Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <976510d2-c7ad-2108-27e0-4c3b82c210f1@redhat.com>
Date:   Thu, 14 Jul 2022 13:36:22 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Thunderbird/91.10.0
Subject: Re: [PATCH] x86/kvm: fix FASTOP_SIZE when return thunks are enabled
Content-Language: en-US
To:     Peter Zijlstra <peterz@infradead.org>,
        Thadeu Lima de Souza Cascardo <cascardo@canonical.com>
Cc:     kvm@vger.kernel.org, linux-kernel@vger.kernel.org, x86@kernel.org,
        stable@vger.kernel.org,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Naresh Kamboju <naresh.kamboju@linaro.org>,
        Borislav Petkov <bp@suse.de>,
        Josh Poimboeuf <jpoimboe@kernel.org>,
        Linux Kernel Functional Testing <lkft@linaro.org>
References: <20220713171241.184026-1-cascardo@canonical.com>
 <Ys/ncSnSFEST4fgL@worktop.programming.kicks-ass.net>
From:   Paolo Bonzini <pbonzini@redhat.com>
In-Reply-To: <Ys/ncSnSFEST4fgL@worktop.programming.kicks-ass.net>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 7/14/22 11:52, Peter Zijlstra wrote:
> On Wed, Jul 13, 2022 at 02:12:41PM -0300, Thadeu Lima de Souza Cascardo wrote:
>> The return thunk call makes the fastop functions larger, just like IBT
>> does. Consider a 16-byte FASTOP_SIZE when CONFIG_RETHUNK is enabled.
>>
>> Otherwise, functions will be incorrectly aligned and when computing their
>> position for differently sized operators, they will executed in the middle
>> or end of a function, which may as well be an int3, leading to a crash
>> like:
> 
> Bah.. I did the SETcc stuff, but then forgot about the FASTOP :/
> 
>    af2e140f3420 ("x86/kvm: Fix SETcc emulation for return thunks")
> 
>> Fixes: aa3d480315ba ("x86: Use return-thunk in asm code")
>> Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@canonical.com>
>> Cc: Peter Zijlstra (Intel) <peterz@infradead.org>
>> Cc: Borislav Petkov <bp@suse.de>
>> Cc: Josh Poimboeuf <jpoimboe@kernel.org>
>> Cc: Paolo Bonzini <pbonzini@redhat.com>
>> Reported-by: Linux Kernel Functional Testing <lkft@linaro.org>
>> ---
>>   arch/x86/kvm/emulate.c | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/arch/x86/kvm/emulate.c b/arch/x86/kvm/emulate.c
>> index db96bf7d1122..d779eea1052e 100644
>> --- a/arch/x86/kvm/emulate.c
>> +++ b/arch/x86/kvm/emulate.c
>> @@ -190,7 +190,7 @@
>>   #define X16(x...) X8(x), X8(x)
>>   
>>   #define NR_FASTOP (ilog2(sizeof(ulong)) + 1)
>> -#define FASTOP_SIZE (8 * (1 + HAS_KERNEL_IBT))
>> +#define FASTOP_SIZE (8 * (1 + (HAS_KERNEL_IBT | IS_ENABLED(CONFIG_RETHUNK))))
> 
> Would it make sense to do something like this instead?

Yes, definitely.  Applied with a small tweak to make FASTOP_LENGTH
more similar to SETCC_LENGTH:
  
diff --git a/arch/x86/kvm/emulate.c b/arch/x86/kvm/emulate.c
index db96bf7d1122..0a15b0fec6d9 100644
--- a/arch/x86/kvm/emulate.c
+++ b/arch/x86/kvm/emulate.c
@@ -189,8 +189,12 @@
  #define X8(x...) X4(x), X4(x)
  #define X16(x...) X8(x), X8(x)
  
-#define NR_FASTOP (ilog2(sizeof(ulong)) + 1)
-#define FASTOP_SIZE (8 * (1 + HAS_KERNEL_IBT))
+#define NR_FASTOP	(ilog2(sizeof(ulong)) + 1)
+#define RET_LENGTH	(1 + (4 * IS_ENABLED(CONFIG_RETHUNK)) + \
+			 IS_ENABLED(CONFIG_SLS))
+#define FASTOP_LENGTH	(ENDBR_INSN_SIZE + 7 + RET_LENGTH)
+#define FASTOP_SIZE	(8 << ((FASTOP_LENGTH > 8) & 1) << ((FASTOP_LENGTH > 16) & 1))
+static_assert(FASTOP_LENGTH <= FASTOP_SIZE);
  
  struct opcode {
  	u64 flags;
@@ -442,8 +446,6 @@ static int fastop(struct x86_emulate_ctxt *ctxt, fastop_t fop);
   * RET | JMP __x86_return_thunk	[1,5 bytes; CONFIG_RETHUNK]
   * INT3				[1 byte; CONFIG_SLS]
   */
-#define RET_LENGTH	(1 + (4 * IS_ENABLED(CONFIG_RETHUNK)) + \
-			 IS_ENABLED(CONFIG_SLS))
  #define SETCC_LENGTH	(ENDBR_INSN_SIZE + 3 + RET_LENGTH)
  #define SETCC_ALIGN	(4 << ((SETCC_LENGTH > 4) & 1) << ((SETCC_LENGTH > 8) & 1))
  static_assert(SETCC_LENGTH <= SETCC_ALIGN);


Paolo