Message-ID: <4B0A5FA7.405@canonical.com>
Date:	Mon, 23 Nov 2009 02:10:47 -0800
From:	John Johansen <john.johansen@...onical.com>
To:	Tetsuo Handa <penguin-kernel@...ove.SAKURA.ne.jp>
CC:	linux-kernel@...r.kernel.org, linux-security-module@...r.kernel.org
Subject: Re: [AppArmor #3 0/12] AppArmor security module

Tetsuo Handa wrote:
> And the rest of files...
> 
> 
> 
> Regarding match.c
> 
> Why not to start YYTD_ID_something from 0 so that we can avoid "- 1" in
> dfa->tables[table->td_id - 1] ? I think you can do "- 1" at
> 
Right this is a legacy bit, to make a long story short they exactly match
the Flex table mappings which is unnecessary as we explicitly are not
compatible with Flex in other ways.  It will probably wait until after
the next push as I am looking at accept state cleanup for the dfa as well.

> 	th.td_id = be16_to_cpu(*(u16 *) (blob));
that is a possibility as long as it got a good comment to explain what
is going on.

> 
> .
> 
> 
> 
>> int unpack_dfa(struct aa_dfa *dfa, void *blob, size_t size)
>> {
> (...snipped...)
>> fail:
>> 	for (i = 0; i < ARRAY_SIZE(dfa->tables); i++) {
>> 		free_table(dfa->tables[i]);
>> 		dfa->tables[i] = NULL;
>> 	}
> 
> This function is called by only aa_unpack_dfa(), and aa_unpack_dfa() calls
> aa_match_free(). Thus, you don't need to call free_table() here.
> 
Hrmm, yeah it needs to be reworked, I don't particularly like returning a
partial struct to have it cleaned up later.  I think it might be better
to rework aa_unpack_dfa, dropping the aa_match_free and moving the verify
call into the unpack routine, and the above for loop can be replaced
by aa_match_free


>> 	return error;
>> }
> 
> 
> 
> Regarding net.c
> 
>> static void audit_cb(struct audit_buffer *ab, void *va)
>> {
>> 	struct aa_audit_net *sa = va;
> 
> static void audit_cb(struct audit_buffer *ab, struct aa_audit *va)
> {
> 	struct aa_audit_net *sa = container_of(va, struct aa_audit_net, base);
> 
>> static int aa_audit_net(struct aa_profile *profile, struct aa_audit_net *sa)
>> {
> (...snipped...)
>> 	return aa_audit(type, profile, (struct aa_audit *)sa, audit_cb);
> 
> 	return aa_audit(type, profile, &sa->base, audit_cb);
> 
>> }
> 
> 
yep, thanks

> 
> Regarding policy.c
> 
>> struct aa_namespace *alloc_aa_namespace(const char *name)
> 
> This function could be "static". Please try "make namespacecheck".
> 
will do

> 
> 
>> struct aa_namespace *aa_prepare_namespace(const char *name)
> 
> This function could be "static".
> 
>> {
>> 	struct aa_namespace *ns;
>>
>> 	write_lock(&ns_list_lock);
>> 	if (name)
>> 		/* released by caller */
>> 		ns = aa_get_namespace(__aa_find_namespace(&ns_list, name));
>> 	else
>> 		/* released by caller */
>> 		ns = aa_get_namespace(default_namespace);
> 
> alloc_aa_namespace() returns NULL if name == NULL.
> If it is intended behavior, you may do like
> 
> 	else {
>  		/* released by caller */
> 		ns = aa_get_namespace(default_namespace);
> 		write_unlock(&ns_list_lock);
> 		return ns;
> 	}
> 
well at this point name != NULL because in that case ns == default_namespace,
but it should be documented that name can not be null here.

In general I am not happy with prepare_namespace and will take another look
at this.

