cache updates
Pablo Neira Ayuso
maintainerv2 cover cache updates
Hi,
The following patchset contains cache updates for nft:
Patch #1 resets filtering for each new command
Patch #2 accumulates cache flags for each command, recent patches are
relaxing cache requirements which could uncover bugs.
Patch #3 updates objects to use the netlink dump filtering infrastructure
to build the cache (
Patch #4 only dumps rules for the given table
Patch #5 updates reset commands to use the cache infrastructure
Patch #6 and #7 extend tests coverage for reset commands.
Pablo Neira Ayuso (7):
cache: reset filter for each command
cache: accumulate flags in batch
cache: add filtering support for objects
cache: only dump rules for the given table
cache: consolidate reset command
tests: shell: cover anonymous set with reset command
tests: shell: cover reset command with counter and quota
include/cache.h | 12 +- include/netlink.h | 5 - src/cache.c | 201 ++++++++++++++---- src/evaluate.c | 2 + src/mnl.c | 7 +- src/netlink.c | 78 ------- src/parser_bison.y | 8 +- src/rule.c | 48 +---- tests/shell/testcases/listing/reset_objects | 104 +++++++++ .../testcases/rule_management/0011reset_0 | 31 ++- 10 files changed, 305 insertions(+), 191 deletions(-) create mode 100755 tests/shell/testcases/listing/reset_objects
Signature · 1 line
2.30.2
Pablo Neira Ayuso
maintainerPATCH v2 1/7 cache: reset filter for each command
Inconditionally reset filter for each command in the batch, this is safer.
---
v2: new in this series
src/cache.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-)
src/cache.c +6 -2
@@ -400,6 +400,11 @@ err_name_too_long:return -1 ; } +static void reset_filter (struct nft_cache_filter *filter) +{ +memset (&filter->list,0 ,sizeof (filter->list)); +} +int nft_cache_evaluate (struct nft_ctx *nft,struct list_head *cmds,struct list_head *msgs,struct nft_cache_filter *filter,unsigned int *pflags)
@@ -411,8 +416,7 @@ int nft_cache_evaluate(struct nft_ctx *nft, struct list_head *cmds,if (nft_handle_validate (cmd, msgs) <0 )return -1 ; -if (filter->list.table && cmd->op !=CMD_LIST ) -memset (&filter->list,0 ,sizeof (filter->list)); +reset_filter (filter);switch (cmd->op) {case CMD_ADD :
Signature · 1 line
2.30.2
Pablo Neira Ayuso
maintainerPATCH v2 2/7 cache: accumulate flags in batch
In reply to Pablo Neira AyusoRecent updates are relaxing cache requirements:
babc6ee8773c ("cache: populate chains on demand from error path")
Flags describe cache requirements for a given batch, accumulate flags that are inferred from commands in this batch.
---
v2: new in this series
src/cache.c | 12 +++++++----- 1 file changed, 7 insertions(+), 5 deletions(-)
src/cache.c +7 -5
@@ -409,13 +409,14 @@ int nft_cache_evaluate(struct nft_ctx *nft, struct list_head *cmds, struct list_head *msgs, struct nft_cache_filter *filter, unsigned int *pflags) { - unsigned int flags = NFT_CACHE_EMPTY; + unsigned int flags, batch_flags = NFT_CACHE_EMPTY;struct cmd *cmd;list_for_each_entry (cmd, cmds, list) {if (nft_handle_validate (cmd, msgs) <0 )return -1 ; + flags =NFT_CACHE_EMPTY ;reset_filter (filter);switch (cmd->op) {
@@ -439,13 +440,13 @@ int nft_cache_evaluate(struct nft_ctx *nft, struct list_head *cmds, flags =evaluate_cache_get (cmd, flags);break ;case CMD_RESET : - flags |=evaluate_cache_reset (cmd, flags, filter); + flags =evaluate_cache_reset (cmd, flags, filter);break ;case CMD_LIST : - flags |=evaluate_cache_list (nft, cmd, flags, filter); + flags =evaluate_cache_list (nft, cmd, flags, filter);break ;case CMD_MONITOR : - flags |=NFT_CACHE_FULL ; + flags =NFT_CACHE_FULL ;break ;case CMD_FLUSH : flags =evaluate_cache_flush (cmd, flags, filter);
@@ -460,8 +461,9 @@ int nft_cache_evaluate(struct nft_ctx *nft, struct list_head *cmds,default :break ; } + batch_flags |= flags; } - *pflags = flags; + *pflags = batch_flags;return 0 ; }
Signature · 1 line
2.30.2
Pablo Neira Ayuso
maintainerPATCH v2 3/7 cache: add filtering support for objects
In reply to Pablo Neira AyusoCurrently, full ruleset flag is set on to fetch objects.
