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> 193ddd9a80SXiaoming Ni #include <linux/kmemleak.h> 2077b14db5SEric W. Biederman #include "internal.h" 2177b14db5SEric W. Biederman 2218d4b42eSJoel Granados #define list_for_each_table_entry(entry, header) \ 2353f3811dSJoel Granados entry = header->ctl_table; \ 2453f3811dSJoel Granados for (size_t i = 0 ; i < header->ctl_table_size && entry->procname; ++i, entry++) 25cb55f27aSMeng Tang 26d72f71ebSAl Viro static const struct dentry_operations proc_sys_dentry_operations; 2777b14db5SEric W. Biederman static const struct file_operations proc_sys_file_operations; 2803a44825SJan Engelhardt static const struct inode_operations proc_sys_inode_operations; 299043476fSAl Viro static const struct file_operations proc_sys_dir_file_operations; 309043476fSAl Viro static const struct inode_operations proc_sys_dir_operations; 3177b14db5SEric W. Biederman 32f9bd6733SEric W. Biederman /* Support for permanently empty directories */ 337fffbc71STom Rix static struct ctl_table sysctl_mount_point[] = { 342f2665c1SJoel Granados {.type = SYSCTL_TABLE_TYPE_PERMANENTLY_EMPTY } 35f9bd6733SEric W. Biederman }; 36f9bd6733SEric W. Biederman 37ee9efac4SLuis Chamberlain /** 38ee9efac4SLuis Chamberlain * register_sysctl_mount_point() - registers a sysctl mount point 39ee9efac4SLuis Chamberlain * @path: path for the mount point 40ee9efac4SLuis Chamberlain * 41ee9efac4SLuis Chamberlain * Used to create a permanently empty directory to serve as mount point. 42ee9efac4SLuis Chamberlain * There are some subtle but important permission checks this allows in the 43ee9efac4SLuis Chamberlain * case of unprivileged mounts. 44ee9efac4SLuis Chamberlain */ 45ee9efac4SLuis Chamberlain struct ctl_table_header *register_sysctl_mount_point(const char *path) 46ee9efac4SLuis Chamberlain { 479edbfe92SJoel Granados return register_sysctl_sz(path, sysctl_mount_point, 0); 48ee9efac4SLuis Chamberlain } 49ee9efac4SLuis Chamberlain EXPORT_SYMBOL(register_sysctl_mount_point); 50ee9efac4SLuis Chamberlain 512f2665c1SJoel Granados #define sysctl_is_perm_empty_ctl_table(tptr) \ 522f2665c1SJoel Granados (tptr[0].type == SYSCTL_TABLE_TYPE_PERMANENTLY_EMPTY) 532f2665c1SJoel Granados #define sysctl_is_perm_empty_ctl_header(hptr) \ 542f2665c1SJoel Granados (sysctl_is_perm_empty_ctl_table(hptr->ctl_table)) 552f2665c1SJoel Granados #define sysctl_set_perm_empty_ctl_header(hptr) \ 562f2665c1SJoel Granados (hptr->ctl_table[0].type = SYSCTL_TABLE_TYPE_PERMANENTLY_EMPTY) 572f2665c1SJoel Granados #define sysctl_clear_perm_empty_ctl_header(hptr) \ 582f2665c1SJoel Granados (hptr->ctl_table[0].type = SYSCTL_TABLE_TYPE_DEFAULT) 59f9bd6733SEric W. Biederman 60f1ecf068SLucas De Marchi void proc_sys_poll_notify(struct ctl_table_poll *poll) 61f1ecf068SLucas De Marchi { 62f1ecf068SLucas De Marchi if (!poll) 63f1ecf068SLucas De Marchi return; 64f1ecf068SLucas De Marchi 65f1ecf068SLucas De Marchi atomic_inc(&poll->event); 66f1ecf068SLucas De Marchi wake_up_interruptible(&poll->wait); 67f1ecf068SLucas De Marchi } 68f1ecf068SLucas De Marchi 69a194558eSEric W. Biederman static struct ctl_table root_table[] = { 70a194558eSEric W. Biederman { 71a194558eSEric W. Biederman .procname = "", 727ec66d06SEric W. Biederman .mode = S_IFDIR|S_IRUGO|S_IXUGO, 73a194558eSEric W. Biederman }, 74a194558eSEric W. Biederman { } 75a194558eSEric W. Biederman }; 760e47c99dSEric W. Biederman static struct ctl_table_root sysctl_table_root = { 770e47c99dSEric W. Biederman .default_set.dir.header = { 781f87f0b5SEric W. Biederman {{.count = 1, 79938aaa4fSEric W. Biederman .nreg = 1, 80ac13ac6fSEric W. Biederman .ctl_table = root_table }}, 810e47c99dSEric W. Biederman .ctl_table_arg = root_table, 821f87f0b5SEric W. Biederman .root = &sysctl_table_root, 831f87f0b5SEric W. Biederman .set = &sysctl_table_root.default_set, 847ec66d06SEric W. Biederman }, 851f87f0b5SEric W. Biederman }; 861f87f0b5SEric W. Biederman 871f87f0b5SEric W. Biederman static DEFINE_SPINLOCK(sysctl_lock); 881f87f0b5SEric W. Biederman 897ec66d06SEric W. Biederman static void drop_sysctl_table(struct ctl_table_header *header); 900e47c99dSEric W. Biederman static int sysctl_follow_link(struct ctl_table_header **phead, 9113bcc6a2SEric W. Biederman struct ctl_table **pentry); 920e47c99dSEric W. Biederman static int insert_links(struct ctl_table_header *head); 930e47c99dSEric W. Biederman static void put_links(struct ctl_table_header *header); 947ec66d06SEric W. Biederman 956980128fSEric W. Biederman static void sysctl_print_dir(struct ctl_dir *dir) 966980128fSEric W. Biederman { 976980128fSEric W. Biederman if (dir->header.parent) 986980128fSEric W. Biederman sysctl_print_dir(dir->header.parent); 9987ebdc00SAndrew Morton pr_cont("%s/", dir->header.ctl_table[0].procname); 1006980128fSEric W. Biederman } 1016980128fSEric W. Biederman 102076c3eedSEric W. Biederman static int namecmp(const char *name1, int len1, const char *name2, int len2) 103076c3eedSEric W. Biederman { 104076c3eedSEric W. Biederman int cmp; 105076c3eedSEric W. Biederman 106d8fc9b66SMasahiro Yamada cmp = memcmp(name1, name2, min(len1, len2)); 107076c3eedSEric W. Biederman if (cmp == 0) 108076c3eedSEric W. Biederman cmp = len1 - len2; 109076c3eedSEric W. Biederman return cmp; 110076c3eedSEric W. Biederman } 111076c3eedSEric W. Biederman 11260f126d9SEric W. Biederman /* Called under sysctl_lock */ 113076c3eedSEric W. Biederman static struct ctl_table *find_entry(struct ctl_table_header **phead, 1140e47c99dSEric W. Biederman struct ctl_dir *dir, const char *name, int namelen) 115076c3eedSEric W. Biederman { 116076c3eedSEric W. Biederman struct ctl_table_header *head; 117076c3eedSEric W. Biederman struct ctl_table *entry; 118ac13ac6fSEric W. Biederman struct rb_node *node = dir->root.rb_node; 119076c3eedSEric W. Biederman 120ac13ac6fSEric W. Biederman while (node) 121ac13ac6fSEric W. Biederman { 122ac13ac6fSEric W. Biederman struct ctl_node *ctl_node; 123ac13ac6fSEric W. Biederman const char *procname; 124ac13ac6fSEric W. Biederman int cmp; 125ac13ac6fSEric W. Biederman 126ac13ac6fSEric W. Biederman ctl_node = rb_entry(node, struct ctl_node, node); 127ac13ac6fSEric W. Biederman head = ctl_node->header; 128ac13ac6fSEric W. Biederman entry = &head->ctl_table[ctl_node - head->node]; 129ac13ac6fSEric W. Biederman procname = entry->procname; 130ac13ac6fSEric W. Biederman 131ac13ac6fSEric W. Biederman cmp = namecmp(name, namelen, procname, strlen(procname)); 132ac13ac6fSEric W. Biederman if (cmp < 0) 133ac13ac6fSEric W. Biederman node = node->rb_left; 134ac13ac6fSEric W. Biederman else if (cmp > 0) 135ac13ac6fSEric W. Biederman node = node->rb_right; 136ac13ac6fSEric W. Biederman else { 137076c3eedSEric W. Biederman *phead = head; 138076c3eedSEric W. Biederman return entry; 139076c3eedSEric W. Biederman } 140076c3eedSEric W. Biederman } 141076c3eedSEric W. Biederman return NULL; 142076c3eedSEric W. Biederman } 143076c3eedSEric W. Biederman 144ac13ac6fSEric W. Biederman static int insert_entry(struct ctl_table_header *head, struct ctl_table *entry) 145ac13ac6fSEric W. Biederman { 146ac13ac6fSEric W. Biederman struct rb_node *node = &head->node[entry - head->ctl_table].node; 147ac13ac6fSEric W. Biederman struct rb_node **p = &head->parent->root.rb_node; 148ac13ac6fSEric W. Biederman struct rb_node *parent = NULL; 149ac13ac6fSEric W. Biederman const char *name = entry->procname; 150ac13ac6fSEric W. Biederman int namelen = strlen(name); 151ac13ac6fSEric W. Biederman 152ac13ac6fSEric W. Biederman while (*p) { 153ac13ac6fSEric W. Biederman struct ctl_table_header *parent_head; 154ac13ac6fSEric W. Biederman struct ctl_table *parent_entry; 155ac13ac6fSEric W. Biederman struct ctl_node *parent_node; 156ac13ac6fSEric W. Biederman const char *parent_name; 157ac13ac6fSEric W. Biederman int cmp; 158ac13ac6fSEric W. Biederman 159ac13ac6fSEric W. Biederman parent = *p; 160ac13ac6fSEric W. Biederman parent_node = rb_entry(parent, struct ctl_node, node); 161ac13ac6fSEric W. Biederman parent_head = parent_node->header; 162ac13ac6fSEric W. Biederman parent_entry = &parent_head->ctl_table[parent_node - parent_head->node]; 163ac13ac6fSEric W. Biederman parent_name = parent_entry->procname; 164ac13ac6fSEric W. Biederman 165ac13ac6fSEric W. Biederman cmp = namecmp(name, namelen, parent_name, strlen(parent_name)); 166ac13ac6fSEric W. Biederman if (cmp < 0) 167ac13ac6fSEric W. Biederman p = &(*p)->rb_left; 168ac13ac6fSEric W. Biederman else if (cmp > 0) 169ac13ac6fSEric W. Biederman p = &(*p)->rb_right; 170ac13ac6fSEric W. Biederman else { 17187ebdc00SAndrew Morton pr_err("sysctl duplicate entry: "); 172ac13ac6fSEric W. Biederman sysctl_print_dir(head->parent); 173153ee1c4SGeert Uytterhoeven pr_cont("%s\n", entry->procname); 174ac13ac6fSEric W. Biederman return -EEXIST; 175ac13ac6fSEric W. Biederman } 176ac13ac6fSEric W. Biederman } 177ac13ac6fSEric W. Biederman 178ac13ac6fSEric W. Biederman rb_link_node(node, parent, p); 179ea5272f5SMichel Lespinasse rb_insert_color(node, &head->parent->root); 180ac13ac6fSEric W. Biederman return 0; 181ac13ac6fSEric W. Biederman } 182ac13ac6fSEric W. Biederman 183ac13ac6fSEric W. Biederman static void erase_entry(struct ctl_table_header *head, struct ctl_table *entry) 184ac13ac6fSEric W. Biederman { 185ac13ac6fSEric W. Biederman struct rb_node *node = &head->node[entry - head->ctl_table].node; 186ac13ac6fSEric W. Biederman 187ac13ac6fSEric W. Biederman rb_erase(node, &head->parent->root); 188ac13ac6fSEric W. Biederman } 189ac13ac6fSEric W. Biederman 190e0d04529SEric W. Biederman static void init_header(struct ctl_table_header *head, 191e0d04529SEric W. Biederman struct ctl_table_root *root, struct ctl_table_set *set, 192b1f01e2bSJoel Granados struct ctl_node *node, struct ctl_table *table, size_t table_size) 193e0d04529SEric W. Biederman { 1947ec66d06SEric W. Biederman head->ctl_table = table; 195b1f01e2bSJoel Granados head->ctl_table_size = table_size; 196e0d04529SEric W. Biederman head->ctl_table_arg = table; 197e0d04529SEric W. Biederman head->used = 0; 198e0d04529SEric W. Biederman head->count = 1; 199e0d04529SEric W. Biederman head->nreg = 1; 200e0d04529SEric W. Biederman head->unregistering = NULL; 201e0d04529SEric W. Biederman head->root = root; 202e0d04529SEric W. Biederman head->set = set; 203e0d04529SEric W. Biederman head->parent = NULL; 204ac13ac6fSEric W. Biederman head->node = node; 2052fd1d2c4SEric W. Biederman INIT_HLIST_HEAD(&head->inodes); 206ac13ac6fSEric W. Biederman if (node) { 207ac13ac6fSEric W. Biederman struct ctl_table *entry; 208cb55f27aSMeng Tang 20918d4b42eSJoel Granados list_for_each_table_entry(entry, head) { 210ac13ac6fSEric W. Biederman node->header = head; 211cb55f27aSMeng Tang node++; 212cb55f27aSMeng Tang } 213ac13ac6fSEric W. Biederman } 214ac13ac6fSEric W. Biederman } 215e0d04529SEric W. Biederman 2168425d6aaSEric W. Biederman static void erase_header(struct ctl_table_header *head) 2178425d6aaSEric W. Biederman { 218ac13ac6fSEric W. Biederman struct ctl_table *entry; 219cb55f27aSMeng Tang 22018d4b42eSJoel Granados list_for_each_table_entry(entry, head) 221ac13ac6fSEric W. Biederman erase_entry(head, entry); 2228425d6aaSEric W. Biederman } 2238425d6aaSEric W. Biederman 2240e47c99dSEric W. Biederman static int insert_header(struct ctl_dir *dir, struct ctl_table_header *header) 2258425d6aaSEric W. Biederman { 226ac13ac6fSEric W. Biederman struct