*tables-restore: Enforce correct counters syntax if present
Phil Sutter
reviewerPATCH 1/4 *tables-restore: Enforce correct counters syntax if present
If '--counters' option was not given, restore parsers would ignore anything following the policy word. Make them more strict, rejecting anything in that spot which does not look like counter values even if not restoring counters.
--- iptables/iptables-restore.c | 20 +++++++++---------- .../ipt-restore/0008-restore-counters_0 | 7 +++++++ iptables/xtables-restore.c | 18 ++++++++--------- 3 files changed, 24 insertions(+), 21 deletions(-)
iptables/iptables-restore.c +9 -11
@@ -283,23 +283,21 @@ ip46tables_restore_main(const struct iptables_restore_cb *cb, xt_params->program_name, line);if (strcmp (policy,"-" ) !=0 ) { +char *ctrs =strtok (NULL ," \t\n" );struct xt_counters count = {}; -if (counters) { -char *ctrs; - ctrs =strtok (NULL ," \t\n" ); - -if (!ctrs || !parse_counters (ctrs, &count)) -xtables_error (PARAMETER_PROBLEM , -"invalid policy counters for chain '%s'" , - chain); - } +if ((!ctrs && counters) || + (ctrs && !parse_counters (ctrs, &count))) +xtables_error (PARAMETER_PROBLEM , +"invalid policy counters for chain '%s'" , + chain);DEBUGP ("Setting policy of chain %s to %s\n" , chain, policy); -if (!cb->ops->set_policy (chain, policy, &count, - handle)) +if (!cb->ops->set_policy (chain, policy, + counters ? &count :NULL , + handle)) xtables_error(OTHER_PROBLEM, "Can't set policy `%s' on `%s' line %u: %s", policy, chain, line,
iptables/tests/shell/testcases/ipt-restore/0008-restore-counters_0 +7 -0
@@ -20,3 +20,10 @@ EXPECT=":foo - [0:0] $XT_MULTI iptables-restore --counters <<< "$DUMP" diff -u -Z <(echo -e "$EXPECT") <($XT_MULTI iptables-save --counters | grep foo) + +# if present, counters must be in proper format +! $XT_MULTI iptables-restore <<EOF +*filter +:FORWARD ACCEPT bar +COMMIT +EOF
iptables/xtables-restore.c +8 -10
@@ -166,19 +166,17 @@ static void xtables_restore_parse_line(struct nft_handle *h, xt_params->program_name, line);if (nft_chain_builtin_find (state->curtable, chain)) { -if (counters) { -char *ctrs; - ctrs =strtok (NULL ," \t\n" ); +char *ctrs =strtok (NULL ," \t\n" ); -if (!ctrs || !parse_counters (ctrs, &count)) -xtables_error (PARAMETER_PROBLEM , -"invalid policy counters for chain '%s'" , - chain); - - } +if ((!ctrs && counters) || + (ctrs && !parse_counters (ctrs, &count))) +xtables_error (PARAMETER_PROBLEM , +"invalid policy counters for chain '%s'" , + chain);if (cb->chain_set && cb->chain_set (h, state->curtable->name, - chain, policy, &count) <0 ) { + chain, policy, + counters ? &count :NULL ) <0 ) { xtables_error(OTHER_PROBLEM, "Can't set policy `%s' on `%s' line %u: %s", policy, chain, line,
Signature · 1 line
2.40.0
Phil Sutter
reviewerPATCH 2/4 *tables: Reject invalid chain names when renaming
While given chain name was sanity checked with --new-chain command, --rename-chain command allowed to choose an invalid name. Keep things consistent by adding the missing check.
--- iptables/xshared.c | 1 + 1 file changed, 1 insertion(+)
iptables/xshared.c +1 -0
@@ -1511,6 +1511,7 @@ void do_parse(int argc, char *argv[],"-%c requires old-chain-name and " "new-chain-name" , cmd2char(CMD_RENAME_CHAIN)); +assert_valid_chain_name (p->newname);break ;case 'P' :
Signature · 1 line
2.40.0
Phil Sutter
reviewerPATCH 3/4 ebtables: Improve invalid chain name detection
In reply to Phil SutterFix several issues:
- Most importantly, --new-chain command accepted any name. Introduce ebt_assert_valid_chain_name() for use with both --new-chain and --rename-chain.
- Restrict maximum name length to what legacy ebtables allows - this is a bit more than iptables-nft, subject to be unified.
- Like iptables, legacy ebtables rejects names prefixed by '-' or '!'.
- Use xs_has_arg() for consistency, keep the check for extra args for now.
