lkml.org 
[lkml]   [2015]   [Apr]   [24]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] x86_64, asm: Work around AMD SYSRET SS descriptor attribute issue
On Fri, Apr 24, 2015 at 11:59:42AM +0200, Denys Vlasenko wrote:
> I propose a more conservative check:
>
> if (ss_sel != __KERNEL_DS)
> loadsegment(ss, __KERNEL_DS);
>
> I would propose this even if I would see no real case where it matters...
> but I even do see such a case.

...

> As in legacy mode, it is desirable to keep the stack-segment requestor privilege-level (SS.RPL)
> equal to the current privilege-level (CPL). When using a call gate to change privilege levels, the
> SS.RPL is updated to reflect the new CPL. The SS.RPL is restored from the return-target CS.RPL
> on the subsequent privilege-level-changing far return.
> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ THIS ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>
> 3. The old values of the SS and RSP registers are pushed onto the stack pointed to by the new RSP.
> ...
> ...
> """
>
> Thus, the NULL selector in SS may actually be not all zeros. Its RPL may be nonzero,
> if we are not in ring 0.

Yeah, that makes more sense. So I tested Andy's patch but changed it as
above and I get

$ taskset -c 0 ./sysret_ss_attrs_32
[RUN] Syscalls followed by SS validation
[OK] We survived

And this is on an AMD F16h and it used to fail before the patch. So
yeah, I think we can call this misfeature "architectural".

Tested-by: Borislav Petkov <bp@suse.de>

Thanks.

--
Regards/Gruss,
Boris.

ECO tip #101: Trim your mails when you reply.
--


\
 
 \ /
  Last update: 2015-04-24 13:41    [W:0.278 / U:1.832 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site