[RFC] netlink: check for NULL attribute in nla_parse_nested_deprecated

[RFC] netlink: check for NULL attribute in nla_parse_nested_deprecated

[Stephen Hemminger]

Lots of code in network schedulers has the pattern:
	if (!nla) {
		NL_SET_ERR_MSG_MOD(extack, "missing attributes");
		return -EINVAL;
	}

	err = nla_parse_nested_deprecated(tb, TCA_CSUM_MAX, nla, csum_policy,
					  NULL);
	if (err < 0)
		return err;

The check for nla being NULL can be moved into nla_parse_nested_deprecated().
Which simplifies lots of places.

This is safer and won't break other places since:
	err = nla_parse_nested_deprecated(tb, TCA_XXXX, NULL, some_policy, NULL);
would have crashed kernel because nla_parse_deprecated derefences the nla
argument before calling __nla_parse.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
---
 include/net/netlink.h | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/include/net/netlink.h b/include/net/netlink.h
index c19ff921b661..05d137283ab0 100644
--- a/include/net/netlink.h
+++ b/include/net/netlink.h
@@ -1322,6 +1322,11 @@ static inline int nla_parse_nested_deprecated(struct nlattr *tb[], int maxtype,
 					      const struct nla_policy *policy,
 					      struct netlink_ext_ack *extack)
 {
+	if (!nla) {
+		NL_SET_ERR_MSG_MOD(extack, "missing attributes");
+		return -EINVAL;
+	}
+
 	return __nla_parse(tb, maxtype, nla_data(nla), nla_len(nla), policy,
 			   NL_VALIDATE_LIBERAL, extack);
 }
-- 
2.43.0

Re: [RFC] netlink: check for NULL attribute in nla_parse_nested_deprecated

[Simon Horman]

On Thu, Feb 15, 2024 at 05:52:48PM -0800, Stephen Hemminger wrote:
> Lots of code in network schedulers has the pattern:
> 	if (!nla) {
> 		NL_SET_ERR_MSG_MOD(extack, "missing attributes");
> 		return -EINVAL;
> 	}
> 
> 	err = nla_parse_nested_deprecated(tb, TCA_CSUM_MAX, nla, csum_policy,
> 					  NULL);
> 	if (err < 0)
> 		return err;
> 
> The check for nla being NULL can be moved into nla_parse_nested_deprecated().
> Which simplifies lots of places.
> 
> This is safer and won't break other places since:
> 	err = nla_parse_nested_deprecated(tb, TCA_XXXX, NULL, some_policy, NULL);
> would have crashed kernel because nla_parse_deprecated derefences the nla
> argument before calling __nla_parse.
> 
> Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>

Hi Stephen,

this seems like a good approach to me.
Would you also plan to update the schedules at some point?

Re: [RFC] netlink: check for NULL attribute in nla_parse_nested_deprecated

[Jakub Kicinski]

On Thu, 15 Feb 2024 17:52:48 -0800 Stephen Hemminger wrote:
> Lots of code in network schedulers has the pattern:
> 	if (!nla) {
> 		NL_SET_ERR_MSG_MOD(extack, "missing attributes");
> 		return -EINVAL;
> 	}
> 
> 	err = nla_parse_nested_deprecated(tb, TCA_CSUM_MAX, nla, csum_policy,
> 					  NULL);
> 	if (err < 0)
> 		return err;
> 
> The check for nla being NULL can be moved into nla_parse_nested_deprecated().
> Which simplifies lots of places.

If it's mainly TC which has this problem the fix belongs in TC,
not in the generic code.