Re: [PATCH v10 1/3] x86/syscalls: Check address limit on user-mode return

From: Thomas Garnier
Date: Thu Jul 06 2017 - 16:52:30 EST


On Thu, Jul 6, 2017 at 1:48 PM, Thomas Gleixner <tglx@xxxxxxxxxxxxx> wrote:
> On Thu, 6 Jul 2017, Thomas Garnier wrote:
>> On Wed, Jun 28, 2017 at 10:52 AM, Kees Cook <keescook@xxxxxxxxxxxx> wrote:
>> >
>> > On Tue, Jun 20, 2017 at 1:24 PM, Kees Cook <keescook@xxxxxxxxxxxx> wrote:
>> > > On Wed, Jun 14, 2017 at 6:12 PM, Thomas Garnier <thgarnie@xxxxxxxxxx> wrote:
>> > >> Ensure the address limit is a user-mode segment before returning to
>> > >> user-mode. Otherwise a process can corrupt kernel-mode memory and elevate
>> > >> privileges [1].
>> > >>
>> > >> The set_fs function sets the TIF_SETFS flag to force a slow path on
>> > >> return. In the slow path, the address limit is checked to be USER_DS if
>> > >> needed.
>> > >>
>> > >> The addr_limit_user_check function is added as a cross-architecture
>> > >> function to check the address limit.
>> > >>
>> > >> [1] https://bugs.chromium.org/p/project-zero/issues/detail?id=990
>> > >>
>> > >> Signed-off-by: Thomas Garnier <thgarnie@xxxxxxxxxx>
>> > >
>> > > Thanks for reworking this series!
>> > >
>> > > The bad state correctly BUGs under the LKDTM test:
>> > >
>> > > [ 21.171586] lkdtm: Performing direct entry CORRUPT_USER_DS
>> > > [ 21.172791] lkdtm: setting bad task size limit
>> > > [ 21.173742] ------------[ cut here ]------------
>> > > [ 21.174641] kernel BUG at ./include/linux/syscalls.h:220!
>> > > ...
>> > > [ 21.193166] Call Trace:
>> > > [ 21.193617] ? trace_hardirqs_on_thunk+0x1a/0x1c
>> > > [ 21.194443] entry_SYSCALL64_slow_path+0x25/0x25
>> > >
>> > >
>> > > Tested-by: Kees Cook <keescook@xxxxxxxxxxxx>
>> >
>> > Is everyone happy with this patch for x86? Does this need anything
>> > more/different?
>>
>> Asking again. Additional feedback? Anyone wants to pick-it up?
>
> Can do. This needs to be a combo of all 3 I assume as the x86 one contains
> the function used by all of them, right?

That is correct.

>
> Thanks,
>
> tglx



--
Thomas