diff options
Diffstat (limited to 'src')
| -rw-r--r-- | src/basic/c-rbtree.c | 679 | ||||
| -rw-r--r-- | src/basic/c-rbtree.h | 297 | ||||
| -rw-r--r-- | src/libudev/libudev-device-private.c | 8 | ||||
| -rw-r--r-- | src/resolve/resolved-def.h | 6 | ||||
| -rw-r--r-- | src/resolve/resolved-dns-cache.c | 40 | ||||
| -rw-r--r-- | src/resolve/resolved-dns-cache.h | 3 | ||||
| -rw-r--r-- | src/resolve/resolved-dns-packet.c | 55 | ||||
| -rw-r--r-- | src/resolve/resolved-dns-packet.h | 6 | ||||
| -rw-r--r-- | src/resolve/resolved-dns-rr.h | 8 | ||||
| -rw-r--r-- | src/resolve/resolved-dns-scope.c | 56 | ||||
| -rw-r--r-- | src/resolve/resolved-dns-scope.h | 1 | ||||
| -rw-r--r-- | src/resolve/resolved-dns-transaction.c | 337 | ||||
| -rw-r--r-- | src/resolve/resolved-dns-transaction.h | 8 | ||||
| -rw-r--r-- | src/resolve/resolved-link.c | 24 | ||||
| -rw-r--r-- | src/resolve/resolved-link.h | 3 | ||||
| -rw-r--r-- | src/resolve/resolved-manager.c | 23 | ||||
| -rw-r--r-- | src/resolve/resolved-manager.h | 8 | ||||
| -rw-r--r-- | src/resolve/resolved-mdns.c | 288 | ||||
| -rw-r--r-- | src/resolve/resolved-mdns.h | 32 | ||||
| -rw-r--r-- | src/test/test-rbtree.c | 362 | ||||
| -rw-r--r-- | src/test/test-rlimit-util.c | 69 | ||||
| -rw-r--r-- | src/udev/udev-event.c | 8 |
22 files changed, 2208 insertions, 113 deletions
diff --git a/src/basic/c-rbtree.c b/src/basic/c-rbtree.c new file mode 100644 index 0000000000..914d7e5229 --- /dev/null +++ b/src/basic/c-rbtree.c @@ -0,0 +1,679 @@ +/*** + This file is part of systemd. See COPYING for details. + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +/* + * RB-Tree Implementation + * This implements the insertion/removal of elements in RB-Trees. You're highly + * recommended to have an RB-Tree documentation at hand when reading this. Both + * insertion and removal can be split into a handful of situations that can + * occur. Those situations are enumerated as "Case 1" to "Case n" here, and + * follow closely the cases described in most RB-Tree documentations. This file + * does not explain why it is enough to handle just those cases, nor does it + * provide a proof of correctness. Dig out your algorithm 101 handbook if + * you're interested. + * + * This implementation is *not* straightforward. Usually, a handful of + * rotation, reparent, swap and link helpers can be used to implement the + * rebalance operations. However, those often perform unnecessary writes. + * Therefore, this implementation hard-codes all the operations. You're highly + * recommended to look at the two basic helpers before reading the code: + * c_rbtree_swap_child() + * c_rbtree_set_parent_and_color() + * Those are the only helpers used, hence, you should really know what they do + * before digging into the code. + * + * For a highlevel documentation of the API, see the header file and docbook + * comments. + */ + +#include <assert.h> +#include <stddef.h> +#include "c-rbtree.h" + +enum { + C_RBNODE_RED = 0, + C_RBNODE_BLACK = 1, +}; + +static inline unsigned long c_rbnode_color(CRBNode *n) { + return (unsigned long)n->__parent_and_color & 1UL; +} + +static inline _Bool c_rbnode_is_red(CRBNode *n) { + return c_rbnode_color(n) == C_RBNODE_RED; +} + +static inline _Bool c_rbnode_is_black(CRBNode *n) { + return c_rbnode_color(n) == C_RBNODE_BLACK; +} + +/** + * c_rbnode_leftmost() - return leftmost child + * @n: current node, or NULL + * + * This returns the leftmost child of @n. If @n is NULL, this will return NULL. + * In all other cases, this function returns a valid pointer. That is, if @n + * does not have any left children, this returns @n. + * + * Worst case runtime (n: number of elements in tree): O(log(n)) + * + * Return: Pointer to leftmost child, or NULL. + */ +CRBNode *c_rbnode_leftmost(CRBNode *n) { + if (n) + while (n->left) + n = n->left; + return n; +} + +/** + * c_rbnode_rightmost() - return rightmost child + * @n: current node, or NULL + * + * This returns the rightmost child of @n. If @n is NULL, this will return + * NULL. In all other cases, this function returns a valid pointer. That is, if + * @n does not have any right children, this returns @n. + * + * Worst case runtime (n: number of elements in tree): O(log(n)) + * + * Return: Pointer to rightmost child, or NULL. + */ +CRBNode *c_rbnode_rightmost(CRBNode *n) { + if (n) + while (n->right) + n = n->right; + return n; +} + +/** + * c_rbnode_next() - return next node + * @n: current node, or NULL + * + * An RB-Tree always defines a linear order of its elements. This function + * returns the logically next node to @n. If @n is NULL, the last node or + * unlinked, this returns NULL. + * + * Worst case runtime (n: number of elements in tree): O(log(n)) + * + * Return: Pointer to next node, or NULL. + */ +CRBNode *c_rbnode_next(CRBNode *n) { + CRBNode *p; + + if (!c_rbnode_is_linked(n)) + return NULL; + if (n->right) + return c_rbnode_leftmost(n->right); + + while ((p = c_rbnode_parent(n)) && n == p->right) + n = p; + + return p; +} + +/** + * c_rbnode_prev() - return previous node + * @n: current node, or NULL + * + * An RB-Tree always defines a linear order of its elements. This function + * returns the logically previous node to @n. If @n is NULL, the first node or + * unlinked, this returns NULL. + * + * Worst case runtime (n: number of elements in tree): O(log(n)) + * + * Return: Pointer to previous node, or NULL. + */ +CRBNode *c_rbnode_prev(CRBNode *n) { + CRBNode *p; + + if (!c_rbnode_is_linked(n)) + return NULL; + if (n->left) + return c_rbnode_rightmost(n->left); + + while ((p = c_rbnode_parent(n)) && n == p->left) + n = p; + + return p; +} + +/** + * c_rbtree_first() - return first node + * @t: tree to operate on + * + * An RB-Tree always defines a linear order of its elements. This function + * returns the logically first node in @t. If @t is empty, NULL is returned. + * + * Fixed runtime (n: number of elements in tree): O(log(n)) + * + * Return: Pointer to first node, or NULL. + */ +CRBNode *c_rbtree_first(CRBTree *t) { + assert(t); + return c_rbnode_leftmost(t->root); +} + +/** + * c_rbtree_last() - return last node + * @t: tree to operate on + * + * An RB-Tree always defines a linear order of its elements. This function + * returns the logically last node in @t. If @t is empty, NULL is returned. + * + * Fixed runtime (n: number of elements in tree): O(log(n)) + * + * Return: Pointer to last node, or NULL. + */ +CRBNode *c_rbtree_last(CRBTree *t) { + assert(t); + return c_rbnode_rightmost(t->root); +} + +/* + * Set the color and parent of a node. This should be treated as a simple + * assignment of the 'color' and 'parent' fields of the node. No other magic is + * applied. But since both fields share its backing memory, this helper + * function is provided. + */ +static inline void c_rbnode_set_parent_and_color(CRBNode *n, CRBNode *p, unsigned long c) { + assert(!((unsigned long)p & 1)); + assert(c < 2); + n->__parent_and_color = (CRBNode*)((unsigned long)p | c); +} + +/* same as c_rbnode_set_parent_and_color(), but keeps the current parent */ +static inline void c_rbnode_set_color(CRBNode *n, unsigned long c) { + c_rbnode_set_parent_and_color(n, c_rbnode_parent(n), c); +} + +/* same as c_rbnode_set_parent_and_color(), but keeps the current color */ +static inline void c_rbnode_set_parent(CRBNode *n, CRBNode *p) { + c_rbnode_set_parent_and_color(n, p, c_rbnode_color(n)); +} + +/* + * This function partially replaces an existing child pointer to a new one. The + * existing child must be given as @old, the new child as @new. @p must be the + * parent of @old (or NULL if it has no parent). + * This function ensures that the parent of @old now points to @new. However, + * it does *NOT* change the parent pointer of @new. The caller must ensure + * this. + * If @p is NULL, this function ensures that the root-pointer is adjusted + * instead (given as @t). + */ +static inline void c_rbtree_swap_child(CRBTree *t, CRBNode *p, CRBNode *old, CRBNode *new) { + if (p) { + if (p->left == old) + p->left = new; + else + p->right = new; + } else { + t->root = new; + } +} + +static inline CRBNode *c_rbtree_paint_one(CRBTree *t, CRBNode *n) { + CRBNode *p, *g, *gg, *u, *x; + + /* + * Paint a single node according to RB-Tree rules. The node must + * already be linked into the tree and painted red. + * We repaint the node or rotate the tree, if required. In case a + * recursive repaint is required, the next node to be re-painted + * is returned. + * p: parent + * g: grandparent + * gg: grandgrandparent + * u: uncle + * x: temporary + */ + + /* node is red, so we can access the parent directly */ + p = n->__parent_and_color; + + if (!p) { + /* Case 1: + * We reached the root. Mark it black and be done. As all + * leaf-paths share the root, the ratio of black nodes on each + * path stays the same. */ + c_rbnode_set_parent_and_color(n, p, C_RBNODE_BLACK); + n = NULL; + } else if (c_rbnode_is_black(p)) { + /* Case 2: + * The parent is already black. As our node is red, we did not + * change the number of black nodes on any path, nor do we have + * multiple consecutive red nodes. */ + n = NULL; + } else if (p == p->__parent_and_color->left) { /* parent is red, so grandparent exists */ + g = p->__parent_and_color; + gg = c_rbnode_parent(g); + u = g->right; + + if (u && c_rbnode_is_red(u)) { + /* Case 3: + * Parent and uncle are both red. We know the + * grandparent must be black then. Repaint parent and + * uncle black, the grandparent red and recurse into + * the grandparent. */ + c_rbnode_set_parent_and_color(p, g, C_RBNODE_BLACK); + c_rbnode_set_parent_and_color(u, g, C_RBNODE_BLACK); + c_rbnode_set_parent_and_color(g, gg, C_RBNODE_RED); + n = g; + } else { + /* parent is red, uncle is black */ + + if (n == p->right) { + /* Case 4: + * We're the right child. Rotate on parent to + * become left child, so we can handle it the + * same as case 5. */ + x = n->left; + p->right = n->left; + n->left = p; + if (x) + c_rbnode_set_parent_and_color(x, p, C_RBNODE_BLACK); + c_rbnode_set_parent_and_color(p, n, C_RBNODE_RED); + p = n; + } + + /* 'n' is invalid from here on! */ + n = NULL; + + /* Case 5: + * We're the red left child or a red parent, black + * grandparent and uncle. Rotate on grandparent and + * switch color with parent. Number of black nodes on + * each path stays the same, but we got rid of the + * double red path. As the grandparent is still black, + * we're done. */ + x = p->right; + g->left = x; + p->right = g; + if (x) + c_rbnode_set_parent_and_color(x, g, C_RBNODE_BLACK); + c_rbnode_set_parent_and_color(p, gg, C_RBNODE_BLACK); + c_rbnode_set_parent_and_color(g, p, C_RBNODE_RED); + c_rbtree_swap_child(t, gg, g, p); + } + } else /* if (p == p->__parent_and_color->left) */ { /* same as above, but mirrored */ + g = p->__parent_and_color; + gg = c_rbnode_parent(g); + u = g->left; + + if (u && c_rbnode_is_red(u)) { + c_rbnode_set_parent_and_color(p, g, C_RBNODE_BLACK); + c_rbnode_set_parent_and_color(u, g, C_RBNODE_BLACK); + c_rbnode_set_parent_and_color(g, gg, C_RBNODE_RED); + n = g; + } else { + if (n == p->left) { + x = n->right; + p->left = n->right; + n->right = p; + if (x) + c_rbnode_set_parent_and_color(x, p, C_RBNODE_BLACK); + c_rbnode_set_parent_and_color(p, n, C_RBNODE_RED); + p = n; + } + + n = NULL; + + x = p->left; + g->right = x; + p->left = g; + if (x) + c_rbnode_set_parent_and_color(x, g, C_RBNODE_BLACK); + c_rbnode_set_parent_and_color(p, gg, C_RBNODE_BLACK); + c_rbnode_set_parent_and_color(g, p, C_RBNODE_RED); + c_rbtree_swap_child(t, gg, g, p); + } + } + + return n; +} + +static inline void c_rbtree_paint(CRBTree *t, CRBNode *n) { + assert(t); + assert(n); + + while (n) + n = c_rbtree_paint_one(t, n); +} + +/** + * c_rbtree_add() - add node to tree + * @t: tree to operate one + * @p: parent node to link under, or NULL + * @l: left/right slot of @p (or root) to link at + * @n: node to add + * + * This links @n into the tree given as @t. The caller must provide the exact + * spot where to link the node. That is, the caller must traverse the tree + * based on their search order. Once they hit a leaf where to insert the node, + * call this function to link it and rebalance the tree. + * + * A typical insertion would look like this (@t is your tree, @n is your node): + * + * CRBNode **i, *p; + * + * i = &t->root; + * p = NULL; + * while (*i) { + * p = *i; + * if (compare(n, *i) < 0) + * i = &(*i)->left; + * else + * i = &(*i)->right; + * } + * + * c_rbtree_add(t, p, i, n); + * + * Once the node is linked into the tree, a simple lookup on the same tree can + * be coded like this: + * + * CRBNode *i; + * + * i = t->root; + * while (i) { + * int v = compare(n, i); + * if (v < 0) + * i = (*i)->left; + * else if (v > 0) + * i = (*i)->right; + * else + * break; + * } + * + * When you add nodes to a tree, the memory contents of the node do not matter. + * That is, there is no need to initialize the node via c_rbnode_init(). + * However, if you relink nodes multiple times during their lifetime, it is + * usually very convenient to use c_rbnode_init() and c_rbtree_remove_init(). + * In those cases, you should validate that a node is unlinked before you call + * c_rbtree_add(). + */ +void c_rbtree_add(CRBTree *t, CRBNode *p, CRBNode **l, CRBNode *n) { + assert(t); + assert(l); + assert(n); + assert(!p || l == &p->left || l == &p->right); + assert(p || l == &t->root); + + c_rbnode_set_parent_and_color(n, p, C_RBNODE_RED); + n->left = n->right = NULL; + *l = n; + + c_rbtree_paint(t, n); +} + +static inline CRBNode *c_rbtree_rebalance_one(CRBTree *t, CRBNode *p, CRBNode *n) { + CRBNode *s, *x, *y, *g; + + /* + * Rebalance tree after a node was removed. This happens only if you + * remove a black node and one path is now left with an unbalanced + * number or black nodes. + * This function assumes all paths through p and n have one black node + * less than all other paths. If recursive fixup is required, the + * current node is returned. + */ + + if (n == p->left) { + s = p->right; + if (c_rbnode_is_red(s)) { + /* Case 3: + * We have a red node as sibling. Rotate it onto our + * side so we can later on turn it black. This way, we + * gain the additional black node in our path. */ + g = c_rbnode_parent(p); + x = s->left; + p->right = x; + s->left = p; + c_rbnode_set_parent_and_color(x, p, C_RBNODE_BLACK); + c_rbnode_set_parent_and_color(s, g, c_rbnode_color(p)); + c_rbnode_set_parent_and_color(p, s, C_RBNODE_RED); + c_rbtree_swap_child(t, g, p, s); + s = x; + } + + x = s->right; + if (!x || c_rbnode_is_black(x)) { + y = s->left; + if (!y || c_rbnode_is_black(y)) { + /* Case 4: + * Our sibling is black and has only black + * children. Flip it red and turn parent black. + * This way we gained a black node in our path, + * or we fix it recursively one layer up, which + * will rotate the red sibling as parent. */ + c_rbnode_set_parent_and_color(s, p, C_RBNODE_RED); + if (c_rbnode_is_black(p)) + return p; + + c_rbnode_set_parent_and_color(p, c_rbnode_parent(p), C_RBNODE_BLACK); + return NULL; + } + + /* Case 5: + * Left child of our sibling is red, right one is black. + * Rotate on parent so the right child of our sibling is + * now red, and we can fall through to case 6. */ + x = y->right; + s->left = y->right; + y->right = s; + p->right = y; + if (x) + c_rbnode_set_parent_and_color(x, s, C_RBNODE_BLACK); + x = s; + s = y; + } + + /* Case 6: + * The right child of our sibling is red. Rotate left and flip + * colors, which gains us an additional black node in our path, + * that was previously on our sibling. */ + g = c_rbnode_parent(p); + y = s->left; + p->right = y; + s->left = p; + c_rbnode_set_parent_and_color(x, s, C_RBNODE_BLACK); + if (y) + c_rbnode_set_parent_and_color(y, p, c_rbnode_color(y)); + c_rbnode_set_parent_and_color(s, g, c_rbnode_color(p)); + c_rbnode_set_parent_and_color(p, s, C_RBNODE_BLACK); + c_rbtree_swap_child(t, g, p, s); + } else /* if (!n || n == p->right) */ { /* same as above, but mirrored */ + s = p->left; + if (c_rbnode_is_red(s)) { + g = c_rbnode_parent(p); + x = s->right; + p->left = x; + s->right = p; + c_rbnode_set_parent_and_color(x, p, C_RBNODE_BLACK); + c_rbnode_set_parent_and_color(s, g, C_RBNODE_BLACK); + c_rbnode_set_parent_and_color(p, s, C_RBNODE_RED); + c_rbtree_swap_child(t, g, p, s); + s = x; + } + + x = s->left; + if (!x || c_rbnode_is_black(x)) { + y = s->right; + if (!y || c_rbnode_is_black(y)) { + c_rbnode_set_parent_and_color(s, p, C_RBNODE_RED); + if (c_rbnode_is_black(p)) + return p; + + c_rbnode_set_parent_and_color(p, c_rbnode_parent(p), C_RBNODE_BLACK); + return NULL; + } + + x = y->left; + s->right = y->left; + y->left = s; + p->left = y; + if (x) + c_rbnode_set_parent_and_color(x, s, C_RBNODE_BLACK); + x = s; + s = y; + } + + g = c_rbnode_parent(p); + y = s->right; + p->left = y; + s->right = p; + c_rbnode_set_parent_and_color(x, s, C_RBNODE_BLACK); + if (y) + c_rbnode_set_parent_and_color(y, p, c_rbnode_color(y)); + c_rbnode_set_parent_and_color(s, g, c_rbnode_color(p)); + c_rbnode_set_parent_and_color(p, s, C_RBNODE_BLACK); + c_rbtree_swap_child(t, g, p, s); + } + + return NULL; +} + +static inline void c_rbtree_rebalance(CRBTree *t, CRBNode *p) { + CRBNode *n = NULL; + + assert(t); + assert(p); + + do { + n = c_rbtree_rebalance_one(t, p, n); + p = n ? c_rbnode_parent(n) : NULL; + } while (p); +} + +/** + * c_rbtree_remove() - remove node from tree + * @t: tree to operate one + * @n: node to remove + * + * This removes the given node from its tree. Once unlinked, the tree is + * rebalanced. + * The caller *must* ensure that the given tree is actually the tree it is + * linked on. Otherwise, behavior is undefined. + * + * This does *NOT* reset @n to being unlinked (for performance reason, this + * function *never* modifies @n at all). If you need this, use + * c_rbtree_remove_init(). + */ +void c_rbtree_remove(CRBTree *t, CRBNode *n) { + CRBNode *p, *s, *gc, *x, *next = NULL; + unsigned long c; + + assert(t); + assert(n); + assert(c_rbnode_is_linked(n)); + + /* + * There are three distinct cases during node removal of a tree: + * * The node has no children, in which case it can simply be removed. + * * The node has exactly one child, in which case the child displaces + * its parent. + * * The node has two children, in which case there is guaranteed to + * be a successor to the node (successor being the node ordered + * directly after it). This successor cannot have two children by + * itself (two interior nodes can never be successive). Therefore, + * we can simply swap the node with its successor (including color) + * and have reduced this case to either of the first two. + * + * Whenever the node we removed was black, we have to rebalance the + * tree. Note that this affects the actual node we _remove_, not @n (in + * case we swap it). + * + * p: parent + * s: successor + * gc: grand-...