>> 	if (!ns) {
>> 		struct aa_namespace *new_ns;
>> 		write_unlock(&ns_list_lock);
>> 		new_ns = alloc_aa_namespace(name);
>> 		if (!new_ns)
>> 			return NULL;
>> 		write_lock(&ns_list_lock);
>> 		/* test for race when new_ns was allocated */
>> 		ns = __aa_find_namespace(&ns_list, name);
>> 		if (!ns) {
>> 			list_add(&new_ns->base.list, &ns_list);
>> 			/* add list ref */
>> 			ns = aa_get_namespace(new_ns);
>> 		} else {
>> 			/* raced so free the new one */
>> 			free_aa_namespace(new_ns);
>> 			/* get reference on namespace */
>> 			aa_get_namespace(ns);
>> 		}
>> 	}
>> 	write_unlock(&ns_list_lock);
>>
>> 	/* return ref */
>> 	return ns;
>> }
> 
> 
> 
>> void __aa_replace_profile(struct aa_profile *old,
>> 			  struct aa_profile *new)
> 
> This function could be "static".
> 
> 
> 
>> void __aa_profile_list_release(struct list_head *head)
> 
> This function could be "static".
> 
> 
> 
>> void __aa_remove_namespace(struct aa_namespace *ns)
> 
> This function could be "static".
> 
>> {
>> 	struct aa_profile *unconfined = ns->unconfined;
>> 	/* remove ns from namespace list */
>> 	list_del_init(&ns->base.list);
>>
>> 	/*
>> 	 * break the ns, unconfined profile cyclic reference and forward
>> 	 * all new unconfined profiles requests to the default namespace
>> 	 * This will result in all confined tasks that have a profile
>> 	 * being removed inheriting the default->unconfined profile.
>> 	 */
>> 	ns->unconfined = aa_get_profile(default_namespace->unconfined);
>> 	__aa_profile_list_release(&ns->base.profiles);
>> 	/* release original ns->unconfined ref */
>> 	aa_put_profile(unconfined);
>> 	/* release ns->base.list ref, from removal above */
>> 	aa_put_namespace(ns);
> 
> aa_put_profile() and aa_put_namespace() may call write_lock() inside
> free_aa_profile(). Are you sure that these calls do not dead lock?
> 
Yes, though I will give it another run through and reverify and add better
comments.  The locking is such that the profile should be removed from
the list before free_aa_profile is called, and once in free_aa_profile
the lock is taken and released before any put_ is done.

>> }
> 
> 
> 
>> ssize_t aa_interface_remove_profiles(char *name, size_t size)
> (...snipped...)
>> 	write_lock(&ns_list_lock);
>> 	if (name[0] == ':') {
>> 	char *ns_name;
>> 	name = aa_split_name_from_ns(name, &ns_name);
>> 	/* released below */
>> 	ns = aa_get_namespace(__aa_find_namespace(&ns_list, ns_name));
> 
> aa_split_name_from_ns() may set ns_name to NULL but __aa_find_namespace() can't
> handle ns_name == NULL case. I think you should check ns_name != NULL.
> 
yep

> 
> 
> Regarding policy_unpack.c
> 
> Please use bool for functions that return 0 or 1.
> 
> 
> 
>> static void audit_cb(struct audit_buffer *ab, void *va)
>> {
>> 	struct aa_audit_iface *sa = va;
> 
> static void audit_cb(struct audit_buffer *ab, struct aa_audit *va)
> {
> 	struct aa_audit_iface *sa = container_of(va, struct aa_audit_iface,
> 						 base);
> 
> 
> 
>> static int aa_unpack_trans_table(struct aa_ext *e, struct aa_profile *profile)
> (...snipped...)
>> 		for (i = 0; i < size; i++) {
>> 			char *tmp;
>> 			if (!unpack_dynstring(e, &tmp, NULL))
>> 				goto fail;
>> 			/*
>> 			 * note: strings beginning with a : have an embedded
>> 			 * \0 seperating the profile ns name from the profile
>> 			 * name
>> 			 */
>> 			profile->file.trans.table[i] = tmp;
> 
> unpack_dynstring() returns string duplicated by kstrdup(). Thus, "tmp" can't
> have an embedded \0 seperating the profile ns name from the profile name
> even if tmp[0] == ':' is true, can it?
> 
indeed, I should not have switched to kstrdup.

>> 		}
> 
> 
> 
> Regarding resource.c
> 
>> static void audit_cb(struct audit_buffer *ab, void *va)
>> {
>> 	struct aa_audit_resource *sa = va;
> 
> static void audit_cb(struct audit_buffer *ab, struct aa_audit *va)
> {
> 	struct aa_audit_resource *sa = container_of(va, struct aa_audit_resource,
> 						    base);
> 
> 
> 
>> static int aa_audit_resource(struct aa_profile *profile,
>> 			     struct aa_audit_resource *sa)
>> {
>> 	return aa_audit(AUDIT_APPARMOR_AUTO, profile, (struct aa_audit *)sa,
>> 			audit_cb);
> 
> 	return aa_audit(AUDIT_APPARMOR_AUTO, profile, &sa->base, audit_cb);
> 
>> }
> 
> 
> 
> Regarding sid.c
> 
>> int aa_add_sid_profile(u32 sid, struct aa_profile *profile)
> 
> This function is not used.
> 
> 
> 
>> int aa_replace_sid_profile(u32 sid, struct aa_profile *profile)
> 
> This function is not used.
> 
> 
> 
>> struct aa_profile *aa_get_sid_profile(u32 sid)
> 
> This function is not used.

right will fix

thanks
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists