric W. Biederman 12260e47c99dSEric W. Biederman if (head->set == root_set) 12270e47c99dSEric W. Biederman return 0; 12280e47c99dSEric W. Biederman 12290e47c99dSEric W. Biederman core_parent = xlate_dir(root_set, head->parent); 12300e47c99dSEric W. Biederman if (IS_ERR(core_parent)) 12310e47c99dSEric W. Biederman return 0; 12320e47c99dSEric W. Biederman 12330e47c99dSEric W. Biederman if (get_links(core_parent, head->ctl_table, head->root)) 12340e47c99dSEric W. Biederman return 0; 12350e47c99dSEric W. Biederman 12360e47c99dSEric W. Biederman core_parent->header.nreg++; 12370e47c99dSEric W. Biederman spin_unlock(&sysctl_lock); 12380e47c99dSEric W. Biederman 12390e47c99dSEric W. Biederman links = new_links(core_parent, head->ctl_table, head->root); 12400e47c99dSEric W. Biederman 12410e47c99dSEric W. Biederman spin_lock(&sysctl_lock); 12420e47c99dSEric W. Biederman err = -ENOMEM; 12430e47c99dSEric W. Biederman if (!links) 12440e47c99dSEric W. Biederman goto out; 12450e47c99dSEric W. Biederman 12460e47c99dSEric W. Biederman err = 0; 12470e47c99dSEric W. Biederman if (get_links(core_parent, head->ctl_table, head->root)) { 12480e47c99dSEric W. Biederman kfree(links); 12490e47c99dSEric W. Biederman goto out; 12500e47c99dSEric W. Biederman } 12510e47c99dSEric W. Biederman 12520e47c99dSEric W. Biederman err = insert_header(core_parent, links); 12530e47c99dSEric W. Biederman if (err) 12540e47c99dSEric W. Biederman kfree(links); 12550e47c99dSEric W. Biederman out: 12560e47c99dSEric W. Biederman drop_sysctl_table(&core_parent->header); 12570e47c99dSEric W. Biederman return err; 12580e47c99dSEric W. Biederman } 12590e47c99dSEric W. Biederman 12601f87f0b5SEric W. Biederman /** 1261f728019bSEric W. Biederman * __register_sysctl_table - register a leaf sysctl table 126260a47a2eSEric W. Biederman * @set: Sysctl tree to register on 12631f87f0b5SEric W. Biederman * @path: The path to the directory the sysctl table is in. 12641f87f0b5SEric W. Biederman * @table: the top-level table structure 12651f87f0b5SEric W. Biederman * 12661f87f0b5SEric W. Biederman * Register a sysctl table hierarchy. @table should be a filled in ctl_table 12671f87f0b5SEric W. Biederman * array. A completely 0 filled entry terminates the table. 12681f87f0b5SEric W. Biederman * 12691f87f0b5SEric W. Biederman * The members of the &struct ctl_table structure are used as follows: 12701f87f0b5SEric W. Biederman * 12711f87f0b5SEric W. Biederman * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not 12721f87f0b5SEric W. Biederman * enter a sysctl file 12731f87f0b5SEric W. Biederman * 12741f87f0b5SEric W. Biederman * data - a pointer to data for use by proc_handler 12751f87f0b5SEric W. Biederman * 12761f87f0b5SEric W. Biederman * maxlen - the maximum size in bytes of the data 12771f87f0b5SEric W. Biederman * 1278f728019bSEric W. Biederman * mode - the file permissions for the /proc/sys file 12791f87f0b5SEric W. Biederman * 1280f728019bSEric W. Biederman * child - must be %NULL. 12811f87f0b5SEric W. Biederman * 12821f87f0b5SEric W. Biederman * proc_handler - the text handler routine (described below) 12831f87f0b5SEric W. Biederman * 12841f87f0b5SEric W. Biederman * extra1, extra2 - extra pointers usable by the proc handler routines 12851f87f0b5SEric W. Biederman * 12861f87f0b5SEric W. Biederman * Leaf nodes in the sysctl tree will be represented by a single file 12871f87f0b5SEric W. Biederman * under /proc; non-leaf nodes will be represented by directories. 12881f87f0b5SEric W. Biederman * 1289f728019bSEric W. Biederman * There must be a proc_handler routine for any terminal nodes. 1290f728019bSEric W. Biederman * Several default handlers are available to cover common cases - 12911f87f0b5SEric W. Biederman * 12921f87f0b5SEric W. Biederman * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(), 12931f87f0b5SEric W. Biederman * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), 12941f87f0b5SEric W. Biederman * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax() 12951f87f0b5SEric W. Biederman * 12961f87f0b5SEric W. Biederman * It is the handler's job to read the input buffer from user memory 12971f87f0b5SEric W. Biederman * and process it. The handler should return 0 on success. 12981f87f0b5SEric W. Biederman * 12991f87f0b5SEric W. Biederman * This routine returns %NULL on a failure to register, and a pointer 13001f87f0b5SEric W. Biederman * to the table header on success. 