diff options
author | Eric W. Biederman <ebiederm@xmission.com> | 2012-01-06 07:07:15 -0500 |
---|---|---|
committer | Eric W. Biederman <ebiederm@xmission.com> | 2012-01-24 19:37:54 -0500 |
commit | 1f87f0b52b1d6581168cb80f86746bc4df918d01 (patch) | |
tree | 73576c0872c61c526ade225d6441a6610beb666b /kernel/sysctl.c | |
parent | de4e83bd6b5e16d491ec068cd22801d5d063b07a (diff) |
sysctl: Move the implementation into fs/proc/proc_sysctl.c
Move the core sysctl code from kernel/sysctl.c and kernel/sysctl_check.c
into fs/proc/proc_sysctl.c.
Currently sysctl maintenance is hampered by the sysctl implementation
being split across 3 files with artificial layering between them.
Consolidate the entire sysctl implementation into 1 file so that
it is easier to see what is going on and hopefully allowing for
simpler maintenance.
For functions that are now only used in fs/proc/proc_sysctl.c remove
their declarations from sysctl.h and make them static in fs/proc/proc_sysctl.c
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Diffstat (limited to 'kernel/sysctl.c')
-rw-r--r-- | kernel/sysctl.c | 464 |
1 files changed, 0 insertions, 464 deletions
diff --git a/kernel/sysctl.c b/kernel/sysctl.c index ad460248acc7..b774909ed46c 100644 --- a/kernel/sysctl.c +++ b/kernel/sysctl.c | |||
@@ -192,20 +192,6 @@ static int sysrq_sysctl_handler(ctl_table *table, int write, | |||
192 | 192 | ||
193 | #endif | 193 | #endif |
194 | 194 | ||
195 | static struct ctl_table root_table[1]; | ||
196 | static struct ctl_table_root sysctl_table_root; | ||
197 | static struct ctl_table_header root_table_header = { | ||
198 | {{.count = 1, | ||
199 | .ctl_table = root_table, | ||
200 | .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}}, | ||
201 | .root = &sysctl_table_root, | ||
202 | .set = &sysctl_table_root.default_set, | ||
203 | }; | ||
204 | static struct ctl_table_root sysctl_table_root = { | ||
205 | .root_list = LIST_HEAD_INIT(sysctl_table_root.root_list), | ||
206 | .default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry), | ||
207 | }; | ||
208 | |||
209 | static struct ctl_table kern_table[]; | 195 | static struct ctl_table kern_table[]; |
210 | static struct ctl_table vm_table[]; | 196 | static struct ctl_table vm_table[]; |
211 | static struct ctl_table fs_table[]; | 197 | static struct ctl_table fs_table[]; |
@@ -1559,459 +1545,12 @@ static struct ctl_table dev_table[] = { | |||
1559 | { } | 1545 | { } |
1560 | }; | 1546 | }; |
1561 | 1547 | ||
1562 | static DEFINE_SPINLOCK(sysctl_lock); | ||
1563 | |||
1564 | /* called under sysctl_lock */ | ||
1565 | static int use_table(struct ctl_table_header *p) | ||
1566 | { | ||
1567 | if (unlikely(p->unregistering)) | ||
1568 | return 0; | ||
1569 | p->used++; | ||
1570 | return 1; | ||
1571 | } | ||
1572 | |||
1573 | /* called under sysctl_lock */ | ||
1574 | static void unuse_table(struct ctl_table_header *p) | ||
1575 | { | ||
1576 | if (!--p->used) | ||
1577 | if (unlikely(p->unregistering)) | ||
1578 | complete(p->unregistering); | ||
1579 | } | ||
1580 | |||
1581 | /* called under sysctl_lock, will reacquire if has to wait */ | ||
1582 | static void start_unregistering(struct ctl_table_header *p) | ||
1583 | { | ||
1584 | /* | ||
1585 | * if p->used is 0, nobody will ever touch that entry again; | ||
1586 | * we'll eliminate all paths to it before dropping sysctl_lock | ||
1587 | */ | ||
1588 | if (unlikely(p->used)) { | ||
1589 | struct completion wait; | ||
1590 | init_completion(&wait); | ||
1591 | p->unregistering = &wait; | ||
1592 | spin_unlock(&sysctl_lock); | ||
1593 | wait_for_completion(&wait); | ||
1594 | spin_lock(&sysctl_lock); | ||
1595 | } else { | ||
1596 | /* anything non-NULL; we'll never dereference it */ | ||
1597 | p->unregistering = ERR_PTR(-EINVAL); | ||
1598 | } | ||
1599 | /* | ||
1600 | * do not remove from the list until nobody holds it; walking the | ||
1601 | * list in do_sysctl() relies on that. | ||
