lockd: Add helper to sanity check incoming NOTIFY requests
[safe/jmp/linux-2.6] / kernel / sysctl.c
index 6b16e16..4588b2c 100644 (file)
@@ -43,6 +43,7 @@
 #include <linux/limits.h>
 #include <linux/dcache.h>
 #include <linux/syscalls.h>
+#include <linux/vmstat.h>
 #include <linux/nfs_fs.h>
 #include <linux/acpi.h>
 #include <linux/reboot.h>
@@ -80,7 +81,6 @@ extern int sysctl_drop_caches;
 extern int percpu_pagelist_fraction;
 extern int compat_log;
 extern int maps_protect;
-extern int sysctl_stat_interval;
 extern int latencytop_enabled;
 extern int sysctl_nr_open_min, sysctl_nr_open_max;
 #ifdef CONFIG_RCU_TORTURE_TEST
@@ -88,15 +88,16 @@ extern int rcutorture_runnable;
 #endif /* #ifdef CONFIG_RCU_TORTURE_TEST */
 
 /* Constants used for minimum and  maximum */
-#if defined(CONFIG_DETECT_SOFTLOCKUP) || defined(CONFIG_HIGHMEM)
+#if defined(CONFIG_HIGHMEM) || defined(CONFIG_DETECT_SOFTLOCKUP)
 static int one = 1;
 #endif
 
 #ifdef CONFIG_DETECT_SOFTLOCKUP
 static int sixty = 60;
+static int neg_one = -1;
 #endif
 
-#ifdef CONFIG_MMU
+#if defined(CONFIG_MMU) && defined(CONFIG_FILE_LOCKING)
 static int two = 2;
 #endif
 
@@ -110,7 +111,7 @@ static int min_percpu_pagelist_fract = 8;
 
 static int ngroups_max = NGROUPS_MAX;
 
-#ifdef CONFIG_KMOD
+#ifdef CONFIG_MODULES
 extern char modprobe_path[];
 #endif
 #ifdef CONFIG_CHR_DEV_SG
@@ -158,13 +159,15 @@ static int proc_dointvec_taint(struct ctl_table *table, int write, struct file *
 static struct ctl_table root_table[];
 static struct ctl_table_root sysctl_table_root;
 static struct ctl_table_header root_table_header = {
+       .count = 1,
        .ctl_table = root_table,
-       .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.header_list),
+       .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),
        .root = &sysctl_table_root,
+       .set = &sysctl_table_root.default_set,
 };
 static struct ctl_table_root sysctl_table_root = {
        .root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
-       .header_list = LIST_HEAD_INIT(root_table_header.ctl_entry),
+       .default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
 };
 