13011f87f0b5SEric W. Biederman */ 13026e9d5164SEric W. Biederman struct ctl_table_header *__register_sysctl_table( 130360a47a2eSEric W. Biederman struct ctl_table_set *set, 13046e9d5164SEric W. Biederman const char *path, struct ctl_table *table) 13051f87f0b5SEric W. Biederman { 130660a47a2eSEric W. Biederman struct ctl_table_root *root = set->dir.header.root; 13071f87f0b5SEric W. Biederman struct ctl_table_header *header; 13086e9d5164SEric W. Biederman const char *name, *nextname; 13097ec66d06SEric W. Biederman struct ctl_dir *dir; 1310ac13ac6fSEric W. Biederman struct ctl_table *entry; 1311ac13ac6fSEric W. Biederman struct ctl_node *node; 1312ac13ac6fSEric W. Biederman int nr_entries = 0; 13131f87f0b5SEric W. Biederman 1314ac13ac6fSEric W. Biederman for (entry = table; entry->procname; entry++) 1315ac13ac6fSEric W. Biederman nr_entries++; 1316ac13ac6fSEric W. Biederman 1317ac13ac6fSEric W. Biederman header = kzalloc(sizeof(struct ctl_table_header) + 1318ac13ac6fSEric W. Biederman sizeof(struct ctl_node)*nr_entries, GFP_KERNEL); 13191f87f0b5SEric W. Biederman if (!header) 13201f87f0b5SEric W. Biederman return NULL; 13211f87f0b5SEric W. Biederman 1322ac13ac6fSEric W. Biederman node = (struct ctl_node *)(header + 1); 1323ac13ac6fSEric W. Biederman init_header(header, root, set, node, table); 13247c60c48fSEric W. Biederman if (sysctl_check_table(path, table)) 13257c60c48fSEric W. Biederman goto fail; 13268d6ecfccSEric W. Biederman 13271f87f0b5SEric W. Biederman spin_lock(&sysctl_lock); 13280e47c99dSEric W. Biederman dir = &set->dir; 132960f126d9SEric W. Biederman /* Reference moved down the diretory tree get_subdir */ 13307ec66d06SEric W. Biederman dir->header.nreg++; 13317ec66d06SEric W. Biederman spin_unlock(&sysctl_lock); 13327ec66d06SEric W. Biederman 13337ec66d06SEric W. Biederman /* Find the directory for the ctl_table */ 13347ec66d06SEric W. Biederman for (name = path; name; name = nextname) { 13357ec66d06SEric W. Biederman int namelen; 13367ec66d06SEric W. Biederman nextname = strchr(name, '/'); 13377ec66d06SEric W. Biederman if (nextname) { 13387ec66d06SEric W. Biederman namelen = nextname - name; 13397ec66d06SEric W. Biederman nextname++; 13407ec66d06SEric W. Biederman } else { 13417ec66d06SEric W. Biederman namelen = strlen(name); 13427ec66d06SEric W. Biederman } 13437ec66d06SEric W. Biederman if (namelen == 0) 13441f87f0b5SEric W. Biederman continue; 13457ec66d06SEric W. Biederman 13460e47c99dSEric W. Biederman dir = get_subdir(dir, name, namelen); 13477ec66d06SEric W. Biederman if (IS_ERR(dir)) 13487ec66d06SEric W. Biederman goto fail; 13491f87f0b5SEric W. Biederman } 13500e47c99dSEric W. Biederman 13517ec66d06SEric W. Biederman spin_lock(&sysctl_lock); 13520e47c99dSEric W. Biederman if (insert_header(dir, header)) 13530e47c99dSEric W. Biederman goto fail_put_dir_locked; 13540e47c99dSEric W. Biederman 13557ec66d06SEric W. Biederman drop_sysctl_table(&dir->header); 13561f87f0b5SEric W. Biederman spin_unlock(&sysctl_lock); 13571f87f0b5SEric W. Biederman 13581f87f0b5SEric W. Biederman return header; 13590e47c99dSEric W. Biederman 13607ec66d06SEric W. Biederman fail_put_dir_locked: 13617ec66d06SEric W. Biederman drop_sysctl_table(&dir->header); 13627c60c48fSEric W. Biederman spin_unlock(&sysctl_lock); 13637c60c48fSEric W. Biederman fail: 13647c60c48fSEric W. Biederman kfree(header); 13657c60c48fSEric W. Biederman dump_stack(); 13667c60c48fSEric W. Biederman return NULL; 13671f87f0b5SEric W. Biederman } 13681f87f0b5SEric W. Biederman 1369fea478d4SEric W. Biederman /** 1370fea478d4SEric W. Biederman * register_sysctl - register a sysctl table 1371fea478d4SEric W. Biederman * @path: The path to the directory the sysctl table is in. 1372fea478d4SEric W. Biederman * @table: the table structure 1373fea478d4SEric W. Biederman * 1374fea478d4SEric W. Biederman * Register a sysctl table. @table should be a filled in ctl_table 1375fea478d4SEric W. Biederman * array. A completely 0 filled entry terminates the table. 1376fea478d4SEric W. Biederman * 1377fea478d4SEric W. Biederman * See __register_sysctl_table for more details. 