* [PATCH RFC 0/3] nf_tables string match support
@ 2022-07-29 9:31 Pablo Neira Ayuso
2022-07-29 9:31 ` [PATCH RFC 1/3] netfilter: add Aho-Corasick string match implementation Pablo Neira Ayuso
` (2 more replies)
0 siblings, 3 replies; 4+ messages in thread
From: Pablo Neira Ayuso @ 2022-07-29 9:31 UTC (permalink / raw)
To: netfilter-devel
Hi,
The following patchset contains nf_tables string match support. This new
infrastructure is based on the Aho-Corasick pattern match algorithm,
which allows for linear search of a dictionary (a "string set" composed
of patterns). The implementation is lockless by performing updates on a
cloned copy of the Aho-Corasick "tree" use the existing 2-phase commit
protocol to atomically expose the new tree to the packet patch.
I decided to add new netlink API for "string set" rather than reusing the
existing set API for simplicity. There is a new Kconfig knob CONFIG_NFT_STRING
to enable built-in support into nf_tables to avoid an indirection between
nf_tables_api and nft_string given that the Aho-Corasick API (see ac_*()
functions) are invoked from the nf_tables netlink frontend.
The implementation of Aho-Corasick comes as a separated file, it is
relatively small (~600 LoC), and a dictionary of 370105 English words
consumes ~150 Mbytes. Maximum string size at this stage is 128 bytes.
The implementation has been validated from userspace via ASAN and
valgrind with testsuites consisting simple tests combined with random
feeding the dictionary with words and autogenerated text patched with a
matching at a random offset to validate correct matching. The userspace
implementation (rather similarly to the one coming in this batch) and
the testsuite is not posted in this batch.
This algorithm is described in "Efficient string matching: An aid to
bibliographic search" by Alfred V. Aho and Margaret J. Corasick (published in
June 1975) at Communications of the ACM 18 (6): 333–340.
There is a few aspect I would like to revisit after this RFC, eg. netlink
notifications are not yet supported.
Please, see specific patch descriptions for implementation details.
Comments welcome.
P.S: Patch 2 reports 200 deletions on nf_tables_api.c. For some reason
diff is removing 200 LoC and adding them again after the new netlink
string API, there are not real line removals, it is just noise.
Pablo Neira Ayuso (3):
netfilter: add Aho-Corasick string match implementation
netfilter: nf_tables: add string set API
netfilter: nf_tables: add string expression
include/net/netfilter/ahocorasick.h | 27 +
include/net/netfilter/nf_tables.h | 37 +
include/net/netfilter/nf_tables_core.h | 1 +
include/uapi/linux/netfilter/nf_tables.h | 65 ++
net/netfilter/Kconfig | 7 +
net/netfilter/Makefile | 3 +
net/netfilter/ahocorasick.c | 677 ++++++++++++
net/netfilter/nf_tables_api.c | 1287 ++++++++++++++++++----
net/netfilter/nf_tables_core.c | 1 +
net/netfilter/nft_string.c | 254 +++++
10 files changed, 2158 insertions(+), 201 deletions(-)
create mode 100644 include/net/netfilter/ahocorasick.h
create mode 100644 net/netfilter/ahocorasick.c
create mode 100644 net/netfilter/nft_string.c
--
2.30.2
^ permalink raw reply [flat|nested] 4+ messages in thread
* [PATCH RFC 1/3] netfilter: add Aho-Corasick string match implementation
2022-07-29 9:31 [PATCH RFC 0/3] nf_tables string match support Pablo Neira Ayuso
@ 2022-07-29 9:31 ` Pablo Neira Ayuso
2022-07-29 9:31 ` [PATCH RFC 2/3] netfilter: nf_tables: add string set API Pablo Neira Ayuso
2022-07-29 9:31 ` [PATCH RFC 3/3] netfilter: nf_tables: add string expression Pablo Neira Ayuso
2 siblings, 0 replies; 4+ messages in thread
From: Pablo Neira Ayuso @ 2022-07-29 9:31 UTC (permalink / raw)
To: netfilter-devel
This implementation uses a tree of arrays for constant access time at
pattern search time. The tree is composed of two type of state nodes:
- Normal state, that has an array of edge pointers to the next state.
The current matching character is used as the array index to access
the next state.
- Final state has no edge pointers, it represents the last state for a
matching pattern. Final state provides an output field which specifies
the number of bytes that have been matched.
Additionally, a normal state might also have a output field. This
implementation allows to store two words with the same prefix pattern in
the tree, the search algorithm stops at the shortest length pattern
match.
Each state also has a fail pointer to backtrack in the tree states in
case of mismatching, so this tree is actually a graph that represents
a state-machine.
The implementation provides no specific locking mechanism, it is up to
the user to protect this data structure from concurrent access.
Nftables follows a lockless approach by allowing updates on a clone,
then use RCU to replace the stale tree by the updated clone from the
commit phase.
An internal stack is used to iterate over the tree nodes to avoid
recursive calls, to delete patterns and to walk the tree.
Maximum string size is 128 bytes.
A dictionary of 370105 English words consumes ~150 Mbytes.
This algorithm is described in "Efficient string matching: An aid to
bibliographic search" by Alfred V. Aho and Margaret J. Corasick (published in
June 1975) at Communications of the ACM 18 (6): 333–340.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
include/net/netfilter/ahocorasick.h | 27 ++
net/netfilter/Kconfig | 7 +
net/netfilter/Makefile | 2 +
net/netfilter/ahocorasick.c | 677 ++++++++++++++++++++++++++++
4 files changed, 713 insertions(+)
create mode 100644 include/net/netfilter/ahocorasick.h
create mode 100644 net/netfilter/ahocorasick.c
diff --git a/include/net/netfilter/ahocorasick.h b/include/net/netfilter/ahocorasick.h
new file mode 100644
index 000000000000..2e7795912558
--- /dev/null
+++ b/include/net/netfilter/ahocorasick.h
@@ -0,0 +1,27 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+#ifndef _NET_NF_AHOCORASICK_H
+#define _NET_NF_AHOCORASICK_H
+
+#define AC_MAXLEN 128
+
+struct ac_tree;
+
+struct ac_tree *ac_create(void);
+void ac_destroy(struct ac_tree *tree);
+
+int ac_add(struct ac_tree *tree, const char *word, uint32_t len, bool create);
+void ac_del(struct ac_tree *tree, const char *word, uint32_t len);
+int ac_resolve_fail(struct ac_tree *tree);
+
+int ac_find(struct ac_tree *tree, const char *text, size_t text_len);
+bool ac_find_word(struct ac_tree *tree, const char *word, uint32_t len);
+
+struct ac_iter;
+
+struct ac_iter *ac_iter_create(struct ac_tree *tree);
+int ac_iterate(struct ac_iter *iter, int (*cb)(const char *word, void *data), void *data);
+void ac_iter_destroy(struct ac_iter *ctx);
+
+struct ac_tree *ac_clone(struct ac_tree *tree);
+
+#endif
diff --git a/net/netfilter/Kconfig b/net/netfilter/Kconfig
index ddc54b6d18ee..383af17f0399 100644
--- a/net/netfilter/Kconfig
+++ b/net/netfilter/Kconfig
@@ -662,6 +662,13 @@ config NFT_TPROXY
help
This makes transparent proxy support available in nftables.
+config NFT_STRING
+ bool "Netfilter nf_tables string expression support"
+ help
+ The string expression allows you to look for pattern matchings in
+ packets using an implementation of Aho-Corasick string matching
+ algorithm.
+
config NFT_SYNPROXY
tristate "Netfilter nf_tables SYNPROXY expression support"
depends on NF_CONNTRACK && NETFILTER_ADVANCED
diff --git a/net/netfilter/Makefile b/net/netfilter/Makefile
index 238b6a620e88..18127c83b88c 100644
--- a/net/netfilter/Makefile
+++ b/net/netfilter/Makefile
@@ -90,6 +90,8 @@ nf_tables-objs += nft_set_pipapo_avx2.o
endif
endif
+nf_tables-$(CONFIG_NFT_STRING) += ahocorasick.o
+
obj-$(CONFIG_NF_TABLES) += nf_tables.o
obj-$(CONFIG_NFT_COMPAT) += nft_compat.o
obj-$(CONFIG_NFT_CONNLIMIT) += nft_connlimit.o
diff --git a/net/netfilter/ahocorasick.c b/net/netfilter/ahocorasick.c
new file mode 100644
index 000000000000..67bdd1b85110
--- /dev/null
+++ b/net/netfilter/ahocorasick.c
@@ -0,0 +1,677 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * (C) 2022 Pablo Neira Ayuso <pablo@netfilter.org>
+ */
+
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/netlink.h>
+#include <net/netfilter/ahocorasick.h>
+
+enum {
+ AC_STATE_NORMAL = 0,
+ AC_STATE_FINAL,
+};
+
+struct ac_state {
+ uint8_t type;
+ uint16_t use;
+ uint16_t output;
+ void *fail;
+ /* same layout as ac_state_final */
+ uint8_t from;
+ uint8_t len;
+ void **edge;
+};
+
+/* this is a subset of struct ac_state, do not change the layout. */
+struct ac_state_final {
+ uint8_t type;
+ uint16_t use;
+ uint16_t output;
+ void *fail;
+};
+
+struct ac_stack {
+ struct ac_state *state;
+ struct ac_state *next;
+ uint8_t pos;
+};
+
+struct ac_tree {
+ struct ac_state root;
+ struct ac_stack stack[AC_MAXLEN];
+ int stack_num;
+};
+
+#define ac_state_fail(__s) ((struct ac_state *)__s)->fail
+#define ac_state_output(__s) ((struct ac_state *)__s)->output
+#define ac_state_use(__s) ((struct ac_state *)__s)->use
+#define ac_state_type(__s) ((struct ac_state *)__s)->type
+
+static int ac_state_pos(const struct ac_state *state, int pos)
+{
+ return pos - state->from;
+}
+
+static int ac_state_edge_len(const struct ac_state *state)
+{
+ return state->from + state->len;
+}
+
+static void *ac_state_edge(const struct ac_state *state, int pos)
+{
+ if (pos < state->from)
+ return NULL;
+ else if (pos >= ac_state_edge_len(state))
+ return NULL;
+
+ return state->edge[ac_state_pos(state, pos)];
+}
+
+static void ac_state_set_edge(struct ac_state *state, int pos, void *edge)
+{
+ state->edge[ac_state_pos(state, pos)] = edge;
+}
+
+static inline uint8_t ac_text_get_char(const char *word, int pos)
+{
+ return (uint8_t)word[pos];
+}
+
+static void ac_state_free(const struct ac_state *state)
+{
+ if (ac_state_type(state) == AC_STATE_NORMAL)
+ kfree((void *)state->edge);
+
+ kfree((void *)state);
+}
+
+void ac_del(struct ac_tree *tree, const char *word, uint32_t len)
+{
+ struct ac_state *cur = &tree->root, *this;
+ struct ac_stack *ss;
+ int i, pos;
+
+ for (i = 0; i < len; i++) {
+ pos = ac_text_get_char(word, i);
+
+ ss = &tree->stack[tree->stack_num++];
+ ss->state = cur;
+ ss->pos = pos;
+
+ cur = ac_state_edge(cur, pos);
+ if (WARN_ON_ONCE(cur == NULL))
+ return;
+ }
+
+ i = 0;
+ while (tree->stack_num > 0) {
+ ss = &tree->stack[--tree->stack_num];
+
+ this = ac_state_edge(ss->state, ss->pos);
+
+ if (i == 0 && ac_state_type(this) == AC_STATE_NORMAL && ac_state_output(this))
+ ac_state_output(this) = 0;
+
+ ac_state_use(this)--;
+ if (ac_state_use(this) == 0) {
+ ac_state_free(this);
+ ac_state_set_edge(ss->state, ss->pos, NULL);
+ }
+ i++;
+ }
+}
+
+bool ac_find_word(struct ac_tree *tree, const char *word, uint32_t len)
+{
+ struct ac_state *cur = &tree->root, *next;
+ int i = 0;
+
+ while (i < len) {
+ if (ac_state_type(cur) == AC_STATE_FINAL)
+ break;
+
+ next = ac_state_edge(cur, ac_text_get_char(word, i));
+ if (next != NULL) {
+ cur = next;
+ i++;
+ continue;
+ }
+
+ return false;
+ }
+
+ if (!cur || !ac_state_output(cur))
+ return false;
+
+ return i == len ? true : false;
+}
+
+static int ac_state_resize(struct ac_state *cur, int pos)
+{
+ int old_from;
+ uint8_t len;
+ void **edge;
+
+ if (cur == NULL || ac_state_type(cur) == AC_STATE_FINAL)
+ return 0;
+
+ if (pos < cur->from) {
+ len = cur->len + (cur->from - pos);
+ edge = kcalloc(len, sizeof(void *), GFP_KERNEL_ACCOUNT);
+ if (!edge)
+ return -ENOMEM;
+
+ old_from = cur->from - pos;
+ memcpy(&edge[old_from], &cur->edge[0], cur->len * sizeof(void *));
+ kfree(cur->edge);
+ cur->edge = edge;
+ cur->len = len;
+ cur->from = pos;
+
+ } else if (pos >= ac_state_edge_len(cur)) {
+ len = pos - cur->from + 1;
+
+ edge = krealloc(cur->edge, len * sizeof(void *), GFP_KERNEL_ACCOUNT);
+ if (!edge)
+ return -ENOMEM;
+
+ memset(&edge[cur->len], 0, (len - cur->len) * sizeof(void *));
+
+ cur->edge = edge;
+ cur->len = len;
+ }
+
+ return 0;
+}
+
+static int ac_add_state(struct ac_state *cur, uint8_t pos, uint8_t next_pos)
+{
+ struct ac_state *tmp = ac_state_edge(cur, pos);
+
+ if (ac_state_resize(cur, pos) < 0)
+ return -ENOMEM;
+
+ if (tmp == NULL) {
+ tmp = kcalloc(1, sizeof(struct ac_state), GFP_KERNEL_ACCOUNT);
+ if (!tmp)
+ return -ENOMEM;
+
+ tmp->from = next_pos;
+ tmp->len = 1;
+ tmp->edge = kcalloc(1, sizeof(void *), GFP_KERNEL_ACCOUNT);
+ if (!tmp->edge) {
+ kfree(tmp);
+ return -ENOMEM;
+ }
+
+ tmp->use = 1;
+ ac_state_set_edge(cur, pos, tmp);
+
+ } else if (tmp->output != 0 && tmp->type == AC_STATE_FINAL) {
+ struct ac_state *this;
+
+ /* There are two words with the same prefix, replace the final
+ * state by a normal state. By now, we only match the word with
+ * the shorter prefix, to match the longest word, we have to
+ * extend the find function.
