From: Andrew Cooper <andrew.cooper3@citrix.com>
To: Peter Zijlstra <peterz@infradead.org>
Cc: Josh Poimboeuf <jpoimboe@kernel.org>,
x86@kernel.org, linux-kernel@vger.kernel.org,
alyssa.milburn@intel.com, scott.d.constable@intel.com,
joao@overdrivepizza.com, jose.marchesi@oracle.com,
hjl.tools@gmail.com, ndesaulniers@google.com,
samitolvanen@google.com, nathan@kernel.org, ojeda@kernel.org,
kees@kernel.org, alexei.starovoitov@gmail.com
Subject: Re: [PATCH 13/14] x86: BHI stubs
Date: Thu, 3 Oct 2024 14:59:11 +0100 [thread overview]
Message-ID: <630467e0-6cd4-441d-a2cd-070a002c6f95@citrix.com> (raw)
In-Reply-To: <20241003121739.GB17263@noisy.programming.kicks-ass.net>
On 03/10/2024 1:17 pm, Peter Zijlstra wrote:
> On Tue, Oct 01, 2024 at 12:20:02PM +0100, Andrew Cooper wrote:
>> On 01/10/2024 12:03 pm, Peter Zijlstra wrote:
>>> * nop4
>>> * call *%r11
>>>
>>> And lets take a random bhi function:
>>>
>>> + .align 16
>>> +SYM_INNER_LABEL(__bhi_args_0_1, SYM_L_LOCAL)
>>> + UNWIND_HINT_FUNC
>>> + cmovne %r10, %rdi
>>> + cmovne %r10, %rsi
>>> + ANNOTATE_UNRET_SAFE
>>> + ret
>>> + int3
>>>
>>> So the case you worry about is SUBL does *not* result in 0, but we
>>> speculate JZ true and end up in CALL, and do CMOVne.
>>>
>>> Since we speculated Z, we must then also not do the CMOV, so the value
>>> of R10 is irrelevant, it will not be used. The thing however is that
>>> CMOV will unconditionally put a store dependency on the target register
>>> (RDI, RSI in the above sequence) and as such any further speculative
>>> code trying to use those registers will stall.
>> How does that help?
>>
>> The write dependency doesn't stop a dependent load from executing in the
>> shadow of a mispredicted branch.
> I've been given to understand CMOVcc will kill any further speculation
> using the target register. So by 'poisoning' all argument registers that
> are involved with loads, we avoid any such load from happening during
> speculation.
IANAPA (I am not a pipeline architect), but AIUI,
CMOVcc establishes a data dependency between flags and the destination
register that doesn't exist in the pipeline if you'd used a conditional
branch instead.
It does prevent a dependent load from executing before the CMOVcc has
executed. But it does not stop that load from executing speculatively
eventually.
So, given the following case:
* SUB is/will results nonzero (ZF=0, %r10=nonzero)
* JZ predicted taken, despite (ZF=0)
we call __bhi_args_XXX wherein:
* CMOVNZ blocks until SUB executes (flags dependency)
* CMOVNZ eventually executes, and because ZF=0, it really does write
%r10 over the target registers
and then we enter the function with all pointers containing the nonzero
residual from the hash check.
Now, because it's a SUBL, the result is < 2^32, a straight deference of
one of these pointers will be blocked by SMAP (noone cares about 32bit,
or pre-SMAP hardware, right?)
Forward references from the pointers will be safe (assuming SIB doesn't
reach the canonical boundary), but backward references may wrap around
back into the kernel space. These will not be blocked by SMAP and will
spill their secrets if suitably provoked.