1378fea478d4SEric W. Biederman */ 1379fea478d4SEric W. Biederman struct ctl_table_header *register_sysctl(const char *path, struct ctl_table *table) 1380fea478d4SEric W. Biederman { 1381fea478d4SEric W. Biederman return __register_sysctl_table(&sysctl_table_root.default_set, 1382fea478d4SEric W. Biederman path, table); 1383fea478d4SEric W. Biederman } 1384fea478d4SEric W. Biederman EXPORT_SYMBOL(register_sysctl); 1385fea478d4SEric W. Biederman 13866e9d5164SEric W. Biederman static char *append_path(const char *path, char *pos, const char *name) 13876e9d5164SEric W. Biederman { 13886e9d5164SEric W. Biederman int namelen; 13896e9d5164SEric W. Biederman namelen = strlen(name); 13906e9d5164SEric W. Biederman if (((pos - path) + namelen + 2) >= PATH_MAX) 13916e9d5164SEric W. Biederman return NULL; 13926e9d5164SEric W. Biederman memcpy(pos, name, namelen); 13936e9d5164SEric W. Biederman pos[namelen] = '/'; 13946e9d5164SEric W. Biederman pos[namelen + 1] = '\0'; 13956e9d5164SEric W. Biederman pos += namelen + 1; 13966e9d5164SEric W. Biederman return pos; 13976e9d5164SEric W. Biederman } 13986e9d5164SEric W. Biederman 1399f728019bSEric W. Biederman static int count_subheaders(struct ctl_table *table) 1400f728019bSEric W. Biederman { 1401f728019bSEric W. Biederman int has_files = 0; 1402f728019bSEric W. Biederman int nr_subheaders = 0; 1403f728019bSEric W. Biederman struct ctl_table *entry; 1404f728019bSEric W. Biederman 1405f728019bSEric W. Biederman /* special case: no directory and empty directory */ 1406f728019bSEric W. Biederman if (!table || !table->procname) 1407f728019bSEric W. Biederman return 1; 1408f728019bSEric W. Biederman 1409f728019bSEric W. Biederman for (entry = table; entry->procname; entry++) { 1410f728019bSEric W. Biederman if (entry->child) 1411f728019bSEric W. Biederman nr_subheaders += count_subheaders(entry->child); 1412f728019bSEric W. Biederman else 1413f728019bSEric W. Biederman has_files = 1; 1414f728019bSEric W. Biederman } 1415f728019bSEric W. Biederman return nr_subheaders + has_files; 1416f728019bSEric W. Biederman } 1417f728019bSEric W. Biederman 1418f728019bSEric W. Biederman static int register_leaf_sysctl_tables(const char *path, char *pos, 141960a47a2eSEric W. Biederman struct ctl_table_header ***subheader, struct ctl_table_set *set, 1420f728019bSEric W. Biederman struct ctl_table *table) 1421f728019bSEric W. Biederman { 1422f728019bSEric W. Biederman struct ctl_table *ctl_table_arg = NULL; 1423f728019bSEric W. Biederman struct ctl_table *entry, *files; 1424f728019bSEric W. Biederman int nr_files = 0; 1425f728019bSEric W. Biederman int nr_dirs = 0; 1426f728019bSEric W. Biederman int err = -ENOMEM; 1427f728019bSEric W. Biederman 1428f728019bSEric W. Biederman for (entry = table; entry->procname; entry++) { 1429f728019bSEric W. Biederman if (entry->child) 1430f728019bSEric W. Biederman nr_dirs++; 1431f728019bSEric W. Biederman else 1432f728019bSEric W. Biederman nr_files++; 1433f728019bSEric W. Biederman } 1434f728019bSEric W. Biederman 1435f728019bSEric W. Biederman files = table; 1436f728019bSEric W. Biederman /* If there are mixed files and directories we need a new table */ 1437f728019bSEric W. Biederman if (nr_dirs && nr_files) { 1438f728019bSEric W. Biederman struct ctl_table *new; 14396396bb22SKees Cook files = kcalloc(nr_files + 1, sizeof(struct ctl_table), 1440f728019bSEric W. Biederman GFP_KERNEL); 1441f728019bSEric W. Biederman if (!files) 1442f728019bSEric W. Biederman goto out; 1443f728019bSEric W. Biederman 1444f728019bSEric W. Biederman ctl_table_arg = files; 1445f728019bSEric W. Biederman for (new = files, entry = table; entry->procname; entry++) { 1446f728019bSEric W. Biederman if (entry->child) 1447f728019bSEric W. Biederman continue; 1448f728019bSEric W. Biederman *new = *entry; 1449f728019bSEric W. Biederman new++; 1450f728019bSEric W. Biederman } 1451f728019bSEric W. Biederman } 1452f728019bSEric W. Biederman 1453f728019bSEric W. Biederman /* Register everything except a directory full of subdirectories */ 1454f728019bSEric W. Biederman if (nr_files || !nr_dirs) { 1455f728019bSEric W. Biederman struct ctl_table_header *header; 145660a47a2eSEric W. Biederman header = __register_sysctl_table(set, path, files); 1457f728019bSEric W. Biederman if (!header) { 1458f728019bSEric W. Biederman kfree(ctl_table_arg); 1459f728019bSEric W. Biederman goto out; 1460f728019bSEric W. Biederman } 1461f728019bSEric W. Biederman 1462f728019bSEric W. Biederman /* Remember if we need to free the file table */ 1463f728019bSEric W. Biederman header->ctl_table_arg = ctl_table_arg; 1464f728019bSEric W. Biederman **subheader = header; 1465f728019bSEric W. Biederman (*subheader)++; 