+ */
+ tmp = kcalloc(1, sizeof(struct ac_state), GFP_KERNEL_ACCOUNT);
+ if (!tmp)
+ return -ENOMEM;
+
+ tmp->from = next_pos;
+ tmp->len = 1;
+ tmp->edge = kcalloc(1, sizeof(void *), GFP_KERNEL_ACCOUNT);
+ if (!tmp->edge) {
+ kfree(tmp);
+ return -ENOMEM;
+ }
+
+ this = ac_state_edge(cur, pos);
+ if (WARN_ON_ONCE(this == NULL))
+ return -EINVAL;
+
+ if (this->use + 1 == 0) {
+ kfree(tmp);
+ return -EOVERFLOW;
+ }
+
+ memcpy(tmp, this, sizeof(struct ac_state_final));
+ tmp->use++;
+ tmp->type = AC_STATE_NORMAL;
+ ac_state_free(this);
+ ac_state_set_edge(cur, pos, tmp);
+ } else {
+ if (tmp->use + 1 == 0)
+ return -EOVERFLOW;
+
+ tmp->use++;
+ }
+ return 0;
+}
+
+static int ac_add_state_final(struct ac_state *cur, uint8_t pos, const char *word, uint32_t len)
+{
+ struct ac_state_final *tmp = ac_state_edge(cur, pos);
+
+ if (ac_state_resize(cur, pos) < 0)
+ return -ENOMEM;
+
+ if (!tmp) {
+ tmp = kcalloc(1, sizeof(struct ac_state_final), GFP_KERNEL_ACCOUNT);
+ if (!tmp)
+ return -ENOMEM;
+
+ tmp->use = 1;
+ ac_state_set_edge(cur, pos, tmp);
+ ac_state_output(tmp) = len;
+ ac_state_type(tmp) = AC_STATE_FINAL;
+ } else {
+ ac_state_output(tmp) = len;
+ tmp->use++;
+ }
+
+ return 0;
+}
+
+static int __ac_add(struct ac_tree *tree, const char *word, uint32_t len)
+{
+ struct ac_state *cur = &tree->root;
+ uint8_t pos, next_pos;
+ int i, err;
+
+ for (i = 0; i < len - 1; i++) {
+ pos = ac_text_get_char(word, i);
+ next_pos = ac_text_get_char(word, i + 1);
+ err = ac_add_state(cur, pos, next_pos);
+ if (err < 0)
+ goto err_add;
+
+ cur = ac_state_edge(cur, pos);
+ if (WARN_ON_ONCE(cur == NULL)) {
+ err = -EINVAL;
+ goto err_add;
+ }
+ }
+ pos = ac_text_get_char(word, i);
+ err = ac_add_state_final(cur, pos, word, len);
+ if (err < 0)
+ goto err_add;
+
+ return 0;
+err_add:
+ if (i > 0)
+ ac_del(tree, word, i);
+
+ return err;
+}
+
+int ac_add(struct ac_tree *tree, const char *word, uint32_t len, bool create)
+{
+ if (len == 0)
+ return -EINVAL;
+ else if (len > AC_MAXLEN)
+ return -E2BIG;
+
+ if (ac_find_word(tree, word, len)) {
+ if (create)
+ return -EEXIST;
+
+ return 0;
+ }
+
+ return __ac_add(tree, word, len);
+}
+
+struct ac_queue_state {
+ struct list_head head;
+ struct ac_state *state;
+};
+
+int ac_resolve_fail(struct ac_tree *tree)
+{
+ struct ac_state *cur = &tree->root, *s, *fail;
+ struct ac_queue_state *q;
+ LIST_HEAD(queue);
+ int i;
+
+ /* first level of the tree. */
+ for (i = cur->from; i < ac_state_edge_len(cur); i++) {
+ if (ac_state_edge(cur, i) != NULL) {
+ struct ac_queue_state *new;
+
+ ac_state_fail(ac_state_edge(cur, i)) = tree;
+
+ new = kmalloc(sizeof(struct ac_queue_state), GFP_KERNEL_ACCOUNT);
+ if (new == NULL)
+ return -ENOMEM;
+
+ new->state = ac_state_edge(cur, i);
+ list_add_tail(&new->head, &queue);
+ }
+ }
+ /* second level of tree and further. */
+ while (!list_empty(&queue)) {
+ q = (struct ac_queue_state *) queue.next;
+
+ /* we reached final state, next in queue. */
+ if (ac_state_output(q->state) != 0 &&
+ ac_state_type(q->state) == AC_STATE_FINAL) {
+ list_del(&q->head);
+ kfree(q);
+ continue;
+ }
+ /* this is a normal state, add edges to queue. */
+ for (i = q->state->from; i < ac_state_edge_len(q->state); i++) {
+ struct ac_queue_state *new;
+
+ s = ac_state_edge(q->state, i);
+ if (s == NULL)
+ continue;
+
+ fail = ac_state_fail(q->state);
+
+ /* our zerostate has fail set to NULL. */
+ while (fail && (ac_state_type(fail) == AC_STATE_FINAL ||
+ ac_state_edge(fail, i) == NULL)) {
+ fail = ac_state_fail(fail);
+ }
+ if (fail)
+ ac_state_fail(s) = ac_state_edge(fail, i);
+ else
+ ac_state_fail(s) = cur;
+
+ new = kmalloc(sizeof(struct ac_queue_state), GFP_KERNEL_ACCOUNT);
+ if (new == NULL)
+ return -ENOMEM;
+
+ new->state = s;
+ list_add_tail(&new->head, &queue);
+ }
+ list_del(&q->head);
+ kfree(q);
+ }
+ return 0;
+}
+
+static void
+ac_add_edges_to_stack(struct ac_tree *tree, struct ac_state *cur, int i)
+{
+ struct ac_stack *ss;
+
+ while (1) {
+ if (ac_state_output(cur) != 0 &&
+ ac_state_type(cur) == AC_STATE_FINAL)
+ return;
+
+ if (i >= cur->from + cur->len)
+ break;
+
+ if (ac_state_edge(cur, i) == NULL) {
+ i++;
+ continue;
+ }
+
+ ss = &tree->stack[tree->stack_num++];
+ if (tree->stack_num >= AC_MAXLEN)
+ return;
+
+ ss->pos = i;
+ ss->state = cur;
+ ss->next = ac_state_edge(cur, i);
+ cur = ac_state_edge(cur, i);
+ i = 0;
+ }
+}
+
+static void ac_free_edges(struct ac_tree *tree)
+{
+ struct ac_state *cur = &tree->root;
+ int i = cur->from;
+
+ while (1) {
+ struct ac_stack *ss;
+
+ ac_add_edges_to_stack(tree, cur, i);
+
+ if (tree->stack_num == 0)
+ break;
+
+ ss = &tree->stack[--tree->stack_num];
+ i = ss->pos + 1;
+ cur = ss->state;
+ ac_state_free(ss->next);
+ }
+}
+
+struct ac_tree *ac_create(void)
+{
+ struct ac_state *root;
+ struct ac_tree *tree;
+
+ tree = kcalloc(1, sizeof(struct ac_tree), GFP_KERNEL_ACCOUNT);
+ if (!tree)
+ return NULL;
+
+ root = &tree->root;
+ root->from = 'a';
+ root->len = 64;
+ root->edge = kcalloc(root->len, sizeof(void *), GFP_KERNEL_ACCOUNT);
+ if (!root->edge) {
+ kfree(tree);
+ return NULL;
+ }
+
+ return tree;
+}
+
+void ac_destroy(struct ac_tree *tree)
+{
+ ac_free_edges(tree);
+ kfree(tree->root.edge);
+ kfree(tree);
+}
+
+int ac_find(struct ac_tree *tree, const char *text, size_t text_len)
+{
+ struct ac_state *cur = &tree->root, *edge;
+ bool found = false;
+ int i = 0, pos;
+
+ while (i < text_len) {
+ pos = ac_text_get_char(text, i);
+
+ edge = ac_state_edge(cur, pos);
+ if (edge != NULL) {
+ if (ac_state_output(edge)) {
+ found = true;
+ break;
+ }
+ WARN_ON_ONCE(ac_state_type(cur) == AC_STATE_FINAL);
+
+ cur = edge;
+ i++;
+
+ /* special case: fail state takes us to output state,
+ * this is a matching infix in the existing pattern.
+ */
+ if (ac_state_fail(edge) &&
+ ac_state_output(ac_state_fail(edge)))
+ return i - ac_state_output(ac_state_fail(edge));
+
+ /* special case: single character word as infix. */
+ edge = ac_state_edge(&tree->root, pos);
+ if (edge != NULL && ac_state_output(edge) != 0)
+ return i - ac_state_output(edge);
+
+ if (i < text_len)
+ continue;
+
+ edge = NULL;
+ }
+ while (edge == NULL) {
+ cur = ac_state_fail(cur);
+
+ /* our zerostate has fail set to NULL or root. */
+ if (cur == NULL || cur == &tree->root) {
+ cur = &tree->root;
+ /* otherwise retry with existing mismatch. */
+ if (ac_state_edge(cur, pos) == NULL)
+ i++;
+ break;
+ }
+
+ if (ac_state_output(cur))
+ return i - ac_state_output(cur);
+
+ edge = ac_state_edge(cur, pos);
+ /* don't increase i, test again the fail case. */
+ }
+ }
+
+ return found ? i - ac_state_output(edge) + 1 : -1;
+}
+
+struct ac_iter {
+ struct ac_tree *tree;
+ struct ac_stack stack[AC_MAXLEN];
+ char word[AC_MAXLEN];
+ bool pending;
+};
+
+struct ac_iter *ac_iter_create(struct ac_tree *tree)
+{
+ struct ac_iter *iter;
+
+ iter = kcalloc(1, sizeof(struct ac_iter), GFP_KERNEL_ACCOUNT);
+ if (!iter)
+ return NULL;
+
+ iter->tree = tree;
+
+ return iter;
+}
+
+int ac_iterate(struct ac_iter *iter, int (*cb)(const char *word, void *data), void *data)
+{
+ struct ac_state *cur = &iter->tree->root, *next;
+ struct ac_stack *stack = iter->stack;
+ char *word = iter->word;
+ int level = 0, i, err;
+ struct ac_stack *ss;
+ int stack_num = 0;
+
+ if (iter->pending) {
+ if (WARN_ON_ONCE(cb(word, data) < 0))
+ return -EINVAL;
+
+ iter->pending = false;
+ }
+
+ i = cur->from;
+ while (1) {
+ if (i >= ac_state_edge_len(cur)) {
+ if (stack_num <= 0)
+ break;
+
+ ss = &stack[--stack_num];
+ cur = ss->state;
+ i = ss->pos;
+ level--;
+ continue;
+ }
+
+ if (ac_state_edge(cur, i) == NULL) {
+ i++;
+ continue;
+ }
+
+ next = ac_state_edge(cur, i);
+ if (ac_state_type(next) == AC_STATE_FINAL) {
+ word[level] = i;
+ word[level + 1] = '\0';
+ i++;
+ err = cb(word, data);
+ if (err == 0) {
+ iter->pending = true;
+ return err;
+ } else if (err < 0) {
+ return err;
+ }
+ continue;
+ }
+
+ if (level >= AC_MAXLEN)
+ break;
+
+ word[level++] = i;
+ ss = &stack[stack_num++];
+ ss->pos = i + 1;
+ ss->state = cur;
+ cur = next;
+ i = next->from;
+
+ if (ac_state_output(next)) {
+ word[level + 1] = '\0';
+ err = cb(word, data);
+ if (err == 0) {
+ iter->pending = true;
+ return err;
+ } else if (err < 0) {
+ return err;
+ }
+ }
+ }
+
+ return 1;
+}
+
+void ac_iter_destroy(struct ac_iter *iter)
+{
+ kfree(iter);
+}
+
+static int ac_clone_cb(const char *word, void *data)
+{
+ if (__ac_add(data, word, strlen(word)) < 0)
+ return -1;
+
+ return 1;
+}
+
+struct ac_tree *ac_clone(struct ac_tree *tree)
+{
+ struct ac_tree *newtree;
+ struct ac_iter *iter;
+ int ret;
+
+ iter = ac_iter_create(tree);
+ if (!iter)
+ return NULL;
+
+ newtree = ac_create();
+ if (!newtree) {
+ ret = -1;
+ goto err_out;
+ }
+
+ ret = ac_iterate(iter, ac_clone_cb, newtree);
+
+err_out:
+ ac_iter_destroy(iter);
+
+ if (ret < 0) {
+ ac_destroy(newtree);
+ return NULL;
+ }
+
+ if (ac_resolve_fail(newtree) < 0) {
+ ac_destroy(newtree);
+ return NULL;
+ }
+
+ return newtree;
+}
--
2.30.2
^ permalink raw reply related [flat|nested] 4+ messages in thread
* [PATCH RFC 2/3] netfilter: nf_tables: add string set API
2022-07-29 9:31 [PATCH RFC 0/3] nf_tables string match support Pablo Neira Ayuso
2022-07-29 9:31 ` [PATCH RFC 1/3] netfilter: add Aho-Corasick string match implementation Pablo Neira Ayuso
@ 2022-07-29 9:31 ` Pablo Neira Ayuso
2022-07-29 9:31 ` [PATCH RFC 3/3] netfilter: nf_tables: add string expression Pablo Neira Ayuso
2 siblings, 0 replies; 4+ messages in thread
From: Pablo Neira Ayuso @ 2022-07-29 9:31 UTC (permalink / raw)
To: netfilter-devel
This patch adds 6 new netlink commands for the nftables API:
- NFT_MSG_NEWSTRSET, to create a new string set.
- NFT_MSG_DELSTRSET, to delete an existing string set.
- NFT_MSG_GETSTRSET, to fetch the list of existing string sets.
- NFT_MSG_NEWSTRING, to add patterns to a string set.
- NFT_MSG_DELSTRING, to delete patterns to a string set.
- NFT_MSG_GETSTRING, to fetch the patterns in an existing string set.
This API uses the Aho-Corasick implementation coming in the previous
patch. String sets stores two versions of the Aho-Corasick tree: the
active tree which is currently used to search for patterns and the clone
which is used for pending updates.
Tree updates are performed under per-netns nftables mutex, the tree
pointer is protected/accessed through RCU. There is a commit_update
field that specifies that there is a updated clone to replace the stale
tree from the commit step of the 2-phase commit protocol.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
include/net/netfilter/nf_tables.h | 33 +
include/uapi/linux/netfilter/nf_tables.h | 39 +
net/netfilter/nf_tables_api.c | 1286 ++++++++++++++++++----
3 files changed, 1158 insertions(+), 200 deletions(-)
diff --git a/include/net/netfilter/nf_tables.h b/include/net/netfilter/nf_tables.h
index 64cf655c818c..c4e9a969122a 100644
--- a/include/net/netfilter/nf_tables.h
+++ b/include/net/netfilter/nf_tables.h
@@ -1158,6 +1158,7 @@ struct nft_table {
struct list_head sets;
struct list_head objects;
struct list_head flowtables;
+ struct list_head strsets;
u64 hgenerator;
u64 handle;
u32 use;
@@ -1339,6 +1340,31 @@ void nf_tables_deactivate_flowtable(const struct nft_ctx *ctx,
void nft_register_flowtable_type(struct nf_flowtable_type *type);
void nft_unregister_flowtable_type(struct nf_flowtable_type *type);
+struct ac_tree;
+
+/**
+ * struct nft_strset - nf_tables string set
+ *
+ * @list: flow table list node in table list
+ * @table: the table the strings is contained in
+ * @name: the name of this string collection
+ * @tree: the aho-corasick tree
+ * @handle: the string collection handle
+ * @genmask: generation mask
+ * @commit_update: data structure needs update on commit
+ * @use: reference counter
+ */
+struct nft_strset {
+ struct list_head list;
+ struct nft_table *table;
+ char *name;
+ struct ac_tree __rcu *tree[2];
+ u64 handle;
+ u8 genmask:2,
+ commit_update:1;
+ u32 use;
+};
+
/**
* struct nft_traceinfo - nft tracing information and state
*
@@ -1607,6 +1633,13 @@ struct nft_trans_flowtable {
#define nft_trans_flowtable_flags(trans) \
(((struct nft_trans_flowtable *)trans->data)->flags)
+struct nft_trans_strset {
+ struct nft_strset *strset;
+};
+
+#define nft_trans_strset(trans) \
+ (((struct nft_trans_strset *)trans->data)->strset)
+
int __init nft_chain_filter_init(void);
void nft_chain_filter_fini(void);
diff --git a/include/uapi/linux/netfilter/nf_tables.h b/include/uapi/linux/netfilter/nf_tables.h
index 466fd3f4447c..d7a668199611 100644
--- a/include/uapi/linux/netfilter/nf_tables.h
+++ b/include/uapi/linux/netfilter/nf_tables.h
@@ -124,6 +124,12 @@ enum nf_tables_msg_types {
NFT_MSG_NEWFLOWTABLE,
NFT_MSG_GETFLOWTABLE,
NFT_MSG_DELFLOWTABLE,
+ NFT_MSG_NEWSTRSET,
+ NFT_MSG_DELSTRSET,
+ NFT_MSG_GETSTRSET,
+ NFT_MSG_NEWSTRING,
+ NFT_MSG_DELSTRING,
+ NFT_MSG_GETSTRING,
NFT_MSG_MAX,
};
@@ -1672,6 +1678,39 @@ enum nft_flowtable_hook_attributes {
};
#define NFTA_FLOWTABLE_HOOK_MAX (__NFTA_FLOWTABLE_HOOK_MAX - 1)
+/**
+ * enum nft_strset_hook_attributes - nf_tables string set netlink attributes
+ *
+ * @NFTA_STRSET_TABLE: table name (NLA_STRING)
+ * @NFTA_STRSET_NAME: string set name (NLA_STRING)
+ * @NFTA_STRSET_LIST: list of string elements (NLA_NESTED)
+ * @NFTA_STRSET_HANDLE: string set handle (NLA_U64)
+ * @NFTA_STRSET_USE: use reference counter (NLA_U32)
+ */
+enum nft_strset_hook_attributes {
+ NFTA_STRSET_UNSPEC,
+ NFTA_STRSET_TABLE,
+ NFTA_STRSET_NAME,
+ NFTA_STRSET_LIST,
+ NFTA_STRSET_HANDLE,
+ NFTA_STRSET_USE,
+ NFTA_STRSET_PAD,
+ __NFTA_STRSET_MAX
+};
+#define NFTA_STRSET_MAX (__NFTA_STRSET_MAX - 1)
+
+/**
+ * enum nft_string_hook_attributes - nf_tables string netlink attributes
+ *
+ * @NFTA_STRING: string (NLA_STRING)
+ */
+enum nft_string_hook_attributes {
+ NFTA_STRING_UNSPEC,
+ NFTA_STRING,
+ __NFTA_STRING_MAX
+};
+#define NFTA_STRING_MAX (__NFTA_STRING_MAX - 1)
+
/**
* enum nft_osf_attributes - nftables osf expression netlink attributes
*
diff --git a/net/netfilter/nf_tables_api.c b/net/netfilter/nf_tables_api.c
index 646d5fd53604..736bdf6b9671 100644
--- a/net/netfilter/nf_tables_api.c
+++ b/net/netfilter/nf_tables_api.c
@@ -20,6 +20,7 @@
#include <net/netfilter/nf_tables_core.h>
#include <net/netfilter/nf_tables.h>
#include <net/netfilter/nf_tables_offload.h>
+#include <net/netfilter/ahocorasick.h>
#include <net/net_namespace.h>
#include <net/sock.h>
@@ -566,6 +567,24 @@ static int nft_delflowtable(struct nft_ctx *ctx,
return err;
}
+static int nft_delstrset(struct nft_ctx *ctx, struct nft_strset *strset)
+{
+ struct nft_trans *trans;
+
+ trans = nft_trans_alloc_gfp(ctx, NFT_MSG_DELSTRSET,
+ sizeof(struct nft_trans_strset),
+ GFP_KERNEL);
+ if (!trans)
+ return -ENOMEM;
+
+ nft_deactivate_next(ctx->net, strset);
+ ctx->table->use--;
+ nft_trans_strset(trans) = strset;