Follow a similar approach to these patches from Phil:
de961b930660 ("cache: Filter set list on server side") and cb4b07d0b628 ("cache: Support filtering for a specific flowtable")
in preparation to update the reset command to use the cache infrastructure.
---
v2: remove check for if (filter) when evaluating cache requirements,
filter must be present in reset command, add assert().
include/cache.h | 2 + src/cache.c | 100 +++++++++++++++++++++++++++++++++++++++++------- 2 files changed, 89 insertions(+), 13 deletions(-)
include/cache.h +2 -0
@@ -54,8 +54,10 @@ struct nft_cache_filter {uint32_t family;const char *table;const char *chain; +const char *obj;const char *set;const char *ft; +int obj_type;uint64_t rule_handle; } list;
src/cache.c +87 -13
162 lines not on this page · raw message
Signature · 1 line
2.30.2
Pablo Neira Ayuso
maintainerPATCH v2 4/7 cache: only dump rules for the given table
In reply to Pablo Neira AyusoOnly family is set on in the dump request, set on table and chain otherwise, rules for the given family are fetched for each existing table.
---
v2: no changes
src/cache.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
src/cache.c +1 -1
@@ -694,7 +694,7 @@ int rule_cache_dump(struct netlink_ctx *ctx, const struct handle *h, bool dump, bool reset) {struct nftnl_rule_list *rule_cache; -const char *table =NULL ; +const char *table = h->table.name;const char *chain =NULL ;uint64_t rule_handle =0 ;
Signature · 1 line
2.30.2
Pablo Neira Ayuso
maintainerPATCH v2 5/7 cache: consolidate reset command
In reply to Pablo Neira AyusoReset command does not utilize the cache infrastructure.
This implicitly fixes a crash with anonymous sets because elements are not fetched. I initially tried to fix it by toggling the missing cache flags, but then ASAN reports memleaks.
To address these issues relies on Phil's list filtering infrastructure which updates is expanded to accomodate filtering requirements of the reset commands, such as 'reset table ip' where only the family is sent to the kernel.
After this update, tests/shell reports a few inconsistencies between reset and list commands:
- reset rules chain t c2
display sets, but it should only list the given chain.
- reset rules table t
reset rules ip
do not list elements in the set. In both cases, these are fully listing a given table and family, elements should be included.
The consolidation also ensures list and reset will not differ.
A few more notes:
- CMD_OBJ_TABLE is used for:
rules family table
from the parser, due to the lack of a better enum, same applies to CMD_OBJ_CHAIN.
- CMD_OBJ_ELEMENTS still does not use the cache, but same occurs in the CMD_GET command case which needs to be consolidated.