-child + * x: temporary + * next: next node to rebalance on + */ + + if (!n->left) { + /* + * Case 1: + * The node has no left child. If it neither has a right child, + * it is a leaf-node and we can simply unlink it. If it also + * was black, we have to rebalance, as always if we remove a + * black node. + * But if the node has a right child, the child *must* be red + * (otherwise, the right path has more black nodes as the + * non-existing left path), and the node to be removed must + * hence be black. We simply replace the node with its child, + * turning the red child black, and thus no rebalancing is + * required. + */ + p = c_rbnode_parent(n); + c = c_rbnode_color(n); + c_rbtree_swap_child(t, p, n, n->right); + if (n->right) + c_rbnode_set_parent_and_color(n->right, p, c); + else + next = (c == C_RBNODE_BLACK) ? p : NULL; + } else if (!n->right) { + /* + * Case 1.1: + * The node has exactly one child, and it is on the left. Treat + * it as mirrored case of Case 1 (i.e., replace the node by its + * child). + */ + p = c_rbnode_parent(n); + c = c_rbnode_color(n); + c_rbtree_swap_child(t, p, n, n->left); + c_rbnode_set_parent_and_color(n->left, p, c); + } else { + /* + * Case 2: + * We are dealing with a full interior node with a child not on + * both sides. Find its successor and swap it. Then remove the + * node similar to Case 1. For performance reasons we don't + * perform the full swap, but skip links that are about to be + * removed, anyway. + */ + s = n->right; + if (!s->left) { + /* right child is next, no need to touch grandchild */ + p = s; + gc = s->right; + } else { + /* find successor and swap partially */ + s = c_rbnode_leftmost(s); + p = c_rbnode_parent(s); + + gc = s->right; + p->left = s->right; + s->right = n->right; + c_rbnode_set_parent(n->right, s); + } + + /* node is partially swapped, now remove as in Case 1 */ + s->left = n->left; + c_rbnode_set_parent(n->left, s); + + x = c_rbnode_parent(n); + c = c_rbnode_color(n); + c_rbtree_swap_child(t, x, n, s); + if (gc) + c_rbnode_set_parent_and_color(gc, p, C_RBNODE_BLACK); + else + next = c_rbnode_is_black(s) ? p : NULL; + c_rbnode_set_parent_and_color(s, x, c); + } + + if (next) + c_rbtree_rebalance(t, next); +} diff --git a/src/basic/c-rbtree.h b/src/basic/c-rbtree.h new file mode 100644 index 0000000000..20c5515ca1 --- /dev/null +++ b/src/basic/c-rbtree.h @@ -0,0 +1,297 @@ +#pragma once + +/*** + This file is part of systemd. See COPYING for details. + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +/* + * Standalone Red-Black-Tree Implementation in Standard ISO-C11 + * + * This header provides an RB-Tree API, that is fully implemented in ISO-C11 + * and has no external dependencies. Furthermore, tree traversal, memory + * allocations, and key comparisons a fully in control of the API user. The + * implementation only provides the RB-Tree specific rebalancing and coloring. + * + * A tree is represented by the "CRBTree" structure. It contains a *singly* + * field, which is a pointer to the root node. If NULL, the tree is empty. If + * non-NULL, there is at least a single element in the tree. + * + * Each node of the tree is represented by the "CRBNode" structure. It has + * three fields. The @left and @right members can be accessed by the API user + * directly to traverse the tree. The third member is an implementation detail + * and encodes the parent pointer and color of the node. + * API users are required to embed the CRBNode object into their own objects + * and then use offsetof() (i.e., container_of() and friends) to turn CRBNode + * pointers into pointers to their own structure. + */ + +#ifdef __cplusplus +extern "C" { +#endif + +typedef struct CRBNode CRBNode; +typedef struct CRBTree CRBTree; + +/** + * struct CRBNode - Node of a Red-Black Tree + * @__parent_and_color: internal state + * @left: left child, or NULL + * @right: right child, or NULL + * + * Each node in an RB-Tree must embed an CRBNode object. This object contains + * pointers to its left and right child, which can be freely accessed by the + * API user at any time. They are NULL, if the node does not have a left/right + * child. + * + * The @__parent_and_color field must never be accessed directly. It encodes + * the pointer to the parent node, and the color of the node. Use the accessor + * functions instead. + * + * There is no reason to initialize a CRBNode object before linking it. + * However, if you need a boolean state that tells you whether the node is + * linked or not, you should initialize the node via c_rbnode_init() or + * C_RBNODE_INIT. + */ +struct CRBNode { + CRBNode *__parent_and_color; + CRBNode *left; + CRBNode *right; +}; + +#define C_RBNODE_INIT(_var) { .__parent_and_color = &(_var) } + +CRBNode *c_rbnode_leftmost(CRBNode *n); +CRBNode *c_rbnode_rightmost(CRBNode *n); +CRBNode *c_rbnode_next(CRBNode *n); +CRBNode *c_rbnode_prev(CRBNode *n); + +/** + * struct CRBTree - Red-Black Tree + * @root: pointer to the root node, or NULL + * + * Each Red-Black Tree is rooted in an CRBTree object. This object contains a + * pointer to the root node of the tree. The API user is free to access the + * @root member at any time, and use it to traverse the tree. + * + * To initialize an RB-Tree, set it to NULL / all zero. + */ +struct CRBTree { + CRBNode *root; +}; + +CRBNode *c_rbtree_first(CRBTree *t); +CRBNode *c_rbtree_last(CRBTree *t); + +void c_rbtree_add(CRBTree *t, CRBNode *p, CRBNode **l, CRBNode *n); +void c_rbtree_remove(CRBTree *t, CRBNode *n); + +/** + * c_rbnode_init() - mark a node as unlinked + * @n: node to operate on + * + * This marks the node @n as unlinked. The node will be set to a valid state + * that can never happen if the node is linked in a tree. Furthermore, this + * state is fully known to the implementation, and as such handled gracefully + * in all cases. + * + * You are *NOT* required to call this on your node. c_rbtree_add() can handle + * uninitialized nodes just fine. However, calling this allows to use + * c_rbnode_is_linked() to check for the state of a node. Furthermore, + * iterators and accessors can be called on initialized (yet unlinked) nodes. + * + * Use the C_RBNODE_INIT macro if you want to initialize static variables. + */ +static inline void c_rbnode_init(CRBNode *n) { + *n = (CRBNode)C_RBNODE_INIT(*n); +} + +/** + * c_rbnode_is_linked() - check whether a node is linked + * @n: node to check, or NULL + * + * This checks whether the passed node is linked. If you pass NULL, or if the + * node is not linked into a tree, this will return false. Otherwise, this + * returns true. + * + * Note that you must have either linked the node or initialized it, before + * calling this function. Never call this function on uninitialized nodes. + * Furthermore, removing a node via c_rbtree_remove() does *NOT* mark the node + * as unlinked. You have to call c_rbnode_init() yourself after removal, or use + * the c_rbtree_remove_init() helper. + * + * Return: true if the node is linked, false if not. + */ +static inline _Bool c_rbnode_is_linked(CRBNode *n) { + return n && n->__parent_and_color != n; +} + +/** + * c_rbnode_parent() - return parent pointer + * @n node to access + * + * This returns a pointer to the parent of the given node @n. If @n does not + * have a parent, NULL is returned. If @n is not linked, @n itself is returned. + * + * You should not call this on unlinked or uninitialized nodes! If you do, you + * better know how its semantics. + * + * Return: Pointer to parent. + */ +static inline CRBNode *c_rbnode_parent(CRBNode *n) { + return (CRBNode*)((unsigned long)n->__parent_and_color & ~1UL); +} + +/** + * c_rbtree_remove_init() - safely remove node from tree and reinitialize it + * @t: tree to operate on + * @n: node to remove, or NULL + * + * This is almost the same as c_rbtree_remove(), but extends it slightly, to be + * more convenient to use in many cases: + * - if @n is unlinked or NULL, this is a no-op + * - @n is reinitialized after being removed + */ +static inline void c_rbtree_remove_init(CRBTree *t, CRBNode *n) { + if (c_rbnode_is_linked(n)) { + c_rbtree_remove(t, n); + c_rbnode_init(n); + } +} + +/** + * CRBCompareFunc - compare a node to a key + * @t: tree where the node is linked to + * @k: key to compare + * @n: node to compare + * + * If you use the tree-traversal helpers (which are optional), you need to + * provide this callback so they can compare nodes in a tree to the key you + * look for. + * + * The tree @t is provided as optional context to this callback. The key you + * look for is provided as @k, the current node that should be compared to is + * provided as @n. This function should work like strcmp(), that is, return -1 + * if @key orders before @n, 0 if both compare equal, and 1 if it orders after + * @n. + */ +typedef int (*CRBCompareFunc) (CRBTree *t, void *k, CRBNode *n); + +/** + * c_rbtree_find_node() - find node + * @t: tree to search through + * @f: comparison function + * @k: key to search for + * + * This searches through @t for a node that compares equal to @k. The function + * @f must be provided by the caller, which is used to compare nodes to @k. See + * the documentation of CRBCompareFunc for details. + * + * If there are multiple entries that compare equal to @k, this will return a + * pseudo-randomly picked node. If you need stable lookup functions for trees + * where duplicate entries are allowed, you better code your own lookup. + * + * Return: Pointer to matching node, or NULL. + */ +static inline CRBNode *c_rbtree_find_node(CRBTree *t, CRBCompareFunc f, const void *k) { + CRBNode *i; + + assert(t); + assert(f); + + i = t->root; + while (i) { + int v = f(t, (void *)k, i); + if (v < 0) + i = i->left; + else if (v > 0) + i = i->right; + else + return i; + } + + return NULL; +} + +/** + * c_rbtree_find_entry() - find entry + * @_t: tree to search through + * @_f: comparison function + * @_k: key to search for + * @_t: type of the structure that embeds the nodes + * @_o: name of the node-member in type @_t + * + * This is very similar to c_rbtree_find_node(), but instead of returning a + * pointer to the CRBNode, it returns a pointer to the surrounding object. This + * object must embed the CRBNode object. The type of the surrounding object + * must be given as @_t, and the name of the embedded CRBNode member as @_o. + * + * See c_rbtree_find_node() for more details. + * + * Return: Pointer to found entry, NULL if not found. + */ +#define c_rbtree_find_entry(_m, _f, _k, _t, _o) \ + ((_t *)(((char *)c_rbtree_find_node((_m), (_f), (_k)) ?: \ + (char *)NULL + offsetof(_t, _o)) - offsetof(_t, _o))) + +/** + * c_rbtree_find_slot() - find slot to insert new node + * @t: tree to search through + * @f: comparison function + * @k: key to search for + * @p: output storage for parent pointer + * + * This searches through @t just like c_rbtree_find_node() does. However, + * instead of returning a pointer to a node that compares equal to @k, this + * searches for a slot to insert a node with key @k. A pointer to the slot is + * returned, and a pointer to the parent of the slot is stored in @p. Both + * can be passed directly to c_rbtree_add(), together with your node to insert. + * + * If there already is a node in the tree, that compares equal to @k, this will + * return NULL and store the conflicting node in @p. In all other cases, + * this will return a pointer (non-NULL) to the empty slot to insert the node + * at. @p will point to the parent node of that slot. + * + * If you want trees that allow duplicate nodes, you better code your own + * insertion function. + * + * Return: Pointer to slot to insert node, or NULL on conflicts. + */ +static inline CRBNode **c_rbtree_find_slot(CRBTree *t, CRBCompareFunc f, const void *k, CRBNode **p) { + CRBNode **i; + + assert(t); + assert(f); + assert(p); + + i = &t->root; + *p = NULL; + while (*i) { + int v = f(t, (void *)k, *i); + *p = *i; + if (v < 0) + i = &(*i)->left; + else if (v > 0) + i = &(*i)->right; + else + return NULL; + } + + return i; +} + +#ifdef __cplusplus +} +#endif diff --git a/src/libudev/libudev-device-private.c b/src/libudev/libudev-device-private.c index 2d3e62410c..2aae0726c1 100644 --- a/src/libudev/libudev-device-private.c +++ b/src/libudev/libudev-device-private.c @@ -137,14 +137,10 @@ gid_t udev_device_get_devnode_gid(struct udev_device *udev_device) { } void udev_device_ensure_usec_initialized(struct udev_device *udev_device, struct udev_device *udev_device_old) { - sd_device *device_old = NULL; - assert(udev_device); - if (udev_device_old) - device_old = udev_device_old->device; - - device_ensure_usec_initialized(udev_device->device, device_old); + device_ensure_usec_initialized(udev_device->device, + udev_device_old ? udev_device_old->device : NULL); } char **udev_device_get_properties_envp(struct udev_device *udev_device) { diff --git a/src/resolve/resolved-def.h b/src/resolve/resolved-def.h index db5ee57b51..6014d345f3 100644 --- a/src/resolve/resolved-def.h +++ b/src/resolve/resolved-def.h @@ -24,6 +24,8 @@ #define SD_RESOLVED_DNS (UINT64_C(1) << 0) #define SD_RESOLVED_LLMNR_IPV4 (UINT64_C(1) << 1) #define SD_RESOLVED_LLMNR_IPV6 (UINT64_C(1) << 2) +#define SD_RESOLVED_MDNS_IPV4 (UINT64_C(1) << 3) +#define SD_RESOLVED_MDNS_IPV6 (UINT64_C(1) << 4) #define SD_RESOLVED_NO_CNAME (UINT64_C(1) << 5) #define SD_RESOLVED_NO_TXT (UINT64_C(1) << 6) #define SD_RESOLVED_NO_ADDRESS (UINT64_C(1) << 7) @@ -31,4 +33,6 @@ #define SD_RESOLVED_AUTHENTICATED (UINT64_C(1) << 9) #define SD_RESOLVED_LLMNR (SD_RESOLVED_LLMNR_IPV4|SD_RESOLVED_LLMNR_IPV6) -#define SD_RESOLVED_PROTOCOLS_ALL (SD_RESOLVED_LLMNR|SD_RESOLVED_DNS) +#define SD_RESOLVED_MDNS (SD_RESOLVED_MDNS_IPV4|SD_RESOLVED_MDNS_IPV6) + +#define SD_RESOLVED_PROTOCOLS_ALL (SD_RESOLVED_MDNS|SD_RESOLVED_LLMNR|SD_RESOLVED_DNS) diff --git a/src/resolve/resolved-dns-cache.c b/src/resolve/resolved-dns-cache.c index bcb9994a8c..6124ff659c 100644 --- a/src/resolve/resolved-dns-cache.c +++ b/src/resolve/resolved-dns-cache.c @@ -459,7 +459,12 @@ int dns_cache_put( /* Second, add in positive entries for all contained RRs */ for (i = 0; i < MIN(max_rrs, answer->n_rrs); i++) { - r = dns_cache_put_positive(c, answer->items[i].rr, authenticated, timestamp, owner_family, owner_address); + DnsResourceRecord *rr = answer->items[i].rr; + + if (rr->key->cache_flush) + dns_cache_remove(c, rr->key); + + r = dns_cache_put_positive(c, rr, authenticated, timestamp, owner_family, owner_address); if (r < 0) goto fail; } @@ -726,6 +731,39 @@ int dns_cache_check_conflicts(DnsCache *cache, DnsResourceRecord *rr, int owner_ return 1; } +int dns_cache_export_shared_to_packet(DnsCache *cache, DnsPacket *p) { + unsigned ancount = 0; + Iterator iterator; + DnsCacheItem *i; + int r; + + assert(cache); + + HASHMAP_FOREACH(i, cache->by_key, iterator) { + DnsCacheItem *j; + + LIST_FOREACH(by_key, j, i) { + _cleanup_free_ char *t = NULL; + + if (!j->rr) + continue; + + if (!dns_key_is_shared(j->rr->key)) + continue; + + r = dns_packet_append_rr(p, j->rr, NULL, NULL); + if (r < 0) + return r; + + ancount ++; + } + } + + DNS_PACKET_HEADER(p)->ancount = htobe16(ancount); + + return 0; +} + void dns_cache_dump(DnsCache *cache, FILE *f) { Iterator iterator; DnsCacheItem *i; diff --git a/src/resolve/resolved-dns-cache.h b/src/resolve/resolved-dns-cache.h index 5f91164785..0f28bbe543 100644 --- a/src/resolve/resolved-dns-cache.h +++ b/src/resolve/resolved-dns-cache.h @@ -32,6 +32,7 @@ typedef struct DnsCache { } DnsCache; #include "resolved-dns-answer.h" +#include "resolved-dns-packet.h" #include "resolved-dns-question.h" #include "resolved-dns-rr.h" @@ -45,3 +46,5 @@ int dns_cache_check_conflicts(DnsCache *cache, DnsResourceRecord *rr, int owner_ void dns_cache_dump(DnsCache *cache, FILE *f); bool dns_cache_is_empty(DnsCache *cache); + +int dns_cache_export_shared_to_packet(DnsCache *cache, DnsPacket *p); diff --git a/src/resolve/resolved-dns-packet.c b/src/resolve/resolved-dns-packet.c index ea776f7916..9bd08eeec2 100644 --- a/src/resolve/resolved-dns-packet.c +++ b/src/resolve/resolved-dns-packet.c @@ -88,6 +88,16 @@ int dns_packet_new_query(DnsPacket **ret, DnsProtocol protocol, size_t mtu, bool 0 /* ad */, 0 /* cd */, 0 /* rcode */)); + else if (protocol == DNS_PROTOCOL_MDNS) + h->flags = htobe16(DNS_PACKET_MAKE_FLAGS(0 /* qr */, + 0 /* opcode */, + 0 /* aa */, + 0 /* tc */, + 0 /* rd (ask for recursion) */, + 0 /* ra */, + 0 /* ad */, + 0 /* cd */, + 0 /* rcode */)); else h->flags = htobe16(DNS_PACKET_MAKE_FLAGS(0 /* qr */, 0 /* opcode */, @@ -182,6 +192,13 @@ int dns_packet_validate_reply(DnsPacket *p) { break; + case DNS_PROTOCOL_MDNS: + /* RFC 6762, Section 18 */ + if (DNS_PACKET_RCODE(p) != 0) + return -EBADMSG; + + break; + default: break; } @@ -223,6 +240,18 @@ int dns_packet_validate_query(DnsPacket *p) { break; + case DNS_PROTOCOL_MDNS: + /* RFC 6762, Section 18 */ + if (DNS_PACKET_AA(p) != 0 || + DNS_PACKET_RD(p) != 0 || + DNS_PACKET_RA(p) != 0 || + DNS_PACKET_AD(p) != 0 || + DNS_PACKET_CD(p) != 0 || + DNS_PACKET_RCODE(p) != 0) + return -EBADMSG; + + break; + default: break; } @@ -1392,6 +1421,7 @@ fail: int dns_packet_read_key(DnsPacket *p, DnsResourceKey **ret, size_t *start) { _cleanup_free_ char *name = NULL; + bool cache_flush = false; uint16_t class, type; DnsResourceKey *key; size_t saved_rindex; @@ -1414,12 +1444,23 @@ int dns_packet_read_key(DnsPacket *p, DnsResourceKey **ret, size_t *start) { if (r < 0) goto fail; + if (p->protocol == DNS_PROTOCOL_MDNS) { + /* See RFC6762, Section 10.2 */ + + if (class & MDNS_RR_CACHE_FLUSH) { + class &= ~MDNS_RR_CACHE_FLUSH; + cache_flush = true; + } + } + key = dns_resource_key_new_consume(class, type, name); if (!key) { r = -ENOMEM; goto fail; } + key->cache_flush = cache_flush; + name = NULL; *ret = key; @@ -1778,8 +1819,16 @@ int dns_packet_read_rr(DnsPacket *p, DnsResourceRecord **ret, size_t *start) { break; - case DNS_TYPE_NSEC: - r = dns_packet_read_name(p, &rr->nsec.next_domain_name, false, NULL); + case DNS_TYPE_NSEC: { + + /* + * RFC6762, section 18.14 explicly states mDNS should use name compression. + * This contradicts RFC3845, section 2.1.1 + */ + + bool allow_compressed = p->protocol == DNS_PROTOCOL_MDNS; + + r = dns_packet_read_name(p, &rr->nsec.next_domain_name, allow_compressed, NULL); if (r < 0) goto fail; @@ -1792,7 +1841,7 @@ int dns_packet_read_rr(DnsPacket *p, DnsResourceRecord **ret, size_t *start) { * without the NSEC bit set. */ break; - + } case DNS_TYPE_NSEC3: { uint8_t size; diff --git a/src/resolve/resolved-dns-packet.h b/src/resolve/resolved-dns-packet.h index aa2823cfb9..48b3572cb4 100644 --- a/src/resolve/resolved-dns-packet.h +++ b/src/resolve/resolved-dns-packet.h @@ -225,6 +225,9 @@ DnsProtocol dns_protocol_from_string(const char *s) _pure_; #define LLMNR_MULTICAST_IPV4_ADDRESS ((struct in_addr) { .s_addr = htobe32(224U << 24 | 252U) }) #define LLMNR_MULTICAST_IPV6_ADDRESS ((struct in6_addr) { .s6_addr = { 0xFF, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x03 } }) +#define MDNS_MULTICAST_IPV4_ADDRESS ((struct in_addr) { .s_addr = htobe32(224U << 24 | 251U) }) +#define MDNS_MULTICAST_IPV6_ADDRESS ((struct in6_addr) { .s6_addr = { 0xFF, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0xfb } }) + static inline uint64_t SD_RESOLVED_FLAGS_MAKE(DnsProtocol protocol, int family, bool authenticated) { uint64_t f; @@ -239,6 +242,9 @@ static inline uint64_t SD_RESOLVED_FLAGS_MAKE(DnsProtocol protocol, int family, case DNS_PROTOCOL_LLMNR: return f|(family == AF_INET6 ? SD_RESOLVED_LLMNR_IPV6 : SD_RESOLVED_LLMNR_IPV4); + case DNS_PROTOCOL_MDNS: + return family == AF_INET6 ? SD_RESOLVED_MDNS_IPV6 : SD_RESOLVED_MDNS_IPV4; + default: break; } diff --git a/src/resolve/resolved-dns-rr.h b/src/resolve/resolved-dns-rr.h index b82fa77562..5c2306ba96 100644 --- a/src/resolve/resolved-dns-rr.h +++ b/src/resolve/resolved-dns-rr.h @@ -45,6 +45,9 @@ enum { #define DNSKEY_FLAG_ZONE_KEY (UINT16_C(1) << 8) #define DNSKEY_FLAG_SEP (UINT16_C(1) << 0) +/* mDNS RR flags */ +#define MDNS_RR_CACHE_FLUSH (UINT16_C(1) << 15) + /* DNSSEC algorithm identifiers, see * http://tools.ietf.org/html/rfc4034#appendix-A.1 and * https://www.iana.org/assignments/dns-sec-alg-numbers/dns-sec-alg-numbers.xhtml */ @@ -76,6 +79,7 @@ struct DnsResourceKey { unsigned n_ref; uint16_t class, type; char *_name; /* don't access directy, use DNS_RESOURCE_KEY_NAME()! */ + bool cache_flush:1; }; /* Creates a temporary resource key. This is only useful to quickly @@ -246,6 +250,10 @@ int dns_resource_key_match_cname(const DnsResourceKey *key, const DnsResourceRec int dns_resource_key_to_string(const DnsResourceKey *key, char **ret); DEFINE_TRIVIAL_CLEANUP_FUNC(DnsResourceKey*, dns_resource_key_unref); +static inline bool dns_key_is_shared(const DnsResourceKey *key) { + return IN_SET(key->type, DNS_TYPE_PTR); +} + DnsResourceRecord* dns_resource_record_new(DnsResourceKey *key); DnsResourceRecord* dns_resource_record_new_full(uint16_t class, uint16_t type, const char *name); DnsResourceRecord* dns_resource_record_ref(DnsResourceRecord *rr); diff --git a/src/resolve/resolved-dns-scope.c