Re: [PATCH 0/7 v21] LSM: Multiple concurrent LSMs
From: Stephen Smalley
Date: Wed Mar 11 2015 - 12:36:45 EST
On 03/11/2015 12:25 PM, Stephen Smalley wrote:
> On 03/09/2015 09:13 PM, Casey Schaufler wrote:
>> Subject: [PATCH 0/7 v21] LSM: Multiple concurrent LSMs
>> Replace the current ad hoc stacking of the capabilities
>> and Yama security modules with a generalized stacking scheme.
>> The old structure had a single set of module hooks contained
>> in a security_operations structure. This structure was initialized
>> with a set of stubs referred to as the "capabilities" module.
>> In fact only a few of these hooks actually did anything useful.
>> When a module replaced the capabilities module the entries
>> supplied replaced those from the capabilities module. The
>> new hook was expected to call the replaced capability code
>> if "stacking" was desired, which it usually was. Yama stacking
>> is done by ifdefs in the security infrastructure.
>> The new structure provides a list of module hooks for each
>> interface. The non-trivial functions from the capabilities
>> module are add to the list first. If Yama stacking is configured
>> the Yama functions are added next. If a module is specified as
>> the "default" module, or is specified on the command line, it
>> is added next.
>> Functions are called in the order added to the list. The
>> security interfaces stop when a function indicates an access
>> denial. It is possible for a list to be empty. That is treated
>> as a success in most cases.
>> Each security module provides an array of function list entries.
>> This is initialized with the information needed to properly add
>> the entries to the function lists.
>> The sheer size of this patch set is somewhat frightening. This
>> is an artifact of the number of security interfaces involved and
>> except for a few cases the changes are mechanical in nature.
>> Except for the removal of some information specific to the security
>> module infrastructure itself, the change is transparent to the rest
>> of the kernel.
>> This is going to break out-of-tree security modules. It's easy to
>> update a module to the new scheme, and I'd be happy to do it for
>> any module I know about, but if it isn't in the tree, I don't know
>> about it.
>> The stacking of modules that use the security blob pointers
>> cred->security, inode->i_security, etc has not been addressed.
>> That is future work with a delightful set of issues.
>> This patch set is based on James Morris' security-next tree,
>> which is itself based on Linus' 4.0-rc1. It reflects the 11
>> patches of v20.
>> Signed-off-by: Casey Schaufler <casey@xxxxxxxxxxxxxxxx>
> Acked-by: Stephen Smalley <sds@xxxxxxxxxxxxx>
> Passes the selinux-testsuite with SELINUX alone or
Hmm..sorry, I missed something. Rescinding my ACK; comments to follow.
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/