---
v2: no changes
include/cache.h | 10 ++- include/netlink.h | 5 -- src/cache.c | 81 +++++++++++++------ src/evaluate.c | 2 + src/mnl.c | 7 +- src/netlink.c | 78 ------------------ src/parser_bison.y | 8 +- src/rule.c | 48 +---------- .../testcases/rule_management/0011reset_0 | 10 +-- 9 files changed, 78 insertions(+), 171 deletions(-)
include/cache.h +6 -4
@@ -64,6 +64,12 @@ struct nft_cache_filter {struct {struct list_head head; } obj[NFT_CACHE_HSIZE ]; + +struct { +bool obj; +bool rule; +bool elem; + } reset; };struct nft_cache ;
@@ -149,8 +155,4 @@ struct netlink_ctx;void nft_chain_cache_update (struct netlink_ctx *ctx,struct table *table,const char *chain); -int rule_cache_dump (struct netlink_ctx *ctx,const struct handle *h, -const struct nft_cache_filter *filter, -bool dump,bool reset); -#endif /* _NFT_CACHE_H_ */
include/netlink.h +0 -5
@@ -176,8 +176,6 @@ extern int netlink_delinearize_setelem(struct nftnl_set_elem *nlse, struct nft_cache *cache);extern int netlink_list_objs (struct netlink_ctx *ctx,const struct handle *h); -extern int netlink_reset_objs (struct netlink_ctx *ctx,const struct cmd *cmd, -uint32_t type,bool dump);extern struct obj *netlink_delinearize_obj (struct netlink_ctx *ctx,struct nftnl_obj *nlo);
@@ -186,9 +184,6 @@ extern int netlink_list_flowtables(struct netlink_ctx *ctx,extern struct flowtable *netlink_delinearize_flowtable (struct netlink_ctx *ctx,struct nftnl_flowtable *nlo); -extern int netlink_reset_rules (struct netlink_ctx *ctx,const struct cmd *cmd, -bool dump); -extern void netlink_dump_chain (const struct nftnl_chain *nlc,struct netlink_ctx *ctx); extern void netlink_dump_rule(const struct nftnl_rule *nlr,
src/cache.c +57 -24
141 lines not on this page · raw message
src/evaluate.c +2 -0
@@ -5848,6 +5848,8 @@ static int cmd_evaluate_reset(struct eval_ctx *ctx, struct cmd *cmd)return 0 ;case CMD_OBJ_ELEMENTS :return setelem_evaluate (ctx, cmd); +case CMD_OBJ_TABLE : +case CMD_OBJ_CHAIN :case CMD_OBJ_SET :case CMD_OBJ_MAP :return cmd_evaluate_list (ctx, cmd);
src/mnl.c +5 -2
@@ -1163,8 +1163,11 @@ struct nftnl_table_list *mnl_nft_table_dump(struct netlink_ctx *ctx,if (!nlt)memory_allocation_error (); -nftnl_table_set_u32 (nlt,NFTNL_TABLE_FAMILY , family); -nftnl_table_set_str (nlt,NFTNL_TABLE_NAME , table); +if (family !=NFPROTO_UNSPEC ) +nftnl_table_set_u32 (nlt,NFTNL_TABLE_FAMILY , family); +if (table) +nftnl_table_set_str (nlt,NFTNL_TABLE_NAME , table); + flags =NLM_F_ACK ; }
src/netlink.c +0 -78
84 lines not on this page · raw message
src/parser_bison.y +4 -4
@@ -1757,21 +1757,21 @@ reset_cmd : COUNTERS ruleset_spec } | RULES table_spec { - $$ = cmd_alloc(CMD_RESET, CMD_OBJ_RULES, &$2, &@$, NULL); + $$ = cmd_alloc(CMD_RESET, CMD_OBJ_TABLE, &$2, &@$, NULL); } | RULES TABLE table_spec { /* alias of previous rule. */ - $$ = cmd_alloc(CMD_RESET, CMD_OBJ_RULES, &$3, &@$, NULL); + $$ = cmd_alloc(CMD_RESET, CMD_OBJ_TABLE, &$3, &@$, NULL); } | RULES chain_spec { - $$ = cmd_alloc(CMD_RESET, CMD_OBJ_RULES, &$2, &@$, NULL); + $$ = cmd_alloc(CMD_RESET, CMD_OBJ_CHAIN, &$2, &@$, NULL); } | RULES CHAIN chain_spec { /* alias of previous rule. */ - $$ = cmd_alloc(CMD_RESET, CMD_OBJ_RULES, &$3, &@$, NULL); + $$ = cmd_alloc(CMD_RESET, CMD_OBJ_CHAIN, &$3, &@$, NULL); } | RULE ruleid_spec {
src/rule.c +1 -47
59 lines not on this page · raw message
tests/shell/testcases/rule_management/0011reset_0 +3 -7
@@ -74,13 +74,6 @@ $DIFF -u <(echo "$EXPECT") <($NFT list ruleset) echo "resetting specific chain" EXPECT='table ip t { - set s { - type ipv4_addr - size 65535 - flags dynamic - counter - } - chain c2 { counter packets 3 bytes 13 accept counter packets 4 bytes 14 drop
@@ -95,6 +88,7 @@ EXPECT='table ip t { size 65535 flags dynamic counter + elements = { 1.1.1.1 counter packets 1 bytes 11 } } chain c {
@@ -116,6 +110,7 @@ EXPECT='table ip t { size 65535 flags dynamic counter + elements = { 1.1.1.1 counter packets 1 bytes 11 } } chain c {
@@ -143,6 +138,7 @@ EXPECT='table ip t { size 65535 flags dynamic counter + elements = { 1.1.1.1 counter packets 1 bytes 11 } } chain c {
Signature · 1 line
2.30.2
Phil Sutter
reviewerRe: [PATCH nft,v2 5/7] cache: consolidate reset command
Hi Pablo,
Quoted 1 line · Pablo Neira Ayuso
Reset command does not utilize the cache infrastructure.