ctl_table *entry; 2272f2665c1SJoel Granados struct ctl_table_header *dir_h = &dir->header; 2280e47c99dSEric W. Biederman int err; 2290e47c99dSEric W. Biederman 2302f2665c1SJoel Granados 231f9bd6733SEric W. Biederman /* Is this a permanently empty directory? */ 2322f2665c1SJoel Granados if (sysctl_is_perm_empty_ctl_header(dir_h)) 233f9bd6733SEric W. Biederman return -EROFS; 234f9bd6733SEric W. Biederman 235f9bd6733SEric W. Biederman /* Am I creating a permanently empty directory? */ 2362f2665c1SJoel Granados if (sysctl_is_perm_empty_ctl_table(header->ctl_table)) { 237f9bd6733SEric W. Biederman if (!RB_EMPTY_ROOT(&dir->root)) 238f9bd6733SEric W. Biederman return -EINVAL; 2392f2665c1SJoel Granados sysctl_set_perm_empty_ctl_header(dir_h); 240f9bd6733SEric W. Biederman } 241f9bd6733SEric W. Biederman 2422f2665c1SJoel Granados dir_h->nreg++; 2437ec66d06SEric W. Biederman header->parent = dir; 2440e47c99dSEric W. Biederman err = insert_links(header); 2450e47c99dSEric W. Biederman if (err) 2460e47c99dSEric W. Biederman goto fail_links; 24718d4b42eSJoel Granados list_for_each_table_entry(entry, header) { 248ac13ac6fSEric W. Biederman err = insert_entry(header, entry); 249ac13ac6fSEric W. Biederman if (err) 250ac13ac6fSEric W. Biederman goto fail; 251ac13ac6fSEric W. Biederman } 2520e47c99dSEric W. Biederman return 0; 253ac13ac6fSEric W. Biederman fail: 254ac13ac6fSEric W. Biederman erase_header(header); 255ac13ac6fSEric W. Biederman put_links(header); 2560e47c99dSEric W. Biederman fail_links: 257f9bd6733SEric W. Biederman if (header->ctl_table == sysctl_mount_point) 2582f2665c1SJoel Granados sysctl_clear_perm_empty_ctl_header(dir_h); 2590e47c99dSEric W. Biederman header->parent = NULL; 2602f2665c1SJoel Granados drop_sysctl_table(dir_h); 2610e47c99dSEric W. Biederman return err; 2628425d6aaSEric W. Biederman } 2638425d6aaSEric W. Biederman 2641f87f0b5SEric W. Biederman /* called under sysctl_lock */ 2651f87f0b5SEric W. Biederman static int use_table(struct ctl_table_header *p) 2661f87f0b5SEric W. Biederman { 2671f87f0b5SEric W. Biederman if (unlikely(p->unregistering)) 2681f87f0b5SEric W. Biederman return 0; 2691f87f0b5SEric W. Biederman p->used++; 2701f87f0b5SEric W. Biederman return 1; 2711f87f0b5SEric W. Biederman } 2721f87f0b5SEric W. Biederman 2731f87f0b5SEric W. Biederman /* called under sysctl_lock */ 2741f87f0b5SEric W. Biederman static void unuse_table(struct ctl_table_header *p) 2751f87f0b5SEric W. Biederman { 2761f87f0b5SEric W. Biederman if (!--p->used) 2771f87f0b5SEric W. Biederman if (unlikely(p->unregistering)) 2781f87f0b5SEric W. Biederman complete(p->unregistering); 2791f87f0b5SEric W. Biederman } 2801f87f0b5SEric W. Biederman 281f90f3cafSEric W. Biederman static void proc_sys_invalidate_dcache(struct ctl_table_header *head) 282d6cffbbeSKonstantin Khlebnikov { 283f90f3cafSEric W. Biederman proc_invalidate_siblings_dcache(&head->inodes, &sysctl_lock); 284d6cffbbeSKonstantin Khlebnikov } 285d6cffbbeSKonstantin Khlebnikov 2861f87f0b5SEric W. Biederman /* called under sysctl_lock, will reacquire if has to wait */ 2871f87f0b5SEric W. Biederman static void start_unregistering(struct ctl_table_header *p) 2881f87f0b5SEric W. Biederman { 2891f87f0b5SEric W. Biederman /* 2901f87f0b5SEric W. Biederman * if p->used is 0, nobody will ever touch that entry again; 2911f87f0b5SEric W. Biederman * we'll eliminate all paths to it before dropping sysctl_lock 2921f87f0b5SEric W. Biederman */ 2931f87f0b5SEric W. Biederman if (unlikely(p->used)) { 2941f87f0b5SEric W. Biederman struct completion wait; 2951f87f0b5SEric W. Biederman init_completion(&wait); 2961f87f0b5SEric W. Biederman p->unregistering = &wait; 2971f87f0b5SEric W. Biederman spin_unlock(&sysctl_lock); 2981f87f0b5SEric W. Biederman wait_for_completion(&wait); 2991f87f0b5SEric W. Biederman } else { 3001f87f0b5SEric W. Biederman /* anything non-NULL; we'll never dereference it */ 3011f87f0b5SEric W. Biederman p->unregistering = ERR_PTR(-EINVAL); 302ace0c791SEric W. Biederman spin_unlock(&sysctl_lock); 3031f87f0b5SEric W. Biederman } 3041f87f0b5SEric W. Biederman /* 305f90f3cafSEric W. Biederman * Invalidate dentries for unregistered sysctls: namespaced sysctls 306d6cffbbeSKonstantin Khlebnikov * can have duplicate names and contaminate dcache very badly. 307d6cffbbeSKonstantin Khlebnikov */ 308f90f3cafSEric W. Biederman proc_sys_invalidate_dcache(p); 309d6cffbbeSKonstantin Khlebnikov /* 3101f87f0b5SEric W. Biederman * do not remove from the list until nobody holds it; walking the 3111f87f0b5SEric W. Biederman * list in do_sysctl() relies on that. 3121f87f0b5SEric W. Biederman */ 313ace0c791SEric W. Biederman spin_lock(&sysctl_lock); 3148425d6aaSEric W. Biederman erase_header(p); 3151f87f0b5SEric W. Biederman } 3161f87f0b5SEric W. Biederman 3171f87f0b5SEric W. Biederman static struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head) 3181f87f0b5SEric W. Biederman { 319ab4a1f24SPrasad Joshi BUG_ON(!head); 3201f87f0b5SEric W. Biederman spin_lock(&sysctl_lock); 3211f87f0b5SEric W. Biederman if (!use_table(head)) 3221f87f0b5SEric W. Biederman head = ERR_PTR(-ENOENT); 3231f87f0b5SEric W. Biederman spin_unlock(&sysctl_lock); 3241f87f0b5SEric W. Biederman return head; 3251f87f0b5SEric W. Biederman } 3261f87f0b5SEric W. Biederman 3271f87f0b5SEric W. Biederman static void sysctl_head_finish(struct ctl_table_header *head) 3281f87f0b5SEric W. Biederman { 3291f87f0b5SEric W. Biederman if (!head) 3301f87f0b5SEric W. Biederman return; 3311f87f0b5SEric W. Biederman spin_lock(&sysctl_lock); 3321f87f0b5SEric W. Biederman unuse_table(head); 3331f87f0b5SEric W. Biederman spin_unlock(&sysctl_lock); 3341f87f0b5SEric W. Biederman } 3351f87f0b5SEric W. Biederman 3361f87f0b5SEric W. Biederman static struct ctl_table_set * 33713bcc6a2SEric W. Biederman lookup_header_set(struct ctl_table_root *root) 3381f87f0b5SEric W. Biederman { 3391f87f0b5SEric W. Biederman struct ctl_table_set *set = &root->default_set; 3401f87f0b5SEric W. Biederman if (root->lookup) 34113bcc6a2SEric W. Biederman set = root->lookup(root); 3421f87f0b5SEric W. Biederman return set; 3431f87f0b5SEric W. Biederman } 3441f87f0b5SEric W. Biederman 345076c3eedSEric W. Biederman static struct ctl_table *lookup_entry(struct ctl_table_header **phead, 3467ec66d06SEric W. Biederman struct ctl_dir *dir, 347076c3eedSEric W. Biederman const char *name, int namelen) 348076c3eedSEric W. Biederman { 349076c3eedSEric W. Biederman struct ctl_table_header *head; 350076c3eedSEric W. Biederman struct ctl_table *entry; 351076c3eedSEric W. Biederman 352076c3eedSEric W. Biederman spin_lock(&sysctl_lock); 3530e47c99dSEric W. Biederman entry = find_entry(&head, dir, name, namelen); 354076c3eedSEric W. Biederman if (entry && use_table(head)) 355076c3eedSEric W. Biederman *phead = head; 356076c3eedSEric W. Biederman else 357076c3eedSEric W. Biederman entry = NULL; 358076c3eedSEric W. Biederman spin_unlock(&sysctl_lock); 359076c3eedSEric W. Biederman return entry; 360076c3eedSEric W. Biederman } 361076c3eedSEric W. Biederman 362ac13ac6fSEric W. Biederman static struct ctl_node *first_usable_entry(struct rb_node *node) 3631f87f0b5SEric W. Biederman { 364ac13ac6fSEric W. Biederman struct ctl_node *ctl_node; 3651f87f0b5SEric W. Biederman 366ac13ac6fSEric W. Biederman for (;node; node = rb_next(node)) { 367ac13ac6fSEric W. Biederman ctl_node = rb_entry(node, struct ctl_node, node); 368ac13ac6fSEric W. Biederman if (use_table(ctl_node->header)) 369ac13ac6fSEric W. Biederman return ctl_node; 3701f87f0b5SEric W. Biederman } 3711f87f0b5SEric W. Biederman return NULL; 3721f87f0b5SEric W. Biederman } 3731f87f0b5SEric W. Biederman 3747ec66d06SEric W. Biederman static void first_entry(struct ctl_dir *dir, 3756a75ce16SEric W. Biederman struct ctl_table_header **phead, struct ctl_table **pentry) 3761f87f0b5SEric W. Biederman { 377ac13ac6fSEric W. Biederman struct ctl_table_header *head = NULL; 3787ec66d06SEric W. Biederman struct ctl_table *entry = NULL; 379ac13ac6fSEric W. Biederman struct ctl_node *ctl_node; 3806a75ce16SEric W. Biederman 3816a75ce16SEric W. Biederman spin_lock(&sysctl_lock); 382ac13ac6fSEric W. Biederman ctl_node = first_usable_entry(rb_first(&dir->root)); 3836a75ce16SEric W. Biederman spin_unlock(&sysctl_lock); 384ac13ac6fSEric W. Biederman if (ctl_node) { 385ac13ac6fSEric W. Biederman head = ctl_node->header; 386ac13ac6fSEric W. Biederman entry = &head->ctl_table[ctl_node - head->node]; 387ac13ac6fSEric W. Biederman } 3886a75ce16SEric W. Biederman *phead = head; 3896a75ce16SEric W. Biederman *pentry = entry; 3906a75ce16SEric W. Biederman } 3916a75ce16SEric W. Biederman 3927ec66d06SEric W. Biederman static void next_entry(struct ctl_table_header **phead, struct ctl_table **pentry) 3936a75ce16SEric W. Biederman { 3946a75ce16SEric W. Biederman struct ctl_table_header *head = *phead; 3956a75ce16SEric W. Biederman struct ctl_table *entry = *pentry; 396ac13ac6fSEric W. Biederman struct ctl_node *ctl_node = &head->node[entry - head->ctl_table]; 3976a75ce16SEric W. Biederman 3986a75ce16SEric W. Biederman spin_lock(&sysctl_lock); 3996a75ce16SEric W. Biederman unuse_table(head); 400ac13ac6fSEric W. Biederman 401ac13ac6fSEric W. Biederman ctl_node = first_usable_entry(rb_next(&ctl_node->node)); 4026a75ce16SEric W. Biederman spin_unlock(&sysctl_lock); 403ac13ac6fSEric W. Biederman head = NULL; 404ac13ac6fSEric W. Biederman if (ctl_node) { 405ac13ac6fSEric W. Biederman head = ctl_node->header; 406ac13ac6fSEric W. Biederman entry = &head->ctl_table[ctl_node - head->node]; 4076a75ce16SEric W. Biederman } 4086a75ce16SEric W. Biederman *phead = head; 4096a75ce16SEric W. Biederman *pentry = entry; 4101f87f0b5SEric W. Biederman } 4111f87f0b5SEric W. Biederman 4121f87f0b5SEric W. Biederman /* 4131f87f0b5SEric W. Biederman * sysctl_perm does NOT grant the superuser all rights automatically, because 4141f87f0b5SEric W. Biederman * some sysctl variables are readonly even to root. 4151f87f0b5SEric W. Biederman */ 4161f87f0b5SEric W. Biederman 4171f87f0b5SEric W. Biederman static int test_perm(int mode, int op) 4181f87f0b5SEric W. Biederman { 419091bd3eaSEric W. Biederman if (uid_eq(current_euid(), GLOBAL_ROOT_UID)) 4201f87f0b5SEric W. Biederman mode >>= 6; 421091bd3eaSEric W. Biederman else if (in_egroup_p(GLOBAL_ROOT_GID)) 4221f87f0b5SEric W. Biederman mode >>= 3; 4231f87f0b5SEric W. Biederman if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0) 4241f87f0b5SEric W. Biederman return 0; 4251f87f0b5SEric W. Biederman return -EACCES; 4261f87f0b5SEric W. Biederman } 4271f87f0b5SEric W. Biederman 42873f7ef43SEric W. Biederman static int sysctl_perm(struct ctl_table_header *head, struct ctl_table *table, int op) 4291f87f0b5SEric W. Biederman { 43073f7ef43SEric W. Biederman struct ctl_table_root *root = head->root; 4311f87f0b5SEric W. Biederman int mode; 4321f87f0b5SEric W. Biederman 4331f87f0b5SEric W. Biederman if (root->permissions) 43473f7ef43SEric W. Biederman mode = root->permissions(head, table); 4351f87f0b5SEric W. Biederman else 4361f87f0b5SEric W. Biederman mode = table->mode; 4371f87f0b5SEric W. Biederman 4381f87f0b5SEric W. Biederman return test_perm(mode, op); 4391f87f0b5SEric W. Biederman } 4401f87f0b5SEric W. Biederman 4419043476fSAl Viro static struct inode *proc_sys_make_inode(struct super_block *sb, 4429043476fSAl Viro struct ctl_table_header *head, struct ctl_table *table) 44377b14db5SEric W. Biederman { 444e79c6a4fSDmitry Torokhov struct ctl_table_root *root = head->root; 44577b14db5SEric W. Biederman struct inode *inode; 4469043476fSAl Viro struct proc_inode *ei; 44777b14db5SEric W. Biederman 4489043476fSAl Viro inode = new_inode(sb); 44977b14db5SEric W. Biederman if (!inode) 450ea5751ccSIvan Delalande return ERR_PTR(-ENOMEM); 45177b14db5SEric W. Biederman 45285fe4025SChristoph Hellwig inode->i_ino = get_next_ino(); 45385fe4025SChristoph Hellwig 45477b14db5SEric W. Biederman ei = PROC_I(inode); 4559043476fSAl Viro 456d6cffbbeSKonstantin Khlebnikov spin_lock(&sysctl_lock); 457ace0c791SEric W. Biederman if (unlikely(head->unregistering)) { 458ace0c791SEric W. Biederman spin_unlock(&sysctl_lock); 459ace0c791SEric W. Biederman iput(inode); 460ea5751ccSIvan Delalande return ERR_PTR(-ENOENT); 461ace0c791SEric W. Biederman } 462ace0c791SEric W. Biederman ei->sysctl = head; 463ace0c791SEric W. Biederman ei->sysctl_entry = table; 4640afa5ca8SEric W. Biederman hlist_add_head_rcu(&ei->sibling_inodes, &head->inodes); 465d6cffbbeSKonstantin Khlebnikov head->count++; 466d6cffbbeSKonstantin Khlebnikov spin_unlock(&sysctl_lock); 467d6cffbbeSKonstantin Khlebnikov 468e9d7d3cbSJeff Layton inode->i_mtime = inode->i_atime = inode_set_ctime_current(inode); 4699043476fSAl Viro inode->i_mode = table->mode; 4707ec66d06SEric W. Biederman if (!S_ISDIR(table->mode)) { 4719043476fSAl Viro inode->i_mode |= S_IFREG; 47277b14db5SEric W. Biederman inode->i_op = &proc_sys_inode_operations; 47377b14db5SEric W. Biederman inode->i_fop = &proc_sys_file_operations; 4749043476fSAl Viro } else { 4759043476fSAl Viro inode->i_mode |= S_IFDIR; 4769043476fSAl Viro inode->i_op = &proc_sys_dir_operations; 4779043476fSAl Viro inode->i_fop = &proc_sys_dir_file_operations; 4782f2665c1SJoel Granados if (sysctl_is_perm_empty_ctl_header(head)) 479f9bd6733SEric W. Biederman make_empty_dir_inode(inode); 4809043476fSAl Viro } 481e79c6a4fSDmitry Torokhov 482e79c6a4fSDmitry Torokhov if (root->set_ownership) 483e79c6a4fSDmitry Torokhov root->set_ownership(head, table, &inode->i_uid, &inode->i_gid); 4845ec27ec7SRadoslaw Burny else { 4855ec27ec7SRadoslaw Burny inode->i_uid = GLOBAL_ROOT_UID; 4865ec27ec7SRadoslaw Burny inode->i_gid = GLOBAL_ROOT_GID; 4875ec27ec7SRadoslaw Burny } 488e79c6a4fSDmitry Torokhov 48977b14db5SEric W. Biederman return inode; 49077b14db5SEric W. Biederman } 49177b14db5SEric W. Biederman 492d6cffbbeSKonstantin Khlebnikov void proc_sys_evict_inode(struct inode *inode, struct ctl_table_header *head) 493d6cffbbeSKonstantin Khlebnikov { 494d6cffbbeSKonstantin Khlebnikov spin_lock(&sysctl_lock); 4950afa5ca8SEric W. Biederman hlist_del_init_rcu(&PROC_I(inode)->sibling_inodes); 496d6cffbbeSKonstantin Khlebnikov if (!--head->count) 497d6cffbbeSKonstantin Khlebnikov kfree_rcu(head, rcu); 498d6cffbbeSKonstantin Khlebnikov spin_unlock(&sysctl_lock); 499d6cffbbeSKonstantin Khlebnikov } 500d6cffbbeSKonstantin Khlebnikov 50181324364SAdrian Bunk static struct ctl_table_header *grab_header(struct inode *inode) 50277b14db5SEric W. Biederman { 5033cc3e046SEric W. Biederman struct ctl_table_header *head = PROC_I(inode)->sysctl; 5043cc3e046SEric W. Biederman if (!head) 5050e47c99dSEric W. Biederman head = &sysctl_table_root.default_set.dir.header; 5063cc3e046SEric W. Biederman return sysctl_head_grab(head); 50777b14db5SEric W. Biederman } 50877b14db5SEric W. Biederman 50977b14db5SEric W. Biederman static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry, 51000cd8dd3SAl Viro unsigned int flags) 51177b14db5SEric W. Biederman { 5129043476fSAl Viro struct ctl_table_header *head = grab_header(dir); 5139043476fSAl Viro struct ctl_table_header *h = NULL; 514dc12e909SAl Viro const struct qstr *name = &dentry->d_name; 5159043476fSAl Viro struct ctl_table *p; 51677b14db5SEric W. Biederman struct inode *inode; 5179043476fSAl Viro struct dentry *err = ERR_PTR(-ENOENT); 5187ec66d06SEric W. Biederman struct ctl_dir *ctl_dir; 5190e47c99dSEric W. Biederman int ret; 52077b14db5SEric W. Biederman 5219043476fSAl Viro if (IS_ERR(head)) 5229043476fSAl Viro return ERR_CAST(head); 5239043476fSAl Viro 5247ec66d06SEric W. Biederman ctl_dir = container_of(head, struct ctl_dir, header); 5259043476fSAl Viro 5267ec66d06SEric W. Biederman p = lookup_entry(&h, ctl_dir, name->name, name->len); 5279043476fSAl Viro if (!p) 52877b14db5SEric W. Biederman goto out; 52977b14db5SEric W. Biederman 5304e757320SEric W. Biederman if (S_ISLNK(p->mode)) { 53113bcc6a2SEric W. Biederman ret = sysctl_follow_link(&h, &p); 5320e47c99dSEric W. Biederman err = ERR_PTR(ret); 5330e47c99dSEric W. Biederman if (ret) 5340e47c99dSEric W. Biederman goto out; 5354e757320SEric W. Biederman } 5360e47c99dSEric W. Biederman 5379043476fSAl Viro inode = proc_sys_make_inode(dir->i_sb, h ? h : head, p); 538ea5751ccSIvan Delalande if (IS_ERR(inode)) { 539ea5751ccSIvan Delalande err = ERR_CAST(inode); 54077b14db5SEric W. Biederman goto out; 541ea5751ccSIvan Delalande } 54277b14db5SEric W. Biederman 543fb045adbSNick Piggin d_set_d_op(dentry, &proc_sys_dentry_operations); 544888e2b03SAl Viro err = d_splice_alias(inode, dentry); 54577b14db5SEric W. Biederman 54677b14db5SEric W. Biederman out: 5476bf61045SFrancesco Ruggeri if (h) 5486bf61045SFrancesco Ruggeri sysctl_head_finish(h); 54977b14db5SEric W. Biederman sysctl_head_finish(head); 55077b14db5SEric W. Biederman return err; 55177b14db5SEric W. Biederman } 55277b14db5SEric W. Biederman 5534bd6a735SMatthew Wilcox (Oracle) static ssize_t proc_sys_call_handler(struct kiocb *iocb, struct iov_iter *iter, 5544bd6a735SMatthew Wilcox (Oracle) int write) 55577b14db5SEric W. Biederman { 5564bd6a735SMatthew Wilcox (Oracle) struct inode *inode = file_inode(iocb->ki_filp); 5579043476fSAl Viro struct ctl_table_header *head = grab_header(inode); 5589043476fSAl Viro struct ctl_table *table = PROC_I(inode)->sysctl_entry; 5594bd6a735SMatthew Wilcox (Oracle) size_t count = iov_iter_count(iter); 5604bd6a735SMatthew Wilcox (Oracle) char *kbuf; 5612a2da53bSDavid Howells ssize_t error; 56277b14db5SEric W. Biederman 5639043476fSAl Viro if (IS_ERR(head)) 5649043476fSAl Viro return PTR_ERR(head); 56577b14db5SEric W. Biederman 56677b14db5SEric W. Biederman /* 56777b14db5SEric W. Biederman * At this point we know that the sysctl was not unregistered 56877b14db5SEric W. Biederman * and won't be until we finish. 56977b14db5SEric W. Biederman */ 57077b14db5SEric W. Biederman error = -EPERM; 57173f7ef43SEric W. Biederman if (sysctl_perm(head, table, write ? MAY_WRITE : MAY_READ)) 57277b14db5SEric W. Biederman goto out; 57377b14db5SEric W. Biederman 5749043476fSAl Viro /* if that can happen at all, it should be -EINVAL, not -EISDIR */ 5759043476fSAl Viro error = -EINVAL; 5769043476fSAl Viro if (!table->proc_handler) 5779043476fSAl Viro goto out; 5789043476fSAl Viro 579ef9d965bSChristoph Hellwig /* don't even try if the size is too large */ 580d4d80e69SMatthew Wilcox (Oracle) error = -ENOMEM; 581d4d80e69SMatthew Wilcox (Oracle) if (count >= KMALLOC_MAX_SIZE) 582d4d80e69SMatthew Wilcox (Oracle) goto out; 58345089437SJosef Bacik kbuf = kvzalloc(count + 1, GFP_KERNEL); 58432927393SChristoph Hellwig if (!kbuf) 58532927393SChristoph Hellwig goto out; 5864bd6a735SMatthew Wilcox (Oracle) 5874bd6a735SMatthew Wilcox (Oracle) if (write) { 5884bd6a735SMatthew Wilcox (Oracle) error = -EFAULT; 5894bd6a735SMatthew Wilcox (Oracle) if (!copy_from_iter_full(kbuf, count, iter)) 5904bd6a735SMatthew Wilcox (Oracle) goto out_free_buf; 5914bd6a735SMatthew Wilcox (Oracle) kbuf[count] = '\0'; 5924e63acdfSAndrey Ignatov } 5934e63acdfSAndrey Ignatov 59432927393SChristoph Hellwig error = BPF_CGROUP_RUN_PROG_SYSCTL(head, table, write, &kbuf, &count, 5954bd6a735SMatthew Wilcox (Oracle) &iocb->ki_pos); 59632927393SChristoph Hellwig if (error) 59732927393SChristoph Hellwig goto out_free_buf; 59832927393SChristoph Hellwig 59932927393SChristoph Hellwig /* careful: calling conventions are nasty here */ 6004bd6a735SMatthew Wilcox (Oracle) error = table->proc_handler(table, write, kbuf, &count, &iocb->ki_pos); 60132927393SChristoph Hellwig if (error) 60232927393SChristoph Hellwig goto out_free_buf; 60332927393SChristoph Hellwig 60432927393SChristoph Hellwig if (!write) { 60532927393SChristoph Hellwig error = -EFAULT; 6064bd6a735SMatthew Wilcox (Oracle) if (copy_to_iter(kbuf, count, iter) < count) 60732927393SChristoph Hellwig goto out_free_buf; 60832927393SChristoph Hellwig } 60932927393SChristoph Hellwig 6104e63acdfSAndrey Ignatov error = count; 61132927393SChristoph Hellwig out_free_buf: 61245089437SJosef Bacik kvfree(kbuf); 61377b14db5SEric W. Biederman out: 61477b14db5SEric W. Biederman sysctl_head_finish(head); 61577b14db5SEric W. Biederman 61677b14db5SEric W. Biederman return error; 61777b14db5SEric W. Biederman } 61877b14db5SEric W. Biederman 6194bd6a735SMatthew Wilcox (Oracle) static ssize_t proc_sys_read(struct kiocb *iocb, struct iov_iter *iter) 6207708bfb1SPavel Emelyanov { 6214bd6a735SMatthew Wilcox (Oracle) return proc_sys_call_handler(iocb, iter, 0); 6227708bfb1SPavel Emelyanov } 6237708bfb1SPavel Emelyanov 6244bd6a735SMatthew Wilcox (Oracle) static ssize_t proc_sys_write(struct kiocb *iocb, struct iov_iter *iter) 62577b14db5SEric W. Biederman { 6264bd6a735SMatthew Wilcox (Oracle) return proc_sys_call_handler(iocb, iter, 1); 62777b14db5SEric W. Biederman } 62877b14db5SEric W. Biederman 629f1ecf068SLucas De Marchi static int proc_sys_open(struct inode *inode, struct file *filp) 630f1ecf068SLucas De Marchi { 6314e474a00SLucas De Marchi struct ctl_table_header *head = grab_header(inode); 632f1ecf068SLucas De Marchi struct ctl_table *table = PROC_I(inode)->sysctl_entry; 633f1ecf068SLucas De Marchi 6344e474a00SLucas De Marchi /* sysctl was unregistered */ 6354e474a00SLucas De Marchi if (IS_ERR(head)) 6364e474a00SLucas De Marchi return PTR_ERR(head); 6374e474a00SLucas De Marchi 638f1ecf068SLucas De Marchi if (table->poll) 639f1ecf068SLucas De Marchi filp->private_data = proc_sys_poll_event(table->poll); 640f1ecf068SLucas De Marchi 6414e474a00SLucas De Marchi sysctl_head_finish(head); 6424e474a00SLucas De Marchi 643f1ecf068SLucas De Marchi return 0; 644f1ecf068SLucas De Marchi } 645f1ecf068SLucas De Marchi 646076ccb76SAl Viro static __poll_t proc_sys_poll(struct file *filp, poll_table *wait) 647f1ecf068SLucas De Marchi { 648496ad9aaSAl Viro struct inode *inode = file_inode(filp); 6494e474a00SLucas De Marchi struct ctl_table_header *head = grab_header(inode); 650f1ecf068SLucas De Marchi struct ctl_table *table = PROC_I(inode)->sysctl_entry; 651076ccb76SAl Viro __poll_t ret = DEFAULT_POLLMASK; 6524e474a00SLucas De Marchi unsigned long event; 6534e474a00SLucas De Marchi 6544e474a00SLucas De Marchi /* sysctl was unregistered */ 6554e474a00SLucas De Marchi if (IS_ERR(head)) 656a9a08845SLinus Torvalds return EPOLLERR | EPOLLHUP; 657f1ecf068SLucas De Marchi 658f1ecf068SLucas De Marchi if (!table->proc_handler) 659f1ecf068SLucas De Marchi goto out; 660f1ecf068SLucas De Marchi 661f1ecf068SLucas De Marchi if (!table->poll) 662f1ecf068SLucas De Marchi goto out; 663f1ecf068SLucas De Marchi 6644e474a00SLucas De Marchi event = (unsigned long)filp->private_data; 665f1ecf068SLucas De Marchi poll_wait(filp, &table->poll->wait, wait); 666f1ecf068SLucas De Marchi 667f1ecf068SLucas De Marchi if (event != atomic_read(&table->poll->event)) { 668f1ecf068SLucas De Marchi