1602 | */ | ||
1603 | list_del_init(&p->ctl_entry); | ||
1604 | } | ||
1605 | |||
1606 | void sysctl_head_get(struct ctl_table_header *head) | ||
1607 | { | ||
1608 | spin_lock(&sysctl_lock); | ||
1609 | head->count++; | ||
1610 | spin_unlock(&sysctl_lock); | ||
1611 | } | ||
1612 | |||
1613 | void sysctl_head_put(struct ctl_table_header *head) | ||
1614 | { | ||
1615 | spin_lock(&sysctl_lock); | ||
1616 | if (!--head->count) | ||
1617 | kfree_rcu(head, rcu); | ||
1618 | spin_unlock(&sysctl_lock); | ||
1619 | } | ||
1620 | |||
1621 | struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head) | ||
1622 | { | ||
1623 | if (!head) | ||
1624 | BUG(); | ||
1625 | spin_lock(&sysctl_lock); | ||
1626 | if (!use_table(head)) | ||
1627 | head = ERR_PTR(-ENOENT); | ||
1628 | spin_unlock(&sysctl_lock); | ||
1629 | return head; | ||
1630 | } | ||
1631 | |||
1632 | void sysctl_head_finish(struct ctl_table_header *head) | ||
1633 | { | ||
1634 | if (!head) | ||
1635 | return; | ||
1636 | spin_lock(&sysctl_lock); | ||
1637 | unuse_table(head); | ||
1638 | spin_unlock(&sysctl_lock); | ||
1639 | } | ||
1640 | |||
1641 | static struct ctl_table_set * | ||
1642 | lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces) | ||
1643 | { | ||
1644 | struct ctl_table_set *set = &root->default_set; | ||
1645 | if (root->lookup) | ||
1646 | set = root->lookup(root, namespaces); | ||
1647 | return set; | ||
1648 | } | ||
1649 | |||
1650 | static struct list_head * | ||
1651 | lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces) | ||
1652 | { | ||
1653 | struct ctl_table_set *set = lookup_header_set(root, namespaces); | ||
1654 | return &set->list; | ||
1655 | } | ||
1656 | |||
1657 | struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces, | ||
1658 | struct ctl_table_header *prev) | ||
1659 | { | ||
1660 | struct ctl_table_root *root; | ||
1661 | struct list_head *header_list; | ||
1662 | struct ctl_table_header *head; | ||
1663 | struct list_head *tmp; | ||
1664 | |||
1665 | spin_lock(&sysctl_lock); | ||
1666 | if (prev) { | ||
1667 | head = prev; | ||
1668 | tmp = &prev->ctl_entry; | ||
1669 | unuse_table(prev); | ||
1670 | goto next; | ||
1671 | } | ||
1672 | tmp = &root_table_header.ctl_entry; | ||
1673 | for (;;) { | ||
1674 | head = list_entry(tmp, struct ctl_table_header, ctl_entry); | ||
1675 | |||
1676 | if (!use_table(head)) | ||
1677 | goto next; | ||
1678 | spin_unlock(&sysctl_lock); | ||
1679 | return head; | ||
1680 | next: | ||
1681 | root = head->root; | ||
1682 | tmp = tmp->next; | ||
1683 | header_list = lookup_header_list(root, namespaces); | ||
1684 | if (tmp != header_list) | ||
1685 | continue; | ||
1686 | |||
1687 | do { | ||
1688 | root = list_entry(root->root_list.next, | ||
1689 | struct ctl_table_root, root_list); | ||
1690 | if (root == &sysctl_table_root) | ||
1691 | goto out; | ||
1692 | header_list = lookup_header_list(root, namespaces); | ||
1693 | } while (list_empty(header_list)); | ||
1694 | tmp = header_list->next; | ||
1695 | } | ||
1696 | out: | ||
1697 | spin_unlock(&sysctl_lock); | ||
1698 | return NULL; | ||
1699 | } | ||
1700 | |||
1701 | struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev) | ||
1702 | { | ||
1703 | return __sysctl_head_next(current->nsproxy, prev); | ||
1704 | } | ||
1705 | |||
1706 | void register_sysctl_root(struct ctl_table_root *root) | ||
1707 | { | ||
1708 | spin_lock(&sysctl_lock); | ||
1709 | list_add_tail(&root->root_list, &sysctl_table_root.root_list); | ||
1710 | spin_unlock(&sysctl_lock); | ||
1711 | } | ||
1712 | |||
1713 | /* | ||
1714 | * sysctl_perm does NOT grant the superuser all rights automatically, because | ||
1715 | * some sysctl variables are readonly even to root. | ||
1716 | */ | ||
1717 | |||
1718 | static int test_perm(int mode, int op) | ||
1719 | { | ||
1720 | if (!current_euid()) | ||
1721 | mode >>= 6; | ||
1722 | else if (in_egroup_p(0)) | ||
1723 | mode >>= 3; | ||