This commit changes audit log output for some reason. At least I see tools/testing/selftests/net/netfilter/nft_audit.sh failing and git bisect pointed at it. The relevant kselftest output is:
# testing for cmd: nft reset rules ... FAIL # table=t1 family=2 entries=3 op=nft_reset_rule # table=t2 family=2 entries=3 op=nft_reset_rule # table=t2 family=2 entries=3 op=nft_reset_rule # -table=t2 family=2 entries=180 op=nft_reset_rule # +table=t2 family=2 entries=186 op=nft_reset_rule # table=t2 family=2 entries=188 op=nft_reset_rule # -table=t2 family=2 entries=135 op=nft_reset_rule # +table=t2 family=2 entries=129 op=nft_reset_rule
I don't know why entries value changes and whether it is expected or not. Could you perhaps have a look?
Cheers, Phil
Pablo Neira Ayuso
maintainerRe: [PATCH nft,v2 5/7] cache: consolidate reset command
Quoted 19 lines · Phil Sutter
Hi Pablo,
Quoted 1 line · Pablo Neira Ayuso
Reset command does not utilize the cache infrastructure.
This commit changes audit log output for some reason. At least I see
tools/testing/selftests/net/netfilter/nft_audit.shfailing and git bisect pointed at it. The relevant kselftest output is:# testing for cmd: nft reset rules ... FAIL # table=t1 family=2 entries=3 op=nft_reset_rule # table=t2 family=2 entries=3 op=nft_reset_rule # table=t2 family=2 entries=3 op=nft_reset_rule # -table=t2 family=2 entries=180 op=nft_reset_rule # +table=t2 family=2 entries=186 op=nft_reset_rule # table=t2 family=2 entries=188 op=nft_reset_rule # -table=t2 family=2 entries=135 op=nft_reset_rule # +table=t2 family=2 entries=129 op=nft_reset_rule
Quoted 2 lines
I don't know why entries value changes and whether it is expected or not. Could you perhaps have a look?
Before my patch, there is a single dump request to the kernel:
rule_cache_dump table (null) chain (null) rule_handle 0 dump 1 reset 1
the skbuff already contains 6 entries for t1/c1 and t1/c2, which why 180 entries of t2 fit into the skbuff is delivered to userspace:
table=t2 family=2 entries=180 op=nft_reset_rule
(it seems 186 rule entries can fit into the skbuff).
after my patch, there is one for each table:
rule_cache_dump table t1 chain (null) rule_handle 0 dump 1 reset 1 rule_cache_dump table t2 chain (null) rule_handle 0 dump 1 reset 1
the skbuff is empty when dumping t2, so we can fit 6 more entries:
table=t2 family=2 entries=186 op=nft_reset_rule
If you take the number, before patch:
180+188+135=503
after patch:
186+188+129=503
show that is the same number of entries. Behaviour is correct.
I don't know how to fix this test without removing the check for 'reset rules', because it will break between different nftables versions (due to the different strategies to dump all rules vs. dump rules per table).
And I don't think it makes sense to revert my userspace update just to make this test happy.
Phil Sutter
reviewerRe: [PATCH nft,v2 5/7] cache: consolidate reset command
Quoted 45 lines · Pablo Neira Ayuso
Quoted 19 lines · Phil Sutter
Hi Pablo,
Quoted 1 line · Pablo Neira Ayuso
Reset command does not utilize the cache infrastructure.