filp->private_data = proc_sys_poll_event(table->poll); 669a9a08845SLinus Torvalds ret = EPOLLIN | EPOLLRDNORM | EPOLLERR | EPOLLPRI; 670f1ecf068SLucas De Marchi } 671f1ecf068SLucas De Marchi 672f1ecf068SLucas De Marchi out: 6734e474a00SLucas De Marchi sysctl_head_finish(head); 6744e474a00SLucas De Marchi 675f1ecf068SLucas De Marchi return ret; 676f1ecf068SLucas De Marchi } 67777b14db5SEric W. Biederman 678f0c3b509SAl Viro static bool proc_sys_fill_cache(struct file *file, 679f0c3b509SAl Viro struct dir_context *ctx, 6809043476fSAl Viro struct ctl_table_header *head, 6819043476fSAl Viro struct ctl_table *table) 68277b14db5SEric W. Biederman { 683f0c3b509SAl Viro struct dentry *child, *dir = file->f_path.dentry; 68477b14db5SEric W. Biederman struct inode *inode; 68577b14db5SEric W. Biederman struct qstr qname; 68677b14db5SEric W. Biederman ino_t ino = 0; 68777b14db5SEric W. Biederman unsigned type = DT_UNKNOWN; 68877b14db5SEric W. Biederman 68977b14db5SEric W. Biederman qname.name = table->procname; 69077b14db5SEric W. Biederman qname.len = strlen(table->procname); 6918387ff25SLinus Torvalds qname.hash = full_name_hash(dir, qname.name, qname.len); 69277b14db5SEric W. Biederman 69377b14db5SEric W. Biederman child = d_lookup(dir, &qname); 69477b14db5SEric W. Biederman if (!child) { 69576aab3abSAl Viro DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); 69676aab3abSAl Viro child = d_alloc_parallel(dir, &qname, &wq); 69776aab3abSAl Viro if (IS_ERR(child)) 69876aab3abSAl Viro return false; 69976aab3abSAl Viro if (d_in_lookup(child)) { 700888e2b03SAl Viro struct dentry *res; 7019043476fSAl Viro inode = proc_sys_make_inode(dir->d_sb, head, table); 702ea5751ccSIvan Delalande if (IS_ERR(inode)) { 70376aab3abSAl Viro d_lookup_done(child); 7049043476fSAl Viro dput(child); 705f0c3b509SAl Viro return false; 70676aab3abSAl Viro } 707fb045adbSNick Piggin d_set_d_op(child, &proc_sys_dentry_operations); 708888e2b03SAl Viro res = d_splice_alias(inode, child); 709888e2b03SAl Viro d_lookup_done(child); 710888e2b03SAl Viro if (unlikely(res)) { 711888e2b03SAl Viro if (IS_ERR(res)) { 712888e2b03SAl Viro dput(child); 713888e2b03SAl Viro return false; 714888e2b03SAl Viro } 715888e2b03SAl Viro dput(child); 716888e2b03SAl Viro child = res; 717888e2b03SAl Viro } 71877b14db5SEric W. Biederman } 71977b14db5SEric W. Biederman } 7202b0143b5SDavid Howells inode = d_inode(child); 72177b14db5SEric W. Biederman ino = inode->i_ino; 72277b14db5SEric W. Biederman type = inode->i_mode >> 12; 72377b14db5SEric W. Biederman dput(child); 724f0c3b509SAl Viro return dir_emit(ctx, qname.name, qname.len, ino, type); 7259043476fSAl Viro } 7269043476fSAl Viro 727f0c3b509SAl Viro static bool proc_sys_link_fill_cache(struct file *file, 728f0c3b509SAl Viro struct dir_context *ctx, 7290e47c99dSEric W. Biederman struct ctl_table_header *head, 7300e47c99dSEric W. Biederman struct ctl_table *table) 7310e47c99dSEric W. Biederman { 732f0c3b509SAl Viro bool ret = true; 733a0b0d1c3SDanilo Krummrich 7340e47c99dSEric W. Biederman head = sysctl_head_grab(head); 735a0b0d1c3SDanilo Krummrich if (IS_ERR(head)) 736a0b0d1c3SDanilo Krummrich return false; 7370e47c99dSEric W. Biederman 7380e47c99dSEric W. Biederman /* It is not an error if we can not follow the link ignore it */ 739835b94e0SDanilo Krummrich if (sysctl_follow_link(&head, &table)) 7400e47c99dSEric W. Biederman goto out; 7410e47c99dSEric W. Biederman 742f0c3b509SAl Viro ret = proc_sys_fill_cache(file, ctx, head, table); 7430e47c99dSEric W. Biederman out: 7440e47c99dSEric W. Biederman sysctl_head_finish(head); 7450e47c99dSEric W. Biederman return ret; 7460e47c99dSEric W. Biederman } 7470e47c99dSEric W. Biederman 748e5eea098SJoe Perches static int scan(struct ctl_table_header *head, struct ctl_table *table, 7499043476fSAl Viro unsigned long *pos, struct file *file, 750f0c3b509SAl Viro struct dir_context *ctx) 7519043476fSAl Viro { 752f0c3b509SAl Viro bool res; 7539043476fSAl Viro 754f0c3b509SAl Viro if ((*pos)++ < ctx->pos) 755f0c3b509SAl Viro return true; 7569043476fSAl Viro 7570e47c99dSEric W. Biederman if (unlikely(S_ISLNK(table->mode))) 758f0c3b509SAl Viro res = proc_sys_link_fill_cache(file, ctx, head, table); 7590e47c99dSEric W. Biederman else 760f0c3b509SAl Viro res = proc_sys_fill_cache(file, ctx, head, table); 7619043476fSAl Viro 762f0c3b509SAl Viro if (res) 763f0c3b509SAl Viro ctx->pos = *pos; 7646a75ce16SEric W. Biederman 7656a75ce16SEric W. Biederman return res; 76677b14db5SEric W. Biederman } 76777b14db5SEric W. Biederman 768f0c3b509SAl Viro static int proc_sys_readdir(struct file *file, struct dir_context *ctx) 76977b14db5SEric W. Biederman { 770f0c3b509SAl Viro struct ctl_table_header *head = grab_header(file_inode(file)); 7719043476fSAl Viro struct ctl_table_header *h = NULL; 7726a75ce16SEric W. Biederman struct ctl_table *entry; 7737ec66d06SEric W. Biederman struct ctl_dir *ctl_dir; 77477b14db5SEric W. Biederman unsigned long pos; 77577b14db5SEric W. Biederman 7769043476fSAl Viro if (IS_ERR(head)) 7779043476fSAl Viro return PTR_ERR(head); 7789043476fSAl Viro 7797ec66d06SEric W. Biederman ctl_dir = container_of(head, struct ctl_dir, header); 78077b14db5SEric W. Biederman 781f0c3b509SAl Viro if (!dir_emit_dots(file, ctx)) 78293362fa4SZhou Chengming goto out; 783f0c3b509SAl Viro 78477b14db5SEric W. Biederman pos = 2; 78577b14db5SEric W. Biederman 7867ec66d06SEric W. Biederman for (first_entry(ctl_dir, &h, &entry); h; next_entry(&h, &entry)) { 787f0c3b509SAl Viro if (!scan(h, entry, &pos, file, ctx)) { 7889043476fSAl Viro sysctl_head_finish(h); 7899043476fSAl Viro break; 79077b14db5SEric W. Biederman } 79177b14db5SEric W. Biederman } 79293362fa4SZhou Chengming out: 79377b14db5SEric W. Biederman sysctl_head_finish(head); 794f0c3b509SAl Viro return 0; 79577b14db5SEric W. Biederman } 79677b14db5SEric W. Biederman 7974609e1f1SChristian Brauner static int proc_sys_permission(struct mnt_idmap *idmap, 798549c7297SChristian Brauner struct inode *inode, int mask) 79977b14db5SEric W. Biederman { 80077b14db5SEric W. Biederman /* 80177b14db5SEric W. Biederman * sysctl entries that are not writeable, 80277b14db5SEric W. Biederman * are _NOT_ writeable, capabilities or not. 80377b14db5SEric W. Biederman */ 804f696a365SMiklos Szeredi struct ctl_table_header *head; 805f696a365SMiklos Szeredi struct ctl_table *table; 80677b14db5SEric W. Biederman int error; 80777b14db5SEric W. Biederman 808f696a365SMiklos Szeredi /* Executable files are not allowed under /proc/sys/ */ 809f696a365SMiklos Szeredi if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode)) 810f696a365SMiklos Szeredi return -EACCES; 811f696a365SMiklos Szeredi 812f696a365SMiklos Szeredi head = grab_header(inode); 8139043476fSAl Viro if (IS_ERR(head)) 8149043476fSAl Viro return PTR_ERR(head); 81577b14db5SEric W. Biederman 816f696a365SMiklos Szeredi table = PROC_I(inode)->sysctl_entry; 8179043476fSAl Viro if (!table) /* global root - r-xr-xr-x */ 8189043476fSAl Viro error = mask & MAY_WRITE ? -EACCES : 0; 8199043476fSAl Viro else /* Use the permissions on the sysctl table entry */ 82073f7ef43SEric W. Biederman error = sysctl_perm(head, table, mask & ~MAY_NOT_BLOCK); 8219043476fSAl Viro 82277b14db5SEric W. Biederman sysctl_head_finish(head); 82377b14db5SEric W. Biederman return error; 82477b14db5SEric W. Biederman } 82577b14db5SEric W. Biederman 826c1632a0fSChristian Brauner static int proc_sys_setattr(struct mnt_idmap *idmap, 827549c7297SChristian Brauner struct dentry *dentry, struct iattr *attr) 82877b14db5SEric W. Biederman { 8292b0143b5SDavid Howells struct inode *inode = d_inode(dentry); 83077b14db5SEric W. Biederman int error; 83177b14db5SEric W. Biederman 83277b14db5SEric W. Biederman if (attr->ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID)) 83377b14db5SEric W. Biederman return -EPERM; 83477b14db5SEric W. Biederman 835c1632a0fSChristian Brauner error = setattr_prepare(&nop_mnt_idmap, dentry, attr); 8361025774cSChristoph Hellwig if (error) 83777b14db5SEric W. Biederman return error; 8381025774cSChristoph Hellwig 839c1632a0fSChristian Brauner setattr_copy(&nop_mnt_idmap, inode, attr); 8401025774cSChristoph Hellwig return 0; 84177b14db5SEric W. Biederman } 84277b14db5SEric W. Biederman 843b74d24f7SChristian Brauner static int proc_sys_getattr(struct mnt_idmap *idmap, 844549c7297SChristian Brauner const struct path *path, struct kstat *stat, 845a528d35eSDavid Howells u32 request_mask, unsigned int query_flags) 8469043476fSAl Viro { 847a528d35eSDavid Howells struct inode *inode = d_inode(path->dentry); 8489043476fSAl Viro struct ctl_table_header *head = grab_header(inode); 8499043476fSAl Viro struct ctl_table *table = PROC_I(inode)->sysctl_entry; 8509043476fSAl Viro 8519043476fSAl Viro if (IS_ERR(head)) 8529043476fSAl Viro return PTR_ERR(head); 8539043476fSAl Viro 8540d72b928SJeff Layton generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat); 8559043476fSAl Viro if (table) 8569043476fSAl Viro stat->mode = (stat->mode & S_IFMT) | table->mode; 8579043476fSAl Viro 8589043476fSAl Viro sysctl_head_finish(head); 8599043476fSAl Viro return 0; 8609043476fSAl Viro } 8619043476fSAl Viro 86277b14db5SEric W. Biederman static const struct file_operations proc_sys_file_operations = { 863f1ecf068SLucas De Marchi .open = proc_sys_open, 864f1ecf068SLucas De Marchi .poll = proc_sys_poll, 8654bd6a735SMatthew Wilcox (Oracle) .read_iter = proc_sys_read, 8664bd6a735SMatthew Wilcox (Oracle) .write_iter = proc_sys_write, 867b0072734SDavid Howells .splice_read = copy_splice_read, 8684bd6a735SMatthew Wilcox (Oracle) .splice_write = iter_file_splice_write, 8696038f373SArnd Bergmann .llseek = default_llseek, 8709043476fSAl Viro }; 8719043476fSAl Viro 8729043476fSAl Viro static const struct file_operations proc_sys_dir_file_operations = { 873887df078SPavel Emelyanov .read = generic_read_dir, 874f50752eaSAl Viro .iterate_shared = proc_sys_readdir, 8753222a3e5SChristoph Hellwig .llseek = generic_file_llseek, 87677b14db5SEric W. Biederman }; 87777b14db5SEric W. Biederman 87803a44825SJan Engelhardt static const struct inode_operations proc_sys_inode_operations = { 8799043476fSAl Viro .permission = proc_sys_permission, 8809043476fSAl Viro .setattr = proc_sys_setattr, 8819043476fSAl Viro .getattr = proc_sys_getattr, 8829043476fSAl Viro }; 8839043476fSAl Viro 8849043476fSAl Viro static const struct inode_operations proc_sys_dir_operations = { 88577b14db5SEric W. Biederman .lookup = proc_sys_lookup, 88677b14db5SEric W. Biederman .permission = proc_sys_permission, 88777b14db5SEric W. Biederman .setattr = proc_sys_setattr, 8889043476fSAl Viro .getattr = proc_sys_getattr, 88977b14db5SEric W. Biederman }; 89077b14db5SEric W. Biederman 8910b728e19SAl Viro static int proc_sys_revalidate(struct dentry *dentry, unsigned int flags) 89277b14db5SEric W. Biederman { 8930b728e19SAl Viro if (flags & LOOKUP_RCU) 89434286d66SNick Piggin return -ECHILD; 8952b0143b5SDavid Howells return !PROC_I(d_inode(dentry))->sysctl->unregistering; 8969043476fSAl Viro } 8979043476fSAl Viro 898fe15ce44SNick Piggin static int proc_sys_delete(const struct dentry *dentry) 8999043476fSAl Viro { 9002b0143b5SDavid Howells return !!PROC_I(d_inode(dentry))->sysctl->unregistering; 9019043476fSAl Viro } 9029043476fSAl Viro 9031f87f0b5SEric W. Biederman static int sysctl_is_seen(struct ctl_table_header *p) 9041f87f0b5SEric W. Biederman { 9051f87f0b5SEric