1466f728019bSEric W. Biederman } 1467f728019bSEric W. Biederman 1468f728019bSEric W. Biederman /* Recurse into the subdirectories. */ 1469f728019bSEric W. Biederman for (entry = table; entry->procname; entry++) { 1470f728019bSEric W. Biederman char *child_pos; 1471f728019bSEric W. Biederman 1472f728019bSEric W. Biederman if (!entry->child) 1473f728019bSEric W. Biederman continue; 1474f728019bSEric W. Biederman 1475f728019bSEric W. Biederman err = -ENAMETOOLONG; 1476f728019bSEric W. Biederman child_pos = append_path(path, pos, entry->procname); 1477f728019bSEric W. Biederman if (!child_pos) 1478f728019bSEric W. Biederman goto out; 1479f728019bSEric W. Biederman 1480f728019bSEric W. Biederman err = register_leaf_sysctl_tables(path, child_pos, subheader, 148160a47a2eSEric W. Biederman set, entry->child); 1482f728019bSEric W. Biederman pos[0] = '\0'; 1483f728019bSEric W. Biederman if (err) 1484f728019bSEric W. Biederman goto out; 1485f728019bSEric W. Biederman } 1486f728019bSEric W. Biederman err = 0; 1487f728019bSEric W. Biederman out: 1488f728019bSEric W. Biederman /* On failure our caller will unregister all registered subheaders */ 1489f728019bSEric W. Biederman return err; 1490f728019bSEric W. Biederman } 1491f728019bSEric W. Biederman 14926e9d5164SEric W. Biederman /** 14936e9d5164SEric W. Biederman * __register_sysctl_paths - register a sysctl table hierarchy 149460a47a2eSEric W. Biederman * @set: Sysctl tree to register on 14956e9d5164SEric W. Biederman * @path: The path to the directory the sysctl table is in. 14966e9d5164SEric W. Biederman * @table: the top-level table structure 14976e9d5164SEric W. Biederman * 14986e9d5164SEric W. Biederman * Register a sysctl table hierarchy. @table should be a filled in ctl_table 14996e9d5164SEric W. Biederman * array. A completely 0 filled entry terminates the table. 15006e9d5164SEric W. Biederman * 15016e9d5164SEric W. Biederman * See __register_sysctl_table for more details. 15026e9d5164SEric W. Biederman */ 15036e9d5164SEric W. Biederman struct ctl_table_header *__register_sysctl_paths( 150460a47a2eSEric W. Biederman struct ctl_table_set *set, 15056e9d5164SEric W. Biederman const struct ctl_path *path, struct ctl_table *table) 15066e9d5164SEric W. Biederman { 1507ec6a5266SEric W. Biederman struct ctl_table *ctl_table_arg = table; 1508f728019bSEric W. Biederman int nr_subheaders = count_subheaders(table); 1509f728019bSEric W. Biederman struct ctl_table_header *header = NULL, **subheaders, **subheader; 15106e9d5164SEric W. Biederman const struct ctl_path *component; 15116e9d5164SEric W. Biederman char *new_path, *pos; 15126e9d5164SEric W. Biederman 15136e9d5164SEric W. Biederman pos = new_path = kmalloc(PATH_MAX, GFP_KERNEL); 15146e9d5164SEric W. Biederman if (!new_path) 15156e9d5164SEric W. Biederman return NULL; 15166e9d5164SEric W. Biederman 15176e9d5164SEric W. Biederman pos[0] = '\0'; 15186e9d5164SEric W. Biederman for (component = path; component->procname; component++) { 15196e9d5164SEric W. Biederman pos = append_path(new_path, pos, component->procname); 15206e9d5164SEric W. Biederman if (!pos) 15216e9d5164SEric W. Biederman goto out; 15226e9d5164SEric W. Biederman } 1523ec6a5266SEric W. Biederman while (table->procname && table->child && !table[1].procname) { 1524ec6a5266SEric W. Biederman pos = append_path(new_path, pos, table->procname); 1525ec6a5266SEric W. Biederman if (!pos) 1526ec6a5266SEric W. Biederman goto out; 1527ec6a5266SEric W. Biederman table = table->child; 1528ec6a5266SEric W. Biederman } 1529f728019bSEric W. Biederman if (nr_subheaders == 1) { 153060a47a2eSEric W. Biederman header = __register_sysctl_table(set, new_path, table); 1531ec6a5266SEric W. Biederman if (header) 1532ec6a5266SEric W. Biederman header->ctl_table_arg = ctl_table_arg; 1533f728019bSEric W. Biederman } else { 1534f728019bSEric W. Biederman header = kzalloc(sizeof(*header) + 1535f728019bSEric W. Biederman sizeof(*subheaders)*nr_subheaders, GFP_KERNEL); 1536f728019bSEric W. Biederman if (!header) 1537f728019bSEric W. Biederman goto out; 1538f728019bSEric W. Biederman 1539f728019bSEric W. Biederman subheaders = (struct ctl_table_header **) (header + 1); 1540f728019bSEric W. Biederman subheader = subheaders; 1541f728019bSEric W. Biederman header->ctl_table_arg = ctl_table_arg; 1542f728019bSEric W. Biederman 1543f728019bSEric W. Biederman if (register_leaf_sysctl_tables(new_path, pos, &subheader, 