Re: [PATCH 11/12] AppArmor: LSM interface, and security module initialization

[John Johansen <john.johansen@canonical.com>]

Eric Paris wrote:
> On Tue, Nov 3, 2009 at 6:48 PM, John Johansen
> <john.johansen@canonical.com> wrote:
>> AppArmor hooks to interface with the LSM, and module parameters and
>> initialization.
>>
>> Signed-off-by: John Johansen <john.johansen@canonical.com>
>> ---
> 
> 
>> +static int apparmor_file_mmap(struct file *file, unsigned long reqprot,
>> +                             unsigned long prot, unsigned long flags,
>> +                             unsigned long addr, unsigned long addr_only)
>> +{
>> +       int rc = 0;
>> +       struct aa_profile *profile = aa_current_profile_wupd();
>> +       /*
>> +        * test before cap_file_mmap.  For confined tasks AppArmor will
>> +        * enforce the mmap value set in the profile or default
>> +        * to LSM_MMAP_MIN_ADDR
>> +        */
>> +       if (profile) {
>> +               if (profile->flags & PFLAG_MMAP_MIN_ADDR) {
>> +                       if (addr < profile->mmap_min_addr)
>> +                               rc = -EACCES;
>> +               } else if (addr < CONFIG_LSM_MMAP_MIN_ADDR) {
>> +                       rc = -EACCES;
>> +               }
>> +               if (rc) {
>> +                       struct aa_audit sa = {
>> +                               .operation = "file_mmap",
>> +                               .gfp_mask = GFP_KERNEL,
>> +                               .info = "addr < mmap_min_addr",
>> +                               .error = rc,
>> +                       };
>> +                       return aa_audit(AUDIT_APPARMOR_DENIED, profile, &sa,
>> +                                       NULL);
>> +               }
>> +       }
>> +       rc = cap_file_mmap(file, reqprot, prot, flags, addr, addr_only);
>> +       if (rc || addr_only)
>> +               return rc;
>> +
>> +       return common_mmap(file, "file_mmap", prot, flags);
>> +}
> 
> There is a reason we do the round_hint_to_min() stuff in the vm and we
> recalculate that value every time dac_mmap_min_addr is change.  It's
> because mmap (NOT MAP_FIXED) with a hint < profile->mmap_min_addr is
> going to end up getting denied here since the VM is going to assign it
> the address it wanted instead of find a new address and you are going
> to deny that task.
> 
> If profile() is a per task thing, I think you are in a failed
> situation and can't solve the problem wtihout intrusive VFS hooks.  If
> profile is a global thing just update that global value.  In either
> case, this code is wrong....
> 
yep, thanks for pointing this out.  I will look into it.

john