 static struct ctl_table kern_table[];
@@ -475,7 +478,7 @@ static struct ctl_table kern_table[] = {
                .proc_handler   = &ftrace_enable_sysctl,
        },
 #endif
-#ifdef CONFIG_KMOD
+#ifdef CONFIG_MODULES
        {
                .ctl_name       = KERN_MODPROBE,
                .procname       = "modprobe",
@@ -623,7 +626,7 @@ static struct ctl_table kern_table[] = {
        {
                .ctl_name       = KERN_PRINTK_RATELIMIT,
                .procname       = "printk_ratelimit",
-               .data           = &printk_ratelimit_jiffies,
+               .data           = &printk_ratelimit_state.interval,
                .maxlen         = sizeof(int),
                .mode           = 0644,
                .proc_handler   = &proc_dointvec_jiffies,
@@ -632,7 +635,7 @@ static struct ctl_table kern_table[] = {
        {
                .ctl_name       = KERN_PRINTK_RATELIMIT_BURST,
                .procname       = "printk_ratelimit_burst",
-               .data           = &printk_ratelimit_burst,
+               .data           = &printk_ratelimit_state.burst,
                .maxlen         = sizeof(int),
                .mode           = 0644,
                .proc_handler   = &proc_dointvec,
@@ -739,13 +742,24 @@ static struct ctl_table kern_table[] = {
 #ifdef CONFIG_DETECT_SOFTLOCKUP
        {
                .ctl_name       = CTL_UNNUMBERED,
+               .procname       = "softlockup_panic",
+               .data           = &softlockup_panic,
+               .maxlen         = sizeof(int),
+               .mode           = 0644,
+               .proc_handler   = &proc_dointvec_minmax,
+               .strategy       = &sysctl_intvec,
+               .extra1         = &zero,
+               .extra2         = &one,
+       },
+       {
+               .ctl_name       = CTL_UNNUMBERED,
                .procname       = "softlockup_thresh",
                .data           = &softlockup_thresh,
-               .maxlen         = sizeof(unsigned long),
+               .maxlen         = sizeof(int),
                .mode           = 0644,
-               .proc_handler   = &proc_doulongvec_minmax,
+               .proc_handler   = &proc_dointvec_minmax,
                .strategy       = &sysctl_intvec,
-               .extra1         = &one,
+               .extra1         = &neg_one,
                .extra2         = &sixty,
        },
        {
@@ -947,7 +961,7 @@ static struct ctl_table vm_table[] = {
 #ifdef CONFIG_HUGETLB_PAGE
         {
                .procname       = "nr_hugepages",
-               .data           = &max_huge_pages,
+               .data           = NULL,
                .maxlen         = sizeof(unsigned long),
                .mode           = 0644,
                .proc_handler   = &hugetlb_sysctl_handler,
@@ -973,10 +987,12 @@ static struct ctl_table vm_table[] = {
        {
                .ctl_name       = CTL_UNNUMBERED,
                .procname       = "nr_overcommit_hugepages",
-               .data           = &sysctl_overcommit_huge_pages,
-               .maxlen         = sizeof(sysctl_overcommit_huge_pages),
+               .data           = NULL,
+               .maxlen         = sizeof(unsigned long),
                .mode           = 0644,
                .proc_handler   = &hugetlb_overcommit_handler,
+               .extra1         = (void *)&hugetlb_zero,
+               .extra2         = (void *)&hugetlb_infinity,
        },
 #endif
        {
@@ -1245,6 +1261,7 @@ static struct ctl_table fs_table[] = {
                .extra1         = &minolduid,
                .extra2         = &maxolduid,
        },
+#ifdef CONFIG_FILE_LOCKING
        {
                .ctl_name       = FS_LEASES,
                .procname       = "leases-enable",
@@ -1253,6 +1270,7 @@ static struct ctl_table fs_table[] = {
                .mode           = 0644,
                .proc_handler   = &proc_dointvec,
        },
+#endif
 #ifdef CONFIG_DNOTIFY
        {
                .ctl_name       = FS_DIR_NOTIFY,
@@ -1264,6 +1282,7 @@ static struct ctl_table fs_table[] = {
        },
 #endif
 #ifdef CONFIG_MMU
+#ifdef CONFIG_FILE_LOCKING
        {
                .ctl_name       = FS_LEASE_TIME,
                .procname       = "lease-break-time",
@@ -1275,6 +1294,7 @@ static struct ctl_table fs_table[] = {
                .extra1         = &zero,
                .extra2         = &two,
        },
+#endif
        {
                .procname       = "aio-nr",
                .data           = &aio_nr,
@@ -1372,6 +1392,9 @@ static void start_unregistering(struct ctl_table_header *p)
                spin_unlock(&sysctl_lock);
                wait_for_completion(&wait);
                spin_lock(&sysctl_lock);
+       } else {
+               /* anything non-NULL; we'll never dereference it */
+               p->unregistering = ERR_PTR(-EINVAL);
        }
        /*
         * do not remove from the list until nobody holds it; walking the
@@ -1380,6 +1403,32 @@ static void start_unregistering(struct ctl_table_header *p)
        list_del_init(&p->ctl_entry);
 }
 
+void sysctl_head_get(struct ctl_table_header *head)
+{
+       spin_lock(&sysctl_lock);
+       head->count++;
+       spin_unlock(&sysctl_lock);
+}
+
+void sysctl_head_put(struct ctl_table_header *head)
+{
+       spin_lock(&sysctl_lock);
+       if (!--head->count)
+               kfree(head);
+       spin_unlock(&sysctl_lock);
+}
+
+struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
+{
+       if (!head)
+               BUG();
+       spin_lock(&sysctl_lock);
+       if (!use_table(head))
+               head = ERR_PTR(-ENOENT);
+       spin_unlock(&sysctl_lock);
+       return head;
+}
+
 void sysctl_head_finish(struct ctl_table_header *head)
 {
        if (!head)
@@ -1389,14 +1438,20 @@ void sysctl_head_finish(struct ctl_table_header *head)
        spin_unlock(&sysctl_lock);
 }
 
+static struct ctl_table_set *
+lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
+{
+       struct ctl_table_set *set = &root->default_set;
+       if (root->lookup)
+               set = root->lookup(root, namespaces);
+       return set;
+}
+
 static struct list_head *
 lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
 {
-       struct list_head *header_list;
-       header_list = &root->header_list;
-       if (root->lookup)
-               header_list = root->lookup(root, namespaces);
-       return header_list;
+       struct ctl_table_set *set = lookup_header_set(root, namespaces);
+       return &set->list;
 }
 
 struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
@@ -1466,9 +1521,9 @@ static int do_sysctl_strategy(struct ctl_table_root *root,
        int op = 0, rc;
 
        if (oldval)
-               op |= 004;
+               op |= MAY_READ;
        if (newval)
-               op |= 002;
+               op |= MAY_WRITE;
        if (sysctl_perm(root, table, op))
                return -EPERM;
 
@@ -1510,7 +1565,7 @@ repeat:
                if (n == table->ctl_name) {
                        int error;
                        if (table->child) {
-                               if (sysctl_perm(root, table, 001))
+                               if (sysctl_perm(root, table, MAY_EXEC))
                                        return -EPERM;
                                name++;
                                nlen--;
@@ -1585,7 +1640,7 @@ static int test_perm(int mode, int op)
                mode >>= 6;
        else if (in_egroup_p(0))
                mode >>= 3;
-       if ((mode & op & 0007) == op)
+       if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
                return 0;
        return -EACCES;
 }
@@ -1595,7 +1650,7 @@ int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
        int error;
        int mode;
 