154460a47a2eSEric W. Biederman set, table)) 1545f728019bSEric W. Biederman goto err_register_leaves; 1546f728019bSEric W. Biederman } 1547f728019bSEric W. Biederman 15486e9d5164SEric W. Biederman out: 15496e9d5164SEric W. Biederman kfree(new_path); 15506e9d5164SEric W. Biederman return header; 1551f728019bSEric W. Biederman 1552f728019bSEric W. Biederman err_register_leaves: 1553f728019bSEric W. Biederman while (subheader > subheaders) { 1554f728019bSEric W. Biederman struct ctl_table_header *subh = *(--subheader); 1555f728019bSEric W. Biederman struct ctl_table *table = subh->ctl_table_arg; 1556f728019bSEric W. Biederman unregister_sysctl_table(subh); 1557f728019bSEric W. Biederman kfree(table); 1558f728019bSEric W. Biederman } 1559f728019bSEric W. Biederman kfree(header); 1560f728019bSEric W. Biederman header = NULL; 1561f728019bSEric W. Biederman goto out; 15626e9d5164SEric W. Biederman } 15636e9d5164SEric W. Biederman 15641f87f0b5SEric W. Biederman /** 15651f87f0b5SEric W. Biederman * register_sysctl_table_path - register a sysctl table hierarchy 15661f87f0b5SEric W. Biederman * @path: The path to the directory the sysctl table is in. 15671f87f0b5SEric W. Biederman * @table: the top-level table structure 15681f87f0b5SEric W. Biederman * 15691f87f0b5SEric W. Biederman * Register a sysctl table hierarchy. @table should be a filled in ctl_table 15701f87f0b5SEric W. Biederman * array. A completely 0 filled entry terminates the table. 15711f87f0b5SEric W. Biederman * 15721f87f0b5SEric W. Biederman * See __register_sysctl_paths for more details. 15731f87f0b5SEric W. Biederman */ 15741f87f0b5SEric W. Biederman struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path, 15751f87f0b5SEric W. Biederman struct ctl_table *table) 15761f87f0b5SEric W. Biederman { 157760a47a2eSEric W. Biederman return __register_sysctl_paths(&sysctl_table_root.default_set, 15781f87f0b5SEric W. Biederman path, table); 15791f87f0b5SEric W. Biederman } 15801f87f0b5SEric W. Biederman EXPORT_SYMBOL(register_sysctl_paths); 15811f87f0b5SEric W. Biederman 15821f87f0b5SEric W. Biederman /** 15831f87f0b5SEric W. Biederman * register_sysctl_table - register a sysctl table hierarchy 15841f87f0b5SEric W. Biederman * @table: the top-level table structure 15851f87f0b5SEric W. Biederman * 15861f87f0b5SEric W. Biederman * Register a sysctl table hierarchy. @table should be a filled in ctl_table 15871f87f0b5SEric W. Biederman * array. A completely 0 filled entry terminates the table. 15881f87f0b5SEric W. Biederman * 15891f87f0b5SEric W. Biederman * See register_sysctl_paths for more details. 15901f87f0b5SEric W. Biederman */ 15911f87f0b5SEric W. Biederman struct ctl_table_header *register_sysctl_table(struct ctl_table *table) 15921f87f0b5SEric W. Biederman { 15931f87f0b5SEric W. Biederman static const struct ctl_path null_path[] = { {} }; 15941f87f0b5SEric W. Biederman 15951f87f0b5SEric W. Biederman return register_sysctl_paths(null_path, table); 15961f87f0b5SEric W. Biederman } 15971f87f0b5SEric W. Biederman EXPORT_SYMBOL(register_sysctl_table); 15981f87f0b5SEric W. Biederman 15990e47c99dSEric W. Biederman static void put_links(struct ctl_table_header *header) 16000e47c99dSEric W. Biederman { 16010e47c99dSEric W. Biederman struct ctl_table_set *root_set = &sysctl_table_root.default_set; 16020e47c99dSEric W. Biederman struct ctl_table_root *root = header->root; 16030e47c99dSEric W. Biederman struct ctl_dir *parent = header->parent; 16040e47c99dSEric W. Biederman struct ctl_dir *core_parent; 16050e47c99dSEric W. Biederman struct ctl_table *entry; 16060e47c99dSEric W. Biederman 16070e47c99dSEric W. Biederman if (header->set == root_set) 16080e47c99dSEric W. Biederman return; 16090e47c99dSEric W. Biederman 16100e47c99dSEric W. Biederman core_parent = xlate_dir(root_set, parent); 16110e47c99dSEric W. Biederman if (IS_ERR(core_parent)) 16120e47c99dSEric W. Biederman return; 16130e47c99dSEric W. Biederman 16140e47c99dSEric W. Biederman for (entry = header->ctl_table; entry->procname; entry++) { 16150e47c99dSEric W. Biederman struct ctl_table_header *link_head; 16160e47c99dSEric W. Biederman struct ctl_table *link; 16170e47c99dSEric W. Biederman const char *name = entry->procname; 16180e47c99dSEric W. Biederman 16190e47c99dSEric W. Biederman link = find_entry(&link_head, core_parent, name, strlen(name)); 16200e47c99dSEric W. Biederman if (link && 16210e47c99dSEric