+ nft_trans_commit_list_add_tail(ctx->net, trans);
+
+ return 0;
+}
+
static void __nft_reg_track_clobber(struct nft_regs_track *track, u8 dreg)
{
int i;
@@ -1232,6 +1251,7 @@ static int nf_tables_newtable(struct sk_buff *skb, const struct nfnl_info *info,
INIT_LIST_HEAD(&table->sets);
INIT_LIST_HEAD(&table->objects);
INIT_LIST_HEAD(&table->flowtables);
+ INIT_LIST_HEAD(&table->strsets);
table->family = family;
table->flags = flags;
table->handle = ++table_handle;
@@ -1260,6 +1280,7 @@ static int nf_tables_newtable(struct sk_buff *skb, const struct nfnl_info *info,
static int nft_flush_table(struct nft_ctx *ctx)
{
struct nft_flowtable *flowtable, *nft;
+ struct nft_strset *strset, *nst;
struct nft_chain *chain, *nc;
struct nft_object *obj, *ne;
struct nft_set *set, *ns;
@@ -1301,6 +1322,15 @@ static int nft_flush_table(struct nft_ctx *ctx)
goto out;
}
+ list_for_each_entry_safe(strset, nst, &ctx->table->strsets, list) {
+ if (!nft_is_active_next(ctx->net, strset))
+ continue;
+
+ err = nft_delstrset(ctx, strset);
+ if (err < 0)
+ goto out;
+ }
+
list_for_each_entry_safe(obj, ne, &ctx->table->objects, list) {
if (!nft_is_active_next(ctx->net, obj))
continue;
@@ -8074,225 +8104,999 @@ static struct notifier_block nf_tables_flowtable_notifier = {
.notifier_call = nf_tables_flowtable_event,
};
-static void nf_tables_gen_notify(struct net *net, struct sk_buff *skb,
- int event)
-{
- struct nlmsghdr *nlh = nlmsg_hdr(skb);
- struct sk_buff *skb2;
- int err;
+/*
+ * String set
+ */
- if (!nlmsg_report(nlh) &&
- !nfnetlink_has_listeners(net, NFNLGRP_NFTABLES))
- return;
+static const struct nla_policy nft_strset_policy[NFTA_STRSET_MAX + 1] = {
+ [NFTA_STRSET_TABLE] = { .type = NLA_STRING,
+ .len = NFT_NAME_MAXLEN - 1 },
+ [NFTA_STRSET_NAME] = { .type = NLA_STRING,
+ .len = NFT_NAME_MAXLEN - 1 },
+ [NFTA_STRSET_LIST] = { .type = NLA_NESTED },
+ [NFTA_STRSET_HANDLE] = { .type = NLA_U64 },
+};
- skb2 = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
- if (skb2 == NULL)
- goto err;
+static struct nft_strset *nft_strset_lookup(const struct net *net,
+ struct nft_table *table,
+ const struct nlattr *nla,
+ u8 genmask)
+{
+ struct nft_strset *strset;
- err = nf_tables_fill_gen_info(skb2, net, NETLINK_CB(skb).portid,
- nlh->nlmsg_seq);
- if (err < 0) {
- kfree_skb(skb2);
- goto err;
+ if (nla == NULL)
+ return ERR_PTR(-EINVAL);
+
+ list_for_each_entry_rcu(strset, &table->strsets, list,
+ lockdep_is_held(&nft_net->commit_mutex)) {
+ if (!nla_strcmp(nla, strset->name) &&
+ nft_active_genmask(strset, genmask))
+ return strset;
}
- nfnetlink_send(skb2, net, NETLINK_CB(skb).portid, NFNLGRP_NFTABLES,
- nlmsg_report(nlh), GFP_KERNEL);
- return;
-err:
- nfnetlink_set_err(net, NETLINK_CB(skb).portid, NFNLGRP_NFTABLES,
- -ENOBUFS);
+ return ERR_PTR(-ENOENT);
}
-static int nf_tables_getgen(struct sk_buff *skb, const struct nfnl_info *info,
- const struct nlattr * const nla[])
+static struct nft_strset *
+nft_strset_lookup_byhandle(const struct nft_table *table, u64 handle, u8 genmask)
{
- struct sk_buff *skb2;
- int err;
-
- skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
- if (skb2 == NULL)
- return -ENOMEM;
-
- err = nf_tables_fill_gen_info(skb2, info->net, NETLINK_CB(skb).portid,
- info->nlh->nlmsg_seq);
- if (err < 0)
- goto err_fill_gen_info;
+ struct nft_strset *strset;
- return nfnetlink_unicast(skb2, info->net, NETLINK_CB(skb).portid);
+ list_for_each_entry(strset, &table->strsets, list) {
+ if (strset->handle == handle &&
+ nft_active_genmask(strset, genmask))
+ return strset;
+ }
-err_fill_gen_info:
- kfree_skb(skb2);
- return err;
+ return ERR_PTR(-ENOENT);
}
-static const struct nfnl_callback nf_tables_cb[NFT_MSG_MAX] = {
- [NFT_MSG_NEWTABLE] = {
- .call = nf_tables_newtable,
- .type = NFNL_CB_BATCH,
- .attr_count = NFTA_TABLE_MAX,
- .policy = nft_table_policy,
- },
- [NFT_MSG_GETTABLE] = {
- .call = nf_tables_gettable,
- .type = NFNL_CB_RCU,
- .attr_count = NFTA_TABLE_MAX,
- .policy = nft_table_policy,
- },
- [NFT_MSG_DELTABLE] = {
- .call = nf_tables_deltable,
- .type = NFNL_CB_BATCH,
- .attr_count = NFTA_TABLE_MAX,
- .policy = nft_table_policy,
- },
- [NFT_MSG_NEWCHAIN] = {
- .call = nf_tables_newchain,
- .type = NFNL_CB_BATCH,
- .attr_count = NFTA_CHAIN_MAX,
- .policy = nft_chain_policy,
- },
- [NFT_MSG_GETCHAIN] = {
- .call = nf_tables_getchain,
- .type = NFNL_CB_RCU,
- .attr_count = NFTA_CHAIN_MAX,
- .policy = nft_chain_policy,
- },
- [NFT_MSG_DELCHAIN] = {
- .call = nf_tables_delchain,
- .type = NFNL_CB_BATCH,
- .attr_count = NFTA_CHAIN_MAX,
- .policy = nft_chain_policy,
- },
- [NFT_MSG_NEWRULE] = {
- .call = nf_tables_newrule,
- .type = NFNL_CB_BATCH,
- .attr_count = NFTA_RULE_MAX,
- .policy = nft_rule_policy,
- },
- [NFT_MSG_GETRULE] = {
- .call = nf_tables_getrule,
- .type = NFNL_CB_RCU,
- .attr_count = NFTA_RULE_MAX,
- .policy = nft_rule_policy,
- },
- [NFT_MSG_DELRULE] = {
- .call = nf_tables_delrule,
- .type = NFNL_CB_BATCH,
- .attr_count = NFTA_RULE_MAX,
- .policy = nft_rule_policy,
- },
- [NFT_MSG_NEWSET] = {
- .call = nf_tables_newset,
- .type = NFNL_CB_BATCH,
- .attr_count = NFTA_SET_MAX,
- .policy = nft_set_policy,
- },
- [NFT_MSG_GETSET] = {
- .call = nf_tables_getset,
- .type = NFNL_CB_RCU,
- .attr_count = NFTA_SET_MAX,
- .policy = nft_set_policy,
- },
- [NFT_MSG_DELSET] = {
- .call = nf_tables_delset,
- .type = NFNL_CB_BATCH,
- .attr_count = NFTA_SET_MAX,
- .policy = nft_set_policy,
- },
- [NFT_MSG_NEWSETELEM] = {
- .call = nf_tables_newsetelem,
- .type = NFNL_CB_BATCH,
- .attr_count = NFTA_SET_ELEM_LIST_MAX,
- .policy = nft_set_elem_list_policy,
- },
- [NFT_MSG_GETSETELEM] = {
- .call = nf_tables_getsetelem,
- .type = NFNL_CB_RCU,
- .attr_count = NFTA_SET_ELEM_LIST_MAX,
- .policy = nft_set_elem_list_policy,
- },
- [NFT_MSG_DELSETELEM] = {
- .call = nf_tables_delsetelem,
- .type = NFNL_CB_BATCH,
- .attr_count = NFTA_SET_ELEM_LIST_MAX,
- .policy = nft_set_elem_list_policy,
- },
- [NFT_MSG_GETGEN] = {
- .call = nf_tables_getgen,
- .type = NFNL_CB_RCU,
- },
- [NFT_MSG_NEWOBJ] = {
- .call = nf_tables_newobj,
- .type = NFNL_CB_BATCH,
- .attr_count = NFTA_OBJ_MAX,
- .policy = nft_obj_policy,
- },
- [NFT_MSG_GETOBJ] = {
- .call = nf_tables_getobj,
- .type = NFNL_CB_RCU,
- .attr_count = NFTA_OBJ_MAX,
- .policy = nft_obj_policy,
- },
- [NFT_MSG_DELOBJ] = {
- .call = nf_tables_delobj,
- .type = NFNL_CB_BATCH,
- .attr_count = NFTA_OBJ_MAX,
- .policy = nft_obj_policy,
- },
- [NFT_MSG_GETOBJ_RESET] = {
- .call = nf_tables_getobj,
- .type = NFNL_CB_RCU,
- .attr_count = NFTA_OBJ_MAX,
- .policy = nft_obj_policy,
- },
- [NFT_MSG_NEWFLOWTABLE] = {
- .call = nf_tables_newflowtable,
- .type = NFNL_CB_BATCH,
- .attr_count = NFTA_FLOWTABLE_MAX,
- .policy = nft_flowtable_policy,
- },
- [NFT_MSG_GETFLOWTABLE] = {
- .call = nf_tables_getflowtable,
- .type = NFNL_CB_RCU,
- .attr_count = NFTA_FLOWTABLE_MAX,
- .policy = nft_flowtable_policy,
- },
- [NFT_MSG_DELFLOWTABLE] = {
- .call = nf_tables_delflowtable,
- .type = NFNL_CB_BATCH,
- .attr_count = NFTA_FLOWTABLE_MAX,
- .policy = nft_flowtable_policy,
- },
-};
-
-static int nf_tables_validate(struct net *net)
+static int nf_tables_newstrset(struct sk_buff *skb, const struct nfnl_info *info,
+ const struct nlattr * const nla[])
{
- struct nftables_pernet *nft_net = nft_pernet(net);
+ struct nftables_pernet *nft_net = nft_pernet(info->net);
+ struct netlink_ext_ack *extack = info->extack;
+ u8 genmask = nft_genmask_next(info->net);
+ u8 family = info->nfmsg->nfgen_family;
+ struct net *net = info->net;
+ struct nft_strset *strset;
+ const struct nlattr *attr;
struct nft_table *table;
+ struct nft_trans *trans;
+ struct ac_tree *tree;
+ struct nft_ctx ctx;
+ int err;
- switch (nft_net->validate_state) {
- case NFT_VALIDATE_SKIP:
- break;
- case NFT_VALIDATE_NEED:
- nft_validate_state_update(net, NFT_VALIDATE_DO);
- fallthrough;
- case NFT_VALIDATE_DO:
- list_for_each_entry(table, &nft_net->tables, list) {
- if (nft_table_validate(net, table) < 0)
- return -EAGAIN;
+ lockdep_assert_held(&nft_net->commit_mutex);
+
+ table = nft_table_lookup(net, nla[NFTA_STRSET_TABLE], family, genmask,
+ NETLINK_CB(skb).portid);
+ if (IS_ERR(table)) {
+ NL_SET_BAD_ATTR(extack, nla[NFTA_STRSET_TABLE]);
+ return PTR_ERR(table);
+ }
+
+ if (nla[NFTA_STRSET_NAME]) {
+ attr = nla[NFTA_STRSET_NAME];
+ strset = nft_strset_lookup(net, table, attr, genmask);
+ if (IS_ERR(strset)) {
+ if (PTR_ERR(strset) != -ENOENT) {
+ NL_SET_BAD_ATTR(extack, attr);
+ return PTR_ERR(strset);
+ }
+ strset = NULL;
}
- break;
+ } else {
+ return -EINVAL;
}
- return 0;
-}
+ if (strset != NULL) {
+ if (info->nlh->nlmsg_flags & NLM_F_EXCL) {
+ NL_SET_BAD_ATTR(extack, attr);
+ return -EEXIST;
+ }
+ if (info->nlh->nlmsg_flags & NLM_F_REPLACE)
+ return -EOPNOTSUPP;
-/* a drop policy has to be deferred until all rules have been activated,
- * otherwise a large ruleset that contains a drop-policy base chain will
- * cause all packets to get dropped until the full transaction has been
- * processed.
- *
- * We defer the drop policy until the transaction has been finalized.