b/src/resolve/resolved-dns-scope.c index a90692cdf4..eae903526b 100644 --- a/src/resolve/resolved-dns-scope.c +++ b/src/resolve/resolved-dns-scope.c @@ -30,6 +30,7 @@ #include "random-util.h" #include "resolved-dns-scope.h" #include "resolved-llmnr.h" +#include "resolved-mdns.h" #include "socket-util.h" #include "strv.h" @@ -59,6 +60,7 @@ int dns_scope_new(Manager *m, DnsScope **ret, Link *l, DnsProtocol protocol, int LIST_PREPEND(scopes, m->dns_scopes, s); dns_scope_llmnr_membership(s, true); + dns_scope_mdns_membership(s, true); log_debug("New scope on link %s, protocol %s, family %s", l ? l->name : "*", dns_protocol_to_string(protocol), family == AF_UNSPEC ? "*" : af_to_name(family)); @@ -95,6 +97,7 @@ DnsScope* dns_scope_free(DnsScope *s) { log_debug("Removing scope on link %s, protocol %s, family %s", s->link ? s->link->name : "*", dns_protocol_to_string(s->protocol), s->family == AF_UNSPEC ? "*" : af_to_name(s->family)); dns_scope_llmnr_membership(s, false); + dns_scope_mdns_membership(s, false); dns_scope_abort_transactions(s); while (s->query_candidates) @@ -162,7 +165,6 @@ int dns_scope_emit(DnsScope *s, int fd, DnsServer *server, DnsPacket *p) { union in_addr_union addr; int ifindex = 0, r; int family; - uint16_t port; uint32_t mtu; size_t saved_size = 0; @@ -228,7 +230,6 @@ int dns_scope_emit(DnsScope *s, int fd, DnsServer *server, DnsPacket *p) { return -EBUSY; family = s->family; - port = LLMNR_PORT; if (family == AF_INET) { addr.in = LLMNR_MULTICAST_IPV4_ADDRESS; @@ -241,7 +242,30 @@ int dns_scope_emit(DnsScope *s, int fd, DnsServer *server, DnsPacket *p) { if (fd < 0) return fd; - r = manager_send(s->manager, fd, ifindex, family, &addr, port, p); + r = manager_send(s->manager, fd, ifindex, family, &addr, LLMNR_PORT, p); + if (r < 0) + return r; + + break; + + case DNS_PROTOCOL_MDNS: + if (!ratelimit_test(&s->ratelimit)) + return -EBUSY; + + family = s->family; + + if (family == AF_INET) { + addr.in = MDNS_MULTICAST_IPV4_ADDRESS; + fd = manager_mdns_ipv4_fd(s->manager); + } else if (family == AF_INET6) { + addr.in6 = MDNS_MULTICAST_IPV6_ADDRESS; + fd = manager_mdns_ipv6_fd(s->manager); + } else + return -EAFNOSUPPORT; + if (fd < 0) + return fd; + + r = manager_send(s->manager, fd, ifindex, family, &addr, MDNS_PORT, p); if (r < 0) return r; @@ -477,19 +501,15 @@ int dns_scope_good_key(DnsScope *s, DnsResourceKey *key) { return true; } -int dns_scope_llmnr_membership(DnsScope *s, bool b) { +static int dns_scope_multicast_membership(DnsScope *s, bool b, struct in_addr in, struct in6_addr in6) { int fd; assert(s); - - if (s->protocol != DNS_PROTOCOL_LLMNR) - return 0; - assert(s->link); if (s->family == AF_INET) { struct ip_mreqn mreqn = { - .imr_multiaddr = LLMNR_MULTICAST_IPV4_ADDRESS, + .imr_multiaddr = in, .imr_ifindex = s->link->ifindex, }; @@ -508,7 +528,7 @@ int dns_scope_llmnr_membership(DnsScope *s, bool b) { } else if (s->family == AF_INET6) { struct ipv6_mreq mreq = { - .ipv6mr_multiaddr = LLMNR_MULTICAST_IPV6_ADDRESS, + .ipv6mr_multiaddr = in6, .ipv6mr_interface = s->link->ifindex, }; @@ -527,6 +547,22 @@ int dns_scope_llmnr_membership(DnsScope *s, bool b) { return 0; } +int dns_scope_llmnr_membership(DnsScope *s, bool b) { + + if (s->protocol != DNS_PROTOCOL_LLMNR) + return 0; + + return dns_scope_multicast_membership(s, b, LLMNR_MULTICAST_IPV4_ADDRESS, LLMNR_MULTICAST_IPV6_ADDRESS); +} + +int dns_scope_mdns_membership(DnsScope *s, bool b) { + + if (s->protocol != DNS_PROTOCOL_MDNS) + return 0; + + return dns_scope_multicast_membership(s, b, MDNS_MULTICAST_IPV4_ADDRESS, MDNS_MULTICAST_IPV6_ADDRESS); +} + static int dns_scope_make_reply_packet( DnsScope *s, uint16_t id, diff --git a/src/resolve/resolved-dns-scope.h b/src/resolve/resolved-dns-scope.h index 15d9a1fd6f..2fc2e07deb 100644 --- a/src/resolve/resolved-dns-scope.h +++ b/src/resolve/resolved-dns-scope.h @@ -93,6 +93,7 @@ DnsServer *dns_scope_get_dns_server(DnsScope *s); void dns_scope_next_dns_server(DnsScope *s); int dns_scope_llmnr_membership(DnsScope *s, bool b); +int dns_scope_mdns_membership(DnsScope *s, bool b); void dns_scope_process_query(DnsScope *s, DnsStream *stream, DnsPacket *p); diff --git a/src/resolve/resolved-dns-transaction.c b/src/resolve/resolved-dns-transaction.c index 1103a34c6f..90f07e6c4b 100644 --- a/src/resolve/resolved-dns-transaction.c +++ b/src/resolve/resolved-dns-transaction.c @@ -24,6 +24,7 @@ #include "dns-domain.h" #include "fd-util.h" #include "random-util.h" +#include "resolved-dns-cache.h" #include "resolved-dns-transaction.h" #include "resolved-llmnr.h" #include "string-table.h" @@ -243,7 +244,7 @@ static int on_stream_complete(DnsStream *s, int error) { } if (dns_packet_validate_reply(p) <= 0) { - log_debug("Invalid LLMNR TCP packet."); + log_debug("Invalid TCP reply packet."); dns_transaction_complete(t, DNS_TRANSACTION_INVALID_REPLY); return 0; } @@ -384,6 +385,18 @@ void dns_transaction_process_reply(DnsTransaction *t, DnsPacket *p) { break; + case DNS_PROTOCOL_MDNS: + assert(t->scope->link); + + /* For mDNS we will not accept any packets from other interfaces */ + if (p->ifindex != t->scope->link->ifindex) + return; + + if (p->family != t->scope->family) + return; + + break; + case DNS_PROTOCOL_DNS: break; @@ -446,6 +459,13 @@ void dns_transaction_process_reply(DnsTransaction *t, DnsPacket *p) { } if (DNS_PACKET_TC(p)) { + + /* Truncated packets for mDNS are not allowed. Give up immediately. */ + if (t->scope->protocol == DNS_PROTOCOL_MDNS) { + dns_transaction_complete(t, DNS_TRANSACTION_INVALID_REPLY); + return; + } + /* Response was truncated, let's try again with good old TCP */ r = dns_transaction_open_tcp(t); if (r == -ESRCH) { @@ -454,7 +474,7 @@ void dns_transaction_process_reply(DnsTransaction *t, DnsPacket *p) { return; } if (r < 0) { - /* On LLMNR, if we cannot connect to the host, + /* On LLMNR and mDNS, if we cannot connect to the host, * we immediately give up */ if (t->scope->protocol == DNS_PROTOCOL_LLMNR) { dns_transaction_complete(t, DNS_TRANSACTION_RESOURCES); @@ -481,29 +501,31 @@ void dns_transaction_process_reply(DnsTransaction *t, DnsPacket *p) { return; } - /* Only consider responses with equivalent query section to the request */ - if (p->question->n_keys != 1 || dns_resource_key_equal(p->question->keys[0], t->key) <= 0) { - dns_transaction_complete(t, DNS_TRANSACTION_INVALID_REPLY); - return; - } + if (t->scope->protocol == DNS_PROTOCOL_DNS) { + /* Only consider responses with equivalent query section to the request */ + if (p->question->n_keys != 1 || dns_resource_key_equal(p->question->keys[0], t->key) <= 0) { + dns_transaction_complete(t, DNS_TRANSACTION_INVALID_REPLY); + return; + } - /* Install the answer as answer to the transaction */ - dns_answer_unref(t->answer); - t->answer = dns_answer_ref(p->answer); - t->answer_rcode = DNS_PACKET_RCODE(p); - t->answer_authenticated = t->scope->dnssec_mode == DNSSEC_TRUST && DNS_PACKET_AD(p); - - /* According to RFC 4795, section 2.9. only the RRs from the answer section shall be cached */ - if (DNS_PACKET_SHALL_CACHE(p)) - dns_cache_put(&t->scope->cache, - t->key, - DNS_PACKET_RCODE(p), - p->answer, - DNS_PACKET_ANCOUNT(p), - t->answer_authenticated, - 0, - p->family, - &p->sender); + /* Install the answer as answer to the transaction */ + dns_answer_unref(t->answer); + t->answer = dns_answer_ref(p->answer); + t->answer_rcode = DNS_PACKET_RCODE(p); + t->answer_authenticated = t->scope->dnssec_mode == DNSSEC_TRUST && DNS_PACKET_AD(p); + + /* According to RFC 4795, section 2.9. only the RRs from the answer section shall be cached */ + if (DNS_PACKET_SHALL_CACHE(p)) + dns_cache_put(&t->scope->cache, + t->key, + DNS_PACKET_RCODE(p), + p->answer, + DNS_PACKET_ANCOUNT(p), + t->answer_authenticated, + 0, + p->family, + &p->sender); + } if (DNS_PACKET_RCODE(p) == DNS_RCODE_SUCCESS) dns_transaction_complete(t, DNS_TRANSACTION_SUCCESS); @@ -571,21 +593,26 @@ static int on_transaction_timeout(sd_event_source *s, usec_t usec, void *userdat assert(s); assert(t); - /* Timeout reached? Increase the timeout for the server used */ - switch (t->scope->protocol) { - case DNS_PROTOCOL_DNS: - assert(t->server); + if (!t->initial_jitter_scheduled || t->initial_jitter_elapsed) { + /* Timeout reached? Increase the timeout for the server used */ + switch (t->scope->protocol) { + case DNS_PROTOCOL_DNS: + assert(t->server); - dns_server_packet_lost(t->server, t->current_features, usec - t->start_usec); + dns_server_packet_lost(t->server, t->current_features, usec - t->start_usec); - break; - case DNS_PROTOCOL_LLMNR: - case DNS_PROTOCOL_MDNS: - dns_scope_packet_lost(t->scope, usec - t->start_usec); + break; + case DNS_PROTOCOL_LLMNR: + case DNS_PROTOCOL_MDNS: + dns_scope_packet_lost(t->scope, usec - t->start_usec); - break; - default: - assert_not_reached("Invalid DNS protocol."); + break; + default: + assert_not_reached("Invalid DNS protocol."); + } + + if (t->initial_jitter_scheduled) + t->initial_jitter_elapsed = true; } /* ...and try again with a new server */ @@ -598,38 +625,6 @@ static int on_transaction_timeout(sd_event_source *s, usec_t usec, void *userdat return 0; } -static int dns_transaction_make_packet(DnsTransaction *t) { - _cleanup_(dns_packet_unrefp) DnsPacket *p = NULL; - int r; - - assert(t); - - if (t->sent) - return 0; - - r = dns_packet_new_query(&p, t->scope->protocol, 0, t->scope->dnssec_mode == DNSSEC_YES); - if (r < 0) - return r; - - r = dns_scope_good_key(t->scope, t->key); - if (r < 0) - return r; - if (r == 0) - return -EDOM; - - r = dns_packet_append_key(p, t->key, NULL); - if (r < 0) - return r; - - DNS_PACKET_HEADER(p)->qdcount = htobe16(1); - DNS_PACKET_HEADER(p)->id = t->id; - - t->sent = p; - p = NULL; - - return 0; -} - static usec_t transaction_get_resend_timeout(DnsTransaction *t) { assert(t); assert(t->scope); @@ -639,17 +634,18 @@ static usec_t transaction_get_resend_timeout(DnsTransaction *t) { assert(t->server); return t->server->resend_timeout; - case DNS_PROTOCOL_LLMNR: case DNS_PROTOCOL_MDNS: + assert(t->n_attempts > 0); + return (1 << (t->n_attempts - 1)) * USEC_PER_SEC; + case DNS_PROTOCOL_LLMNR: return t->scope->resend_timeout; default: assert_not_reached("Invalid DNS protocol."); } } -int dns_transaction_go(DnsTransaction *t) { +static int dns_transaction_prepare_next_attempt(DnsTransaction *t, usec_t ts) { bool had_stream; - usec_t ts; int r; assert(t); @@ -658,11 +654,6 @@ int dns_transaction_go(DnsTransaction *t) { dns_transaction_stop(t); - log_debug("Excercising transaction on scope %s on %s/%s", - dns_protocol_to_string(t->scope->protocol), - t->scope->link ? t->scope->link->name : "*", - t->scope->family == AF_UNSPEC ? "*" : af_to_name(t->scope->family)); - if (t->n_attempts >= TRANSACTION_ATTEMPTS_MAX(t->scope->protocol)) { dns_transaction_complete(t, DNS_TRANSACTION_ATTEMPTS_MAX_REACHED); return 0; @@ -675,8 +666,6 @@ int dns_transaction_go(DnsTransaction *t) { return 0; } - assert_se(sd_event_now(t->scope->manager->event, clock_boottime_or_monotonic(), &ts) >= 0); - t->n_attempts++; t->start_usec = ts; t->received = dns_packet_unref(t->received); @@ -739,31 +728,203 @@ int dns_transaction_go(DnsTransaction *t) { } } - if (t->scope->protocol == DNS_PROTOCOL_LLMNR && !t->initial_jitter) { - usec_t jitter; + return 1; +} + +static int dns_transaction_make_packet_mdns(DnsTransaction *t) { + + _cleanup_(dns_packet_unrefp) DnsPacket *p = NULL; + bool add_known_answers = false; + DnsTransaction *other; + unsigned qdcount; + usec_t ts; + int r; + + assert(t); + assert(t->scope->protocol == DNS_PROTOCOL_MDNS); + + /* Discard any previously prepared packet, so we can start over and coaleasce again */ + t->sent = dns_packet_unref(t->sent); + + r = dns_packet_new_query(&p, t->scope->protocol, 0, false); + if (r < 0) + return r; + + r = dns_packet_append_key(p, t->key, NULL); + if (r < 0) + return r; + + qdcount = 1; + + if (dns_key_is_shared(t->key)) + add_known_answers = true; + + /* + * For mDNS, we want to coalesce as many open queries in pending transactions into one single + * query packet on the wire as possible. To achieve that, we iterate through all pending transactions + * in our current scope, and see whether their timing contraints allow them to be sent. + */ + + assert_se(sd_event_now(t->scope->manager->event, clock_boottime_or_monotonic(), &ts) >= 0); + + LIST_FOREACH(transactions_by_scope, other, t->scope->transactions) { + + /* Skip ourselves */ + if (other == t) + continue; + + if (other->state != DNS_TRANSACTION_PENDING) + continue; + + if (other->next_attempt_after > ts) + continue; + + if (qdcount >= UINT16_MAX) + break; + + r = dns_packet_append_key(p, other->key, NULL); + + /* + * If we can't stuff more questions into the packet, just give up. + * One of the 'other' transactions will fire later and take care of the rest. + */ + if (r == -EMSGSIZE) + break; + + if (r < 0) + return r; + + r = dns_transaction_prepare_next_attempt(other, ts); + if (r <= 0) + continue; + + ts += transaction_get_resend_timeout(other); + + r = sd_event_add_time( + other->scope->manager->event, + &other->timeout_event_source, + clock_boottime_or_monotonic(), + ts, 0, + on_transaction_timeout, other); + if (r < 0) + return r; + + other->state = DNS_TRANSACTION_PENDING; + other->next_attempt_after = ts; + + qdcount ++; + + if (dns_key_is_shared(other->key)) + add_known_answers = true; + } + + DNS_PACKET_HEADER(p)->qdcount = htobe16(qdcount); + DNS_PACKET_HEADER(p)->id = t->id; + + /* Append known answer section if we're asking for any shared record */ + if (add_known_answers) { + r = dns_cache_export_shared_to_packet(&t->scope->cache, p); + if (r < 0) + return r; + } + + t->sent = p; + p = NULL; + + return 0; +} + +static int dns_transaction_make_packet(DnsTransaction *t) { + _cleanup_(dns_packet_unrefp) DnsPacket *p = NULL; + int r; + + assert(t); + + if (t->scope->protocol == DNS_PROTOCOL_MDNS) + return dns_transaction_make_packet_mdns(t); + + if (t->sent) + return 0; + + r = dns_packet_new_query(&p, t->scope->protocol, 0, t->scope->dnssec_mode == DNSSEC_YES); + if (r < 0) + return r; + + r = dns_scope_good_key(t->scope, t->key); + if (r < 0) + return r; + if (r == 0) + return -EDOM; + + r = dns_packet_append_key(p, t->key, NULL); + if (r < 0) + return r; + + DNS_PACKET_HEADER(p)->qdcount = htobe16(1); + DNS_PACKET_HEADER(p)->id = t->id; + + t->sent = p; + p = NULL; + + return 0; +} + +int dns_transaction_go(DnsTransaction *t) { + usec_t ts; + int r; + + assert(t); + + assert_se(sd_event_now(t->scope->manager->event, clock_boottime_or_monotonic(), &ts) >= 0); + r = dns_transaction_prepare_next_attempt(t, ts); + if (r <= 0) + return r; + + log_debug("Excercising transaction on scope %s on %s/%s", + dns_protocol_to_string(t->scope->protocol), + t->scope->link ? t->scope->link->name : "*", + t->scope->family == AF_UNSPEC ? "*" : af_to_name(t->scope->family)); + + if (!t->initial_jitter_scheduled && + (t->scope->protocol == DNS_PROTOCOL_LLMNR || + t->scope->protocol == DNS_PROTOCOL_MDNS)) { + usec_t jitter, accuracy; /* RFC 4795 Section 2.7 suggests all queries should be * delayed by a random time from 0 to JITTER_INTERVAL. */ - t->initial_jitter = true; + t->initial_jitter_scheduled = true; random_bytes(&jitter, sizeof(jitter)); - jitter %= LLMNR_JITTER_INTERVAL_USEC; + + switch (t->scope->protocol) { + case DNS_PROTOCOL_LLMNR: + jitter %= LLMNR_JITTER_INTERVAL_USEC; + accuracy = LLMNR_JITTER_INTERVAL_USEC; + break; + case DNS_PROTOCOL_MDNS: + jitter %= MDNS_JITTER_RANGE_USEC; + jitter += MDNS_JITTER_MIN_USEC; + accuracy = MDNS_JITTER_RANGE_USEC; + break; + default: + assert_not_reached("bad protocol"); + } r = sd_event_add_time( t->scope->manager->event, &t->timeout_event_source, clock_boottime_or_monotonic(), - ts + jitter, - LLMNR_JITTER_INTERVAL_USEC, + ts + jitter, accuracy, on_transaction_timeout, t); if (r < 0) return r; t->n_attempts = 0; + t->next_attempt_after = ts; t->state = DNS_TRANSACTION_PENDING; - log_debug("Delaying LLMNR transaction for " USEC_FMT "us.", jitter); + log_debug("Delaying %s transaction for " USEC_FMT "us.", dns_protocol_to_string(t->scope->protocol), jitter); return 0; } @@ -810,16 +971,20 @@ int dns_transaction_go(DnsTransaction *t) { return dns_transaction_go(t); } + ts += transaction_get_resend_timeout(t); + r = sd_event_add_time( t->scope->manager->event, &t->timeout_event_source, clock_boottime_or_monotonic(), - ts + transaction_get_resend_timeout(t), 0, + ts, 0, on_transaction_timeout, t); if (r < 0) return r; t->state = DNS_TRANSACTION_PENDING; + t->next_attempt_after = ts; + return 1; } diff --git a/src/resolve/resolved-dns-transaction.h b/src/resolve/resolved-dns-transaction.h index a3058ce6e8..af08b20e44 100644 --- a/src/resolve/resolved-dns-transaction.h +++ b/src/resolve/resolved-dns-transaction.h @@ -62,7 +62,8 @@ struct DnsTransaction { DnsTransactionState state; uint16_t id; - bool initial_jitter; + bool initial_jitter_scheduled; + bool initial_jitter_elapsed; DnsPacket *sent, *received; @@ -72,6 +73,7 @@ struct DnsTransaction { bool answer_authenticated; usec_t start_usec; + usec_t next_attempt_after; sd_event_source *timeout_event_source; unsigned n_attempts; @@ -119,6 +121,10 @@ DnsTransactionSource dns_transaction_source_from_string(const char *s) _pure_; /* LLMNR Jitter interval, see RFC 4795 Section 7 */ #define LLMNR_JITTER_INTERVAL_USEC (100 * USEC_PER_MSEC) +/* mDNS Jitter interval, see RFC 6762 Section 5.2 */ +#define MDNS_JITTER_MIN_USEC (20 * USEC_PER_MSEC) +#define MDNS_JITTER_RANGE_USEC (100 * USEC_PER_MSEC) + /* Maximum attempts to send DNS requests, across all DNS servers */ #define DNS_TRANSACTION_ATTEMPTS_MAX 16 diff --git a/src/resolve/resolved-link.c b/src/resolve/resolved-link.c index ddd9427dab..84100bd988 100644 --- a/src/resolve/resolved-link.c +++ b/src/resolve/resolved-link.c @@ -77,6 +77,8 @@ Link *link_free(Link *l) { dns_scope_free(l->unicast_scope); dns_scope_free(l->llmnr_ipv4_scope); dns_scope_free(l->llmnr_ipv6_scope); + dns_scope_free(l->mdns_ipv4_scope); + dns_scope_free(l->mdns_ipv6_scope); free(l); return NULL; @@ -118,6 +120,28 @@ static void link_allocate_scopes(Link *l) { } } else l->llmnr_ipv6_scope = dns_scope_free(l->llmnr_ipv6_scope); + + if (link_relevant(l, AF_INET) && + l->mdns_support != SUPPORT_NO && + l->manager->mdns_support != SUPPORT_NO) { + if (!l->mdns_ipv4_scope) { + r = dns_scope_new(l->manager, &l->mdns_ipv4_scope, l, DNS_PROTOCOL_MDNS, AF_INET); + if (r < 0) + log_warning_errno(r, "Failed to allocate mDNS IPv4 scope: %m"); + } + } else + l->mdns_ipv4_scope = dns_scope_free(l->mdns_ipv4_scope); + + if (link_relevant(l, AF_INET6) && + l->mdns_support != SUPPORT_NO && + l->manager->mdns_support != SUPPORT_NO) { + if (!l->mdns_ipv6_scope) { + r = dns_scope_new(l->manager, &l->mdns_ipv6_scope, l, DNS_PROTOCOL_MDNS, AF_INET6); + if (r < 0) + log_warning_errno(r, "Failed to allocate mDNS IPv6 scope: %m"); + } + } else + l->mdns_ipv6_scope = dns_scope_free(l->mdns_ipv6_scope); } void link_add_rrs(Link *l, bool force_remove) { diff --git a/src/resolve/resolved-link.h b/src/resolve/resolved-link.h index eb00015bd5..a3b406bbc2 100644 --- a/src/resolve/resolved-link.h +++ b/src/resolve/resolved-link.h @@ -67,10 +67,13 @@ struct Link { unsigned n_search_domains; Support llmnr_support; + Support mdns_support; DnsScope *unicast_scope; DnsScope *llmnr_ipv4_scope; DnsScope *llmnr_ipv6_scope; + DnsScope *mdns_ipv4_scope; + DnsScope *mdns_ipv6_scope; char name[IF_NAMESIZE]; uint32_t mtu; diff --git a/src/resolve/resolved-manager.c b/src/resolve/resolved-manager.c index 5a3696ccb0..a2677f442a 100644 --- a/src/resolve/resolved-manager.c +++ b/src/resolve/resolved-manager.c @@ -40,6 +40,7 @@ #include "resolved-llmnr.h" #include "resolved-manager.h" #include "resolved-resolv-conf.h" +#include "resolved-mdns.h" #include "socket-util.h" #include "string-table.h" #include "string-util.h" @@ -472,6 +473,7 @@ int manager_new(Manager **ret) { m->llmnr_ipv4_udp_fd = m->llmnr_ipv6_udp_fd = -1; m->llmnr_ipv4_tcp_fd = m->llmnr_ipv6_tcp_fd = -1; + m->mdns_ipv4_fd = m->mdns_ipv6_fd = -1; m->hostname_fd = -1; m->llmnr_support = SUPPORT_YES; @@ -528,6 +530,10 @@ int manager_start(Manager *m) { if (r < 0) return r; + r = manager_mdns_start(m); + if (r < 0) + return r; + return 0; } @@ -559,6 +565,7 @@ Manager *manager_free(Manager *m) { sd_event_source_unref(m->rtnl_event_source); manager_llmnr_stop(m); + manager_mdns_stop(m); sd_bus_slot_unref(m->prepare_for_sleep_slot); sd_event_source_unref(m->bus_retry_event_source); @@ -1024,11 +1031,25 @@ DnsScope* manager_find_scope(Manager *m, DnsPacket *p) { if (!l) return NULL; - if (p->protocol == DNS_PROTOCOL_LLMNR) { + switch (p->protocol) { + case DNS_PROTOCOL_LLMNR: if (p->family == AF_INET) return l->llmnr_ipv4_scope; else if (p->family == AF_INET6) return l->llmnr_ipv6_scope; + + break; + + case DNS_PROTOCOL_MDNS: + if (p->family == AF_INET) + return l->mdns_ipv4_scope; + else if (p->family == AF_INET6) + return l->mdns_ipv6_scope; + + break; + + default: + break; } return NULL; diff --git a/src/resolve/resolved-manager.h b/src/resolve/resolved-manager.h index 1056f23ab7..b52273403a 100644 --- a/src/resolve/resolved-manager.h +++ b/src/resolve/resolved-manager.h @@ -54,6 +54,7 @@ struct Manager { sd_event *event; Support llmnr_support; + Support mdns_support; /* Network */ Hashmap *links; @@ -102,6 +103,13 @@ struct Manager { sd_event_source *llmnr_ipv4_tcp_event_source; sd_event_source *llmnr_ipv6_tcp_event_source; + /* mDNS */ + int mdns_ipv4_fd; + int mdns_ipv6_fd; + + sd_event_source *mdns_ipv4_event_source; + sd_event_source *mdns_ipv6_event_source; + /* dbus */ sd_bus *bus; sd_event_source *bus_retry_event_source; diff --git a/src/resolve/resolved-mdns.c b/src/resolve/resolved-mdns.c new file mode 100644 index 0000000000..096a4b1fe5 --- /dev/null +++ b/src/resolve/resolved-mdns.c @@ -0,0 +1,288 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2015 Daniel Mack + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. + ***/ + +#include <resolv.h> +#include <netinet/in.h> +#include <arpa/inet.h> + +#include "fd-util.h" +#include "resolved-manager.h" +#include "resolved-mdns.h" + +void manager_mdns_stop(Manager *m) { + assert(m); + + m->mdns_ipv4_event_source = sd_event_source_unref(m->mdns_ipv4_event_source); + m->mdns_ipv4_fd = safe_close(m->mdns_ipv4_fd); + + m->mdns_ipv6_event_source = sd_event_source_unref(m->mdns_ipv6_event_source); + m->mdns_ipv6_fd = safe_close(m->mdns_ipv6_fd); +} + +int manager_mdns_start(Manager *m) { + int r; + + assert(m); + + if (m->mdns_support == SUPPORT_NO) + return 0; + + r = manager_mdns_ipv4_fd(m); + if (r == -EADDRINUSE) + goto eaddrinuse; + if (r < 0) + return r; + + if (socket_ipv6_is_supported()) { + r = manager_mdns_ipv6_fd(m); + if (r == -EADDRINUSE) + goto eaddrinuse; + if (r < 0) + return r; + } + + return 0; + +eaddrinuse: + log_warning("There appears to be another mDNS responder running. Turning off mDNS support."); + m->mdns_support = SUPPORT_NO; + manager_mdns_stop(m); + + return 0; +} + +static int on_mdns_packet(sd_event_source *s, int fd, uint32_t revents, void *userdata) { + _cleanup_(dns_packet_unrefp) DnsPacket *p = NULL; + Manager *m = userdata; + DnsScope *scope; + int r; + + r = manager_recv(m, fd, DNS_PROTOCOL_MDNS, &p); + if (r <= 0) + return r; + + scope = manager_find_scope(m, p); + if (!scope) { + log_warning("Got mDNS UDP packet on unknown scope. Ignoring."); + return 0; + } + + if (dns_packet_validate_reply(p) > 0) { + unsigned i; + + log_debug("Got mDNS reply packet"); + + /* + * mDNS is different from regular DNS and LLMNR with regard to handling responses. + * While on other protocols, we can ignore every answer that doesn't match a question + * we broadcast earlier, RFC6762, section 18.1 recommends looking at and caching all + * incoming information, regardless of the DNS packet ID. + * + * Hence, extract the packet here, and try to find a transaction for answer the we got + * and complete it. Also store the new information in scope's cache. + */ + r = dns_packet_extract(p); + if (r < 0) { + log_debug("mDNS packet extraction failed."); + return 0; + } + + dns_scope_check_conflicts(scope, p); + + for (i = 0; i < p->answer->n_rrs; i++) { + DnsResourceRecord *rr; + DnsTransaction *t; + + rr = p->answer->items[i].rr; + + t = dns_scope_find_transaction(scope, rr->key, false); + if (t) + dns_transaction_process_reply(t, p); + } + + dns_cache_put(&scope->cache, NULL, DNS_PACKET_RCODE(p), p->answer, + p->answer->n_rrs, false, 0, p->family, &p->sender); + + } else if (dns_packet_validate_query(p) > 0) { + log_debug("Got mDNS query packet for id %u", DNS_PACKET_ID(p)); + + dns_scope_process_query(scope, NULL, p); + } else + log_debug("Invalid mDNS UDP packet."); + + return 0; +} + +int manager_mdns_ipv4_fd(Manager *m) { + union sockaddr_union sa = { + .in.sin_family = AF_INET, + .in.sin_port = htobe16(MDNS_PORT), + }; + static const int one = 1, pmtu = IP_PMTUDISC_DONT, ttl = 255; + int r; + + assert(m); + + if (m->mdns_ipv4_fd >= 0) + return m->mdns_ipv4_fd; + + m->mdns_ipv4_fd = socket(AF_INET, SOCK_DGRAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0); + if (m->mdns_ipv4_fd < 0) + return -errno; + + r = setsockopt(m->mdns_ipv4_fd, IPPROTO_IP, IP_TTL, &ttl, sizeof(ttl)); + if (r < 0) { + r = -errno; + goto fail; + } + + r = setsockopt(m->mdns_ipv4_fd, IPPROTO_IP, IP_MULTICAST_TTL, &ttl, sizeof(ttl)); + if (r < 0) { + r = -errno; + goto fail; + } + + r = setsockopt(m->mdns_ipv4_fd, IPPROTO_IP, IP_MULTICAST_LOOP, &one, sizeof(one)); + if (r < 0) { + r = -errno; + goto fail; + } + + r = setsockopt(m->mdns_ipv4_fd, SOL_SOCKET, SO_REUSEADDR, &one, sizeof(one)); + if (r < 0) { + r = -errno; + goto fail; + } + + r = setsockopt(m->mdns_ipv4_fd, IPPROTO_IP, IP_PKTINFO, &one, sizeof(one)); + if (r < 0) { + r = -errno; + goto fail; + } + + r = setsockopt(m->mdns_ipv4_fd, IPPROTO_IP, IP_RECVTTL, &one, sizeof(one)); + if (r < 0) { + r = -errno; + goto fail; + } + + /* Disable Don't-Fragment bit in the IP header */ + r = setsockopt(m->mdns_ipv4_fd, IPPROTO_IP, IP_MTU_DISCOVER, &pmtu, sizeof(pmtu)); + if (r < 0) { + r = -errno; + goto fail; + } + + r = bind(m->mdns_ipv4_fd, &sa.sa, sizeof(sa.in)); + if (r < 0) { + r = -errno; + goto fail; + } + + r = sd_event_add_io(m->event, &m->mdns_ipv4_event_source, m->mdns_ipv4_fd, EPOLLIN, on_mdns_packet, m); + if (r < 0) + goto fail; + + return m->mdns_ipv4_fd; + +fail: + m->mdns_ipv4_fd = safe_close(m->mdns_ipv4_fd); + return r; +} + +int manager_mdns_ipv6_fd(Manager *m) { + union sockaddr_union sa = { + .in6.sin6_family = AF_INET6, + .in6.sin6_port = htobe16(MDNS_PORT), + }; + static const int one = 1, ttl = 255; + int r; + + assert(m); + + if (m->mdns_ipv6_fd >= 0) + return m->mdns_ipv6_fd; + + m->mdns_ipv6_fd = socket(AF_INET6, SOCK_DGRAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0); + if (m->mdns_ipv6_fd < 0) + return -errno; + + r = setsockopt(m->mdns_ipv6_fd, IPPROTO_IPV6, IPV6_UNICAST_HOPS, &ttl, sizeof(ttl)); + if (r < 0) { + r = -errno; + goto fail; + } + + /* RFC 4795, section 2.5 recommends setting the TTL of UDP packets to 255. */ + r = setsockopt(m->mdns_ipv6_fd, IPPROTO_IPV6, IPV6_MULTICAST_HOPS, &ttl, sizeof(ttl)); + if (r < 0) { + r = -errno; + goto fail; + } + + r = setsockopt(m->mdns_ipv6_fd, IPPROTO_IPV6, IPV6_MULTICAST_LOOP, &one, sizeof(one)); + if (r < 0) { + r = -errno; + goto fail; + } + + r = setsockopt(m->mdns_ipv6_fd, IPPROTO_IPV6, IPV6_V6ONLY, &one, sizeof(one)); + if (r < 0) { + r = -errno; + goto fail; + } + + r = setsockopt(m->mdns_ipv6_fd, SOL_SOCKET, SO_REUSEADDR, &one, sizeof(one)); + if (r < 0) { + r = -errno; + goto fail; + } + + r = setsockopt(m->mdns_ipv6_fd, IPPROTO_IPV6, IPV6_RECVPKTINFO, &one, sizeof(one)); + if (r < 0) { + r = -errno; + goto fail; + } + + r = setsockopt(m->mdns_ipv6_fd, IPPROTO_IPV6, IPV6_RECVHOPLIMIT, &one, sizeof(one)); + if (r < 0) { + r = -errno; + goto fail; + } + + r = bind(m->mdns_ipv6_fd, &sa.sa, sizeof(sa.in6)); + if (r < 0) { + r = -errno; + goto fail; + } + + r = sd_event_add_io(m->event, &m->mdns_ipv6_event_source, m->mdns_ipv6_fd, EPOLLIN, on_mdns_packet, m); + if (r < 0) { + r = -errno; + goto fail; + } + + return m->mdns_ipv6_fd; + +fail: + m->mdns_ipv6_fd = safe_close(m->mdns_ipv6_fd); + return r; +} diff --git a/src/resolve/resolved-mdns.h b/src/resolve/resolved-mdns.h new file mode 100644 index 0000000000..8a84010615 --- /dev/null +++ b/src/resolve/resolved-mdns.h @@ -0,0 +1,32 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2015 Daniel Mack + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "resolved-manager.h" + +#define MDNS_PORT 5353 + +int manager_mdns_ipv4_fd(Manager *m); +int manager_mdns_ipv6_fd(Manager *m); + +void manager_mdns_stop(Manager *m); +int manager_mdns_start(Manager *m); diff --git a/src/test/test-rbtree.c b/src/test/test-rbtree.c new file mode 100644 index 0000000000..8ae416c557 --- /dev/null +++ b/src/test/test-rbtree.c @@ -0,0 +1,362 @@ +/*** + This file is part of systemd. See COPYING for details. + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +/* + * Tests for RB-Tree + */ + +#undef NDEBUG +#include <assert.h> +#include <stddef.h> +#include <stdlib.h> +#include "c-rbtree.h" + +/* verify that all API calls are exported */ +static void test_api(void) { + CRBTree t = {}; + CRBNode n = C_RBNODE_INIT(n); + + assert(!c_rbnode_is_linked(&n)); + + /* init, is_linked, add, remove, remove_init */ + + c_rbtree_add(&t, NULL, &t.root, &n); + assert(c_rbnode_is_linked(&n)); + + c_rbtree_remove_init(&t, &n); + assert(!c_rbnode_is_linked(&n)); + + c_rbtree_add(&t, NULL, &t.root, &n); + assert(c_rbnode_is_linked(&n)); + + c_rbtree_remove(&t, &n); + assert(c_rbnode_is_linked(&n)); /* @n wasn't touched */ + + c_rbnode_init(&n); + assert(!c_rbnode_is_linked(&n)); + + /* first, last, leftmost, rightmost, next, prev */ + + assert(!c_rbtree_first(&t)); + assert(!c_rbtree_last(&t)); + assert(&n == c_rbnode_leftmost(&n)); + assert(&n == c_rbnode_rightmost(&n)); + assert(!c_rbnode_next(&n)); + assert(!c_rbnode_prev(&n)); +} + +/* copied from c-rbtree.c, relies on internal representation */ +static inline _Bool c_rbnode_is_red(CRBNode *n) { + return !