-       error = security_sysctl(table, op);
+       error = security_sysctl(table, op & (MAY_READ | MAY_WRITE | MAY_EXEC));
        if (error)
                return error;
 
@@ -1630,6 +1685,54 @@ static __init int sysctl_init(void)
 
 core_initcall(sysctl_init);
 
+static struct ctl_table *is_branch_in(struct ctl_table *branch,
+                                     struct ctl_table *table)
+{
+       struct ctl_table *p;
+       const char *s = branch->procname;
+
+       /* branch should have named subdirectory as its first element */
+       if (!s || !branch->child)
+               return NULL;
+
+       /* ... and nothing else */
+       if (branch[1].procname || branch[1].ctl_name)
+               return NULL;
+
+       /* table should contain subdirectory with the same name */
+       for (p = table; p->procname || p->ctl_name; p++) {
+               if (!p->child)
+                       continue;
+               if (p->procname && strcmp(p->procname, s) == 0)
+                       return p;
+       }
+       return NULL;
+}
+
+/* see if attaching q to p would be an improvement */
+static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
+{
+       struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
+       struct ctl_table *next;
+       int is_better = 0;
+       int not_in_parent = !p->attached_by;
+
+       while ((next = is_branch_in(by, to)) != NULL) {
+               if (by == q->attached_by)
+                       is_better = 1;
+               if (to == p->attached_by)
+                       not_in_parent = 1;
+               by = by->child;
+               to = next->child;
+       }
+
+       if (is_better && not_in_parent) {
+               q->attached_by = by;
+               q->attached_to = to;
+               q->parent = p;
+       }
+}
+
 /**
  * __register_sysctl_paths - register a sysctl hierarchy
  * @root: List of sysctl headers to register on
@@ -1706,10 +1809,10 @@ struct ctl_table_header *__register_sysctl_paths(
        struct nsproxy *namespaces,
        const struct ctl_path *path, struct ctl_table *table)
 {
-       struct list_head *header_list;
        struct ctl_table_header *header;
        struct ctl_table *new, **prevp;
        unsigned int n, npath;
+       struct ctl_table_set *set;
 
        /* Count the path components */
        for (npath = 0; path[npath].ctl_name || path[npath].procname; ++npath)
@@ -1751,6 +1854,7 @@ struct ctl_table_header *__register_sysctl_paths(
        header->unregistering = NULL;
        header->root = root;
        sysctl_set_parent(NULL, header->ctl_table);
+       header->count = 1;
 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
        if (sysctl_check_table(namespaces, header->ctl_table)) {
                kfree(header);
@@ -1758,8 +1862,20 @@ struct ctl_table_header *__register_sysctl_paths(
        }
 #endif
        spin_lock(&sysctl_lock);
-       header_list = lookup_header_list(root, namespaces);
-       list_add_tail(&header->ctl_entry, header_list);
+       header->set = lookup_header_set(root, namespaces);
+       header->attached_by = header->ctl_table;
+       header->attached_to = root_table;
+       header->parent = &root_table_header;
+       for (set = header->set; set; set = set->parent) {
+               struct ctl_table_header *p;
+               list_for_each_entry(p, &set->list, ctl_entry) {
+                       if (p->unregistering)
+                               continue;
+                       try_attach(p, header);
+               }
+       }
+       header->parent->count++;
+       list_add_tail(&header->ctl_entry, &header->set->list);
        spin_unlock(&sysctl_lock);
 
        return header;
@@ -1814,8 +1930,37 @@ void unregister_sysctl_table(struct ctl_table_header * header)
 
        spin_lock(&sysctl_lock);
        start_unregistering(header);
+       if (!--header->parent->count) {
+               WARN_ON(1);
+               kfree(header->parent);
+       }
+       if (!--header->count)
+               kfree(header);
+       spin_unlock(&sysctl_lock);
+}
+
+int sysctl_is_seen(struct ctl_table_header *p)
+{
+       struct ctl_table_set *set = p->set;
+       int res;
+       spin_lock(&sysctl_lock);
+       if (p->unregistering)
+               res = 0;
+       else if (!set->is_seen)
+               res = 1;
+       else
+               res = set->is_seen(set);
        spin_unlock(&sysctl_lock);
-       kfree(header);
+       return res;
+}
+
+void setup_sysctl_set(struct ctl_table_set *p,
+       struct ctl_table_set *parent,
+       int (*is_seen)(struct ctl_table_set *))
+{
+       INIT_LIST_HEAD(&p->list);
+       p->parent = parent ? parent : &sysctl_table_root.default_set;
+       p->is_seen = is_seen;
 }
 
 #else /* !CONFIG_SYSCTL */
@@ -1834,6 +1979,16 @@ void unregister_sysctl_table(struct ctl_table_header * table)
 {
 }
 
+void setup_sysctl_set(struct ctl_table_set *p,
+       struct ctl_table_set *parent,
+       int (*is_seen)(struct ctl_table_set *))
+{
+}
+
+void sysctl_head_put(struct ctl_table_header *head)
+{
+}
+
 #endif /* CONFIG_SYSCTL */
 
 /*