+ return 0;
+ }
+
+ strset = kmalloc(sizeof(*strset), GFP_KERNEL_ACCOUNT);
+ if (!strset)
+ return -ENOMEM;
+
+ strset->use = 0;
+ strset->table = table;
+
+ strset->name = nla_strdup(nla[NFTA_STRSET_NAME], GFP_KERNEL_ACCOUNT);
+ if (!strset->name) {
+ err = -ENOMEM;
+ goto err_name;
+ }
+
+ tree = ac_create();
+ if (!tree) {
+ err = -ENOMEM;
+ goto err_ac_create;
+ }
+ RCU_INIT_POINTER(strset->tree[0], tree);
+ RCU_INIT_POINTER(strset->tree[1], tree);
+
+ nft_ctx_init(&ctx, net, skb, info->nlh, family, table, NULL, nla);
+
+ trans = nft_trans_alloc_gfp(&ctx, NFT_MSG_NEWSTRSET,
+ sizeof(struct nft_trans_strset),
+ GFP_KERNEL);
+ if (!trans) {
+ err = -ENOMEM;
+ goto err_trans_alloc;
+ }
+
+ nft_trans_strset(trans) = strset;
+ nft_activate_next(info->net, strset);
+ nft_trans_commit_list_add_tail(info->net, trans);
+
+ list_add_tail_rcu(&strset->list, &table->strsets);
+ table->use++;
+
+ return 0;
+
+err_trans_alloc:
+ ac_destroy(tree);
+err_ac_create:
+ kfree(strset->name);
+err_name:
+ kfree(strset);
+
+ return err;
+}
+
+static int nft_strset_fill_info(struct sk_buff *skb, struct net *net,
+ u32 portid, u32 seq, int event, u32 flags,
+ int family, const struct nft_strset *strset)
+{
+ struct nlmsghdr *nlh;
+
+ event = nfnl_msg_type(NFNL_SUBSYS_NFTABLES, event);
+ nlh = nfnl_msg_put(skb, portid, seq, event, flags, family,
+ NFNETLINK_V0, nft_base_seq(net));
+ if (!nlh)
+ goto nla_put_failure;
+
+ if (nla_put_string(skb, NFTA_STRSET_TABLE, strset->table->name) ||
+ nla_put_string(skb, NFTA_STRSET_NAME, strset->name) ||
+ nla_put_be32(skb, NFTA_STRSET_USE, htonl(strset->use)) ||
+ nla_put_be64(skb, NFTA_STRSET_HANDLE, cpu_to_be64(strset->handle),
+ NFTA_STRSET_PAD))
+ goto nla_put_failure;
+
+ nlmsg_end(skb, nlh);
+
+ return 0;
+
+nla_put_failure:
+ nlmsg_trim(skb, nlh);
+ return -1;
+}
+
+static int nf_tables_dump_strsets(struct sk_buff *skb, struct netlink_callback *cb)
+{
+ const struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh);
+ unsigned int idx = 0, s_idx = cb->args[0];
+ struct net *net = sock_net(skb->sk);
+ int family = nfmsg->nfgen_family;
+ struct nftables_pernet *nft_net;
+ struct nft_strset *strset;
+ struct nft_table *table;
+
+ rcu_read_lock();
+ nft_net = nft_pernet(net);
+ list_for_each_entry_rcu(table, &nft_net->tables, list) {
+ if (family != NFPROTO_UNSPEC && family != table->family)
+ continue;
+ list_for_each_entry_rcu(strset, &table->strsets, list) {
+ if (idx < s_idx)
+ goto cont;
+ if (idx > s_idx)
+ memset(&cb->args[1], 0,
+ sizeof(cb->args) - sizeof(cb->args[0]));
+ if (!nft_is_active(net, strset))
+ continue;
+ if (nft_strset_fill_info(skb, net, NETLINK_CB(cb->skb).portid,
+ cb->nlh->nlmsg_seq, NFT_MSG_NEWSTRSET, 0,
+ table->family, strset) < 0)
+ goto done;
+cont:
+ idx++;
+ }
+ }
+done:
+ rcu_read_unlock();
+ cb->args[0] = idx;
+ return skb->len;
+}
+
+static int nf_tables_getstrset(struct sk_buff *skb, const struct nfnl_info *info,
+ const struct nlattr * const nla[])
+{
+ if (nla[NFTA_STRSET_TABLE] ||
+ nla[NFTA_STRSET_NAME] ||
+ nla[NFTA_STRSET_HANDLE])
+ return -EOPNOTSUPP;
+
+ if (info->nlh->nlmsg_flags & NLM_F_DUMP) {
+ struct netlink_dump_control c = {
+ .dump = nf_tables_dump_strsets,
+ .module = THIS_MODULE,
+ };
+
+ return nft_netlink_dump_start_rcu(info->sk, skb, info->nlh, &c);
+ }
+
+ return -EOPNOTSUPP;
+}
+
+static void nft_strset_destroy(struct nft_ctx *ctx, struct nft_strset *strset)
+{
+ struct ac_tree *tree_0, *tree_1;
+
+ tree_0 = rcu_dereference_protected(strset->tree[0],
+ lockdep_commit_lock_is_held(ctx->net));
+ tree_1 = rcu_dereference_protected(strset->tree[1],
+ lockdep_commit_lock_is_held(ctx->net));
+ ac_destroy(tree_0);
+ ac_destroy(tree_1);
+ kfree(strset->name);
+ kfree(strset);
+}
+
+static int nf_tables_delstrset(struct sk_buff *skb, const struct nfnl_info *info,
+ const struct nlattr * const nla[])
+{
+ struct nftables_pernet *nft_net = nft_pernet(info->net);
+ struct netlink_ext_ack *extack = info->extack;
+ u8 genmask = nft_genmask_next(info->net);
+ u8 family = info->nfmsg->nfgen_family;
+ struct net *net = info->net;
+ struct nft_strset *strset;
+ const struct nlattr *attr;
+ struct nft_table *table;
+ struct nft_ctx ctx;
+ u64 handle = 0;
+
+ lockdep_assert_held(&nft_net->commit_mutex);
+
+ table = nft_table_lookup(net, nla[NFTA_STRSET_TABLE], family, genmask,
+ NETLINK_CB(skb).portid);
+ if (IS_ERR(table)) {
+ NL_SET_BAD_ATTR(extack, nla[NFTA_STRSET_TABLE]);
+ return PTR_ERR(table);
+ }
+
+ if (nla[NFTA_STRSET_HANDLE]) {
+ attr = nla[NFTA_STRSET_HANDLE];
+ handle = be64_to_cpu(nla_get_be64(attr));
+ strset = nft_strset_lookup_byhandle(table, handle, genmask);
+ } else if (nla[NFTA_STRSET_NAME]) {
+ attr = nla[NFTA_STRSET_NAME];
+ strset = nft_strset_lookup(net, table, attr, genmask);
+ } else
+ return -EINVAL;
+
+ if (IS_ERR(strset)) {
+ NL_SET_BAD_ATTR(extack, attr);
+ return PTR_ERR(strset);
+ }
+
+ if (strset->use > 0) {
+ NL_SET_BAD_ATTR(extack, attr);
+ return -EBUSY;
+ }
+
+ nft_ctx_init(&ctx, net, skb, info->nlh, family, table, NULL, nla);
+
+ return nft_delstrset(&ctx, strset);
+}
+
+static const struct nla_policy nft_string_policy[NFTA_STRING_MAX + 1] = {
+ [NFTA_STRING] = { .type = NLA_STRING,
+ .len = NFT_NAME_MAXLEN - 1 },
+};
+
+static struct nlattr *nft_parse_string(struct nlattr *attr)
+{
+ struct nlattr *tb[NFTA_STRING_MAX + 1];
+ int err;
+
+ err = nla_parse_nested_deprecated(tb, NFTA_STRING_MAX, attr,
+ nft_string_policy, NULL);
+ if (err < 0)
+ return ERR_PTR(err);
+
+ if (!tb[NFTA_STRING])
+ return ERR_PTR(-EINVAL);
+
+ return tb[NFTA_STRING];
+}
+
+static int nft_string_add(struct nft_ctx *ctx, struct nft_strset *strset,
+ const struct nlattr *attr,
+ struct netlink_ext_ack *extack, bool create)
+{
+ u8 next_genbit = nft_gencursor_next(ctx->net);
+ struct nlattr *tmp, *str;
+ struct ac_tree *tree;
+ int err, rem;
+
+ nla_for_each_nested(tmp, attr, rem) {
+ err = -EINVAL;
+ if (nla_type(tmp) != NFTA_LIST_ELEM)
+ return err;
+
+ str = nft_parse_string(tmp);
+ if (IS_ERR(str)) {
+ NL_SET_BAD_ATTR(extack, tmp);
+ return PTR_ERR(str);
+ }
+
+ tree = rcu_dereference_protected(strset->tree[next_genbit],
+ lockdep_commit_lock_is_held(ctx->net));
+
+ err = ac_add(tree, nla_data(str), nla_len(str), create);
+ if (err < 0) {
+ NL_SET_BAD_ATTR(extack, tmp);
+ return err;
+ }
+ }
+
+ return 0;
+}
+
+static int nf_tables_newstr(struct sk_buff *skb, const struct nfnl_info *info,
+ const struct nlattr * const nla[])
+{
+ struct nftables_pernet *nft_net = nft_pernet(info->net);
+ u8 next_genbit = nft_gencursor_next(info->net);
+ struct netlink_ext_ack *extack = info->extack;
+ u8 cur_genbit = info->net->nft.gencursor;
+ u8 genmask = nft_genmask_next(info->net);
+ u8 family = info->nfmsg->nfgen_family;
+ struct ac_tree *tree, *clone;
+ struct net *net = info->net;
+ const struct nlattr *attr;
+ struct nft_strset *strset;
+ struct nft_trans *trans;
+ struct nft_table *table;
+ struct nft_ctx ctx;
+ u64 handle = 0;
+
+ lockdep_assert_held(&nft_net->commit_mutex);
+
+ table = nft_table_lookup(net, nla[NFTA_STRSET_TABLE], family, genmask,
+ NETLINK_CB(skb).portid);
+ if (IS_ERR(table)) {
+ NL_SET_BAD_ATTR(extack, nla[NFTA_STRSET_TABLE]);
+ return PTR_ERR(table);
+ }
+
+ if (nla[NFTA_STRSET_HANDLE]) {
+ attr = nla[NFTA_STRSET_HANDLE];
+ handle = be64_to_cpu(nla_get_be64(attr));
+ strset = nft_strset_lookup_byhandle(table, handle, genmask);
+ } else if (nla[NFTA_STRSET_NAME]) {
+ attr = nla[NFTA_STRSET_NAME];
+ strset = nft_strset_lookup(net, table, attr, genmask);
+ } else
+ return -EINVAL;
+
+ if (IS_ERR(strset)) {
+ NL_SET_BAD_ATTR(extack, attr);
+ return PTR_ERR(strset);
+ }
+
+ nft_ctx_init(&ctx, net, skb, info->nlh, family, table, NULL, NULL);
+
+ if (!strset->commit_update) {
+ tree = rcu_dereference_protected(strset->tree[cur_genbit],
+ lockdep_commit_lock_is_held(net));
+ clone = ac_clone(tree);
+ if (!clone)
+ return -ENOMEM;
+
+ rcu_assign_pointer(strset->tree[next_genbit], clone);
+ strset->commit_update = 1;
+
+ trans = nft_trans_alloc_gfp(&ctx, NFT_MSG_NEWSTRING,
+ sizeof(struct nft_trans_strset),
+ GFP_KERNEL);
+ if (!trans) {
+ ac_destroy(clone);
+ return -ENOMEM;
+ }
+
+ nft_trans_strset(trans) = strset;
+ nft_trans_commit_list_add_tail(net, trans);
+ }
+
+ return nft_string_add(&ctx, strset, nla[NFTA_STRSET_LIST],
+ extack, info->nlh->nlmsg_flags & NLM_F_EXCL);
+}
+
+static int nft_string_del(struct nft_ctx *ctx, struct nft_strset *strset,
+ const struct nlattr *attr,
+ struct netlink_ext_ack *extack)
+{
+ u8 next_genbit = nft_gencursor_next(ctx->net);
+ struct nlattr *tmp, *str;
+ struct ac_tree *tree;
+ int err, rem;
+
+ nla_for_each_nested(tmp, attr, rem) {
+ err = -EINVAL;
+ if (nla_type(tmp) != NFTA_LIST_ELEM)
+ return err;
+
+ str = nft_parse_string(tmp);
+ if (IS_ERR(str)) {
+ NL_SET_BAD_ATTR(extack, tmp);
+ return err;
+ }
+
+ err = -E2BIG;
+ if (nla_len(str) == 0 ||
+ nla_len(str) > AC_MAXLEN) {
+ NL_SET_BAD_ATTR(extack, str);
+ return err;
+ }
+
+ tree = rcu_dereference_protected(strset->tree[next_genbit],
+ lockdep_commit_lock_is_held(ctx->net));
+ ac_del(tree, nla_data(str), nla_len(str));
+ }