~Andrew
next prev parent reply other threads:[~2024-10-03 13:59 UTC|newest]
Thread overview: 60+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-09-27 19:48 [PATCH 00/14] x86/ibt: FineIBT-BHI Peter Zijlstra
2024-09-27 19:48 ` [PATCH 01/14] x86/cfi: Wreck things Peter Zijlstra
2024-09-27 23:15 ` Josh Poimboeuf
2024-09-28 13:31 ` Peter Zijlstra
2024-09-30 21:42 ` Josh Poimboeuf
2024-09-27 19:48 ` [PATCH 02/14] x86/boot: Mark start_secondary() with __noendbr Peter Zijlstra
2024-09-27 19:48 ` [PATCH 03/14] x86/alternative: Simplify callthunk patching Peter Zijlstra
2024-09-27 23:27 ` Josh Poimboeuf
2024-09-27 19:49 ` [PATCH 04/14] objtool/x86: Add .tail_call_sites Peter Zijlstra
2024-09-27 23:42 ` Josh Poimboeuf
2024-10-09 15:25 ` Peter Zijlstra
2024-10-10 4:55 ` Josh Poimboeuf
2024-09-27 19:49 ` [PATCH 05/14] objtool: Rename the skylake hack to --direct-call Peter Zijlstra
2024-09-27 19:49 ` [PATCH 06/14] x86/traps: Prepare for ENDBR poison UD1 usage Peter Zijlstra
2024-09-27 19:49 ` [PATCH 07/14] x86/ibt: Clean up is_endbr() Peter Zijlstra
2024-09-28 0:04 ` Josh Poimboeuf
2024-09-28 13:08 ` Peter Zijlstra
2024-09-29 17:32 ` Alexei Starovoitov
2024-09-30 8:30 ` Peter Zijlstra
2024-09-30 9:33 ` Peter Zijlstra
2024-09-30 16:43 ` Alexei Starovoitov
2024-09-30 20:58 ` Andrii Nakryiko
2024-09-27 19:49 ` [PATCH 08/14] x86/ibt: Clean up poison_endbr() Peter Zijlstra
2024-09-27 19:49 ` [PATCH 09/14] x86/ibt: Implement IBT+ Peter Zijlstra
2024-09-28 1:07 ` Josh Poimboeuf
2024-09-28 13:12 ` Peter Zijlstra
2024-09-29 17:38 ` Alexei Starovoitov
2024-09-30 8:23 ` Peter Zijlstra
2024-09-30 17:00 ` Alexei Starovoitov
2024-11-05 10:40 ` Peter Zijlstra
2024-09-27 19:49 ` [PATCH 10/14] x86/early_printk: Harden early_serial Peter Zijlstra
2024-09-27 19:49 ` [PATCH 11/14] llvm: kCFI pointer stuff Peter Zijlstra
2024-09-29 17:53 ` Alexei Starovoitov
2024-09-30 8:27 ` Peter Zijlstra
2024-09-30 16:59 ` Alexei Starovoitov
2024-10-01 10:21 ` Peter Zijlstra
2024-10-02 16:48 ` Alexei Starovoitov
2024-10-30 6:29 ` Constable, Scott D
2024-10-30 20:07 ` Constable, Scott D
2024-09-27 19:49 ` [PATCH 12/14] x86: Hacks for hacked up llvm Peter Zijlstra
2024-09-27 19:49 ` [PATCH 13/14] x86: BHI stubs Peter Zijlstra
2024-09-28 1:37 ` Josh Poimboeuf
2024-09-28 13:23 ` Peter Zijlstra
2024-09-30 21:30 ` Josh Poimboeuf
2024-09-30 21:46 ` Josh Poimboeuf
2024-09-30 22:23 ` Andrew Cooper
2024-09-30 22:38 ` Josh Poimboeuf
2024-09-30 22:52 ` Andrew Cooper
2024-10-01 11:03 ` Peter Zijlstra
2024-10-01 11:20 ` Andrew Cooper
2024-10-03 12:17 ` Peter Zijlstra
2024-10-03 13:59 ` Andrew Cooper [this message]
2024-10-14 17:50 ` Constable, Scott D
2024-10-14 21:54 ` Andrew Cooper
2024-10-21 15:06 ` Constable, Scott D
2024-10-29 5:59 ` Joao Moreira
2024-09-27 19:49 ` [PATCH 14/14] x86/fineibt: Add FineIBT+BHI mitigation Peter Zijlstra
2024-09-28 1:50 ` Josh Poimboeuf
2024-09-28 13:16 ` Peter Zijlstra
2024-10-28 5:45 ` Constable, Scott D
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=630467e0-6cd4-441d-a2cd-070a002c6f95@citrix.com \
--to=andrew.cooper3@citrix.com \
--cc=alexei.starovoitov@gmail.com \
--cc=alyssa.milburn@intel.com \
--cc=hjl.tools@gmail.com \
--cc=joao@overdrivepizza.com \
--cc=jose.marchesi@oracle.com \
--cc=jpoimboe@kernel.org \
--cc=kees@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=nathan@kernel.org \
--cc=ndesaulniers@google.com \
--cc=ojeda@kernel.org \
--cc=peterz@infradead.org \
--cc=samitolvanen@google.com \
--cc=scott.d.constable@intel.com \
--cc=x86@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®