W. Biederman ((S_ISDIR(link->mode) && S_ISDIR(entry->mode)) || 16220e47c99dSEric W. Biederman (S_ISLNK(link->mode) && (link->data == root)))) { 16230e47c99dSEric W. Biederman drop_sysctl_table(link_head); 16240e47c99dSEric W. Biederman } 16250e47c99dSEric W. Biederman else { 162687ebdc00SAndrew Morton pr_err("sysctl link missing during unregister: "); 16270e47c99dSEric W. Biederman sysctl_print_dir(parent); 162887ebdc00SAndrew Morton pr_cont("/%s\n", name); 16290e47c99dSEric W. Biederman } 16300e47c99dSEric W. Biederman } 16310e47c99dSEric W. Biederman } 16320e47c99dSEric W. Biederman 1633938aaa4fSEric W. Biederman static void drop_sysctl_table(struct ctl_table_header *header) 1634938aaa4fSEric W. Biederman { 16357ec66d06SEric W. Biederman struct ctl_dir *parent = header->parent; 16367ec66d06SEric W. Biederman 1637938aaa4fSEric W. Biederman if (--header->nreg) 1638938aaa4fSEric W. Biederman return; 1639938aaa4fSEric W. Biederman 164089189557SYueHaibing if (parent) { 16410e47c99dSEric W. Biederman put_links(header); 1642938aaa4fSEric W. Biederman start_unregistering(header); 164389189557SYueHaibing } 164489189557SYueHaibing 1645938aaa4fSEric W. Biederman if (!--header->count) 1646938aaa4fSEric W. Biederman kfree_rcu(header, rcu); 16477ec66d06SEric W. Biederman 16487ec66d06SEric W. Biederman if (parent) 16497ec66d06SEric W. Biederman drop_sysctl_table(&parent->header); 1650938aaa4fSEric W. Biederman } 1651938aaa4fSEric W. Biederman 16521f87f0b5SEric W. Biederman /** 16531f87f0b5SEric W. Biederman * unregister_sysctl_table - unregister a sysctl table hierarchy 16541f87f0b5SEric W. Biederman * @header: the header returned from register_sysctl_table 16551f87f0b5SEric W. Biederman * 16561f87f0b5SEric W. Biederman * Unregisters the sysctl table and all children. proc entries may not 16571f87f0b5SEric W. Biederman * actually be removed until they are no longer used by anyone. 16581f87f0b5SEric W. Biederman */ 16591f87f0b5SEric W. Biederman void unregister_sysctl_table(struct ctl_table_header * header) 16601f87f0b5SEric W. Biederman { 1661f728019bSEric W. Biederman int nr_subheaders; 16621f87f0b5SEric W. Biederman might_sleep(); 16631f87f0b5SEric W. Biederman 16641f87f0b5SEric W. Biederman if (header == NULL) 16651f87f0b5SEric W. Biederman return; 16661f87f0b5SEric W. Biederman 1667f728019bSEric W. Biederman nr_subheaders = count_subheaders(header->ctl_table_arg); 1668f728019bSEric W. Biederman if (unlikely(nr_subheaders > 1)) { 1669f728019bSEric W. Biederman struct ctl_table_header **subheaders; 1670f728019bSEric W. Biederman int i; 1671f728019bSEric W. Biederman 1672f728019bSEric W. Biederman subheaders = (struct ctl_table_header **)(header + 1); 1673f728019bSEric W. Biederman for (i = nr_subheaders -1; i >= 0; i--) { 1674f728019bSEric W. Biederman struct ctl_table_header *subh = subheaders[i]; 1675f728019bSEric W. Biederman struct ctl_table *table = subh->ctl_table_arg; 1676f728019bSEric W. Biederman unregister_sysctl_table(subh); 1677f728019bSEric W. Biederman kfree(table); 1678f728019bSEric W. Biederman } 1679f728019bSEric W. Biederman kfree(header); 1680f728019bSEric W. Biederman return; 1681f728019bSEric W. Biederman } 1682f728019bSEric W. Biederman 16831f87f0b5SEric W. Biederman spin_lock(&sysctl_lock); 1684938aaa4fSEric W. Biederman drop_sysctl_table(header); 16851f87f0b5SEric W. Biederman spin_unlock(&sysctl_lock); 16861f87f0b5SEric W. Biederman } 16871f87f0b5SEric W. Biederman EXPORT_SYMBOL(unregister_sysctl_table); 16881f87f0b5SEric W. Biederman 16890e47c99dSEric W. Biederman void setup_sysctl_set(struct ctl_table_set *set, 16909eb47c26SEric W. Biederman struct ctl_table_root *root, 16911f87f0b5SEric W. Biederman int (*is_seen)(struct ctl_table_set *)) 16921f87f0b5SEric W. Biederman { 16931347440dSDan Carpenter memset(set, 0, sizeof(*set)); 16940e47c99dSEric W. Biederman set->is_seen = is_seen; 1695ac13ac6fSEric W. Biederman init_header(&set->dir.header, root, set, NULL, root_table); 16961f87f0b5SEric W. Biederman } 16971f87f0b5SEric W. Biederman 169897324cd8SEric W. Biederman void retire_sysctl_set(struct ctl_table_set *set) 169997324cd8SEric W. Biederman { 1700ac13ac6fSEric W. Biederman WARN_ON(!RB_EMPTY_ROOT(&set->dir.root)); 170197324cd8SEric W. Biederman } 17021f87f0b5SEric W. Biederman 17031e0edd3fSAlexey Dobriyan int __init proc_sys_init(void) 170477b14db5SEric W. Biederman { 1705e1675231SAlexey Dobriyan