+
+ return 0;
+}
+
+static int nf_tables_delstr(struct sk_buff *skb, const struct nfnl_info *info,
+ const struct nlattr * const nla[])
+{
+ struct nftables_pernet *nft_net = nft_pernet(info->net);
+ u8 next_genbit = nft_gencursor_next(info->net);
+ struct netlink_ext_ack *extack = info->extack;
+ u8 cur_genbit = info->net->nft.gencursor;
+ u8 genmask = nft_genmask_next(info->net);
+ u8 family = info->nfmsg->nfgen_family;
+ struct ac_tree *tree, *clone;
+ struct net *net = info->net;
+ const struct nlattr *attr;
+ struct nft_strset *strset;
+ struct nft_trans *trans;
+ struct nft_table *table;
+ struct nft_ctx ctx;
+ u64 handle = 0;
+
+ lockdep_assert_held(&nft_net->commit_mutex);
+
+ table = nft_table_lookup(net, nla[NFTA_STRSET_TABLE], family, genmask,
+ NETLINK_CB(skb).portid);
+ if (IS_ERR(table)) {
+ NL_SET_BAD_ATTR(extack, nla[NFTA_STRSET_TABLE]);
+ return PTR_ERR(table);
+ }
+
+ if (nla[NFTA_STRSET_HANDLE]) {
+ attr = nla[NFTA_STRSET_HANDLE];
+ handle = be64_to_cpu(nla_get_be64(attr));
+ strset = nft_strset_lookup_byhandle(table, handle, genmask);
+ } else if (nla[NFTA_STRSET_NAME]) {
+ attr = nla[NFTA_STRSET_NAME];
+ strset = nft_strset_lookup(net, table, attr, genmask);
+ } else
+ return -EINVAL;
+
+ if (IS_ERR(strset)) {
+ NL_SET_BAD_ATTR(extack, attr);
+ return PTR_ERR(strset);
+ }
+
+ nft_ctx_init(&ctx, net, skb, info->nlh, family, table, NULL, NULL);
+
+ if (!strset->commit_update) {
+ tree = rcu_dereference_protected(strset->tree[cur_genbit],
+ lockdep_commit_lock_is_held(net));
+ clone = ac_clone(tree);
+ if (!clone)
+ return -ENOMEM;
+
+ rcu_assign_pointer(strset->tree[next_genbit], clone);
+ strset->commit_update = 1;
+
+ trans = nft_trans_alloc_gfp(&ctx, NFT_MSG_DELSTRING,
+ sizeof(struct nft_trans_strset),
+ GFP_KERNEL);
+ if (!trans) {
+ ac_destroy(clone);
+ return -ENOMEM;
+ }
+
+ nft_trans_strset(trans) = strset;
+ nft_trans_commit_list_add_tail(net, trans);
+ }
+
+ return nft_string_del(&ctx, strset, nla[NFTA_STRSET_LIST], extack);
+}
+
+struct nft_strset_dump_ctx {
+ const struct nft_strset *strset;
+ struct ac_iter *iter;
+ struct nft_ctx ctx;
+ u32 genid;
+};
+
+static int nf_tables_dump_str_start(struct netlink_callback *cb)
+{
+ struct nft_strset_dump_ctx *dump_ctx = cb->data;
+ struct net *net = sock_net(cb->skb->sk);
+ struct nftables_pernet *nft_net;
+ struct ac_tree *tree;
+ struct ac_iter *iter;
+ u8 cur_genbit;
+
+ cur_genbit = READ_ONCE(net->nft.gencursor);
+
+ tree = rcu_dereference(dump_ctx->strset->tree[cur_genbit]);
+
+ iter = ac_iter_create(tree);
+ if (!iter)
+ return -ENOMEM;
+
+ cb->data = kmemdup(dump_ctx, sizeof(*dump_ctx), GFP_ATOMIC);
+ if (!cb->data) {
+ ac_iter_destroy(iter);
+ return -ENOMEM;
+ }
+
+ nft_net = nft_pernet(net);
+
+ dump_ctx = cb->data;
+ dump_ctx->iter = iter;
+ dump_ctx->genid = nft_net->base_seq;
+
+ return 0;
+}
+
+static int nf_tables_dump_str_done(struct netlink_callback *cb)
+{
+ struct nft_strset_dump_ctx *dump_ctx = cb->data;
+
+ ac_iter_destroy(dump_ctx->iter);
+ kfree(dump_ctx);
+
+ return 0;
+}
+
+static int nft_iter_str_cb(const char *string, void *data)
+{
+ struct sk_buff *skb = data;
+ unsigned char *b = skb_tail_pointer(skb);
+ struct nlattr *nest;
+
+ nest = nla_nest_start_noflag(skb, NFTA_LIST_ELEM);
+ if (!nest)
+ goto nla_put_failure;
+
+ if (nla_put_string(skb, NFTA_STRING, string))
+ goto nla_put_failure;
+
+ nla_nest_end(skb, nest);
+
+ return 1;
+
+nla_put_failure:
+ nlmsg_trim(skb, b);
+
+ return -1;
+}
+
+static int nft_string_fill_info(struct net *net, struct sk_buff *skb,
+ const struct nft_table *table,
+ const struct nft_strset *strset,
+ struct nft_strset_dump_ctx *dump_ctx,
+ struct netlink_callback *cb)
+{
+ struct nlmsghdr *nlh;
+ struct nlattr *nest;
+ u32 portid, seq;
+ int event, ret;
+
+ event = nfnl_msg_type(NFNL_SUBSYS_NFTABLES, NFT_MSG_NEWSTRING);
+ portid = NETLINK_CB(cb->skb).portid;
+ seq = cb->nlh->nlmsg_seq;
+
+ nlh = nfnl_msg_put(skb, portid, seq, event, NLM_F_MULTI,
+ table->family, NFNETLINK_V0, nft_base_seq(net));
+ if (!nlh)
+ goto nla_put_failure;
+
+ if (nla_put_string(skb, NFTA_STRSET_TABLE, table->name))
+ goto nla_put_failure;
+ if (nla_put_string(skb, NFTA_STRSET_NAME, strset->name))
+ goto nla_put_failure;
+
+ nest = nla_nest_start_noflag(skb, NFTA_STRSET_LIST);
+ if (nest == NULL)
+ goto nla_put_failure;
+
+ ret = ac_iterate(dump_ctx->iter, nft_iter_str_cb, skb);
+
+ nla_nest_end(skb, nest);
+ nlmsg_end(skb, nlh);
+
+ return ret;
+
+nla_put_failure:
+ nlmsg_trim(skb, nlh);
+
+ return -1;
+}
+
+static int nf_tables_dump_str(struct sk_buff *skb, struct netlink_callback *cb)
+{
+ struct nft_strset_dump_ctx *dump_ctx = cb->data;
+ struct net *net = sock_net(skb->sk);
+ struct nftables_pernet *nft_net;
+ bool strset_found = false;
+ struct nft_strset *strset;
+ struct nft_table *table;
+ int ret;
+
+ if (cb->args[0])
+ return 0;
+
+ rcu_read_lock();
+ nft_net = nft_pernet(net);
+ if (dump_ctx->genid != READ_ONCE(nft_net->base_seq)) {
+ rcu_read_unlock();
+ return 0;
+ }
+
+ list_for_each_entry_rcu(table, &nft_net->tables, list) {
+ if (dump_ctx->ctx.family != NFPROTO_UNSPEC &&
+ dump_ctx->ctx.family != table->family)
+ continue;
+
+ if (table != dump_ctx->ctx.table)
+ continue;
+
+ list_for_each_entry_rcu(strset, &table->strsets, list) {
+ if (strset == dump_ctx->strset) {
+ strset_found = true;
+ break;
+ }
+ }
+ break;
+ }
+
+ if (!strset_found) {
+ rcu_read_unlock();
+ return -ENOENT;
+ }
+
+ ret = nft_string_fill_info(net, skb, table, strset, dump_ctx, cb);
+ if (ret == 1)
+ cb->args[0] = 1;
+
+ nl_dump_check_consistent(cb, nlmsg_hdr(skb));
+
+ rcu_read_unlock();
+
+ return skb->len;
+}
+
+static int nf_tables_getstr(struct sk_buff *skb, const struct nfnl_info *info,
+ const struct nlattr * const nla[])
+{
+ struct netlink_ext_ack *extack = info->extack;
+ u8 genmask = nft_genmask_cur(info->net);
+ u8 family = info->nfmsg->nfgen_family;
+ struct nft_strset_dump_ctx dump_ctx;
+ struct net *net = info->net;
+ struct nft_strset *strset;
+ struct nft_table *table;
+ struct nft_ctx ctx;
+
+ if (!nla[NFTA_STRSET_TABLE] ||
+ !nla[NFTA_STRSET_NAME])
+ return -EINVAL;
+
+ table = nft_table_lookup(net, nla[NFTA_STRSET_TABLE], family, genmask,
+ NETLINK_CB(skb).portid);
+ if (!table) {
+ NL_SET_BAD_ATTR(extack, nla[NFTA_STRSET_TABLE]);
+ return PTR_ERR(table);
+ }
+
+ strset = nft_strset_lookup(net, table, nla[NFTA_STRSET_NAME], genmask);
+ if (IS_ERR(strset)) {
+ NL_SET_BAD_ATTR(extack, nla[NFTA_STRSET_HANDLE]);
+ return PTR_ERR(strset);
+ }
+
+ nft_ctx_init(&ctx, net, skb, info->nlh, family, table, NULL, NULL);
+
+ dump_ctx.ctx = ctx;
+ dump_ctx.strset = strset;
+
+ if (info->nlh->nlmsg_flags & NLM_F_DUMP) {
+ struct netlink_dump_control c = {
+ .start = nf_tables_dump_str_start,
+ .dump = nf_tables_dump_str,
+ .done = nf_tables_dump_str_done,
+ .data = &dump_ctx,
+ .module = THIS_MODULE,
+ };
+
+ return nft_netlink_dump_start_rcu(info->sk, skb, info->nlh, &c);
+ }
+
+ return 0;
+}
+
+static void nf_tables_gen_notify(struct net *net, struct sk_buff *skb,
+ int event)
+{
+ struct nlmsghdr *nlh = nlmsg_hdr(skb);
+ struct sk_buff *skb2;
+ int err;
+
+ if (!nlmsg_report(nlh) &&
+ !nfnetlink_has_listeners(net, NFNLGRP_NFTABLES))
+ return;
+
+ skb2 = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
+ if (skb2 == NULL)
+ goto err;
+
+ err = nf_tables_fill_gen_info(skb2, net, NETLINK_CB(skb).portid,
+ nlh->nlmsg_seq);
+ if (err < 0) {
+ kfree_skb(skb2);
+ goto err;
+ }
+
+ nfnetlink_send(skb2, net, NETLINK_CB(skb).portid, NFNLGRP_NFTABLES,
+ nlmsg_report(nlh), GFP_KERNEL);
+ return;
+err:
+ nfnetlink_set_err(net, NETLINK_CB(skb).portid, NFNLGRP_NFTABLES,
+ -ENOBUFS);
+}
+
+static int nf_tables_getgen(struct sk_buff *skb, const struct nfnl_info *info,
+ const struct nlattr * const nla[])
+{
+ struct sk_buff *skb2;
+ int err;
+
+ skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
+ if (skb2 == NULL)
+ return -ENOMEM;
+
+ err = nf_tables_fill_gen_info(skb2, info->net, NETLINK_CB(skb).portid,
+ info->nlh->nlmsg_seq);
+ if (err < 0)
+ goto err_fill_gen_info;
+
+ return nfnetlink_unicast(skb2, info->net, NETLINK_CB(skb).portid);
+
+err_fill_gen_info:
+ kfree_skb(skb2);
+ return err;
+}
+
+static const struct nfnl_callback nf_tables_cb[NFT_MSG_MAX] = {
+ [NFT_MSG_NEWTABLE] = {
+ .call = nf_tables_newtable,
+ .type = NFNL_CB_BATCH,
+ .attr_count = NFTA_TABLE_MAX,
+ .policy = nft_table_policy,
+ },
+ [NFT_MSG_GETTABLE] = {
+ .call = nf_tables_gettable,
+ .type = NFNL_CB_RCU,
+ .attr_count = NFTA_TABLE_MAX,
+ .policy = nft_table_policy,
+ },
+ [NFT_MSG_DELTABLE] = {
+ .call = nf_tables_deltable,
+ .type = NFNL_CB_BATCH,
+ .attr_count = NFTA_TABLE_MAX,
+ .policy = nft_table_policy,
+ },
+ [NFT_MSG_NEWCHAIN] = {
+ .call = nf_tables_newchain,
+ .type = NFNL_CB_BATCH,
+ .attr_count = NFTA_CHAIN_MAX,
+ .policy = nft_chain_policy,
+ },
+ [NFT_MSG_GETCHAIN] = {
+ .call = nf_tables_getchain,
+ .type = NFNL_CB_RCU,
+ .attr_count = NFTA_CHAIN_MAX,
+ .policy = nft_chain_policy,
+ },
+ [NFT_MSG_DELCHAIN] = {
+ .call = nf_tables_delchain,
+ .type = NFNL_CB_BATCH,
+ .attr_count = NFTA_CHAIN_MAX,
