Skip to contentSkip to conversation

A round of covscan indicated fixes

PATCH Superseded Phil Sutter 2 participants 9 messages Aug 12, 2016 – Aug 17, 2016 netfilter

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Phil Sutter

author reviewer
#0 lore raw

cover A round of covscan indicated fixes

As for libnftnl, this series aims at fixing a number of issues identified by covscan. And again, due to my limited overview of the code-base, some of them might as well be invalid although I tried to verify the issues as best as I can.

Phil Sutter (4):
evaluate: Fix datalen checks in expr_evaluate_string()
netlink_delinearize: Avoid potential null pointer deref
proto_find_num: Avoid potential null pointer dereference
evaluate: Avoid undefined behaviour in concat_subtype_id()

 src/evaluate.c            | 8 +++++---
 src/netlink_delinearize.c | 2 ++
 src/proto.c               | 2 +-
 3 files changed, 8 insertions(+), 4 deletions(-)
Signature · 1 line
2.8.2
author reviewer
#1 lore raw

PATCH 1/4 evaluate: Fix datalen checks in expr_evaluate_string()

This part of the code is pretty weird due to suboptimal variable name
choice: 'data', 'len', 'datalen', 'data_len'.

But even without understanding all of it, the code checking 'datalen - 1

Quoted 1 line

= 0' assumes 'datalen - 1' may actually become negative, which is not

true since it is unsigned. So make 'datalen' a signed integer instead.

Another issue is the check for "data[datalen] != '*'" which will access unallocated memory if 'strlen(data) == 0'. So make sure 'datalen >= 0' before using it as array index.

---
 src/evaluate.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
src/evaluate.c +4 -2
index 87f5a6d77d485..523eedabe84ac 100644
@@ -211,8 +211,9 @@ static int expr_evaluate_symbol(struct eval_ctx *ctx, struct expr **expr)
 static int expr_evaluate_string(struct eval_ctx *ctx, struct expr **exprp)
 {
 	struct expr *expr = *exprp;
-	unsigned int len = div_round_up(expr->len, BITS_PER_BYTE), datalen;
+	unsigned int len = div_round_up(expr->len, BITS_PER_BYTE);
 	struct expr *value, *prefix;
+	int datalen;
 	int data_len = ctx->ectx.len > 0 ? ctx->ectx.len : len + 1;
 	char data[data_len];
 
@@ -228,7 +229,8 @@ static int expr_evaluate_string(struct eval_ctx *ctx, struct expr **exprp)
 	mpz_export_data(data, expr->value, BYTEORDER_HOST_ENDIAN, len);
 
 	datalen = strlen(data) - 1;
-	if (data[datalen] != '*') {
+	if (datalen >= 0 &&
+	    data[datalen] != '*') {
 		/* We need to reallocate the constant expression with the right
 		 * expression length to avoid problems on big endian.
 		 */
Signature · 1 line
2.8.2
#2 lore raw

Re: [nft PATCH 1/4] evaluate: Fix datalen checks in expr_evaluate_string()

Quoted 10 lines · Phil Sutter

This part of the code is pretty weird due to suboptimal variable name
choice: 'data', 'len', 'datalen', 'data_len'.

But even without understanding all of it, the code checking 'datalen - 1

Quoted 1 line

= 0' assumes 'datalen - 1' may actually become negative, which is not

true since it is unsigned. So make 'datalen' a signed integer instead.

Another issue is the check for "data[datalen] != '*'" which will access unallocated memory if 'strlen(data) == 0'. So make sure 'datalen >= 0' before using it as array index.

We don't allow empty strings from our flex scanner as string, so we assume the string is at least 1.

You can probably add an assert() here instead.

author reviewer
#3 lore raw

PATCH 2/4 netlink_delinearize: Avoid potential null pointer deref

In reply to Phil Sutter

As netlink_get_register() may return NULL, we must not pass the returned data unchecked to expr_set_type() as that will dereference it. Since the parser has failed at that point anyway, by returning early we can skip the useless statement allocation that follows in netlink_parse_ct_stmt().

---
 src/netlink_delinearize.c | 2 ++
 1 file changed, 2 insertions(+)
src/netlink_delinearize.c +2 -0
index 12d0b4a277795..6ac2e9690fd39 100644
@@ -518,6 +518,8 @@ static void netlink_parse_ct_stmt(struct netlink_parse_ctx *ctx,
 
 	sreg = netlink_parse_register(nle, NFTNL_EXPR_CT_SREG);
 	expr = netlink_get_register(ctx, loc, sreg);
+	if (!expr)
+		return;
 
 	key  = nftnl_expr_get_u32(nle, NFTNL_EXPR_CT_KEY);
 	stmt = ct_stmt_alloc(loc, key, expr);
Signature · 1 line
2.8.2
#4 lore raw

Re: [nft PATCH 2/4] netlink_delinearize: Avoid potential null pointer deref

Quoted 16 lines · Phil Sutter

As netlink_get_register() may return NULL, we must not pass the returned data unchecked to expr_set_type() as that will dereference it. Since the parser has failed at that point anyway, by returning early we can skip the useless statement allocation that follows in netlink_parse_ct_stmt().