struct proc_dir_entry *proc_sys_root; 1706e1675231SAlexey Dobriyan 170777b14db5SEric W. Biederman proc_sys_root = proc_mkdir("sys", NULL); 17089043476fSAl Viro proc_sys_root->proc_iops = &proc_sys_dir_operations; 1709d56c0d45SAlexey Dobriyan proc_sys_root->proc_dir_ops = &proc_sys_dir_file_operations; 171077b14db5SEric W. Biederman proc_sys_root->nlink = 0; 1711de4e83bdSEric W. Biederman 1712de4e83bdSEric W. Biederman return sysctl_init(); 171377b14db5SEric W. Biederman } 17143db978d4SVlastimil Babka 17150a477e1aSVlastimil Babka struct sysctl_alias { 17160a477e1aSVlastimil Babka const char *kernel_param; 17170a477e1aSVlastimil Babka const char *sysctl_param; 17180a477e1aSVlastimil Babka }; 17190a477e1aSVlastimil Babka 17200a477e1aSVlastimil Babka /* 17210a477e1aSVlastimil Babka * Historically some settings had both sysctl and a command line parameter. 17220a477e1aSVlastimil Babka * With the generic sysctl. parameter support, we can handle them at a single 17230a477e1aSVlastimil Babka * place and only keep the historical name for compatibility. This is not meant 17240a477e1aSVlastimil Babka * to add brand new aliases. When adding existing aliases, consider whether 17250a477e1aSVlastimil Babka * the possibly different moment of changing the value (e.g. from early_param 17260a477e1aSVlastimil Babka * to the moment do_sysctl_args() is called) is an issue for the specific 17270a477e1aSVlastimil Babka * parameter. 17280a477e1aSVlastimil Babka */ 17290a477e1aSVlastimil Babka static const struct sysctl_alias sysctl_aliases[] = { 1730f117955aSGuilherme G. Piccoli {"hardlockup_all_cpu_backtrace", "kernel.hardlockup_all_cpu_backtrace" }, 1731b467f3efSVlastimil Babka {"hung_task_panic", "kernel.hung_task_panic" }, 1732f117955aSGuilherme G. Piccoli {"numa_zonelist_order", "vm.numa_zonelist_order" }, 1733f117955aSGuilherme G. Piccoli {"softlockup_all_cpu_backtrace", "kernel.softlockup_all_cpu_backtrace" }, 1734f117955aSGuilherme G. Piccoli {"softlockup_panic", "kernel.softlockup_panic" }, 17350a477e1aSVlastimil Babka { } 17360a477e1aSVlastimil Babka }; 17370a477e1aSVlastimil Babka 17380a477e1aSVlastimil Babka static const char *sysctl_find_alias(char *param) 17390a477e1aSVlastimil Babka { 17400a477e1aSVlastimil Babka const struct sysctl_alias *alias; 17410a477e1aSVlastimil Babka 17420a477e1aSVlastimil Babka for (alias = &sysctl_aliases[0]; alias->kernel_param != NULL; alias++) { 17430a477e1aSVlastimil Babka if (strcmp(alias->kernel_param, param) == 0) 17440a477e1aSVlastimil Babka return alias->sysctl_param; 17450a477e1aSVlastimil Babka } 17460a477e1aSVlastimil Babka 17470a477e1aSVlastimil Babka return NULL; 17480a477e1aSVlastimil Babka } 17490a477e1aSVlastimil Babka 17503db978d4SVlastimil Babka /* Set sysctl value passed on kernel command line. */ 17513db978d4SVlastimil Babka static int process_sysctl_arg(char *param, char *val, 17523db978d4SVlastimil Babka const char *unused, void *arg) 17533db978d4SVlastimil Babka { 17543db978d4SVlastimil Babka char *path; 17553db978d4SVlastimil Babka struct vfsmount **proc_mnt = arg; 17563db978d4SVlastimil Babka struct file_system_type *proc_fs_type; 17573db978d4SVlastimil Babka struct file *file; 17583db978d4SVlastimil Babka int len; 17593db978d4SVlastimil Babka int err; 17603db978d4SVlastimil Babka loff_t pos = 0; 17613db978d4SVlastimil Babka ssize_t wret; 17623db978d4SVlastimil Babka 17630a477e1aSVlastimil Babka if (strncmp(param, "sysctl", sizeof("sysctl") - 1) == 0) { 17643db978d4SVlastimil Babka param += sizeof("sysctl") - 1; 17653db978d4SVlastimil Babka 17663db978d4SVlastimil Babka if (param[0] != '/' && param[0] != '.') 17673db978d4SVlastimil Babka return 0; 17683db978d4SVlastimil Babka 17693db978d4SVlastimil Babka param++; 17700a477e1aSVlastimil Babka } else { 17710a477e1aSVlastimil Babka param = (char *) sysctl_find_alias(param); 17720a477e1aSVlastimil Babka if (!param) 17730a477e1aSVlastimil Babka return 0; 17740a477e1aSVlastimil Babka } 17753db978d4SVlastimil Babka 1776697edcb0SXiaoming Ni if (!val) 1777697edcb0SXiaoming Ni return -EINVAL; 1778697edcb0SXiaoming Ni len = strlen(val); 1779697edcb0SXiaoming Ni if (len == 0) 1780697edcb0SXiaoming Ni return -EINVAL; 1781697edcb0SXiaoming Ni 17823db978d4SVlastimil Babka /* 17833db978d4SVlastimil Babka * To set sysctl options, we use a temporary mount of proc, look up the 17843db978d4SVlastimil Babka * respective sys/ file and write to it. To avoid mounting it when no 17853db978d4SVlastimil Babka * options were given, we mount it only when the first sysctl option is 17863db978d4SVlastimil Babka * found. Why not a persistent mount? There are problems with a 17873db978d4SVlastimil Babka * persistent mount of proc in that it forces userspace not to use any 17883db978d4SVlastimil Babka * proc mount options. 