+ .policy = nft_chain_policy,
+ },
+ [NFT_MSG_NEWRULE] = {
+ .call = nf_tables_newrule,
+ .type = NFNL_CB_BATCH,
+ .attr_count = NFTA_RULE_MAX,
+ .policy = nft_rule_policy,
+ },
+ [NFT_MSG_GETRULE] = {
+ .call = nf_tables_getrule,
+ .type = NFNL_CB_RCU,
+ .attr_count = NFTA_RULE_MAX,
+ .policy = nft_rule_policy,
+ },
+ [NFT_MSG_DELRULE] = {
+ .call = nf_tables_delrule,
+ .type = NFNL_CB_BATCH,
+ .attr_count = NFTA_RULE_MAX,
+ .policy = nft_rule_policy,
+ },
+ [NFT_MSG_NEWSET] = {
+ .call = nf_tables_newset,
+ .type = NFNL_CB_BATCH,
+ .attr_count = NFTA_SET_MAX,
+ .policy = nft_set_policy,
+ },
+ [NFT_MSG_GETSET] = {
+ .call = nf_tables_getset,
+ .type = NFNL_CB_RCU,
+ .attr_count = NFTA_SET_MAX,
+ .policy = nft_set_policy,
+ },
+ [NFT_MSG_DELSET] = {
+ .call = nf_tables_delset,
+ .type = NFNL_CB_BATCH,
+ .attr_count = NFTA_SET_MAX,
+ .policy = nft_set_policy,
+ },
+ [NFT_MSG_NEWSETELEM] = {
+ .call = nf_tables_newsetelem,
+ .type = NFNL_CB_BATCH,
+ .attr_count = NFTA_SET_ELEM_LIST_MAX,
+ .policy = nft_set_elem_list_policy,
+ },
+ [NFT_MSG_GETSETELEM] = {
+ .call = nf_tables_getsetelem,
+ .type = NFNL_CB_RCU,
+ .attr_count = NFTA_SET_ELEM_LIST_MAX,
+ .policy = nft_set_elem_list_policy,
+ },
+ [NFT_MSG_DELSETELEM] = {
+ .call = nf_tables_delsetelem,
+ .type = NFNL_CB_BATCH,
+ .attr_count = NFTA_SET_ELEM_LIST_MAX,
+ .policy = nft_set_elem_list_policy,
+ },
+ [NFT_MSG_GETGEN] = {
+ .call = nf_tables_getgen,
+ .type = NFNL_CB_RCU,
+ },
+ [NFT_MSG_NEWOBJ] = {
+ .call = nf_tables_newobj,
+ .type = NFNL_CB_BATCH,
+ .attr_count = NFTA_OBJ_MAX,
+ .policy = nft_obj_policy,
+ },
+ [NFT_MSG_GETOBJ] = {
+ .call = nf_tables_getobj,
+ .type = NFNL_CB_RCU,
+ .attr_count = NFTA_OBJ_MAX,
+ .policy = nft_obj_policy,
+ },
+ [NFT_MSG_DELOBJ] = {
+ .call = nf_tables_delobj,
+ .type = NFNL_CB_BATCH,
+ .attr_count = NFTA_OBJ_MAX,
+ .policy = nft_obj_policy,
+ },
+ [NFT_MSG_GETOBJ_RESET] = {
+ .call = nf_tables_getobj,
+ .type = NFNL_CB_RCU,
+ .attr_count = NFTA_OBJ_MAX,
+ .policy = nft_obj_policy,
+ },
+ [NFT_MSG_NEWFLOWTABLE] = {
+ .call = nf_tables_newflowtable,
+ .type = NFNL_CB_BATCH,
+ .attr_count = NFTA_FLOWTABLE_MAX,
+ .policy = nft_flowtable_policy,
+ },
+ [NFT_MSG_GETFLOWTABLE] = {
+ .call = nf_tables_getflowtable,
+ .type = NFNL_CB_RCU,
+ .attr_count = NFTA_FLOWTABLE_MAX,
+ .policy = nft_flowtable_policy,
+ },
+ [NFT_MSG_DELFLOWTABLE] = {
+ .call = nf_tables_delflowtable,
+ .type = NFNL_CB_BATCH,
+ .attr_count = NFTA_FLOWTABLE_MAX,
+ .policy = nft_flowtable_policy,
+ },
+ [NFT_MSG_NEWSTRSET] = {
+ .call = nf_tables_newstrset,
+ .type = NFNL_CB_BATCH,
+ .attr_count = NFTA_STRSET_MAX,
+ .policy = nft_strset_policy,
+ },
+ [NFT_MSG_GETSTRSET] = {
+ .call = nf_tables_getstrset,
+ .type = NFNL_CB_RCU,
+ .attr_count = NFTA_STRSET_MAX,
+ .policy = nft_strset_policy,
+ },
+ [NFT_MSG_DELSTRSET] = {
+ .call = nf_tables_delstrset,
+ .type = NFNL_CB_BATCH,
+ .attr_count = NFTA_STRSET_MAX,
+ .policy = nft_strset_policy,
+ },
+ [NFT_MSG_NEWSTRING] = {
+ .call = nf_tables_newstr,
+ .type = NFNL_CB_BATCH,
+ .attr_count = NFTA_STRSET_MAX,
+ .policy = nft_strset_policy,
+ },
+ [NFT_MSG_GETSTRING] = {
+ .call = nf_tables_getstr,
+ .type = NFNL_CB_RCU,
+ .attr_count = NFTA_STRSET_MAX,
+ .policy = nft_strset_policy,
+ },
+ [NFT_MSG_DELSTRING] = {
+ .call = nf_tables_delstr,
+ .type = NFNL_CB_BATCH,
+ .attr_count = NFTA_STRSET_MAX,
+ .policy = nft_strset_policy,
+ },
+};
+
+static int nf_tables_validate(struct net *net)
+{
+ struct nftables_pernet *nft_net = nft_pernet(net);
+ struct nft_table *table;
+
+ switch (nft_net->validate_state) {
+ case NFT_VALIDATE_SKIP:
+ break;
+ case NFT_VALIDATE_NEED:
+ nft_validate_state_update(net, NFT_VALIDATE_DO);
+ fallthrough;
+ case NFT_VALIDATE_DO:
+ list_for_each_entry(table, &nft_net->tables, list) {
+ if (nft_table_validate(net, table) < 0)
+ return -EAGAIN;
+ }
+ break;
+ }
+
+ return 0;
+}
+
+/* a drop policy has to be deferred until all rules have been activated,
+ * otherwise a large ruleset that contains a drop-policy base chain will
+ * cause all packets to get dropped until the full transaction has been
+ * processed.
+ *
+ * We defer the drop policy until the transaction has been finalized.
*/
static void nft_chain_commit_drop_policy(struct nft_trans *trans)
{
@@ -8388,6 +9192,9 @@ static void nft_commit_release(struct nft_trans *trans)
else
nf_tables_flowtable_destroy(nft_trans_flowtable(trans));
break;
+ case NFT_MSG_DELSTRSET:
+ nft_strset_destroy(&trans->ctx, nft_trans_strset(trans));
+ break;
}
if (trans->put_net)
@@ -8588,6 +9395,54 @@ static void nf_tables_commit_chain(struct net *net, struct nft_chain *chain)
nf_tables_commit_chain_free_rules_old(g0);
}
+static int nft_strset_commit_update(struct net *net)
+{
+ struct nftables_pernet *nft_net = nft_pernet(net);
+ u8 next_genbit = nft_gencursor_next(net);
+ struct nft_strset *strset;
+ struct nft_table *table;
+ struct ac_tree *tree;
+ int err;
+
+ list_for_each_entry(table, &nft_net->tables, list) {
+ list_for_each_entry(strset, &table->strsets, list) {
+ tree = rcu_dereference(strset->tree[next_genbit]);
+
+ if (!strset->commit_update) {
+ if (strset->tree[0] != strset->tree[1]) {
+ ac_destroy(tree);
+ rcu_assign_pointer(strset->tree[next_genbit],
+ strset->tree[!next_genbit]);
+ }
+ continue;
+ }
+
+ err = ac_resolve_fail(tree);
+ if (err < 0)
+ return err;
+
+ strset->commit_update = 0;
+ }
+ }
+
+ return 0;
+}
+
+static void nft_strset_abort_update(struct net *net)
+{
+ struct nftables_pernet *nft_net = nft_pernet(net);
+ struct nft_strset *strset;
+ struct nft_table *table;
+
+ list_for_each_entry(table, &nft_net->tables, list) {
+ list_for_each_entry(strset, &table->strsets, list) {
+ if (!strset->commit_update)
+ continue;
+ }
+ strset->commit_update = 0;
+ }
+}
+
static void nft_obj_del(struct nft_object *obj)
{
rhltable_remove(&nft_objname_ht, &obj->rhlhead, nft_objname_ht_params);
@@ -8766,6 +9621,10 @@ static int nf_tables_commit(struct net *net, struct sk_buff *skb)
if (nf_tables_validate(net) < 0)
return -EAGAIN;
+ err = nft_strset_commit_update(net);
+ if (err < 0)
+ return err;
+
err = nft_flow_rule_offload_commit(net);
if (err < 0)
return err;
@@ -8966,6 +9825,13 @@ static int nf_tables_commit(struct net *net, struct sk_buff *skb)
&nft_trans_flowtable(trans)->hook_list);
}
break;
+ case NFT_MSG_NEWSTRSET:
+ nft_clear(net, nft_trans_strset(trans));
+ nft_trans_destroy(trans);
+ break;
+ case NFT_MSG_DELSTRSET:
+ list_del_rcu(&nft_trans_strset(trans)->list);
+ break;
}
}
@@ -9021,6 +9887,9 @@ static void nf_tables_abort_release(struct nft_trans *trans)
else
nf_tables_flowtable_destroy(nft_trans_flowtable(trans));
break;
+ case NFT_MSG_NEWSTRSET:
+ nft_strset_destroy(&trans->ctx, nft_trans_strset(trans));
+ break;
}
kfree(trans);
}
@@ -9035,6 +9904,8 @@ static int __nf_tables_abort(struct net *net, enum nfnl_abort_action action)
nf_tables_validate(net) < 0)
return -EAGAIN;
+ nft_strset_abort_update(net);
+
list_for_each_entry_safe_reverse(trans, next, &nft_net->commit_list,
list) {
switch (trans->msg_type) {
@@ -9168,6 +10039,15 @@ static int __nf_tables_abort(struct net *net, enum nfnl_abort_action action)
}
nft_trans_destroy(trans);
break;
+ case NFT_MSG_NEWSTRSET:
+ trans->ctx.table->use--;
+ list_del_rcu(&nft_trans_strset(trans)->list);
+ break;
+ case NFT_MSG_DELSTRSET:
+ trans->ctx.table->use++;
+ nft_clear(trans->ctx.net, nft_trans_strset(trans));
+ nft_trans_destroy(trans);
+ break;
}
}
@@ -9833,6 +10713,7 @@ static void __nft_release_hooks(struct net *net)
static void __nft_release_table(struct net *net, struct nft_table *table)
{
struct nft_flowtable *flowtable, *nf;
+ struct nft_strset *strset, *nst;
struct nft_chain *chain, *nc;
struct nft_object *obj, *ne;
struct nft_rule *rule, *nr;
@@ -9873,6 +10754,11 @@ static void __nft_release_table(struct net *net, struct nft_table *table)
table->use--;
nf_tables_chain_destroy(&ctx);
}
+ list_for_each_entry_safe(strset, nst, &table->strsets, list) {
+ list_del(&strset->list);
+ table->use--;
+ nft_strset_destroy(&ctx, strset);
+ }
nf_tables_table_destroy(&ctx);
}
--
2.30.2
^ permalink raw reply related [flat|nested] 4+ messages in thread
* [PATCH RFC 3/3] netfilter: nf_tables: add string expression
2022-07-29 9:31 [PATCH RFC 0/3] nf_tables string match support Pablo Neira Ayuso
2022-07-29 9:31 ` [PATCH RFC 1/3] netfilter: add Aho-Corasick string match implementation Pablo Neira Ayuso
2022-07-29 9:31 ` [PATCH RFC 2/3] netfilter: nf_tables: add string set API Pablo Neira Ayuso
@ 2022-07-29 9:31 ` Pablo Neira Ayuso
2 siblings, 0 replies; 4+ messages in thread
From: Pablo Neira Ayuso @ 2022-07-29 9:31 UTC (permalink / raw)
To: netfilter-devel
Add string expression to allow to match patterns on packets. These
patterns are specified by an existing string set which uses the
Aho-Corasick tree to store patterns.
This expression allows for a packet base offset to be specified: network,
transport or inner header offset. On top of this, you can specify an offset
range to restrict the pattern search.
This patch uses the existing sequencial read skbuff API to search for
matching patterns.