1724 | if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0) | ||
1725 | return 0; | ||
1726 | return -EACCES; | ||
1727 | } | ||
1728 | |||
1729 | int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op) | ||
1730 | { | ||
1731 | int mode; | ||
1732 | |||
1733 | if (root->permissions) | ||
1734 | mode = root->permissions(root, current->nsproxy, table); | ||
1735 | else | ||
1736 | mode = table->mode; | ||
1737 | |||
1738 | return test_perm(mode, op); | ||
1739 | } | ||
1740 | |||
1741 | static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table) | ||
1742 | { | ||
1743 | for (; table->procname; table++) { | ||
1744 | table->parent = parent; | ||
1745 | if (table->child) | ||
1746 | sysctl_set_parent(table, table->child); | ||
1747 | } | ||
1748 | } | ||
1749 | |||
1750 | int __init sysctl_init(void) | 1548 | int __init sysctl_init(void) |
1751 | { | 1549 | { |
1752 | register_sysctl_table(sysctl_base_table); | 1550 | register_sysctl_table(sysctl_base_table); |
1753 | return 0; | 1551 | return 0; |
1754 | } | 1552 | } |
1755 | 1553 | ||
1756 | static struct ctl_table *is_branch_in(struct ctl_table *branch, | ||
1757 | struct ctl_table *table) | ||
1758 | { | ||
1759 | struct ctl_table *p; | ||
1760 | const char *s = branch->procname; | ||
1761 | |||
1762 | /* branch should have named subdirectory as its first element */ | ||
1763 | if (!s || !branch->child) | ||
1764 | return NULL; | ||
1765 | |||
1766 | /* ... and nothing else */ | ||
1767 | if (branch[1].procname) | ||
1768 | return NULL; | ||
1769 | |||
1770 | /* table should contain subdirectory with the same name */ | ||
1771 | for (p = table; p->procname; p++) { | ||
1772 | if (!p->child) | ||
1773 | continue; | ||
1774 | if (p->procname && strcmp(p->procname, s) == 0) | ||
1775 | return p; | ||
1776 | } | ||
1777 | return NULL; | ||
1778 | } | ||
1779 | |||
1780 | /* see if attaching q to p would be an improvement */ | ||
1781 | static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q) | ||
1782 | { | ||
1783 | struct ctl_table *to = p->ctl_table, *by = q->ctl_table; | ||
1784 | struct ctl_table *next; | ||
1785 | int is_better = 0; | ||
1786 | int not_in_parent = !p->attached_by; | ||
1787 | |||
1788 | while ((next = is_branch_in(by, to)) != NULL) { | ||
1789 | if (by == q->attached_by) | ||
1790 | is_better = 1; | ||
1791 | if (to == p->attached_by) | ||
1792 | not_in_parent = 1; | ||
1793 | by = by->child; | ||
1794 | to = next->child; | ||
1795 | } | ||
1796 | |||
1797 | if (is_better && not_in_parent) { | ||
1798 | q->attached_by = by; | ||
1799 | q->attached_to = to; | ||
1800 | q->parent = p; | ||
1801 | } | ||
1802 | } | ||
1803 | |||
1804 | /** | ||
1805 | * __register_sysctl_paths - register a sysctl hierarchy | ||
1806 | * @root: List of sysctl headers to register on | ||
1807 | * @namespaces: Data to compute which lists of sysctl entries are visible | ||
1808 | * @path: The path to the directory the sysctl table is in. | ||
1809 | * @table: the top-level table structure | ||
1810 | * | ||
1811 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | ||
1812 | * array. A completely 0 filled entry terminates the table. | ||
1813 | * | ||
1814 | * The members of the &struct ctl_table structure are used as follows: | ||
1815 | * | ||
1816 | * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not | ||
1817 | * enter a sysctl file | ||
1818 | * | ||
1819 | * data - a pointer to data for use by proc_handler | ||
1820 | * | ||
1821 | * maxlen - the maximum size in bytes of the data | ||
1822 | * | ||
1823 | * mode - the file permissions for the /proc/sys file, and for sysctl(2) | ||
1824 | * | ||
1825 | * child - a pointer to the child sysctl table if this entry is a directory, or | ||
1826 | * %NULL. | ||
1827 | * | ||
1828 | * proc_handler - the text handler routine (described below) | ||
1829 | * | ||
1830 | * de - for internal use by the sysctl routines | ||
1831 | * | ||