--- iptables/xtables-eb.c | 33 ++++++++++++++++++++++++++++----- 1 file changed, 28 insertions(+), 5 deletions(-)
iptables/xtables-eb.c +28 -5
@@ -42,6 +42,10 @@#include "nft.h" #include "nft-bridge.h" +/* from linux/netfilter_bridge/ebtables.h */ +#define EBT_TABLE_MAXNAMELEN 32 +#define EBT_CHAIN_MAXNAMELEN EBT_TABLE_MAXNAMELEN +/* * From include/ebtables_u.h */
@@ -74,6 +78,26 @@ static int ebt_check_inverse2(const char option[], int argc, char **argv)return ebt_invert; } +/* XXX: merge with assert_valid_chain_name()? */ +static void ebt_assert_valid_chain_name (const char *chainname) +{ +if (strlen (chainname) >=EBT_CHAIN_MAXNAMELEN ) +xtables_error (PARAMETER_PROBLEM , +"Chain name length can't exceed %d" , +EBT_CHAIN_MAXNAMELEN -1 ); + +if (*chainname =='-' || *chainname =='!' ) +xtables_error (PARAMETER_PROBLEM ,"No chain name specified" ); + +if (xtables_find_target (chainname,XTF_TRY_LOAD )) +xtables_error (PARAMETER_PROBLEM , +"Target with name %s exists" , chainname); + +if (strchr (chainname,' ' ) !=NULL ) +xtables_error (PARAMETER_PROBLEM , +"Use of ' ' not allowed in chain names" ); +} +/* * Glue code to use libxtables */
@@ -751,6 +775,7 @@ int do_commandeb(struct nft_handle *h, int argc, char *argv[], char **table, flags |=OPT_COMMAND ;if (c =='N' ) { +ebt_assert_valid_chain_name (chain); ret =nft_cmd_chain_user_add (h, chain, *table);break ; }else if (c =='X' ) {
@@ -764,14 +789,12 @@ int do_commandeb(struct nft_handle *h, int argc, char *argv[], char **table, }if (c =='E' ) { -if (optind >= argc) +if (!xs_has_arg (argc, argv))xtables_error (PARAMETER_PROBLEM ,"No new chain name specified" );else if (optind < argc -1 )xtables_error (PARAMETER_PROBLEM ,"No extra options allowed with -E" ); -else if (strlen (argv[optind]) >=NFT_CHAIN_MAXNAMELEN ) -xtables_error (PARAMETER_PROBLEM ,"Chain name length can't exceed %d"" characters" ,NFT_CHAIN_MAXNAMELEN -1 ); -else if (strchr (argv[optind],' ' ) !=NULL ) -xtables_error (PARAMETER_PROBLEM ,"Use of ' ' not allowed in chain names" ); + +ebt_assert_valid_chain_name (argv[optind]); errno =0 ; ret = nft_cmd_chain_user_rename(h, chain, *table,
Signature · 1 line
2.40.0
Phil Sutter
reviewerPATCH 4/4 tests: shell: Fix and extend chain rename test
In reply to Phil SutterThe old version exited unintentionally before testing ip6tables. Replace it by a more complete variant testing for all tools, creating and renaming of,chains with various illegal names instead of just renaming to a clashing name.
--- .../tests/shell/testcases/chain/0003rename_0 | 40 +++++++++++++++++++ .../tests/shell/testcases/chain/0003rename_1 | 12 ------ 2 files changed, 40 insertions(+), 12 deletions(-) create mode 100755 iptables/tests/shell/testcases/chain/0003rename_0 delete mode 100755 iptables/tests/shell/testcases/chain/0003rename_1
iptables/tests/shell/testcases/chain/0003rename_0 new file +40 -0
@@ -0,0 +1,40 @@ +#!/bin/bash -x + +die() { + echo "E: $@" + exit 1 +} + +cmds="iptables ip6tables" +[[ $XT_MULTI == *xtables-nft-multi ]] && cmds+=" arptables ebtables" + +declare -A invnames +invnames["existing"]="c2" +invnames["spaced"]="foo bar" +invnames["dashed"]="-foo" +invnames["negated"]="!foo" +# XXX: ebtables-nft accepts 255 chars +#invnames["overlong"]="thisisquitealongnameforachain" +invnames["standard target"]="ACCEPT" +invnames["extension target"]="DNAT" + +for cmd in $cmds; do + $XT_MULTI $cmd -N c1 || die "$cmd: can't add chain c1" + $XT_MULTI $cmd -N c2 || die "$cmd: can't add chain c2" + for key in "${!invnames[@]}"; do + val="${invnames[$key]}" + if [[ $key == "extension target" ]]; then + if [[ $cmd == "arptables" ]]; then + val="mangle" + elif [[ $cmd == "ebtables" ]]; then + val="dnat" + fi + fi + $XT_MULTI $cmd -N "$val" && \ + die "$cmd: added chain with $key name" + $XT_MULTI $cmd -E c1 "$val" && \ + die "$cmd: renamed to $key name" + done +done + +exit 0
iptables/tests/shell/testcases/chain/0003rename_1 deleted +0 -12
@@ -1,12 +0,0 @@ -#!/bin/bash - -$XT_MULTI iptables -N c1 || exit 0 -$XT_MULTI iptables -N c2 || exit 0 -$XT_MULTI iptables -E c1 c2 || exit 1 - -$XT_MULTI ip6tables -N c1 || exit 0 -$XT_MULTI ip6tables -N c2 || exit 0 -$XT_MULTI ip6tables -E c1 c2 || exit 1 - -echo "E: Renamed with existing chain" >&2 -exit 0
Signature · 1 line
2.40.0
Phil Sutter
reviewerQuoted 6 lines · Phil Sutter
If '--counters' option was not given, restore parsers would ignore anything following the policy word. Make them more strict, rejecting anything in that spot which does not look like counter values even if not restoring counters.
Signed-off-by Phil Sutter <[email protected]>
Series applied.