17893db978d4SVlastimil Babka */ 17903db978d4SVlastimil Babka if (!*proc_mnt) { 17913db978d4SVlastimil Babka proc_fs_type = get_fs_type("proc"); 17923db978d4SVlastimil Babka if (!proc_fs_type) { 17933db978d4SVlastimil Babka pr_err("Failed to find procfs to set sysctl from command line\n"); 17943db978d4SVlastimil Babka return 0; 17953db978d4SVlastimil Babka } 17963db978d4SVlastimil Babka *proc_mnt = kern_mount(proc_fs_type); 17973db978d4SVlastimil Babka put_filesystem(proc_fs_type); 17983db978d4SVlastimil Babka if (IS_ERR(*proc_mnt)) { 17993db978d4SVlastimil Babka pr_err("Failed to mount procfs to set sysctl from command line\n"); 18003db978d4SVlastimil Babka return 0; 18013db978d4SVlastimil Babka } 18023db978d4SVlastimil Babka } 18033db978d4SVlastimil Babka 18043db978d4SVlastimil Babka path = kasprintf(GFP_KERNEL, "sys/%s", param); 18053db978d4SVlastimil Babka if (!path) 18063db978d4SVlastimil Babka panic("%s: Failed to allocate path for %s\n", __func__, param); 18073db978d4SVlastimil Babka strreplace(path, '.', '/'); 18083db978d4SVlastimil Babka 1809*ffb37ca3SAl Viro file = file_open_root_mnt(*proc_mnt, path, O_WRONLY, 0); 18103db978d4SVlastimil Babka if (IS_ERR(file)) { 18113db978d4SVlastimil Babka err = PTR_ERR(file); 18123db978d4SVlastimil Babka if (err == -ENOENT) 18133db978d4SVlastimil Babka pr_err("Failed to set sysctl parameter '%s=%s': parameter not found\n", 18143db978d4SVlastimil Babka param, val); 18153db978d4SVlastimil Babka else if (err == -EACCES) 18163db978d4SVlastimil Babka pr_err("Failed to set sysctl parameter '%s=%s': permission denied (read-only?)\n", 18173db978d4SVlastimil Babka param, val); 18183db978d4SVlastimil Babka else 18193db978d4SVlastimil Babka pr_err("Error %pe opening proc file to set sysctl parameter '%s=%s'\n", 18203db978d4SVlastimil Babka file, param, val); 18213db978d4SVlastimil Babka goto out; 18223db978d4SVlastimil Babka } 18233db978d4SVlastimil Babka wret = kernel_write(file, val, len, &pos); 18243db978d4SVlastimil Babka if (wret < 0) { 18253db978d4SVlastimil Babka err = wret; 18263db978d4SVlastimil Babka if (err == -EINVAL) 18273db978d4SVlastimil Babka pr_err("Failed to set sysctl parameter '%s=%s': invalid value\n", 18283db978d4SVlastimil Babka param, val); 18293db978d4SVlastimil Babka else 18303db978d4SVlastimil Babka pr_err("Error %pe writing to proc file to set sysctl parameter '%s=%s'\n", 18313db978d4SVlastimil Babka ERR_PTR(err), param, val); 18323db978d4SVlastimil Babka } else if (wret != len) { 18333db978d4SVlastimil Babka pr_err("Wrote only %zd bytes of %d writing to proc file %s to set sysctl parameter '%s=%s\n", 18343db978d4SVlastimil Babka wret, len, path, param, val); 18353db978d4SVlastimil Babka } 18363db978d4SVlastimil Babka 18373db978d4SVlastimil Babka err = filp_close(file, NULL); 18383db978d4SVlastimil Babka if (err) 18393db978d4SVlastimil Babka pr_err("Error %pe closing proc file to set sysctl parameter '%s=%s\n", 18403db978d4SVlastimil Babka ERR_PTR(err), param, val); 18413db978d4SVlastimil Babka out: 18423db978d4SVlastimil Babka kfree(path); 18433db978d4SVlastimil Babka return 0; 18443db978d4SVlastimil Babka } 18453db978d4SVlastimil Babka 18463db978d4SVlastimil Babka void do_sysctl_args(void) 18473db978d4SVlastimil Babka { 18483db978d4SVlastimil Babka char *command_line; 18493db978d4SVlastimil Babka struct vfsmount *proc_mnt = NULL; 18503db978d4SVlastimil Babka 18513db978d4SVlastimil Babka command_line = kstrdup(saved_command_line, GFP_KERNEL); 18523db978d4SVlastimil Babka if (!command_line) 18533db978d4SVlastimil Babka panic("%s: Failed to allocate copy of command line\n", __func__); 18543db978d4SVlastimil Babka 18553db978d4SVlastimil Babka parse_args("Setting sysctl args", command_line, 18563db978d4SVlastimil Babka NULL, 0, -1, -1, &proc_mnt, process_sysctl_arg); 18573db978d4SVlastimil Babka 18583db978d4SVlastimil Babka if (proc_mnt) 18593db978d4SVlastimil Babka kern_unmount(proc_mnt); 18603db978d4SVlastimil Babka 18613db978d4SVlastimil Babka kfree(command_line); 18623db978d4SVlastimil Babka } 1863