Currently, it only supports for the NFT_STR_F_PRESENT flags which is
mandatory to test whether a pattern is matching the payload. This leaves
room to extend this expression to support for more new matching modes in
the future.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
include/net/netfilter/nf_tables.h | 4 +
include/net/netfilter/nf_tables_core.h | 1 +
include/uapi/linux/netfilter/nf_tables.h | 26 +++
net/netfilter/Makefile | 1 +
net/netfilter/nf_tables_api.c | 7 +-
net/netfilter/nf_tables_core.c | 1 +
net/netfilter/nft_string.c | 254 +++++++++++++++++++++++
7 files changed, 290 insertions(+), 4 deletions(-)
create mode 100644 net/netfilter/nft_string.c
diff --git a/include/net/netfilter/nf_tables.h b/include/net/netfilter/nf_tables.h
index c4e9a969122a..20e12a881246 100644
--- a/include/net/netfilter/nf_tables.h
+++ b/include/net/netfilter/nf_tables.h
@@ -1365,6 +1365,10 @@ struct nft_strset {
u32 use;
};
+struct nft_strset *nft_strset_lookup(const struct net *net,
+ struct nft_table *table,
+ const struct nlattr *nla, u8 genmask);
+
/**
* struct nft_traceinfo - nft tracing information and state
*
diff --git a/include/net/netfilter/nf_tables_core.h b/include/net/netfilter/nf_tables_core.h
index 0ea7c55cea4d..f5a4bef2afba 100644
--- a/include/net/netfilter/nf_tables_core.h
+++ b/include/net/netfilter/nf_tables_core.h
@@ -12,6 +12,7 @@ extern struct nft_expr_type nft_lookup_type;
extern struct nft_expr_type nft_bitwise_type;
extern struct nft_expr_type nft_byteorder_type;
extern struct nft_expr_type nft_payload_type;
+extern struct nft_expr_type nft_string_type;
extern struct nft_expr_type nft_dynset_type;
extern struct nft_expr_type nft_range_type;
extern struct nft_expr_type nft_meta_type;
diff --git a/include/uapi/linux/netfilter/nf_tables.h b/include/uapi/linux/netfilter/nf_tables.h
index d7a668199611..8ce2400952c0 100644
--- a/include/uapi/linux/netfilter/nf_tables.h
+++ b/include/uapi/linux/netfilter/nf_tables.h
@@ -1711,6 +1711,32 @@ enum nft_string_hook_attributes {
};
#define NFTA_STRING_MAX (__NFTA_STRING_MAX - 1)
+enum nft_string_flags {
+ NFT_STR_F_PRESENT = (1 << 0),
+};
+
+/**
+ * enum nft_str_hook_attributes - nf_tables string expression netlink attributes
+ *
+ * @NFTA_STR_NAME: string set name (NLA_STRING)
+ * @NFTA_STR_BASE: payload base (NLA_U32: nft_payload_bases)
+ * @NFTA_STR_FROM: offset to start matching from (NLA_U32)
+ * @NFTA_STR_TO: offset to end matching to (NLA_U32)
+ * @NFTA_STR_DREG: destination register (NLA_U32: nft_registers)
+ * @NFTA_STR_FLAGS: flags (NLA_U32)
+ */
+enum nft_str_hook_attributes {
+ NFTA_STR_UNSPEC,
+ NFTA_STR_NAME,
+ NFTA_STR_BASE,
+ NFTA_STR_FROM,
+ NFTA_STR_TO,
+ NFTA_STR_DREG,
+ NFTA_STR_FLAGS,
+ __NFTA_STR_MAX
+};
+#define NFTA_STR_MAX (__NFTA_STR_MAX - 1)
+
/**
* enum nft_osf_attributes - nftables osf expression netlink attributes
*
diff --git a/net/netfilter/Makefile b/net/netfilter/Makefile
index 18127c83b88c..47ed19be4ac6 100644
--- a/net/netfilter/Makefile
+++ b/net/netfilter/Makefile
@@ -91,6 +91,7 @@ endif
endif
nf_tables-$(CONFIG_NFT_STRING) += ahocorasick.o
+nf_tables-$(CONFIG_NFT_STRING) += nft_string.o
obj-$(CONFIG_NF_TABLES) += nf_tables.o
obj-$(CONFIG_NFT_COMPAT) += nft_compat.o
diff --git a/net/netfilter/nf_tables_api.c b/net/netfilter/nf_tables_api.c
index 736bdf6b9671..854ee78884aa 100644
--- a/net/netfilter/nf_tables_api.c
+++ b/net/netfilter/nf_tables_api.c
@@ -8117,10 +8117,9 @@ static const struct nla_policy nft_strset_policy[NFTA_STRSET_MAX + 1] = {
[NFTA_STRSET_HANDLE] = { .type = NLA_U64 },
};
-static struct nft_strset *nft_strset_lookup(const struct net *net,
- struct nft_table *table,
- const struct nlattr *nla,
- u8 genmask)
+struct nft_strset *nft_strset_lookup(const struct net *net,
+ struct nft_table *table,
+ const struct nlattr *nla, u8 genmask)
{
struct nft_strset *strset;
diff --git a/net/netfilter/nf_tables_core.c b/net/netfilter/nf_tables_core.c
index 3ddce24ac76d..ffa413378d7a 100644
--- a/net/netfilter/nf_tables_core.c
+++ b/net/netfilter/nf_tables_core.c
@@ -336,6 +336,7 @@ static struct nft_expr_type *nft_basic_types[] = {
&nft_payload_type,
&nft_dynset_type,
&nft_range_type,
+ &nft_string_type,
&nft_meta_type,
&nft_rt_type,
&nft_exthdr_type,
diff --git a/net/netfilter/nft_string.c b/net/netfilter/nft_string.c
new file mode 100644
index 000000000000..e3bf4be59c7c
--- /dev/null
+++ b/net/netfilter/nft_string.c
@@ -0,0 +1,254 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (c) 2022 Pablo Neira Ayuso <pablo@netfilter.org>
+ */
+
+#include <linux/kernel.h>
+#include <linux/netlink.h>
+#include <linux/netfilter.h>
+#include <linux/netfilter/nf_tables.h>
+#include <net/netfilter/nf_tables.h>
+#include <net/netfilter/nf_tables_core.h>
+#include <net/netfilter/ahocorasick.h>
+#include <linux/tcp.h>
+#include <linux/udp.h>
+
+struct nft_string {
+ struct nft_strset *strset;
+ u8 dreg;
+ u8 base;
+ u8 flags;
+ u32 from;
+ u32 to;
+};
+
+static int __nft_payload_inner_offset(struct nft_pktinfo *pkt)
+{
+ unsigned int thoff = nft_thoff(pkt);
+
+ if (!(pkt->flags & NFT_PKTINFO_L4PROTO) || pkt->fragoff)
+ return -1;
+
+ switch (pkt->tprot) {
+ case IPPROTO_UDP:
+ pkt->inneroff = thoff + sizeof(struct udphdr);
+ break;
+ case IPPROTO_TCP: {
+ struct tcphdr *th, _tcph;
+
+ th = skb_header_pointer(pkt->skb, thoff, sizeof(_tcph), &_tcph);
+ if (!th)
+ return -1;
+
+ pkt->inneroff = thoff + __tcp_hdrlen(th);
+ }
+ break;
+ default:
+ return -1;
+ }
+
+ pkt->flags |= NFT_PKTINFO_INNER;
+
+ return 0;
+}
+
+static int nft_payload_inner_offset(const struct nft_pktinfo *pkt)
+{
+ if (!(pkt->flags & NFT_PKTINFO_INNER) &&
+ __nft_payload_inner_offset((struct nft_pktinfo *)pkt) < 0)
+ return -1;
+
+ return pkt->inneroff;
+}
+
+static int nft_string_base(const struct nft_string *priv,
+ const struct nft_pktinfo *pkt)
+{
+ int base_offset;
+
+ switch (priv->base) {
+ case NFT_PAYLOAD_NETWORK_HEADER:
+ base_offset = skb_network_offset(pkt->skb);
+ break;
+ case NFT_PAYLOAD_TRANSPORT_HEADER:
+ base_offset = skb_transport_offset(pkt->skb);
+ break;
+ case NFT_PAYLOAD_INNER_HEADER:
+ if (nft_payload_inner_offset(pkt))
+ base_offset = pkt->inneroff;
+ else
+ base_offset = 0;
+ break;
+ default:
+ WARN_ON_ONCE(1);
+ break;
+ }
+
+ if (WARN_ON_ONCE(base_offset < 0))
+ base_offset = 0;
+
+ return base_offset;
+}
+
+static void nft_string_eval(const struct nft_expr *expr,
+ struct nft_regs *regs,
+ const struct nft_pktinfo *pkt)
+{
+ const struct nft_string *priv = nft_expr_priv(expr);
+ bool genbit = READ_ONCE(nft_net(pkt)->nft.gencursor);
+ u32 *dest = ®s->data[priv->dreg];
+ unsigned int consumed = 0, len;
+ struct sk_buff *skb = pkt->skb;
+ struct ac_tree __rcu *tree;
+ struct skb_seq_state state;
+ int base_offset, pos = -1;
+ const u8 *data;
+ u32 from, to;
+
+ base_offset = nft_string_base(priv, pkt);
+ if (priv->from + base_offset < priv->from)
+ from = priv->from;
+ else
+ from = priv->from + base_offset;
+
+ if (priv->to + base_offset < priv->to)
+ to = priv->to;
+ else
+ to = priv->to + base_offset;
+
+ skb_prepare_seq_read(skb, from, to, &state);
+
+ tree = rcu_dereference(priv->strset->tree[genbit]);
+
+ len = skb_seq_read(consumed, &data, &state);
+ while (len > 0) {
+ pos = ac_find(tree, data, len);
+ if (pos >= 0)
+ break;
+
+ consumed += len;
+ len = skb_seq_read(consumed, &data, &state);
+ }
+ skb_abort_seq_read(&state);
+
+ if (pos < 0)
+ *dest = 0;
+ else
+ *dest = 1;
+}
+
+static const struct nla_policy nft_string_policy[NFTA_STR_MAX + 1] = {
+ [NFTA_STR_DREG] = { .type = NLA_U32 },
+ [NFTA_STR_NAME] = { .type = NLA_STRING },
+ [NFTA_STR_BASE] = { .type = NLA_U32 },
+ [NFTA_STR_FROM] = { .type = NLA_U32 },
+ [NFTA_STR_TO] = { .type = NLA_U32 },
+ [NFTA_STR_FLAGS] = { .type = NLA_U32 },
+};
+
+static int nft_string_init(const struct nft_ctx *ctx,
+ const struct nft_expr *expr,
+ const struct nlattr * const tb[])
+{
+ struct nft_string *priv = nft_expr_priv(expr);
+ u8 genmask = nft_genmask_next(ctx->net);
+ uint32_t from, to, flags, base;
+ struct nft_strset *strset;
+ int err;
+
+ if (!tb[NFTA_STR_NAME] ||
+ !tb[NFTA_STR_BASE] ||
+ !tb[NFTA_STR_FROM] ||
+ !tb[NFTA_STR_TO] ||
+ !tb[NFTA_STR_DREG] ||
+ !tb[NFTA_STR_FLAGS])
+ return -EINVAL;
+
+ from = ntohl(nla_get_be32(tb[NFTA_STR_FROM]));
+ to = ntohl(nla_get_be32(tb[NFTA_STR_TO]));
+ if (from > to)
+ return -EINVAL;
+
+ flags = ntohl(nla_get_be32(tb[NFTA_STR_FLAGS]));
+ if (flags & ~NFT_STR_F_PRESENT)
+ return -EOPNOTSUPP;
+ else if (flags == 0) /* always NFT_STR_F_PRESENT set on by now. */
+ return -EOPNOTSUPP;
+
+ base = ntohl(nla_get_be32(tb[NFTA_STR_BASE]));
+ switch (base) {
+ case NFT_PAYLOAD_NETWORK_HEADER:
+ case NFT_PAYLOAD_TRANSPORT_HEADER:
+ case NFT_PAYLOAD_INNER_HEADER:
+ break;
+ default:
+ return -EOPNOTSUPP;
+ }
+ priv->base = base;
+
+ err = nft_parse_register_store(ctx, tb[NFTA_STR_DREG], &priv->dreg,
+ NULL, NFT_DATA_VALUE, sizeof(u32));
+ if (err < 0)
+ return err;
+
+ strset = nft_strset_lookup(ctx->net, ctx->table, tb[NFTA_STR_NAME],
+ genmask);
+ if (IS_ERR(strset))
+ return PTR_ERR(strset);
+
+ strset->use++;
+
+ priv->strset = strset;
+ priv->from = from;
+ priv->to = to;
+
+ return 0;
+}
+
+static void nft_string_destroy(const struct nft_ctx *ctx,
+ const struct nft_expr *expr)
+{
+ const struct nft_string *priv = nft_expr_priv(expr);
+
+ priv->strset->use--;
+}
+
+static int nft_string_dump(struct sk_buff *skb,
+ const struct nft_expr *expr)
+{
+ const struct nft_string *priv = nft_expr_priv(expr);
+
+ if (nla_put_string(skb, NFTA_STR_NAME, priv->strset->name))
+ goto nla_put_failure;
+ if (nla_put_be32(skb, NFTA_STR_BASE, htonl(priv->base)))
+ goto nla_put_failure;
+ if (nla_put_be32(skb, NFTA_STR_FROM, htonl(priv->from)))
+ goto nla_put_failure;
+ if (nla_put_be32(skb, NFTA_STR_TO, htonl(priv->to)))
+ goto nla_put_failure;
+ if (nla_put_be32(skb, NFTA_STR_FLAGS, htonl(priv->flags)))
+ goto nla_put_failure;
+ if (nft_dump_register(skb, NFTA_STR_DREG, priv->dreg))
+ goto nla_put_failure;
+ return 0;
+
+nla_put_failure:
+ return -1;
+}
+
+static const struct nft_expr_ops nft_string_get_ops = {
+ .type = &nft_string_type,
+ .size = NFT_EXPR_SIZE(sizeof(struct nft_string)),
+ .eval = nft_string_eval,
+ .init = nft_string_init,
+ .destroy = nft_string_destroy,
+ .dump = nft_string_dump,
+};
+
+struct nft_expr_type nft_string_type __read_mostly = {
+ .name = "string",
+ .ops = &nft_string_get_ops,
+ .policy = nft_string_policy,
+ .maxattr = NFTA_STR_MAX,
+ .owner = THIS_MODULE,
+};
--
2.30.2
^ permalink raw reply related [flat|nested] 4+ messages in thread
end of thread, other threads:[~2022-07-29 9:31 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-07-29 9:31 [PATCH RFC 0/3] nf_tables string match support Pablo Neira Ayuso
2022-07-29 9:31 ` [PATCH RFC 1/3] netfilter: add Aho-Corasick string match implementation Pablo Neira Ayuso
2022-07-29 9:31 ` [PATCH RFC 2/3] netfilter: nf_tables: add string set API Pablo Neira Ayuso
2022-07-29 9:31 ` [PATCH RFC 3/3] netfilter: nf_tables: add string expression Pablo Neira Ayuso
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).