1832 | * extra1, extra2 - extra pointers usable by the proc handler routines | ||
1833 | * | ||
1834 | * Leaf nodes in the sysctl tree will be represented by a single file | ||
1835 | * under /proc; non-leaf nodes will be represented by directories. | ||
1836 | * | ||
1837 | * sysctl(2) can automatically manage read and write requests through | ||
1838 | * the sysctl table. The data and maxlen fields of the ctl_table | ||
1839 | * struct enable minimal validation of the values being written to be | ||
1840 | * performed, and the mode field allows minimal authentication. | ||
1841 | * | ||
1842 | * There must be a proc_handler routine for any terminal nodes | ||
1843 | * mirrored under /proc/sys (non-terminals are handled by a built-in | ||
1844 | * directory handler). Several default handlers are available to | ||
1845 | * cover common cases - | ||
1846 | * | ||
1847 | * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(), | ||
1848 | * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), | ||
1849 | * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax() | ||
1850 | * | ||
1851 | * It is the handler's job to read the input buffer from user memory | ||
1852 | * and process it. The handler should return 0 on success. | ||
1853 | * | ||
1854 | * This routine returns %NULL on a failure to register, and a pointer | ||
1855 | * to the table header on success. | ||
1856 | */ | ||
1857 | struct ctl_table_header *__register_sysctl_paths( | ||
1858 | struct ctl_table_root *root, | ||
1859 | struct nsproxy *namespaces, | ||
1860 | const struct ctl_path *path, struct ctl_table *table) | ||
1861 | { | ||
1862 | struct ctl_table_header *header; | ||
1863 | struct ctl_table *new, **prevp; | ||
1864 | unsigned int n, npath; | ||
1865 | struct ctl_table_set *set; | ||
1866 | |||
1867 | /* Count the path components */ | ||
1868 | for (npath = 0; path[npath].procname; ++npath) | ||
1869 | ; | ||
1870 | |||
1871 | /* | ||
1872 | * For each path component, allocate a 2-element ctl_table array. | ||
1873 | * The first array element will be filled with the sysctl entry | ||
1874 | * for this, the second will be the sentinel (procname == 0). | ||
1875 | * | ||
1876 | * We allocate everything in one go so that we don't have to | ||
1877 | * worry about freeing additional memory in unregister_sysctl_table. | ||
1878 | */ | ||
1879 | header = kzalloc(sizeof(struct ctl_table_header) + | ||
1880 | (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL); | ||
1881 | if (!header) | ||
1882 | return NULL; | ||
1883 | |||
1884 | new = (struct ctl_table *) (header + 1); | ||
1885 | |||
1886 | /* Now connect the dots */ | ||
1887 | prevp = &header->ctl_table; | ||
1888 | for (n = 0; n < npath; ++n, ++path) { | ||
1889 | /* Copy the procname */ | ||
1890 | new->procname = path->procname; | ||
1891 | new->mode = 0555; | ||
1892 | |||
1893 | *prevp = new; | ||
1894 | prevp = &new->child; | ||
1895 | |||
1896 | new += 2; | ||
1897 | } | ||
1898 | *prevp = table; | ||
1899 | header->ctl_table_arg = table; | ||
1900 | |||
1901 | INIT_LIST_HEAD(&header->ctl_entry); | ||
1902 | header->used = 0; | ||
1903 | header->unregistering = NULL; | ||
1904 | header->root = root; | ||
1905 | sysctl_set_parent(NULL, header->ctl_table); | ||
1906 | header->count = 1; | ||
1907 | #ifdef CONFIG_SYSCTL_SYSCALL_CHECK | ||
1908 | if (sysctl_check_table(namespaces, header->ctl_table)) { | ||
1909 | kfree(header); | ||
1910 | return NULL; | ||
1911 | } | ||
1912 | #endif | ||
1913 | spin_lock(&sysctl_lock); | ||
1914 | header->set = lookup_header_set(root, namespaces); | ||
1915 | header->attached_by = header->ctl_table; | ||
1916 | header->attached_to = root_table; | ||
1917 | header->parent = &root_table_header; | ||
1918 | for (set = header->set; set; set = set->parent) { | ||
1919 | struct ctl_table_header *p; | ||
1920 | list_for_each_entry(p, &set->list, ctl_entry) { | ||
1921 | if (p->unregistering) | ||