W. Biederman struct ctl_table_set *set = p->set; 9061f87f0b5SEric W. Biederman int res; 9071f87f0b5SEric W. Biederman spin_lock(&sysctl_lock); 9081f87f0b5SEric W. Biederman if (p->unregistering) 9091f87f0b5SEric W. Biederman res = 0; 9101f87f0b5SEric W. Biederman else if (!set->is_seen) 9111f87f0b5SEric W. Biederman res = 1; 9121f87f0b5SEric W. Biederman else 9131f87f0b5SEric W. Biederman res = set->is_seen(set); 9141f87f0b5SEric W. Biederman spin_unlock(&sysctl_lock); 9151f87f0b5SEric W. Biederman return res; 9161f87f0b5SEric W. Biederman } 9171f87f0b5SEric W. Biederman 9186fa67e70SAl Viro static int proc_sys_compare(const struct dentry *dentry, 919621e155aSNick Piggin unsigned int len, const char *str, const struct qstr *name) 9209043476fSAl Viro { 921dfef6dcdSAl Viro struct ctl_table_header *head; 922da53be12SLinus Torvalds struct inode *inode; 923da53be12SLinus Torvalds 92431e6b01fSNick Piggin /* Although proc doesn't have negative dentries, rcu-walk means 92531e6b01fSNick Piggin * that inode here can be NULL */ 926dfef6dcdSAl Viro /* AV: can it, indeed? */ 9272b0143b5SDavid Howells inode = d_inode_rcu(dentry); 92831e6b01fSNick Piggin if (!inode) 929dfef6dcdSAl Viro return 1; 930621e155aSNick Piggin if (name->len != len) 9319043476fSAl Viro return 1; 932621e155aSNick Piggin if (memcmp(name->name, str, len)) 9339043476fSAl Viro return 1; 934dfef6dcdSAl Viro head = rcu_dereference(PROC_I(inode)->sysctl); 935dfef6dcdSAl Viro return !head || !sysctl_is_seen(head); 93677b14db5SEric W. Biederman } 93777b14db5SEric W. Biederman 938d72f71ebSAl Viro static const struct dentry_operations proc_sys_dentry_operations = { 93977b14db5SEric W. Biederman .d_revalidate = proc_sys_revalidate, 9409043476fSAl Viro .d_delete = proc_sys_delete, 9419043476fSAl Viro .d_compare = proc_sys_compare, 94277b14db5SEric W. Biederman }; 94377b14db5SEric W. Biederman 9440e47c99dSEric W. Biederman static struct ctl_dir *find_subdir(struct ctl_dir *dir, 9457ec66d06SEric W. Biederman const char *name, int namelen) 9461f87f0b5SEric W. Biederman { 9477ec66d06SEric W. Biederman struct ctl_table_header *head; 9487ec66d06SEric W. Biederman struct ctl_table *entry; 9491f87f0b5SEric W. Biederman 9500e47c99dSEric W. Biederman entry = find_entry(&head, dir, name, namelen); 9517ec66d06SEric W. Biederman if (!entry) 9527ec66d06SEric W. Biederman return ERR_PTR(-ENOENT); 95351f72f4aSEric W. Biederman if (!S_ISDIR(entry->mode)) 9547ec66d06SEric W. Biederman return ERR_PTR(-ENOTDIR); 95551f72f4aSEric W. Biederman return container_of(head, struct ctl_dir, header); 9561f87f0b5SEric W. Biederman } 9571f87f0b5SEric W. Biederman 9587ec66d06SEric W. Biederman static struct ctl_dir *new_dir(struct ctl_table_set *set, 9597ec66d06SEric W. Biederman const char *name, int namelen) 9601f87f0b5SEric W. Biederman { 9617ec66d06SEric W. Biederman struct ctl_table *table; 9627ec66d06SEric W. Biederman struct ctl_dir *new; 963ac13ac6fSEric W. Biederman struct ctl_node *node; 9647ec66d06SEric W. Biederman char *new_name; 9651f87f0b5SEric W. Biederman 966ac13ac6fSEric W. Biederman new = kzalloc(sizeof(*new) + sizeof(struct ctl_node) + 967ac13ac6fSEric W. Biederman sizeof(struct ctl_table)*2 + namelen + 1, 968ac13ac6fSEric W. Biederman GFP_KERNEL); 9697ec66d06SEric W. Biederman if (!new) 9707ec66d06SEric W. Biederman return NULL; 9717ec66d06SEric W. Biederman 972ac13ac6fSEric W. Biederman node = (struct ctl_node *)(new + 1); 973ac13ac6fSEric W. Biederman table = (struct ctl_table *)(node + 1); 9747ec66d06SEric W. Biederman new_name = (char *)(table + 2); 9757ec66d06SEric W. Biederman memcpy(new_name, name, namelen); 9767ec66d06SEric W. Biederman table[0].procname = new_name; 9777ec66d06SEric W. Biederman table[0].mode = S_IFDIR|S_IRUGO|S_IXUGO; 978b1f01e2bSJoel Granados init_header(&new->header, set->dir.header.root, set, node, table, 1); 9797ec66d06SEric W. Biederman 9807ec66d06SEric W. Biederman return new; 9811f87f0b5SEric W. Biederman } 9821f87f0b5SEric W. Biederman 98360f126d9SEric W. Biederman /** 98460f126d9SEric W. Biederman * get_subdir - find or create a subdir with the specified name. 98560f126d9SEric W. Biederman * @dir: Directory to create the subdirectory in 98660f126d9SEric W. Biederman * @name: The name of the subdirectory to find or create 98760f126d9SEric W. Biederman * @namelen: The length of name 98860f126d9SEric W. Biederman * 98960f126d9SEric W. Biederman * Takes a directory with an elevated reference count so we know that 99060f126d9SEric W. Biederman * if we drop the lock the directory will not go away. Upon success 99160f126d9SEric W. Biederman * the reference is moved from @dir to the returned subdirectory. 99260f126d9SEric W. Biederman * Upon error an error code is returned and the reference on @dir is 99360f126d9SEric W. Biederman * simply dropped. 99460f126d9SEric W. Biederman */ 9950e47c99dSEric W. Biederman static struct ctl_dir *get_subdir(struct ctl_dir *dir, 9960e47c99dSEric W. Biederman const char *name, int namelen) 9977ec66d06SEric W. Biederman { 9980e47c99dSEric W. Biederman struct ctl_table_set *set = dir->header.set; 9997ec66d06SEric W. Biederman struct ctl_dir *subdir, *new = NULL; 10000eb97f38SEric W. Biederman int err; 10017ec66d06SEric W. Biederman 10027ec66d06SEric W. Biederman spin_lock(&sysctl_lock); 10030e47c99dSEric W. Biederman subdir = find_subdir(dir, name, namelen); 10047ec66d06SEric W. Biederman if (!IS_ERR(subdir)) 10057ec66d06SEric W. Biederman goto found; 10067ec66d06SEric W. Biederman if (PTR_ERR(subdir) != -ENOENT) 10077ec66d06SEric W. Biederman goto failed; 10087ec66d06SEric W. Biederman 10097ec66d06SEric W. Biederman spin_unlock(&sysctl_lock); 10107ec66d06SEric W. Biederman new = new_dir(set, name, namelen); 10117ec66d06SEric W. Biederman spin_lock(&sysctl_lock); 10127ec66d06SEric W. Biederman subdir = ERR_PTR(-ENOMEM); 10137ec66d06SEric W. Biederman if (!new) 10147ec66d06SEric W. Biederman goto failed; 10157ec66d06SEric W. Biederman 101660f126d9SEric W. Biederman /* Was the subdir added while we dropped the lock? */ 10170e47c99dSEric W. Biederman subdir = find_subdir(dir, name, namelen); 10187ec66d06SEric W. Biederman if (!IS_ERR(subdir)) 10197ec66d06SEric W. Biederman goto found; 10207ec66d06SEric W. Biederman if (PTR_ERR(subdir) != -ENOENT) 10217ec66d06SEric W. Biederman goto failed; 10227ec66d06SEric W. Biederman 102360f126d9SEric W. Biederman /* Nope. Use the our freshly made directory entry. */ 10240eb97f38SEric W. Biederman err = insert_header(dir, &new->header); 10250eb97f38SEric W. Biederman subdir = ERR_PTR(err); 10260eb97f38SEric W. Biederman if (err) 10270e47c99dSEric W. Biederman goto failed; 10287ec66d06SEric W. Biederman subdir = new; 10297ec66d06SEric W. Biederman found: 10307ec66d06SEric W. Biederman subdir->header.nreg++; 10317ec66d06SEric W. Biederman failed: 1032a1c83681SViresh Kumar if (IS_ERR(subdir)) { 103387ebdc00SAndrew Morton pr_err("sysctl could not get directory: "); 10346980128fSEric W. Biederman sysctl_print_dir(dir); 1035153ee1c4SGeert Uytterhoeven pr_cont("%*.*s %ld\n", namelen, namelen, name, 1036153ee1c4SGeert Uytterhoeven PTR_ERR(subdir)); 10371f87f0b5SEric W. Biederman } 10387ec66d06SEric W. Biederman drop_sysctl_table(&dir->header); 10397ec66d06SEric W. Biederman if (new) 10407ec66d06SEric W. Biederman drop_sysctl_table(&new->header); 10417ec66d06SEric W. Biederman spin_unlock(&sysctl_lock); 10427ec66d06SEric W. Biederman return subdir; 10431f87f0b5SEric W. Biederman } 10441f87f0b5SEric W. Biederman 10450e47c99dSEric W. Biederman static struct ctl_dir *xlate_dir(struct ctl_table_set *set, struct ctl_dir *dir) 10460e47c99dSEric W. Biederman { 10470e47c99dSEric W. Biederman struct ctl_dir *parent; 10480e47c99dSEric W. Biederman const char *procname; 10490e47c99dSEric W. Biederman if (!dir->header.parent) 10500e47c99dSEric W. Biederman return &set->dir; 10510e47c99dSEric W. Biederman parent = xlate_dir(set, dir->header.parent); 10520e47c99dSEric W. Biederman if (IS_ERR(parent)) 10530e47c99dSEric W. Biederman return parent; 10540e47c99dSEric W. Biederman procname = dir->header.ctl_table[0].procname; 10550e47c99dSEric W. Biederman return find_subdir(parent, procname, strlen(procname)); 10560e47c99dSEric W. Biederman } 10570e47c99dSEric W. Biederman 10580e47c99dSEric W. Biederman static int sysctl_follow_link(struct ctl_table_header **phead, 105913bcc6a2SEric W. Biederman struct ctl_table **pentry) 10600e47c99dSEric W. Biederman { 10610e47c99dSEric W. Biederman struct ctl_table_header *head; 10620e47c99dSEric W. Biederman struct ctl_table_root *root; 10630e47c99dSEric W. Biederman struct ctl_table_set *set; 10640e47c99dSEric W. Biederman struct ctl_table *entry; 10650e47c99dSEric W. Biederman struct ctl_dir *dir; 10660e47c99dSEric W. Biederman int ret; 10670e47c99dSEric W. Biederman 10680e47c99dSEric W. Biederman spin_lock(&sysctl_lock); 10690e47c99dSEric W. Biederman root = (*pentry)->data; 107013bcc6a2SEric W. Biederman set = lookup_header_set(root); 10710e47c99dSEric W. Biederman dir = xlate_dir(set, (*phead)->parent); 10720e47c99dSEric W. Biederman if (IS_ERR(dir)) 10730e47c99dSEric W. Biederman ret = PTR_ERR(dir); 10740e47c99dSEric W. Biederman else { 10750e47c99dSEric W. Biederman const char *procname = (*pentry)->procname; 10760e47c99dSEric W. Biederman head = NULL; 10770e47c99dSEric W. Biederman entry = find_entry(&head, dir, procname, strlen(procname)); 10780e47c99dSEric W. Biederman ret = -ENOENT; 10790e47c99dSEric W. Biederman if (entry && use_table(head)) { 10800e47c99dSEric W. Biederman unuse_table(*phead); 10810e47c99dSEric W. Biederman *phead = head; 10820e47c99dSEric W. Biederman *pentry = entry; 10830e47c99dSEric W. Biederman ret = 0; 10840e47c99dSEric W. Biederman } 10850e47c99dSEric W. Biederman } 10860e47c99dSEric W. Biederman 10870e47c99dSEric W. Biederman spin_unlock(&sysctl_lock); 10880e47c99dSEric W. Biederman return ret; 10890e47c99dSEric W. Biederman } 10900e47c99dSEric W. Biederman 10917c60c48fSEric W. Biederman static int sysctl_err(const char *path, struct ctl_table *table, char *fmt, ...) 