src: Don't parse string as verdict in map
PATCH 1/1 src: Don't parse string as verdict in map
In verdict map, string values are accidentally treated as verdicts.
For example:
table ip t {
map foo {
type mark : verdict
elements = {
0 : bar
}
}
}The value "bar" is sent to kernel as verdict.
Indeed, we don't parse verdicts during evaluation, but only chains, which is of type string rather than verdict.
--- src/datatype.c | 12 ------------ src/evaluate.c | 3 ++- 2 files changed, 2 insertions(+), 13 deletions(-)
src/datatype.c +0 -12
@@ -321,23 +321,11 @@ static void verdict_type_print(const struct expr *expr, struct output_ctx *octx) } } -static struct error_record *verdict_type_parse (struct parse_ctx *ctx, -const struct expr *sym, -struct expr **res) -{ - *res =constant_expr_alloc (&sym->location, &string_type, -BYTEORDER_HOST_ENDIAN , - (strlen (sym->identifier) +1 ) *BITS_PER_BYTE , - sym->identifier); -return NULL ; -} -const struct datatype verdict_type = { .type =TYPE_VERDICT , .name ="verdict" , .desc ="netfilter verdict" , .print = verdict_type_print, - .parse = verdict_type_parse, };static const struct symbol_table nfproto_tbl = {
src/evaluate.c +2 -1
@@ -2575,7 +2575,8 @@ static int stmt_evaluate_verdict(struct eval_ctx *ctx, struct stmt *stmt)if (stmt->expr->verdict !=NFT_CONTINUE ) stmt->flags |=STMT_F_TERMINAL ;if (stmt->expr->chain !=NULL ) { -if (expr_evaluate (ctx, &stmt->expr->chain) <0 ) +if (stmt_evaluate_arg (ctx, stmt, &string_type,0 ,0 , + &stmt->expr->chain) <0 )return -1 ;if (stmt->expr->chain->etype !=EXPR_VALUE ) { return expr_error(ctx->msgs, stmt->expr->chain,
Signature · 1 line
2.37.1
Florian Westphal
maintainerQuoted 11 lines · Xiao Liang
In verdict map, string values are accidentally treated as verdicts.
For example:
table ip t { map foo { type mark : verdict elements = { 0 : bar } } }
Quoted 4 lines
The value "bar" is sent to kernel as verdict.
Indeed, we don't parse verdicts during evaluation, but only chains, which is of type string rather than verdict.
Can you explain what this is fixing?
This reverts the commit that adds support for defines as aliases:
commit c64457cff9673fbb41f613a67e158b4d62235c09
src: Allow goto and jump to a variable
Quoted 2 lines · Florian Westphal
Can you explain what this is fixing?
See this example:
table t {
map foo {
type ipv4_addr : verdict
elements = {
192.168.0.1 : bar
}
}
chain output {
type filter hook output priority mangle;
ip daddr vmap @foo
}
}Though "bar" is not a valid verdict (should be "jump bar" or something), the string is taken as the element value. Then NFTA_DATA_VALUE is sent to the kernel instead of NFTA_DATA_VERDICT. Recent kernel checks the type and returns error, but olders (e.g. v5.4.x) doesn't, causing a warning when the rule is hit:
[5120263.467627] WARNING: CPU: 12 PID: 303303 at
net/netfilter/nf_tables_core.c:229 nft_do_chain+0x394/0x500 [nf_tables]
Quoted 6 lines
This reverts the commit that adds support for defines as aliases:
commit
c64457cff9673fbb41f613a67e158b4d62235c09
src: Allow goto and jump to a variable
This patch fixes it by parsing chain names as strings rather than verdicts, i.e. "jump $var" is a verdict while "$var" is a string.
Florian Westphal
maintainerQuoted 24 lines · Xiao Liang
Quoted 2 lines · Florian Westphal
Can you explain what this is fixing?
See this example:
table t { map foo { type ipv4_addr : verdict elements = { 192.168.0.1 : bar } } chain output { type filter hook output priority mangle; ip daddr vmap @foo } }Though "bar" is not a valid verdict (should be "jump bar" or something), the string is taken as the element value. Then
NFTA_DATA_VALUEis sent to the kernel instead ofNFTA_DATA_VERDICT. Recent kernel checks the type and returns error, but olders (e.g. v5.4.x) doesn't, causing a warning when the rule is hit:
Quoted 3 lines
[5120263.467627] WARNING: CPU: 12 PID: 303303 at
net/netfilter/nf_tables_core.c:229nft_do_chain+0x394/0x500[nf_tables]
Thanks. All of this info should be included in the commit message.
Perhaps adding a test case is warrented as well.