((unsigned long)n->__parent_and_color & 1UL); +} + +/* copied from c-rbtree.c, relies on internal representation */ +static inline _Bool c_rbnode_is_black(CRBNode *n) { + return !!((unsigned long)n->__parent_and_color & 1UL); +} + +static size_t validate(CRBTree *t) { + unsigned int i_black, n_black; + CRBNode *n, *p, *o; + size_t count = 0; + + assert(t); + assert(!t->root || c_rbnode_is_black(t->root)); + + /* traverse to left-most child, count black nodes */ + i_black = 0; + n = t->root; + while (n && n->left) { + if (c_rbnode_is_black(n)) + ++i_black; + n = n->left; + } + n_black = i_black; + + /* + * Traverse tree and verify correctness: + * 1) A node is either red or black + * 2) The root is black + * 3) All leaves are black + * 4) Every red node must have two black child nodes + * 5) Every path to a leaf contains the same number of black nodes + * + * Note that NULL nodes are considered black, which is why we don't + * check for 3). + */ + o = NULL; + while (n) { + ++count; + + /* verify natural order */ + assert(n > o); + o = n; + + /* verify consistency */ + assert(!n->right || c_rbnode_parent(n->right) == n); + assert(!n->left || c_rbnode_parent(n->left) == n); + + /* verify 2) */ + if (!c_rbnode_parent(n)) + assert(c_rbnode_is_black(n)); + + if (c_rbnode_is_red(n)) { + /* verify 4) */ + assert(!n->left || c_rbnode_is_black(n->left)); + assert(!n->right || c_rbnode_is_black(n->right)); + } else { + /* verify 1) */ + assert(c_rbnode_is_black(n)); + } + + /* verify 5) */ + if (!n->left && !n->right) + assert(i_black == n_black); + + /* get next node */ + if (n->right) { + n = n->right; + if (c_rbnode_is_black(n)) + ++i_black; + + while (n->left) { + n = n->left; + if (c_rbnode_is_black(n)) + ++i_black; + } + } else { + while ((p = c_rbnode_parent(n)) && n == p->right) { + n = p; + if (c_rbnode_is_black(p->right)) + --i_black; + } + + n = p; + if (p && c_rbnode_is_black(p->left)) + --i_black; + } + } + + return count; +} + +static void insert(CRBTree *t, CRBNode *n) { + CRBNode **i, *p; + + assert(t); + assert(n); + assert(!c_rbnode_is_linked(n)); + + i = &t->root; + p = NULL; + while (*i) { + p = *i; + if (n < *i) { + i = &(*i)->left; + } else { + assert(n > *i); + i = &(*i)->right; + } + } + + c_rbtree_add(t, p, i, n); +} + +static void shuffle(void **nodes, size_t n_memb) { + unsigned int i, j; + void *t; + + for (i = 0; i < n_memb; ++i) { + j = rand() % n_memb; + t = nodes[j]; + nodes[j] = nodes[i]; + nodes[i] = t; + } +} + +/* run some pseudo-random tests on the tree */ +static void test_shuffle(void) { + CRBNode *nodes[256]; + CRBTree t = {}; + unsigned int i, j; + size_t n; + + /* allocate and initialize all nodes */ + for (i = 0; i < sizeof(nodes) / sizeof(*nodes); ++i) { + nodes[i] = malloc(sizeof(*nodes[i])); + assert(nodes[i]); + c_rbnode_init(nodes[i]); + } + + /* shuffle nodes and validate *empty* tree */ + shuffle((void **)nodes, sizeof(nodes) / sizeof(*nodes)); + n = validate(&t); + assert(n == 0); + + /* add all nodes and validate after each insertion */ + for (i = 0; i < sizeof(nodes) / sizeof(*nodes); ++i) { + insert(&t, nodes[i]); + n = validate(&t); + assert(n == i + 1); + } + + /* shuffle nodes again */ + shuffle((void **)nodes, sizeof(nodes) / sizeof(*nodes)); + + /* remove all nodes (in different order) and validate on each round */ + for (i = 0; i < sizeof(nodes) / sizeof(*nodes); ++i) { + c_rbtree_remove(&t, nodes[i]); + n = validate(&t); + assert(n == sizeof(nodes) / sizeof(*nodes) - i - 1); + c_rbnode_init(nodes[i]); + } + + /* shuffle nodes and validate *empty* tree again */ + shuffle((void **)nodes, sizeof(nodes) / sizeof(*nodes)); + n = validate(&t); + assert(n == 0); + + /* add all nodes again */ + for (i = 0; i < sizeof(nodes) / sizeof(*nodes); ++i) { + insert(&t, nodes[i]); + n = validate(&t); + assert(n == i + 1); + } + + /* 4 times, remove half of the nodes and add them again */ + for (j = 0; j < 4; ++j) { + /* shuffle nodes again */ + shuffle((void **)nodes, sizeof(nodes) / sizeof(*nodes)); + + /* remove half of the nodes */ + for (i = 0; i < sizeof(nodes) / sizeof(*nodes) / 2; ++i) { + c_rbtree_remove(&t, nodes[i]); + n = validate(&t); + assert(n == sizeof(nodes) / sizeof(*nodes) - i - 1); + c_rbnode_init(nodes[i]); + } + + /* shuffle the removed half */ + shuffle((void **)nodes, sizeof(nodes) / sizeof(*nodes) / 2); + + /* add the removed half again */ + for (i = 0; i < sizeof(nodes) / sizeof(*nodes) / 2; ++i) { + insert(&t, nodes[i]); + n = validate(&t); + assert(n == sizeof(nodes) / sizeof(*nodes) / 2 + i + 1); + } + } + + /* shuffle nodes again */ + shuffle((void **)nodes, sizeof(nodes) / sizeof(*nodes)); + + /* remove all */ + for (i = 0; i < sizeof(nodes) / sizeof(*nodes); ++i) { + c_rbtree_remove(&t, nodes[i]); + n = validate(&t); + assert(n == sizeof(nodes) / sizeof(*nodes) - i - 1); + c_rbnode_init(nodes[i]); + } + + /* free nodes again */ + for (i = 0; i < sizeof(nodes) / sizeof(*nodes); ++i) + free(nodes[i]); +} + +typedef struct { + unsigned long key; + CRBNode rb; +} Node; + +#define node_from_rb(_rb) ((Node *)((char *)(_rb) - offsetof(Node, rb))) + +static int compare(CRBTree *t, void *k, CRBNode *n) { + unsigned long key = (unsigned long)k; + Node *node = node_from_rb(n); + + return (key < node->key) ? -1 : (key > node->key) ? 1 : 0; +} + +/* run tests against the c_rbtree_find*() helpers */ +static void test_map(void) { + CRBNode **slot, *p; + CRBTree t = {}; + Node *nodes[2048]; + unsigned long i; + + /* allocate and initialize all nodes */ + for (i = 0; i < sizeof(nodes) / sizeof(*nodes); ++i) { + nodes[i] = malloc(sizeof(*nodes[i])); + assert(nodes[i]); + nodes[i]->key = i; + c_rbnode_init(&nodes[i]->rb); + } + + /* shuffle nodes */ + shuffle((void **)nodes, sizeof(nodes) / sizeof(*nodes)); + + /* add all nodes, and verify that each node is linked */ + for (i = 0; i < sizeof(nodes) / sizeof(*nodes); ++i) { + assert(!c_rbnode_is_linked(&nodes[i]->rb)); + assert(!c_rbtree_find_entry(&t, compare, (void *)nodes[i]->key, Node, rb)); + + slot = c_rbtree_find_slot(&t, compare, (void *)nodes[i]->key, &p); + assert(slot); + c_rbtree_add(&t, p, slot, &nodes[i]->rb); + + assert(c_rbnode_is_linked(&nodes[i]->rb)); + assert(nodes[i] == c_rbtree_find_entry(&t, compare, (void *)nodes[i]->key, Node, rb)); + } + + /* shuffle nodes again */ + shuffle((void **)nodes, sizeof(nodes) / sizeof(*nodes)); + + /* remove all nodes (in different order) */ + for (i = 0; i < sizeof(nodes) / sizeof(*nodes); ++i) { + assert(c_rbnode_is_linked(&nodes[i]->rb)); + assert(nodes[i] == c_rbtree_find_entry(&t, compare, (void *)nodes[i]->key, Node, rb)); + + c_rbtree_remove_init(&t, &nodes[i]->rb); + + assert(!c_rbnode_is_linked(&nodes[i]->rb)); + assert(!c_rbtree_find_entry(&t, compare, (void *)nodes[i]->key, Node, rb)); + } + + /* free nodes again */ + for (i = 0; i < sizeof(nodes) / sizeof(*nodes); ++i) + free(nodes[i]); +} + +int main(int argc, char **argv) { + unsigned int i; + + /* we want stable tests, so use fixed seed */ + srand(0xdeadbeef); + + test_api(); + + /* + * The tests are pseudo random; run them multiple times, each run will + * have different orders and thus different results. + */ + for (i = 0; i < 4; ++i) { + test_shuffle(); + test_map(); + } + + return 0; +} diff --git a/src/test/test-rlimit-util.c b/src/test/test-rlimit-util.c new file mode 100644 index 0000000000..00d3ecc0de --- /dev/null +++ b/src/test/test-rlimit-util.c @@ -0,0 +1,69 @@ +/*** + This file is part of systemd + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <sys/resource.h> + +#include "capability-util.h" +#include "macro.h" +#include "rlimit-util.h" +#include "string-util.h" +#include "util.h" + +int main(int argc, char *argv[]) { + struct rlimit old, new, high; + struct rlimit err = { + .rlim_cur = 10, + .rlim_max = 5, + }; + + log_parse_environment(); + log_open(); + + assert_se(drop_capability(CAP_SYS_RESOURCE) == 0); + + assert_se(getrlimit(RLIMIT_NOFILE, &old) == 0); + new.rlim_cur = MIN(5U, old.rlim_max); + new.rlim_max = MIN(10U, old.rlim_max); + assert_se(setrlimit(RLIMIT_NOFILE, &new) >= 0); + + assert_se(rlimit_from_string("LimitNOFILE") == RLIMIT_NOFILE); + assert_se(rlimit_from_string("DefaultLimitNOFILE") == -1); + + assert_se(streq_ptr(rlimit_to_string(RLIMIT_NOFILE), "LimitNOFILE")); + assert_se(rlimit_to_string(-1) == NULL); + + assert_se(getrlimit(RLIMIT_NOFILE, &old) == 0); + assert_se(setrlimit_closest(RLIMIT_NOFILE, &old) == 0); + assert_se(getrlimit(RLIMIT_NOFILE, &new) == 0); + assert_se(old.rlim_cur == new.rlim_cur); + assert_se(old.rlim_max == new.rlim_max); + + assert_se(getrlimit(RLIMIT_NOFILE, &old) == 0); + high = RLIMIT_MAKE_CONST(old.rlim_max + 1); + assert_se(setrlimit_closest(RLIMIT_NOFILE, &high) == 0); + assert_se(getrlimit(RLIMIT_NOFILE, &new) == 0); + assert_se(new.rlim_max == old.rlim_max); + assert_se(new.rlim_cur == new.rlim_max); + + assert_se(getrlimit(RLIMIT_NOFILE, &old) == 0); + assert_se(setrlimit_closest(RLIMIT_NOFILE, &err) == -EINVAL); + assert_se(getrlimit(RLIMIT_NOFILE, &new) == 0); + assert_se(old.rlim_cur == new.rlim_cur); + assert_se(old.rlim_max == new.rlim_max); + + return 0; +} diff --git a/src/udev/udev-event.c b/src/udev/udev-event.c index 12ba2025f4..c1dcee6c73 100644 --- a/src/udev/udev-event.c +++ b/src/udev/udev-event.c @@ -850,11 +850,11 @@ void udev_event_execute_rules(struct udev_event *event, /* disable watch during event processing */ if (major(udev_device_get_devnum(dev)) != 0) udev_watch_end(event->udev, event->dev_db); - } - if (major(udev_device_get_devnum(dev)) == 0 && - streq(udev_device_get_action(dev), "move")) - udev_device_copy_properties(dev, event->dev_db); + if (major(udev_device_get_devnum(dev)) == 0 && + streq(udev_device_get_action(dev), "move")) + udev_device_copy_properties(dev, event->dev_db); + } udev_rules_apply_to_event(rules, event, timeout_usec, timeout_warn_usec, |