10927c60c48fSEric W. Biederman { 10937c60c48fSEric W. Biederman struct va_format vaf; 10947c60c48fSEric W. Biederman va_list args; 10957c60c48fSEric W. Biederman 10967c60c48fSEric W. Biederman va_start(args, fmt); 10977c60c48fSEric W. Biederman vaf.fmt = fmt; 10987c60c48fSEric W. Biederman vaf.va = &args; 10997c60c48fSEric W. Biederman 110087ebdc00SAndrew Morton pr_err("sysctl table check failed: %s/%s %pV\n", 11017c60c48fSEric W. Biederman path, table->procname, &vaf); 11027c60c48fSEric W. Biederman 11037c60c48fSEric W. Biederman va_end(args); 11047c60c48fSEric W. Biederman return -EINVAL; 11057c60c48fSEric W. Biederman } 11067c60c48fSEric W. Biederman 11074f2fec00SLuis R. Rodriguez static int sysctl_check_table_array(const char *path, struct ctl_table *table) 11084f2fec00SLuis R. Rodriguez { 11094f2fec00SLuis R. Rodriguez int err = 0; 11104f2fec00SLuis R. Rodriguez 111161d9b56aSLuis R. Rodriguez if ((table->proc_handler == proc_douintvec) || 111261d9b56aSLuis R. Rodriguez (table->proc_handler == proc_douintvec_minmax)) { 11134f2fec00SLuis R. Rodriguez if (table->maxlen != sizeof(unsigned int)) 111464a11f3dSWaiman Long err |= sysctl_err(path, table, "array not allowed"); 11154f2fec00SLuis R. Rodriguez } 11164f2fec00SLuis R. Rodriguez 1117cb944413SEric Dumazet if (table->proc_handler == proc_dou8vec_minmax) { 1118cb944413SEric Dumazet if (table->maxlen != sizeof(u8)) 1119cb944413SEric Dumazet err |= sysctl_err(path, table, "array not allowed"); 1120cb944413SEric Dumazet } 1121cb944413SEric Dumazet 1122f1aa2eb5SOndrej Mosnacek if (table->proc_handler == proc_dobool) { 1123f1aa2eb5SOndrej Mosnacek if (table->maxlen != sizeof(bool)) 1124f1aa2eb5SOndrej Mosnacek err |= sysctl_err(path, table, "array not allowed"); 1125f1aa2eb5SOndrej Mosnacek } 1126f1aa2eb5SOndrej Mosnacek 11274f2fec00SLuis R. Rodriguez return err; 11284f2fec00SLuis R. Rodriguez } 11294f2fec00SLuis R. Rodriguez 1130cc9f7ee0SJoel Granados static int sysctl_check_table(const char *path, struct ctl_table_header *header) 11317c60c48fSEric W. Biederman { 1132cb55f27aSMeng Tang struct ctl_table *entry; 11337c60c48fSEric W. Biederman int err = 0; 113418d4b42eSJoel Granados list_for_each_table_entry(entry, header) { 1135cb55f27aSMeng Tang if ((entry->proc_handler == proc_dostring) || 1136f1aa2eb5SOndrej Mosnacek (entry->proc_handler == proc_dobool) || 1137cb55f27aSMeng Tang (entry->proc_handler == proc_dointvec) || 1138cb55f27aSMeng Tang (entry->proc_handler == proc_douintvec) || 1139cb55f27aSMeng Tang (entry->proc_handler == proc_douintvec_minmax) || 1140cb55f27aSMeng Tang (entry->proc_handler == proc_dointvec_minmax) || 1141cb55f27aSMeng Tang (entry->proc_handler == proc_dou8vec_minmax) || 1142cb55f27aSMeng Tang (entry->proc_handler == proc_dointvec_jiffies) || 1143cb55f27aSMeng Tang (entry->proc_handler == proc_dointvec_userhz_jiffies) || 1144cb55f27aSMeng Tang (entry->proc_handler == proc_dointvec_ms_jiffies) || 1145cb55f27aSMeng Tang (entry->proc_handler == proc_doulongvec_minmax) || 1146cb55f27aSMeng Tang (entry->proc_handler == proc_doulongvec_ms_jiffies_minmax)) { 1147cb55f27aSMeng Tang if (!entry->data) 1148cb55f27aSMeng Tang err |= sysctl_err(path, entry, "No data"); 1149cb55f27aSMeng Tang if (!entry->maxlen) 1150cb55f27aSMeng Tang err |= sysctl_err(path, entry, "No maxlen"); 11514f2fec00SLuis R. Rodriguez else 1152cb55f27aSMeng Tang err |= sysctl_check_table_array(path, entry); 11531f87f0b5SEric W. Biederman } 1154cb55f27aSMeng Tang if (!entry->proc_handler) 1155cb55f27aSMeng Tang err |= sysctl_err(path, entry, "No proc_handler"); 11567c60c48fSEric W. Biederman 1157cb55f27aSMeng Tang if ((entry->mode & (S_IRUGO|S_IWUGO)) != entry->mode) 1158cb55f27aSMeng Tang err |= sysctl_err(path, entry, "bogus .mode 0%o", 1159cb55f27aSMeng Tang entry->mode); 11601f87f0b5SEric W. Biederman } 11617c60c48fSEric W. Biederman return err; 11621f87f0b5SEric W. Biederman } 11631f87f0b5SEric W. Biederman 1164cc9f7ee0SJoel Granados static struct ctl_table_header *new_links(struct ctl_dir *dir, struct ctl_table_header *head) 11650e47c99dSEric W. Biederman { 11660e47c99dSEric W. Biederman struct ctl_table *link_table, *entry, *link; 11670e47c99dSEric W. Biederman struct ctl_table_header *links; 1168ac13ac6fSEric W. Biederman struct ctl_node *node; 11690e47c99dSEric W. Biederman char *link_name; 11700e47c99dSEric W. Biederman int nr_entries, name_bytes; 11710e47c99dSEric W. Biederman 11720e47c99dSEric W. Biederman name_bytes = 0; 11730e47c99dSEric W. Biederman nr_entries = 0; 117418d4b42eSJoel Granados list_for_each_table_entry(entry, head) { 11750e47c99dSEric W. Biederman nr_entries++; 11760e47c99dSEric W. Biederman name_bytes += strlen(entry->procname) + 1; 11770e47c99dSEric W. Biederman } 11780e47c99dSEric W. Biederman 11790e47c99dSEric W. Biederman links = kzalloc(sizeof(struct ctl_table_header) + 1180ac13ac6fSEric W. Biederman sizeof(struct ctl_node)*nr_entries + 11810e47c99dSEric W. Biederman sizeof(struct ctl_table)*(nr_entries + 1) + 11820e47c99dSEric W. Biederman name_bytes, 11830e47c99dSEric W. Biederman GFP_KERNEL); 11840e47c99dSEric W. Biederman 11850e47c99dSEric W. Biederman if (!links) 11860e47c99dSEric W. Biederman return NULL; 11870e47c99dSEric W. Biederman 1188ac13ac6fSEric W. Biederman node = (struct ctl_node *)(links + 1); 1189ac13ac6fSEric W. Biederman link_table = (struct ctl_table *)(node + nr_entries); 11900e47c99dSEric W. Biederman link_name = (char *)&link_table[nr_entries + 1]; 1191cb55f27aSMeng Tang link = link_table; 11920e47c99dSEric W. Biederman 119318d4b42eSJoel Granados list_for_each_table_entry(entry, head) { 11940e47c99dSEric W. Biederman int len = strlen(entry->procname) + 1; 11950e47c99dSEric W. Biederman memcpy(link_name, entry->procname, len); 11960e47c99dSEric W. Biederman link->procname = link_name; 11970e47c99dSEric W. Biederman link->mode = S_IFLNK|S_IRWXUGO; 1198cc9f7ee0SJoel Granados link->data = head->root; 11990e47c99dSEric W. Biederman link_name += len; 1200cb55f27aSMeng Tang link++; 12010e47c99dSEric W. Biederman } 1202b1f01e2bSJoel Granados init_header(links, dir->header.root, dir->header.set, node, link_table, 1203b1f01e2bSJoel Granados head->ctl_table_size); 12040e47c99dSEric W. Biederman links->nreg = nr_entries; 12050e47c99dSEric W. Biederman 12060e47c99dSEric W. Biederman return links; 12070e47c99dSEric W. Biederman } 12080e47c99dSEric W. Biederman 12090e47c99dSEric W. Biederman static bool get_links(struct ctl_dir *dir, 1210cc9f7ee0SJoel Granados struct ctl_table_header *header, 1211cc9f7ee0SJoel Granados struct ctl_table_root *link_root) 12120e47c99dSEric W. Biederman { 1213cc9f7ee0SJoel Granados struct ctl_table_header *tmp_head; 12140e47c99dSEric W. Biederman struct ctl_table *entry, *link; 12150e47c99dSEric W. Biederman 12160e47c99dSEric W. Biederman /* Are there links available for every entry in table? */ 121718d4b42eSJoel Granados list_for_each_table_entry(entry, header) { 12180e47c99dSEric W. Biederman const char *procname = entry->procname; 1219cc9f7ee0SJoel Granados link = find_entry(&tmp_head, dir, procname, strlen(procname)); 12200e47c99dSEric W. Biederman if (!link) 12210e47c99dSEric W. Biederman return false; 12220e47c99dSEric W. Biederman if (S_ISDIR(link->mode) && S_ISDIR(entry->mode)) 12230e47c99dSEric W. Biederman continue; 12240e47c99dSEric W. Biederman if (S_ISLNK(link->mode) && (link->data == link_root)) 12250e47c99dSEric W. Biederman continue; 12260e47c99dSEric W. Biederman return false; 12270e47c99dSEric W. Biederman } 12280e47c99dSEric W. Biederman 12290e47c99dSEric W. Biederman /* The checks passed. Increase the registration count on the links */ 123018d4b42eSJoel Granados list_for_each_table_entry(entry, header) { 12310e47c99dSEric W. Biederman const char *procname = entry->procname; 1232cc9f7ee0SJoel Granados link = find_entry(&tmp_head, dir, procname, strlen(procname)); 1233cc9f7ee0SJoel Granados tmp_head->nreg++; 12340e47c99dSEric W. Biederman } 12350e47c99dSEric W. Biederman return true; 12360e47c99dSEric W. Biederman } 12370e47c99dSEric W. Biederman 12380e47c99dSEric W. Biederman static int insert_links(struct ctl_table_header *head) 12390e47c99dSEric W. Biederman { 12400e47c99dSEric W. Biederman struct ctl_table_set *root_set = &sysctl_table_root.default_set; 12419a521359SLi zeming struct ctl_dir *core_parent; 12420e47c99dSEric W. Biederman struct ctl_table_header *links; 12430e47c99dSEric W. Biederman int err; 12440e47c99dSEric W. Biederman 12450e47c99dSEric W. Biederman if (head->set == root_set) 12460e47c99dSEric W. Biederman return 0; 12470e47c99dSEric W. Biederman 12480e47c99dSEric W. Biederman core_parent = xlate_dir(root_set, head->parent); 12490e47c99dSEric W. Biederman if (IS_ERR(core_parent)) 12500e47c99dSEric W. Biederman return 0; 12510e47c99dSEric W. Biederman 1252cc9f7ee0SJoel Granados if (get_links(core_parent, head, head->root)) 12530e47c99dSEric W. Biederman return 0; 12540e47c99dSEric W. Biederman 12550e47c99dSEric W. Biederman core_parent->header.nreg++; 12560e47c99dSEric W. Biederman spin_unlock(&sysctl_lock); 12570e47c99dSEric W. Biederman 1258cc9f7ee0SJoel Granados links = new_links(core_parent, head); 12590e47c99dSEric W. Biederman 12600e47c99dSEric W. Biederman spin_lock(&sysctl_lock); 12610e47c99dSEric W. Biederman err = -ENOMEM; 12620e47c99dSEric W. Biederman if (!links) 12630e47c99dSEric W. Biederman goto out; 12640e47c99dSEric W. Biederman 12650e47c99dSEric W. Biederman err = 0; 1266cc9f7ee0SJoel Granados if (get_links(core_parent, head, head->root)) { 12670e47c99dSEric W. Biederman kfree(links); 12680e47c99dSEric W. Biederman goto out; 12690e47c99dSEric W. Biederman } 12700e47c99dSEric W. Biederman 12710e47c99dSEric W. Biederman err = insert_header(core_parent, links); 12720e47c99dSEric W. Biederman if (err) 12730e47c99dSEric W. Biederman kfree(links); 12740e47c99dSEric W. Biederman out: 12750e47c99dSEric W. Biederman drop_sysctl_table(&core_parent->header); 12760e47c99dSEric W. Biederman return err; 12770e47c99dSEric W. Biederman } 12780e47c99dSEric W. Biederman 1279b2f56e55SLuis Chamberlain /* Find the directory for the ctl_table. If one is not found create it. */ 1280b2f56e55SLuis Chamberlain static struct ctl_dir *sysctl_mkdir_p(struct ctl_dir *dir, const char *path) 1281b2f56e55SLuis Chamberlain { 1282b2f56e55SLuis Chamberlain const char *name, *nextname; 1283b2f56e55SLuis Chamberlain 1284b2f56e55SLuis Chamberlain for (name = path; name; name = nextname) { 1285b2f56e55SLuis Chamberlain int namelen; 1286b2f56e55SLuis Chamberlain nextname = strchr(name, '/'); 1287b2f56e55SLuis Chamberlain if (nextname) { 1288b2f56e55SLuis Chamberlain namelen = nextname - name; 1289b2f56e55SLuis Chamberlain nextname++; 1290b2f56e55SLuis Chamberlain } else { 1291b2f56e55SLuis Chamberlain namelen = strlen(name); 1292b2f56e55SLuis Chamberlain } 1293b2f56e55SLuis Chamberlain if (namelen == 0) 1294b2f56e55SLuis Chamberlain continue; 1295b2f56e55SLuis Chamberlain 1296b2f56e55SLuis Chamberlain /* 1297b2f56e55SLuis Chamberlain * namelen ensures if name is "foo/bar/yay" only foo is 1298b2f56e55SLuis Chamberlain * registered first. We traverse as if using mkdir -p and 1299b2f56e55SLuis Chamberlain * return a ctl_dir for the last directory entry. 1300b2f56e55SLuis Chamberlain */ 1301b2f56e55SLuis Chamberlain dir = get_subdir(dir, name, namelen); 1302b2f56e55SLuis Chamberlain if (IS_ERR(dir)) 1303b2f56e55SLuis Chamberlain break; 1304b2f56e55SLuis Chamberlain } 1305b2f56e55SLuis Chamberlain return dir; 1306b2f56e55SLuis Chamberlain } 1307b2f56e55SLuis Chamberlain 13081f87f0b5SEric W. Biederman /** 1309f728019bSEric W. Biederman * __register_sysctl_table - register a leaf sysctl table 131060a47a2eSEric W. Biederman * @set: Sysctl tree to register on 13111f87f0b5SEric W. Biederman * @path: The path to the directory the sysctl table is in. 13121dc8689eSLuis Chamberlain * @table: the top-level table structure without any child. This table 13131dc8689eSLuis Chamberlain * should not be free'd after registration. So it should not be 13141dc8689eSLuis Chamberlain * used on stack. It can either be a global or dynamically allocated 13151dc8689eSLuis Chamberlain * by the caller and free'd later after sysctl unregistration. 1316bff97cf1SJoel Granados * @table_size : The number of elements in table 13171f87f0b5SEric W. Biederman * 13181f87f0b5SEric W. Biederman * Register a sysctl table hierarchy. @table should be a filled in ctl_table 13191f87f0b5SEric W. Biederman * array. A completely 0 filled entry terminates the table. 