This commit changes audit log output for some reason. At least I see
tools/testing/selftests/net/netfilter/nft_audit.shfailing and git bisect pointed at it. The relevant kselftest output is:# testing for cmd: nft reset rules ... FAIL # table=t1 family=2 entries=3 op=nft_reset_rule # table=t2 family=2 entries=3 op=nft_reset_rule # table=t2 family=2 entries=3 op=nft_reset_rule # -table=t2 family=2 entries=180 op=nft_reset_rule # +table=t2 family=2 entries=186 op=nft_reset_rule # table=t2 family=2 entries=188 op=nft_reset_rule # -table=t2 family=2 entries=135 op=nft_reset_rule # +table=t2 family=2 entries=129 op=nft_reset_ruleQuoted 2 lines
I don't know why entries value changes and whether it is expected or not. Could you perhaps have a look?
Before my patch, there is a single dump request to the kernel:
rule_cache_dumptable (null) chain (null)rule_handle0 dump 1 reset 1the skbuff already contains 6 entries for t1/c1 and t1/c2, which why 180 entries of t2 fit into the skbuff is delivered to userspace:
table=t2 family=2 entries=180 op=nft_reset_rule
(it seems 186 rule entries can fit into the skbuff).
after my patch, there is one for each table:
rule_cache_dumptable t1 chain (null)rule_handle0 dump 1 reset 1rule_cache_dumptable t2 chain (null)rule_handle0 dump 1 reset 1the skbuff is empty when dumping t2, so we can fit 6 more entries:
table=t2 family=2 entries=186 op=nft_reset_rule
If you take the number, before patch:
Quoted 7 lines
180+188+135=503
after patch:
186+188+129=503
show that is the same number of entries. Behaviour is correct.
Thanks for investigating!
Quoted 7 lines
I don't know how to fix this test without removing the check for 'reset rules', because it will break between different nftables versions (due to the different strategies to dump all rules vs. dump rules per table).
And I don't think it makes sense to revert my userspace update just to make this test happy.
Maybe a valid measure is to sum up entries there to make sure the total remains the same (just like you did above). I'll have a look at making the test tolerant to both variants.
Thanks, Phil
Pablo Neira Ayuso
maintainerPATCH v2 6/7 tests: shell: cover anonymous set with reset command
In reply to Pablo Neira AyusoExtend existing test to reset counters for rules with anonymous set.
---
v2: no changes
.../testcases/rule_management/0011reset_0 | 21 +++++++++++++++++++ 1 file changed, 21 insertions(+)
tests/shell/testcases/rule_management/0011reset_0 +21 -0
@@ -4,6 +4,27 @@ set -e +echo "loading ruleset with anonymous set" +$NFT -f - <<EOF +table t { + chain dns-nat-pre { + type nat hook prerouting priority filter; policy accept; + meta l4proto { tcp, udp } th dport 53 ip saddr 10.24.0.0/24 ip daddr != 10.25.0.1 counter packets 1000 bytes 1000 dnat to 10.25.0.1 + } +} +EOF + +echo "resetting ruleset with anonymous set" +$NFT reset rules +EXPECT='table ip t { + chain dns-nat-pre { + type nat hook prerouting priority filter; policy accept; + meta l4proto { tcp, udp } th dport 53 ip saddr 10.24.0.0/24 ip daddr != 10.25.0.1 counter packets 0 bytes 0 dnat to 10.25.0.1 + } +}' +$DIFF -u <(echo "$EXPECT") <($NFT list ruleset) +$NFT flush ruleset + echo "loading ruleset" $NFT -f - <<EOF table ip t {
Signature · 1 line
2.30.2
Pablo Neira Ayuso
maintainerPATCH v2 7/7 tests: shell: cover reset command with counter and quota
In reply to Pablo Neira Ayuso---
v2: no changes
tests/shell/testcases/listing/reset_objects | 104 ++++++++++++++++++++ 1 file changed, 104 insertions(+) create mode 100755 tests/shell/testcases/listing/reset_objects
tests/shell/testcases/listing/reset_objects new file +104 -0