1922 | continue; | ||
1923 | try_attach(p, header); | ||
1924 | } | ||
1925 | } | ||
1926 | header->parent->count++; | ||
1927 | list_add_tail(&header->ctl_entry, &header->set->list); | ||
1928 | spin_unlock(&sysctl_lock); | ||
1929 | |||
1930 | return header; | ||
1931 | } | ||
1932 | |||
1933 | /** | ||
1934 | * register_sysctl_table_path - register a sysctl table hierarchy | ||
1935 | * @path: The path to the directory the sysctl table is in. | ||
1936 | * @table: the top-level table structure | ||
1937 | * | ||
1938 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | ||
1939 | * array. A completely 0 filled entry terminates the table. | ||
1940 | * | ||
1941 | * See __register_sysctl_paths for more details. | ||
1942 | */ | ||
1943 | struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path, | ||
1944 | struct ctl_table *table) | ||
1945 | { | ||
1946 | return __register_sysctl_paths(&sysctl_table_root, current->nsproxy, | ||
1947 | path, table); | ||
1948 | } | ||
1949 | |||
1950 | /** | ||
1951 | * register_sysctl_table - register a sysctl table hierarchy | ||
1952 | * @table: the top-level table structure | ||
1953 | * | ||
1954 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | ||
1955 | * array. A completely 0 filled entry terminates the table. | ||
1956 | * | ||
1957 | * See register_sysctl_paths for more details. | ||
1958 | */ | ||
1959 | struct ctl_table_header *register_sysctl_table(struct ctl_table *table) | ||
1960 | { | ||
1961 | static const struct ctl_path null_path[] = { {} }; | ||
1962 | |||
1963 | return register_sysctl_paths(null_path, table); | ||
1964 | } | ||
1965 | |||
1966 | /** | ||
1967 | * unregister_sysctl_table - unregister a sysctl table hierarchy | ||
1968 | * @header: the header returned from register_sysctl_table | ||
1969 | * | ||
1970 | * Unregisters the sysctl table and all children. proc entries may not | ||
1971 | * actually be removed until they are no longer used by anyone. | ||
1972 | */ | ||
1973 | void unregister_sysctl_table(struct ctl_table_header * header) | ||
1974 | { | ||
1975 | might_sleep(); | ||
1976 | |||
1977 | if (header == NULL) | ||
1978 | return; | ||
1979 | |||
1980 | spin_lock(&sysctl_lock); | ||
1981 | start_unregistering(header); | ||
1982 | if (!--header->parent->count) { | ||
1983 | WARN_ON(1); | ||
1984 | kfree_rcu(header->parent, rcu); | ||
1985 | } | ||
1986 | if (!--header->count) | ||
1987 | kfree_rcu(header, rcu); | ||
1988 | spin_unlock(&sysctl_lock); | ||
1989 | } | ||
1990 | |||
1991 | int sysctl_is_seen(struct ctl_table_header *p) | ||
1992 | { | ||
1993 | struct ctl_table_set *set = p->set; | ||
1994 | int res; | ||
1995 | spin_lock(&sysctl_lock); | ||
1996 | if (p->unregistering) | ||
1997 | res = 0; | ||
1998 | else if (!set->is_seen) | ||
1999 | res = 1; | ||
2000 | else | ||
2001 | res = set->is_seen(set); | ||
2002 | spin_unlock(&sysctl_lock); | ||
2003 | return res; | ||
2004 | } | ||
2005 | |||
2006 | void setup_sysctl_set(struct ctl_table_set *p, | ||
2007 | struct ctl_table_set *parent, | ||
2008 | int (*is_seen)(struct ctl_table_set *)) | ||
2009 | { | ||
2010 | INIT_LIST_HEAD(&p->list); | ||
2011 | p->parent = parent ? parent : &sysctl_table_root.default_set; | ||
2012 | p->is_seen = is_seen; | ||
2013 | } | ||
2014 | |||
2015 | #endif /* CONFIG_SYSCTL */ | 1554 | #endif /* CONFIG_SYSCTL */ |
2016 | 1555 | ||
2017 | /* | 1556 | /* |
@@ -2977,6 +2516,3 @@ EXPORT_SYMBOL(proc_dointvec_ms_jiffies); | |||
2977 | EXPORT_SYMBOL(proc_dostring); | 2516 | EXPORT_SYMBOL(proc_dostring); |
2978 | EXPORT_SYMBOL(proc_doulongvec_minmax); | 2517 | EXPORT_SYMBOL(proc_doulongvec_minmax); |
2979 | EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax); | 2518 | EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax); |
2980 | EXPORT_SYMBOL(register_sysctl_table); | ||
2981 | EXPORT_SYMBOL(register_sysctl_paths); | ||
2982 | EXPORT_SYMBOL(unregister_sysctl_table); | ||