13201f87f0b5SEric W. Biederman * 13211f87f0b5SEric W. Biederman * The members of the &struct ctl_table structure are used as follows: 13221f87f0b5SEric W. Biederman * 13231f87f0b5SEric W. Biederman * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not 13241f87f0b5SEric W. Biederman * enter a sysctl file 13251f87f0b5SEric W. Biederman * 13261f87f0b5SEric W. Biederman * data - a pointer to data for use by proc_handler 13271f87f0b5SEric W. Biederman * 13281f87f0b5SEric W. Biederman * maxlen - the maximum size in bytes of the data 13291f87f0b5SEric W. Biederman * 1330f728019bSEric W. Biederman * mode - the file permissions for the /proc/sys file 13311f87f0b5SEric W. Biederman * 1332f728019bSEric W. Biederman * child - must be %NULL. 13331f87f0b5SEric W. Biederman * 13341f87f0b5SEric W. Biederman * proc_handler - the text handler routine (described below) 13351f87f0b5SEric W. Biederman * 13361f87f0b5SEric W. Biederman * extra1, extra2 - extra pointers usable by the proc handler routines 133767ff3228SLuis Chamberlain * XXX: we should eventually modify these to use long min / max [0] 133867ff3228SLuis Chamberlain * [0] https://lkml.kernel.org/87zgpte9o4.fsf@email.froward.int.ebiederm.org 13391f87f0b5SEric W. Biederman * 13401f87f0b5SEric W. Biederman * Leaf nodes in the sysctl tree will be represented by a single file 134167ff3228SLuis Chamberlain * under /proc; non-leaf nodes (where child is not NULL) are not allowed, 134267ff3228SLuis Chamberlain * sysctl_check_table() verifies this. 13431f87f0b5SEric W. Biederman * 1344f728019bSEric W. Biederman * There must be a proc_handler routine for any terminal nodes. 1345f728019bSEric W. Biederman * Several default handlers are available to cover common cases - 13461f87f0b5SEric W. Biederman * 13471f87f0b5SEric W. Biederman * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(), 13481f87f0b5SEric W. Biederman * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), 13491f87f0b5SEric W. Biederman * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax() 13501f87f0b5SEric W. Biederman * 13511f87f0b5SEric W. Biederman * It is the handler's job to read the input buffer from user memory 13521f87f0b5SEric W. Biederman * and process it. The handler should return 0 on success. 13531f87f0b5SEric W. Biederman * 13541f87f0b5SEric W. Biederman * This routine returns %NULL on a failure to register, and a pointer 13551f87f0b5SEric W. Biederman * to the table header on success. 13561f87f0b5SEric W. Biederman */ 13576e9d5164SEric W. Biederman struct ctl_table_header *__register_sysctl_table( 135860a47a2eSEric W. Biederman struct ctl_table_set *set, 1359bff97cf1SJoel Granados const char *path, struct ctl_table *table, size_t table_size) 13601f87f0b5SEric W. Biederman { 136160a47a2eSEric W. Biederman struct ctl_table_root *root = set->dir.header.root; 13621f87f0b5SEric W. Biederman struct ctl_table_header *header; 13637ec66d06SEric W. Biederman struct ctl_dir *dir; 1364ac13ac6fSEric W. Biederman struct ctl_node *node; 1365ac13ac6fSEric W. Biederman 1366ac13ac6fSEric W. Biederman header = kzalloc(sizeof(struct ctl_table_header) + 1367bff97cf1SJoel Granados sizeof(struct ctl_node)*table_size, GFP_KERNEL_ACCOUNT); 13681f87f0b5SEric W. Biederman if (!header) 13691f87f0b5SEric W. Biederman return NULL; 13701f87f0b5SEric W. Biederman 1371ac13ac6fSEric W. Biederman node = (struct ctl_node *)(header + 1); 1372bff97cf1SJoel Granados init_header(header, root, set, node, table, table_size); 1373cc9f7ee0SJoel Granados if (sysctl_check_table(path, header)) 13747c60c48fSEric W. Biederman goto fail; 13758d6ecfccSEric W. Biederman 13761f87f0b5SEric W. Biederman spin_lock(&sysctl_lock); 13770e47c99dSEric W. Biederman dir = &set->dir; 137867ff3228SLuis Chamberlain /* Reference moved down the directory tree get_subdir */ 13797ec66d06SEric W. Biederman dir->header.nreg++; 13807ec66d06SEric W. Biederman spin_unlock(&sysctl_lock); 13817ec66d06SEric W. Biederman 1382b2f56e55SLuis Chamberlain dir = sysctl_mkdir_p(dir, path); 13837ec66d06SEric W. Biederman if (IS_ERR(dir)) 13847ec66d06SEric W. Biederman goto fail; 13857ec66d06SEric W. Biederman spin_lock(&sysctl_lock); 13860e47c99dSEric W. Biederman if (insert_header(dir, header)) 13870e47c99dSEric W. Biederman goto fail_put_dir_locked; 13880e47c99dSEric W. Biederman 13897ec66d06SEric W. Biederman drop_sysctl_table(&dir->header); 13901f87f0b5SEric W. Biederman spin_unlock(&sysctl_lock); 13911f87f0b5SEric W. Biederman 13921f87f0b5SEric W. Biederman return header; 13930e47c99dSEric W. Biederman 13947ec66d06SEric W. Biederman fail_put_dir_locked: 13957ec66d06SEric W. Biederman drop_sysctl_table(&dir->header); 13967c60c48fSEric W. Biederman spin_unlock(&sysctl_lock); 13977c60c48fSEric W. Biederman fail: 13987c60c48fSEric W. Biederman kfree(header); 13997c60c48fSEric W. Biederman return NULL; 14001f87f0b5SEric W. Biederman } 14011f87f0b5SEric W. Biederman 1402fea478d4SEric W. Biederman /** 14039edbfe92SJoel Granados * register_sysctl_sz - register a sysctl table 14041dc8689eSLuis Chamberlain * @path: The path to the directory the sysctl table is in. If the path 14051dc8689eSLuis Chamberlain * doesn't exist we will create it for you. 14061dc8689eSLuis Chamberlain * @table: the table structure. The calller must ensure the life of the @table 14071dc8689eSLuis Chamberlain * will be kept during the lifetime use of the syctl. It must not be freed 14081dc8689eSLuis Chamberlain * until unregister_sysctl_table() is called with the given returned table 14091dc8689eSLuis Chamberlain * with this registration. If your code is non modular then you don't need 14101dc8689eSLuis Chamberlain * to call unregister_sysctl_table() and can instead use something like 14111dc8689eSLuis Chamberlain * register_sysctl_init() which does not care for the result of the syctl 14121dc8689eSLuis Chamberlain * registration. 14139edbfe92SJoel Granados * @table_size: The number of elements in table. 1414fea478d4SEric W. Biederman * 1415fea478d4SEric W. Biederman * Register a sysctl table. @table should be a filled in ctl_table 1416fea478d4SEric W. Biederman * array. A completely 0 filled entry terminates the table. 1417fea478d4SEric W. Biederman * 1418fea478d4SEric W. Biederman * See __register_sysctl_table for more details. 1419fea478d4SEric W. Biederman */ 14209edbfe92SJoel Granados struct ctl_table_header *register_sysctl_sz(const char *path, struct ctl_table *table, 14219edbfe92SJoel Granados size_t table_size) 1422fea478d4SEric W. Biederman { 1423fea478d4SEric W. Biederman return __register_sysctl_table(&sysctl_table_root.default_set, 14249edbfe92SJoel Granados path, table, table_size); 1425fea478d4SEric W. Biederman } 14269edbfe92SJoel Granados EXPORT_SYMBOL(register_sysctl_sz); 1427fea478d4SEric W. Biederman 14283ddd9a80SXiaoming Ni /** 14293ddd9a80SXiaoming Ni * __register_sysctl_init() - register sysctl table to path 14301dc8689eSLuis Chamberlain * @path: path name for sysctl base. If that path doesn't exist we will create 14311dc8689eSLuis Chamberlain * it for you. 14321dc8689eSLuis Chamberlain * @table: This is the sysctl table that needs to be registered to the path. 14331dc8689eSLuis Chamberlain * The caller must ensure the life of the @table will be kept during the 14341dc8689eSLuis Chamberlain * lifetime use of the sysctl. 14353ddd9a80SXiaoming Ni * @table_name: The name of sysctl table, only used for log printing when 14363ddd9a80SXiaoming Ni * registration fails 14373bc269cfSJoel Granados * @table_size: The number of elements in table 14383ddd9a80SXiaoming Ni * 14393ddd9a80SXiaoming Ni * The sysctl interface is used by userspace to query or modify at runtime 14403ddd9a80SXiaoming Ni * a predefined value set on a variable. These variables however have default 14413ddd9a80SXiaoming Ni * values pre-set. Code which depends on these variables will always work even 14423ddd9a80SXiaoming Ni * if register_sysctl() fails. If register_sysctl() fails you'd just loose the 14433ddd9a80SXiaoming Ni * ability to query or modify the sysctls dynamically at run time. Chances of 14443ddd9a80SXiaoming Ni * register_sysctl() failing on init are extremely low, and so for both reasons 14453ddd9a80SXiaoming Ni * this function does not return any error as it is used by initialization code. 14463ddd9a80SXiaoming Ni * 1447228b09deSLuis Chamberlain * Context: if your base directory does not exist it will be created for you. 14483ddd9a80SXiaoming Ni */ 14493ddd9a80SXiaoming Ni void __init __register_sysctl_init(const char *path, struct ctl_table *table, 14503bc269cfSJoel Granados const char *table_name, size_t table_size) 14513ddd9a80SXiaoming Ni { 14523bc269cfSJoel Granados struct ctl_table_header *hdr = register_sysctl_sz(path, table, table_size); 14533ddd9a80SXiaoming Ni 14543ddd9a80SXiaoming Ni if (unlikely(!hdr)) { 14559edbfe92SJoel Granados pr_err("failed when register_sysctl_sz %s to %s\n", table_name, path); 14563ddd9a80SXiaoming Ni return; 14573ddd9a80SXiaoming Ni } 14583ddd9a80SXiaoming Ni kmemleak_not_leak(hdr); 14593ddd9a80SXiaoming Ni } 14603ddd9a80SXiaoming Ni 14610e47c99dSEric W. Biederman static void put_links(struct ctl_table_header *header) 14620e47c99dSEric W. Biederman { 14630e47c99dSEric W. Biederman struct ctl_table_set *root_set = &sysctl_table_root.default_set; 14640e47c99dSEric W. Biederman struct ctl_table_root *root = header->root; 14650e47c99dSEric W. Biederman struct ctl_dir *parent = header->parent; 14660e47c99dSEric W. Biederman struct ctl_dir *core_parent; 14670e47c99dSEric W. Biederman struct ctl_table *entry; 14680e47c99dSEric W. Biederman 14690e47c99dSEric W. Biederman if (header->set == root_set) 14700e47c99dSEric W. Biederman return; 14710e47c99dSEric W. Biederman 14720e47c99dSEric W. Biederman core_parent = xlate_dir(root_set, parent); 14730e47c99dSEric W. Biederman if (IS_ERR(core_parent)) 14740e47c99dSEric W. Biederman return; 14750e47c99dSEric W. Biederman 147618d4b42eSJoel Granados list_for_each_table_entry(entry, header) { 14770e47c99dSEric W. Biederman struct ctl_table_header *link_head; 14780e47c99dSEric W. Biederman struct ctl_table *link; 14790e47c99dSEric W. Biederman const char *name = entry->procname; 14800e47c99dSEric W. Biederman 14810e47c99dSEric W. Biederman link = find_entry(&link_head, core_parent, name, strlen(name)); 14820e47c99dSEric W. Biederman if (link && 14830e47c99dSEric W. Biederman ((S_ISDIR(link->mode) && S_ISDIR(entry->mode)) || 14840e47c99dSEric W. Biederman (S_ISLNK(link->mode) && (link->data == root)))) { 14850e47c99dSEric W. Biederman drop_sysctl_table(link_head); 14860e47c99dSEric W. Biederman } 14870e47c99dSEric W. Biederman else { 148887ebdc00SAndrew Morton pr_err("sysctl link missing during unregister: "); 14890e47c99dSEric W. Biederman sysctl_print_dir(parent); 1490153ee1c4SGeert Uytterhoeven pr_cont("%s\n", name); 14910e47c99dSEric W. Biederman } 14920e47c99dSEric W. Biederman } 14930e47c99dSEric W. Biederman } 14940e47c99dSEric W. Biederman 1495938aaa4fSEric W. Biederman static void drop_sysctl_table(struct ctl_table_header *header) 1496938aaa4fSEric W. Biederman { 14977ec66d06SEric W. Biederman struct ctl_dir *parent = header->parent; 14987ec66d06SEric W. Biederman 1499938aaa4fSEric W. Biederman if (--header->nreg) 1500938aaa4fSEric W. Biederman return; 1501938aaa4fSEric W. Biederman 150289189557SYueHaibing if (parent) { 15030e47c99dSEric W. Biederman put_links(header); 1504938aaa4fSEric W. Biederman start_unregistering(header); 150589189557SYueHaibing } 150689189557SYueHaibing 1507938aaa4fSEric W. Biederman if (!