@@ -0,0 +1,104 @@ +#!/bin/bash + +set -e + +load_ruleset() +{ + $NFT -f - <<EOF +table ip test { + quota https-quota { + 25 mbytes used 10 mbytes + } + counter https-counter { + packets 10 bytes 4096 + } +} +EOF +} + +check_list_quota() +{ + EXPECT="table ip test { + quota https-quota { + 25 mbytes + } +}" + $DIFF -u <(echo "$EXPECT") <($NFT list quotas) +} + +check_list_counter() +{ + EXPECT="table ip test { + counter https-counter { + packets 0 bytes 0 + } +}" + $DIFF -u <(echo "$EXPECT") <($NFT list counters) +} + +load_ruleset + +EXPECT="table ip test { + quota https-quota { + 25 mbytes used 10 mbytes + } +}" +$DIFF -u <(echo "$EXPECT") <($NFT reset quotas) + +check_list_quota +$NFT flush ruleset +load_ruleset + +EXPECT="table ip test { + quota https-quota { + 25 mbytes used 10 mbytes + } +}" +$DIFF -u <(echo "$EXPECT") <($NFT reset quotas ip) + +check_list_quota +$NFT flush ruleset +load_ruleset + +EXPECT="table ip test { + quota https-quota { + 25 mbytes used 10 mbytes + } +}" +$DIFF -u <(echo "$EXPECT") <($NFT reset quota ip test https-quota) + +check_list_quota +$NFT flush ruleset +load_ruleset + +EXPECT="table ip test { + counter https-counter { + packets 10 bytes 4096 + } +}" +$DIFF -u <(echo "$EXPECT") <($NFT reset counters) + +check_list_counter +$NFT flush ruleset +load_ruleset + +EXPECT="table ip test { + counter https-counter { + packets 10 bytes 4096 + } +}" +$DIFF -u <(echo "$EXPECT") <($NFT reset counters ip) + +check_list_counter +$NFT flush ruleset +load_ruleset + +EXPECT="table ip test { + counter https-counter { + packets 10 bytes 4096 + } +}" +$DIFF -u <(echo "$EXPECT") <($NFT reset counter ip test https-counter) + +check_list_counter +$NFT flush ruleset
Signature · 1 line
2.30.2
Quoted 27 lines · Pablo Neira Ayuso
Hi,
The following patchset contains cache updates for nft:
Patch #1 resets filtering for each new command
Patch #2 accumulates cache flags for each command, recent patches are
relaxing cache requirements which could uncover bugs.Patch #3 updates objects to use the netlink dump filtering infrastructure
to build the cache (Patch #4 only dumps rules for the given table
Patch #5 updates reset commands to use the cache infrastructure
Patch #6 and #7 extend tests coverage for reset commands.
Pablo Neira Ayuso (7):
cache: reset filter for each command
cache: accumulate flags in batch
cache: add filtering support for objects
cache: only dump rules for the given table
cache: consolidate reset command
tests: shell: cover anonymous set with reset command
tests: shell: cover reset command with counter and quota
Quoted 15 lines
include/cache.h | 12 +- include/netlink.h | 5 - src/cache.c | 201 ++++++++++++++---- src/evaluate.c | 2 + src/mnl.c | 7 +- src/netlink.c | 78 ------- src/parser_bison.y | 8 +- src/rule.c | 48 +---- tests/shell/testcases/listing/reset_objects | 104 +++++++++ .../testcases/rule_management/0011reset_0 | 31 ++- 10 files changed, 305 insertions(+), 191 deletions(-) create mode 100755 tests/shell/testcases/listing/reset_objectsSignature · 1 line
2.30.2
I ran this against the firewalld testsuite; lgtm. It does not cover "reset" commands.
Pablo Neira Ayuso
maintainerQuoted 18 lines · Eric Garver
Quoted 17 lines · Pablo Neira Ayuso
Hi,
The following patchset contains cache updates for nft:
Patch #1 resets filtering for each new command
Patch #2 accumulates cache flags for each command, recent patches are
relaxing cache requirements which could uncover bugs.Patch #3 updates objects to use the netlink dump filtering infrastructure
to build the cache (Patch #4 only dumps rules for the given table
Patch #5 updates reset commands to use the cache infrastructure
Patch #6 and #7 extend tests coverage for reset commands.
[...]
Quoted 4 lines
I ran this against the firewalld testsuite; lgtm. It does not cover "reset" commands.
Tested-by Eric Garver <[email protected]>
Pushed out, thanks for testing.
I have one more series cooking for cache refining, I will get on Cc.