[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20190724091438.GA31363@hirez.programming.kicks-ass.net>
Date: Wed, 24 Jul 2019 11:14:38 +0200
From: Peter Zijlstra <peterz@...radead.org>
To: Matt Mullins <mmullins@...com>
Cc: tglx@...utronix.de, luto@...nel.org,
Ingo Molnar <mingo@...hat.com>, Borislav Petkov <bp@...en8.de>,
"H. Peter Anvin" <hpa@...or.com>, x86@...nel.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH] x86/entry/32: pass cr2 to do_async_page_fault
On Tue, Jul 23, 2019 at 09:20:58PM -0700, Matt Mullins wrote:
> Commit a0d14b8909de ("x86/mm, tracing: Fix CR2 corruption") added the
> address parameter to do_async_page_fault, but does not pass it from the
> 32-bit entry point. To plumb it through, factor-out
> common_exception_read_cr2 in the same fashion as common_exception, and
> uses it from both page_fault and async_page_fault.
>
> For a 32-bit KVM guest, this fixes:
>
> [ 1.148669][ T1] Run /sbin/init as init process
> [ 1.153328][ T1] Starting init: /sbin/init exists but couldn't execute it (error -14)
>
> Fixes: a0d14b8909de ("x86/mm, tracing: Fix CR2 corruption")
> Signed-off-by: Matt Mullins <mmullins@...com>
Argh. Sorry for overlooking that one.
Acked-by: Peter Zijlstra (Intel) <peterz@...radead.org>
There's some potential for sharing between common_exception and your
common_exception_read_cr2, I'll see if I can do that without making a
mess of it.
Also; FB uses 32bit guests?!?
> ---
> arch/x86/entry/entry_32.S | 13 +++++++++----
> 1 file changed, 9 insertions(+), 4 deletions(-)
>
> diff --git a/arch/x86/entry/entry_32.S b/arch/x86/entry/entry_32.S
> index 2bb986f305ac..4f86928246e7 100644
> --- a/arch/x86/entry/entry_32.S
> +++ b/arch/x86/entry/entry_32.S
> @@ -1443,8 +1443,12 @@ BUILD_INTERRUPT3(hv_stimer0_callback_vector, HYPERV_STIMER0_VECTOR,
>
> ENTRY(page_fault)
> ASM_CLAC
> - pushl $0; /* %gs's slot on the stack */
> + pushl $do_page_fault
> + jmp common_exception_read_cr2
> +END(page_fault)
>
> +common_exception_read_cr2:
> + /* the function address is in %gs's slot on the stack */
> SAVE_ALL switch_stacks=1 skip_gs=1
>
> ENCODE_FRAME_POINTER
> @@ -1452,6 +1456,7 @@ ENTRY(page_fault)
>
> /* fixup %gs */
> GS_TO_REG %ecx
> + movl PT_GS(%esp), %edi
> REG_TO_PTGS %ecx
> SET_KERNEL_GS %ecx
>
> @@ -1463,9 +1468,9 @@ ENTRY(page_fault)
>
> TRACE_IRQS_OFF
> movl %esp, %eax # pt_regs pointer
> - call do_page_fault
> + CALL_NOSPEC %edi
> jmp ret_from_exception
> -END(page_fault)
> +END(common_exception_read_cr2)
>
> common_exception:
> /* the function address is in %gs's slot on the stack */
> @@ -1595,7 +1600,7 @@ END(general_protection)
> ENTRY(async_page_fault)
> ASM_CLAC
> pushl $do_async_page_fault
> - jmp common_exception
> + jmp common_exception_read_cr2
> END(async_page_fault)
> #endif
>
> --
> 2.17.1
>
Powered by blists - more mailing lists