--header->count) 1508938aaa4fSEric W. Biederman kfree_rcu(header, rcu); 15097ec66d06SEric W. Biederman 15107ec66d06SEric W. Biederman if (parent) 15117ec66d06SEric W. Biederman drop_sysctl_table(&parent->header); 1512938aaa4fSEric W. Biederman } 1513938aaa4fSEric W. Biederman 15141f87f0b5SEric W. Biederman /** 15151f87f0b5SEric W. Biederman * unregister_sysctl_table - unregister a sysctl table hierarchy 151619c4e618SJoel Granados * @header: the header returned from register_sysctl or __register_sysctl_table 15171f87f0b5SEric W. Biederman * 15181f87f0b5SEric W. Biederman * Unregisters the sysctl table and all children. proc entries may not 15191f87f0b5SEric W. Biederman * actually be removed until they are no longer used by anyone. 15201f87f0b5SEric W. Biederman */ 15211f87f0b5SEric W. Biederman void unregister_sysctl_table(struct ctl_table_header * header) 15221f87f0b5SEric W. Biederman { 15231f87f0b5SEric W. Biederman might_sleep(); 15241f87f0b5SEric W. Biederman 15251f87f0b5SEric W. Biederman if (header == NULL) 15261f87f0b5SEric W. Biederman return; 15271f87f0b5SEric W. Biederman 15281f87f0b5SEric W. Biederman spin_lock(&sysctl_lock); 1529938aaa4fSEric W. Biederman drop_sysctl_table(header); 15301f87f0b5SEric W. Biederman spin_unlock(&sysctl_lock); 15311f87f0b5SEric W. Biederman } 15321f87f0b5SEric W. Biederman EXPORT_SYMBOL(unregister_sysctl_table); 15331f87f0b5SEric W. Biederman 15340e47c99dSEric W. Biederman void setup_sysctl_set(struct ctl_table_set *set, 15359eb47c26SEric W. Biederman struct ctl_table_root *root, 15361f87f0b5SEric W. Biederman int (*is_seen)(struct ctl_table_set *)) 15371f87f0b5SEric W. Biederman { 15381347440dSDan Carpenter memset(set, 0, sizeof(*set)); 15390e47c99dSEric W. Biederman set->is_seen = is_seen; 1540b1f01e2bSJoel Granados init_header(&set->dir.header, root, set, NULL, root_table, 1); 15411f87f0b5SEric W. Biederman } 15421f87f0b5SEric W. Biederman 154397324cd8SEric W. Biederman void retire_sysctl_set(struct ctl_table_set *set) 154497324cd8SEric W. Biederman { 1545ac13ac6fSEric W. Biederman WARN_ON(!RB_EMPTY_ROOT(&set->dir.root)); 154697324cd8SEric W. Biederman } 15471f87f0b5SEric W. Biederman 15481e0edd3fSAlexey Dobriyan int __init proc_sys_init(void) 154977b14db5SEric W. Biederman { 1550e1675231SAlexey Dobriyan struct proc_dir_entry *proc_sys_root; 1551e1675231SAlexey Dobriyan 155277b14db5SEric W. Biederman proc_sys_root = proc_mkdir("sys", NULL); 15539043476fSAl Viro proc_sys_root->proc_iops = &proc_sys_dir_operations; 1554d56c0d45SAlexey Dobriyan proc_sys_root->proc_dir_ops = &proc_sys_dir_file_operations; 155577b14db5SEric W. Biederman proc_sys_root->nlink = 0; 1556de4e83bdSEric W. Biederman 1557d8c0418aSLuis Chamberlain return sysctl_init_bases(); 155877b14db5SEric W. Biederman } 15593db978d4SVlastimil Babka 15600a477e1aSVlastimil Babka struct sysctl_alias { 15610a477e1aSVlastimil Babka const char *kernel_param; 15620a477e1aSVlastimil Babka const char *sysctl_param; 15630a477e1aSVlastimil Babka }; 15640a477e1aSVlastimil Babka 15650a477e1aSVlastimil Babka /* 15660a477e1aSVlastimil Babka * Historically some settings had both sysctl and a command line parameter. 15670a477e1aSVlastimil Babka * With the generic sysctl. parameter support, we can handle them at a single 15680a477e1aSVlastimil Babka * place and only keep the historical name for compatibility. This is not meant 15690a477e1aSVlastimil Babka * to add brand new aliases. When adding existing aliases, consider whether 15700a477e1aSVlastimil Babka * the possibly different moment of changing the value (e.g. from early_param 15710a477e1aSVlastimil Babka * to the moment do_sysctl_args() is called) is an issue for the specific 15720a477e1aSVlastimil Babka * parameter. 15730a477e1aSVlastimil Babka */ 15740a477e1aSVlastimil Babka static const struct sysctl_alias sysctl_aliases[] = { 1575f117955aSGuilherme G. Piccoli {"hardlockup_all_cpu_backtrace", "kernel.hardlockup_all_cpu_backtrace" }, 1576b467f3efSVlastimil Babka {"hung_task_panic", "kernel.hung_task_panic" }, 1577f117955aSGuilherme G. Piccoli {"numa_zonelist_order", "vm.numa_zonelist_order" }, 1578f117955aSGuilherme G. Piccoli {"softlockup_all_cpu_backtrace", "kernel.softlockup_all_cpu_backtrace" }, 1579f117955aSGuilherme G. Piccoli {"softlockup_panic", "kernel.softlockup_panic" }, 15800a477e1aSVlastimil Babka { } 15810a477e1aSVlastimil Babka }; 15820a477e1aSVlastimil Babka 15830a477e1aSVlastimil Babka static const char *sysctl_find_alias(char *param) 15840a477e1aSVlastimil Babka { 15850a477e1aSVlastimil Babka const struct sysctl_alias *alias; 15860a477e1aSVlastimil Babka 15870a477e1aSVlastimil Babka for (alias = &sysctl_aliases[0]; alias->kernel_param != NULL; alias++) { 15880a477e1aSVlastimil Babka if (strcmp(alias->kernel_param, param) == 0) 15890a477e1aSVlastimil Babka return alias->sysctl_param; 15900a477e1aSVlastimil Babka } 15910a477e1aSVlastimil Babka 15920a477e1aSVlastimil Babka return NULL; 15930a477e1aSVlastimil Babka } 15940a477e1aSVlastimil Babka 1595*2b1ee516SKrister Johansen bool sysctl_is_alias(char *param) 1596*2b1ee516SKrister Johansen { 1597*2b1ee516SKrister Johansen const char *alias = sysctl_find_alias(param); 1598*2b1ee516SKrister Johansen 1599*2b1ee516SKrister Johansen return alias != NULL; 1600*2b1ee516SKrister Johansen } 1601*2b1ee516SKrister Johansen 16023db978d4SVlastimil Babka /* Set sysctl value passed on kernel command line. */ 16033db978d4SVlastimil Babka static int process_sysctl_arg(char *param, char *val, 16043db978d4SVlastimil Babka const char *unused, void *arg) 16053db978d4SVlastimil Babka { 16063db978d4SVlastimil Babka char *path; 16073db978d4SVlastimil Babka struct vfsmount **proc_mnt = arg; 16083db978d4SVlastimil Babka struct file_system_type *proc_fs_type; 16093db978d4SVlastimil Babka struct file *file; 16103db978d4SVlastimil Babka int len; 16113db978d4SVlastimil Babka int err; 16123db978d4SVlastimil Babka loff_t pos = 0; 16133db978d4SVlastimil Babka ssize_t wret; 16143db978d4SVlastimil Babka 16150a477e1aSVlastimil Babka if (strncmp(param, "sysctl", sizeof("sysctl") - 1) == 0) { 16163db978d4SVlastimil Babka param += sizeof("sysctl") - 1; 16173db978d4SVlastimil Babka 16183db978d4SVlastimil Babka if (param[0] != '/' && param[0] != '.') 16193db978d4SVlastimil Babka return 0; 16203db978d4SVlastimil Babka 16213db978d4SVlastimil Babka param++; 16220a477e1aSVlastimil Babka } else { 16230a477e1aSVlastimil Babka param = (char *) sysctl_find_alias(param); 16240a477e1aSVlastimil Babka if (!param) 16250a477e1aSVlastimil Babka return 0; 16260a477e1aSVlastimil Babka } 16273db978d4SVlastimil Babka 1628697edcb0SXiaoming Ni if (!val) 1629697edcb0SXiaoming Ni return -EINVAL; 1630697edcb0SXiaoming Ni len = strlen(val); 1631697edcb0SXiaoming Ni if (len == 0) 1632697edcb0SXiaoming Ni return -EINVAL; 1633697edcb0SXiaoming Ni 16343db978d4SVlastimil Babka /* 16353db978d4SVlastimil Babka * To set sysctl options, we use a temporary mount of proc, look up the 16363db978d4SVlastimil Babka * respective sys/ file and write to it. To avoid mounting it when no 16373db978d4SVlastimil Babka * options were given, we mount it only when the first sysctl option is 16383db978d4SVlastimil Babka * found. Why not a persistent mount? There are problems with a 16393db978d4SVlastimil Babka * persistent mount of proc in that it forces userspace not to use any 16403db978d4SVlastimil Babka * proc mount options. 16413db978d4SVlastimil Babka */ 16423db978d4SVlastimil Babka if (!*proc_mnt) { 16433db978d4SVlastimil Babka proc_fs_type = get_fs_type("proc"); 16443db978d4SVlastimil Babka if (!proc_fs_type) { 16453db978d4SVlastimil Babka pr_err("Failed to find procfs to set sysctl from command line\n"); 16463db978d4SVlastimil Babka return 0; 16473db978d4SVlastimil Babka } 16483db978d4SVlastimil Babka *proc_mnt = kern_mount(proc_fs_type); 16493db978d4SVlastimil Babka put_filesystem(proc_fs_type); 16503db978d4SVlastimil Babka if (IS_ERR(*proc_mnt)) { 16513db978d4SVlastimil Babka pr_err("Failed to mount procfs to set sysctl from command line\n"); 16523db978d4SVlastimil Babka return 0; 16533db978d4SVlastimil Babka } 16543db978d4SVlastimil Babka } 16553db978d4SVlastimil Babka 16563db978d4SVlastimil Babka path = kasprintf(GFP_KERNEL, "sys/%s", param); 16573db978d4SVlastimil Babka if (!path) 16583db978d4SVlastimil Babka panic("%s: Failed to allocate path for %s\n", __func__, param); 16593db978d4SVlastimil Babka strreplace(path, '.', '/'); 16603db978d4SVlastimil Babka 1661ffb37ca3SAl Viro file = file_open_root_mnt(*proc_mnt, path, O_WRONLY, 0); 16623db978d4SVlastimil Babka if (IS_ERR(file)) { 16633db978d4SVlastimil Babka err = PTR_ERR(file); 16643db978d4SVlastimil Babka if (err == -ENOENT) 16653db978d4SVlastimil Babka pr_err("Failed to set sysctl parameter '%s=%s': parameter not found\n", 16663db978d4SVlastimil Babka param, val); 16673db978d4SVlastimil Babka else if (err == -EACCES) 16683db978d4SVlastimil Babka pr_err("Failed to set sysctl parameter '%s=%s': permission denied (read-only?)\n", 16693db978d4SVlastimil Babka param, val); 16703db978d4SVlastimil Babka else 16713db978d4SVlastimil Babka pr_err("Error %pe opening proc file to set sysctl parameter '%s=%s'\n", 16723db978d4SVlastimil Babka file, param, val); 16733db978d4SVlastimil Babka goto out; 16743db978d4SVlastimil Babka } 16753db978d4SVlastimil Babka wret = kernel_write(file, val, len, &pos); 16763db978d4SVlastimil Babka if (wret < 0) { 16773db978d4SVlastimil Babka err = wret; 16783db978d4SVlastimil Babka if (err == -EINVAL) 16793db978d4SVlastimil Babka pr_err("Failed to set sysctl parameter '%s=%s': invalid value\n", 16803db978d4SVlastimil Babka param, val); 16813db978d4SVlastimil Babka else 16823db978d4SVlastimil Babka pr_err("Error %pe writing to proc file to set sysctl parameter '%s=%s'\n", 16833db978d4SVlastimil Babka ERR_PTR(err), param, val); 16843db978d4SVlastimil Babka } else if (wret != len) { 16853db978d4SVlastimil Babka pr_err("Wrote only %zd bytes of %d writing to proc file %s to set sysctl parameter '%s=%s\n", 16863db978d4SVlastimil Babka wret, len, path, param, val); 16873db978d4SVlastimil Babka } 16883db978d4SVlastimil Babka 16893db978d4SVlastimil Babka err = filp_close(file, NULL); 16903db978d4SVlastimil Babka if (err) 16913db978d4SVlastimil Babka pr_err("Error %pe closing proc file to set sysctl parameter '%s=%s\n", 16923db978d4SVlastimil Babka ERR_PTR(err), param, val); 16933db978d4SVlastimil Babka out: 16943db978d4SVlastimil Babka kfree(path); 16953db978d4SVlastimil Babka return 0; 16963db978d4SVlastimil Babka } 16973db978d4SVlastimil Babka 16983db978d4SVlastimil Babka void do_sysctl_args(void) 16993db978d4SVlastimil Babka { 17003db978d4SVlastimil Babka char *command_line; 17013db978d4SVlastimil Babka struct vfsmount *proc_mnt = NULL; 17023db978d4SVlastimil Babka 17033db978d4SVlastimil Babka command_line = kstrdup(saved_command_line, GFP_KERNEL); 17043db978d4SVlastimil Babka if (!command_line) 17053db978d4SVlastimil Babka panic("%s: Failed to allocate copy of command line\n", __func__); 17063db978d4SVlastimil Babka 17073db978d4SVlastimil Babka parse_args("Setting sysctl args", command_line, 17083db978d4SVlastimil Babka NULL, 0, -1, -1, &proc_mnt, process_sysctl_arg); 17093db978d4SVlastimil Babka 17103db978d4SVlastimil Babka if (proc_mnt) 17113db978d4SVlastimil Babka kern_unmount(proc_mnt); 17123db978d4SVlastimil Babka 17133db978d4SVlastimil Babka kfree(command_line); 17143db978d4SVlastimil Babka } 1715