---
 src/netlink_delinearize.c | 2 ++
 1 file changed, 2 insertions(+)
src/netlink_delinearize.c +0 -0
index 12d0b4a277795..6ac2e9690fd39 100644

src/netlink_delinearize.c static void netlink_parse_ct_stmt(struct netlink_parse_ctx *ctx, · line 522

Quoted 5 lines
src/netlink_delinearize.c +2 -0
@@ -518,6 +518,8 @@ static void netlink_parse_ct_stmt(struct netlink_parse_ctx *ctx,
 
 	sreg = netlink_parse_register(nle, NFTNL_EXPR_CT_SREG);
 	expr = netlink_get_register(ctx, loc, sreg);
+	if (!expr)
+		return;

I would add a netlink_error() by here too.

I can see more spots with this problem, would you send us a patch to resolve the multiple reincarnations of this problem in one go?

Thanks!

author reviewer
#5 lore raw

PATCH 3/4 proto_find_num: Avoid potential null pointer dereference

In reply to Phil Sutter

When being called from stmt_evaluate_reset(), it seems that 'base' might actually be NULL, so better make sure it is not in proto_find_num().

---
This might be invalid in that if 'base' is NULL, ctx->pctx.family is always either NFPROTO_INET or NFPROTO_BRIDGE. But if so, the corresponding check in stmt_evaluate_reset() may be simplified to just having 'base' falling back to &proto_inet_service.

---
 src/proto.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
src/proto.c +1 -1
index 4c12977cef082..d9210afeaf256 100644
@@ -75,7 +75,7 @@ int proto_find_num(const struct proto_desc *base,
 {
 	unsigned int i;
 
-	for (i = 0; i < array_size(base->protocols); i++) {
+	for (i = 0; base && i < array_size(base->protocols); i++) {
 		if (base->protocols[i].desc == desc)
 			return base->protocols[i].num;
 	}
Signature · 1 line
2.8.2
#6 lore raw

Re: [nft PATCH 3/4] proto_find_num: Avoid potential null pointer dereference

Quoted 2 lines · Phil Sutter

When being called from stmt_evaluate_reset(), it seems that 'base' might actually be NULL, so better make sure it is not in proto_find_num().

I would suggest you address this from stmt_evaluate_reset().

author reviewer
#7 lore raw

PATCH 4/4 evaluate: Avoid undefined behaviour in concat_subtype_id()

In reply to Phil Sutter

Looking at expr_evaluate_concat(), 'off' might be zero and the error checks not triggering (by having dtype != NULL and i->dtype->size > 0). Decrementing it will then lead to casting -1 to unsigned during the call to concat_subtype_lookup() will lead to bit-shifting in concat_subtype_id() by a value bigger than the number of bits in 'type' (which is 32bit).

---
This patch is just an ugly sanitization hack and should probably be substituted by an additional error check in expr_evaluate_concat() giving an explanation of what went wrong.

---
 src/evaluate.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
src/evaluate.c +1 -1
index 523eedabe84ac..c8568690f6338 100644
@@ -950,7 +950,7 @@ static int expr_evaluate_concat(struct eval_ctx *ctx, struct expr **expr)
 						 "expressions",
 						 i->dtype->name);
 
-		tmp = concat_subtype_lookup(type, --off);
+		tmp = concat_subtype_lookup(type, off > 0 ? --off : 0);
 		expr_set_context(&ctx->ectx, tmp, tmp->size);
 
 		if (list_member_evaluate(ctx, &i) < 0)
Signature · 1 line
2.8.2
#8 lore raw

Re: [nft PATCH 4/4] evaluate: Avoid undefined behaviour in concat_subtype_id()

Quoted 7 lines · Phil Sutter

Looking at expr_evaluate_concat(), 'off' might be zero and the error checks not triggering (by having dtype != NULL and i->dtype->size > 0). Decrementing it will then lead to casting -1 to unsigned during the call to concat_subtype_lookup() will lead to bit-shifting in concat_subtype_id() by a value bigger than the number of bits in 'type' (which is 32bit).

Quoted 5 lines

---
This patch is just an ugly sanitization hack and should probably be substituted by an additional error check in expr_evaluate_concat() giving an explanation of what went wrong.

I agree. It would be good to look back and see what condition can indeed trigger this instead of papering the problem with this check.

Thanks!

Quoted 19 lines
 src/evaluate.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
src/evaluate.c +1 -1
index 523eedabe84ac..c8568690f6338 100644
@@ -950,7 +950,7 @@ static int expr_evaluate_concat(struct eval_ctx *ctx, struct expr **expr)
 						 "expressions",
 						 i->dtype->name);
 
-		tmp = concat_subtype_lookup(type, --off);
+		tmp = concat_subtype_lookup(type, off > 0 ? --off : 0);
 		expr_set_context(&ctx->ectx, tmp, tmp->size);
 
 		if (list_member_evaluate(ctx, &i) < 0)
Signature · 1 line
2.8.2