Skip to contentSkip to conversation

netfilter: nf_tables: Add a null pointer check in two functions

PATCH In review Markus Elfring 3 participants 4 messages Jan 23, 2024

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 netfilter: nf_tables: Add a null pointer check in two functions

From: Markus Elfring <[email protected]>
Date: Tue, 23 Jan 2024 14:28:31 +0100

The result from a call of the function “kasprintf” was passed to a subsequent function call without checking for a null pointer before (according to a memory allocation failure).
This issue was detected by using the Coccinelle software.

Thus add a null pointer check and a jump target in affected functions.

Fixes 8877393029e76 ("netfilter: nf_tables: Open-code audit log call in nf_tables_getrule()")
Fixes 0854db2aaef3f ("netfilter: nf_tables: use net_generic infra for transaction data")
---
 net/netfilter/nf_tables_api.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)
net/netfilter/nf_tables_api.c +7 -2
index 8438a8922e4a..cb61c7a39a76 100644
@@ -3728,10 +3728,13 @@ static int nf_tables_getrule_reset(struct sk_buff *skb,
 			nla_len(nla[NFTA_RULE_TABLE]),
 			(char *)nla_data(nla[NFTA_RULE_TABLE]),
 			nft_net->base_seq);
+	if (!buf)
+		goto exit;
+
 	audit_log_nfcfg(buf, info->nfmsg->nfgen_family, 1,
 			AUDIT_NFT_OP_RULE_RESET, GFP_ATOMIC);
 	kfree(buf);
-
+exit:
 	return nfnetlink_unicast(skb2, net, portid);
 }
 
@@ -7917,6 +7920,8 @@ static int nf_tables_getobj(struct sk_buff *skb, const struct nfnl_info *info,
 
 		nft_net = nft_pernet(net);
 		buf = kasprintf(GFP_ATOMIC, "%s:%u", table->name, nft_net->base_seq);
+		if (!buf)
+			goto fill_obj_info;
 
 		audit_log_nfcfg(buf,
 				family,
@@ -7925,7 +7930,7 @@ static int nf_tables_getobj(struct sk_buff *skb, const struct nfnl_info *info,
 				GFP_ATOMIC);
 		kfree(buf);
 	}
-
+fill_obj_info:
 	err = nf_tables_fill_obj_info(skb2, net, NETLINK_CB(skb).portid,
 				      info->nlh->nlmsg_seq, NFT_MSG_NEWOBJ, 0,
 				      family, table, obj, reset);

--
2.43.0

reviewer
#1 lore raw

Hi Markus,

Quoted 7 lines · Markus Elfring

From: Markus Elfring <[email protected]>
Date: Tue, 23 Jan 2024 14:28:31 +0100

The result from a call of the function “kasprintf” was passed to a subsequent function call without checking for a null pointer before (according to a memory allocation failure).
This issue was detected by using the Coccinelle software.

This is correct and I'm fine with the patch if it avoids ringing alarm bells somewhere, yet it doesn't fix an actual issue here since the allocated buffer is merely passed to vsnprintf() which detects and sanitizes %s args being NULL.

Cheers, Phil

#2 lore raw
Quoted 9 lines
Quoted 4 lines

The result from a call of the function “kasprintf” was passed to a subsequent function call without checking for a null pointer before (according to a memory allocation failure).
This issue was detected by using the Coccinelle software.

This is correct and I'm fine with the patch if it avoids ringing alarm bells somewhere, yet it doesn't fix an actual issue here since the allocated buffer is merely passed to vsnprintf() which detects and sanitizes %s args being NULL.

Should null pointer tolerance be better indicated for such use cases by any additional means?

Regards,
Markus

reviewer
#3 lore raw
In reply to Markus Elfring
Quoted 8 lines · Markus Elfring

From: Markus Elfring <[email protected]>
Date: Tue, 23 Jan 2024 14:28:31 +0100

The result from a call of the function “kasprintf” was passed to a subsequent function call without checking for a null pointer before (according to a memory allocation failure).
This issue was detected by using the Coccinelle software.

Quoted 5 lines

Thus add a null pointer check and a jump target in affected functions.

Fixes 8877393029e76 ("netfilter: nf_tables: Open-code audit log call in nf_tables_getrule()")
Fixes 0854db2aaef3f ("netfilter: nf_tables: use net_generic infra for transaction data")

For the record:

Markus has posed patches of this nature many times over the past weeks and months. They are not fixing bugs - there is no user visible problem. And moreover they move the code away from the preferred style - avoiding unnecessary checks around code that can handle NULL values.

Markus, please stop posting patches of this nature.