Skip to contentSkip to conversation

genetlink custom attribute type

PATCH In review Johannes Berg 3 participants 7 messages Sep 26, 2006 – Oct 2, 2006

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 genetlink custom attribute type

This patch adds an NLA_CUSTOM_CHECK type for netlink attributes in order to be able to centrally define new attribute structures instead of having to check these special types in each function that uses such an attribute.

nl80211 will benefit from this because it needs an information element attribute with a fixed type|length|value structure defined in IEEE 802.11 which ought to be checked when passing an IE from userspace.

include/net/netlink.h +5 -0
@@ -159,6 +159,7 @@ enum {
 	NLA_MSECS,
 	NLA_NESTED,
 	NLA_NUL_STRING,
+	NLA_CUSTOM_CHECK,
 	__NLA_TYPE_MAX,
 };
 
@@ -176,6 +177,8 @@ enum {
  *    NLA_STRING           Maximum length of string
  *    NLA_NUL_STRING       Maximum length of string (excluding NUL)
  *    NLA_FLAG             Unused
+ *    NLA_CUSTOM_CHECK     Unused, check function must be assigned,
+ *                         it must return 0 if the attribute is valid
  *    All other            Exact length of attribute payload
  *
  * Example:
@@ -183,11 +186,13 @@ enum {
  * 	[ATTR_FOO] = { .type = NLA_U16 },
  *	[ATTR_BAR] = { .type = NLA_STRING, len = BARSIZ },
  *	[ATTR_BAZ] = { .len = sizeof(struct mystruct) },
+ *	[ATTR_CST] = { .type = NLA_CUSTOM_CHECK, .check = my_check },
  * };
  */
 struct nla_policy {
 	u16		type;
 	u16		len;
+	int		(*check)(struct nlattr *nla);
 };
 
 extern void		netlink_run_queue(struct sock *sk, unsigned int *qlen,
net/netlink/attr.c +3 -0
@@ -67,6 +67,9 @@ static int validate_nla(struct nlattr *n
 		}
 		break;
 
+	case NLA_CUSTOM_CHECK:
+		BUG_ON(!pt->check);
+		return pt->check(nla);
 	default:
 		if (pt->len)
 			minlen = pt->len;
#1 lore raw

* Johannes Berg <[email protected]> 2006-09-26 09:25

Quoted 4 lines

This patch adds an NLA_CUSTOM_CHECK type for netlink attributes in order to be able to centrally define new attribute structures instead of having to check these special types in each function that uses such an attribute.

Thinking it over I'm still not completely happy with this. A small subset of all the validation tasks is simply too complex to be put into the policy. The validation of your type value array is such a case, address fields with variable length based on their family is another. I think it's just not worth to blow up struct nla_policy by 12 bytes per entry just to save some code.

#2 lore raw
Quoted 7 lines · Thomas Graf

Thinking it over I'm still not completely happy with this. A small subset of all the validation tasks is simply too complex to be put into the policy. The validation of your type value array is such a case, address fields with variable length based on their family is another. I think it's just not worth to blow up struct nla_policy by 12 bytes per entry just to save some code.

Alright, I can instead just add the validation function call wherever that attribute is going to be used. Just thought it might be nice to have this generically, I'm not really too attached to it :)

I'll change it in the next nl80211 iteration.

johannes

maintainer
#3 lore raw

From: Johannes Berg <[email protected]>
Date: Tue, 26 Sep 2006 12:04:34 +0200

Quoted 10 lines · Thomas Graf
Quoted 8 lines

Thinking it over I'm still not completely happy with this. A small subset of all the validation tasks is simply too complex to be put into the policy. The validation of your type value array is such a case, address fields with variable length based on their family is another. I think it's just not worth to blow up struct nla_policy by 12 bytes per entry just to save some code.

Quoted 5 lines

Alright, I can instead just add the validation function call wherever that attribute is going to be used. Just thought it might be nice to have this generically, I'm not really too attached to it :)

I'll change it in the next nl80211 iteration.

Thanks Johannes.

#4 lore raw
In reply to Thomas Graf
Quoted 7 lines · Thomas Graf

Thinking it over I'm still not completely happy with this. A small subset of all the validation tasks is simply too complex to be put into the policy. The validation of your type value array is such a case, address fields with variable length based on their family is another. I think it's just not worth to blow up struct nla_policy by 12 bytes per entry just to save some code.

Oh, I just had another idea... Feel free to ignore me if you think that having this done in some generic way isn't worth it though. As I said, it doesn't really make a difference to me in the end :)

Currently, we always pass a "struct nla_policy *policy" into things, which really is an array. We could instead pass in a new

struct nla_validation {
	int (*custom_validate)(struct nlattr *nla);
	struct nla_policy *policy;
};

and call custom_validate() whenever we encounter something in the policy that has type NLA_CUSTOM_VALIDATE. Downsides of this approach are that it requires changing all current users, and introduces 16 bytes constant overhead on 64-bit platforms, the size of nla_validation.

johannes

#5 lore raw

* Johannes Berg <[email protected]> 2006-09-27 14:18

Quoted 23 lines · Thomas Graf
Quoted 8 lines

Thinking it over I'm still not completely happy with this. A small subset of all the validation tasks is simply too complex to be put into the policy. The validation of your type value array is such a case, address fields with variable length based on their family is another. I think it's just not worth to blow up struct nla_policy by 12 bytes per entry just to save some code.

Oh, I just had another idea... Feel free to ignore me if you think that having this done in some generic way isn't worth it though. As I said, it doesn't really make a difference to me in the end :)

Currently, we always pass a "struct nla_policy *policy" into things, which really is an array. We could instead pass in a new

struct nla_validation {
	int (*custom_validate)(struct nlattr *nla);
	struct nla_policy *policy;
};
Quoted 4 lines

and call custom_validate() whenever we encounter something in the policy that has type NLA_CUSTOM_VALIDATE. Downsides of this approach are that it requires changing all current users, and introduces 16 bytes constant overhead on 64-bit platforms, the size of nla_validation.

Sorry for the delay.

That's not a bad idea, although it seems cleaner to just allow defining a callback function which gets called foreach unknown attribute. As for generic netlink, this callback could be defined on a per command basis in struct genl_ops.

#6 lore raw
Quoted 2 lines · Thomas Graf

That's not a bad idea, although it seems cleaner to just allow defining a callback function which gets called foreach unknown attribute.

Hm, that'd work too, but it'd force me to leave these 'unknown attributes' at the end of the attribute space, no? Otherwise, something with .len=0,.type=0 is sort of valid as well, not?

Quoted 3 lines

As for
generic netlink, this callback could be defined on a per command basis in struct genl_ops.

I'd have to look at it more closely again, I removed the custom stuff and it's not immediately important to me at this time.

johannes