Re: [PATCH v2 5/5] LSM: LoadPin for kernel file loading restrictions

[Kees Cook]

On Thu, Mar 31, 2016 at 2:24 PM, Mimi Zohar <zohar@xxxxxxxxxxxxxxxxxx> wrote:
> On Mon, 2016-03-28 at 14:14 -0700, Kees Cook wrote:
>
>> +static const char *id_str[READING_MAX_ID] = {
>> +     [READING_FIRMWARE] = "firmware",
>> +     [READING_MODULE] = "kernel module",
>> +     [READING_KEXEC_IMAGE] = "kexec image",
>> +     [READING_KEXEC_INITRAMFS] = "kexec initramfs",
>> +     [READING_POLICY] = "security policy",
>> +};
>> +
>> +static int loadpin_read_file(struct file *file, enum kernel_read_file_id id)
>> +{
>> +     struct super_block *load_root;
>> +     const char *origin;
>> +
>> +     if (id < 0 || id >= READING_MAX_ID)
>> +             return -EINVAL;
>> +
>
> The kernel_read_file_id enumeration starts at 1.

Yeah, I noticed that but decided that since I didn't have a named
value to mark the start, and negative value enums are extremely
unliked, I'd just check against "0" since missing id_str entries will
be correctly identified as "unknown" in the case where id_str and
READING_* enum is out of sync.

>> +     origin = id_str[id];
>> +     if (!origin) {
>> +             origin = "unknown";
>
> Right, all the entries in the kernel_read_file_id enumeration are
> currently listed in id_str.  Checking origin is needed for when id_str[]
> isn't kept in sync with the enumeration.

I wonder if there should be a function that returns a const string for
each kernel_read_file_id enum so users of the enum don't need to do
it?

> Looks good!

Thanks! I'll be sending an updated version shortly...

-Kees

-- 
Kees Cook
Chrome OS & Brillo Security