Re: [PATCH v6 38/76] x86/head/64: Set CR4.FSGSBASE early

From: Borislav Petkov
Date: Sat Aug 29 2020 - 11:55:58 EST


On Mon, Aug 24, 2020 at 10:54:33AM +0200, Joerg Roedel wrote:
> From: Joerg Roedel <jroedel@xxxxxxx>
>
> Early exception handling will use rd/wrgsbase in paranoid_entry/exit.
> Enable the feature to avoid #UD exceptions on boot APs.
>
> Signed-off-by: Joerg Roedel <jroedel@xxxxxxx>
> Link: https://lore.kernel.org/r/20200724160336.5435-38-joro@xxxxxxxxxx
> ---
> arch/x86/kernel/head_64.S | 7 +++++++
> 1 file changed, 7 insertions(+)
>
> diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S
> index 08412f308de3..4622940134a5 100644
> --- a/arch/x86/kernel/head_64.S
> +++ b/arch/x86/kernel/head_64.S
> @@ -153,6 +153,13 @@ SYM_CODE_START(secondary_startup_64)
> orl $X86_CR4_LA57, %ecx
> 1:
> #endif
> +
> + ALTERNATIVE "jmp .Lstartup_write_cr4", "", X86_FEATURE_FSGSBASE
> +
> + /* Early exception handling uses FSGSBASE on APs */
> + orl $X86_CR4_FSGSBASE, %ecx

How is this supposed to work?

Alternatives haven't run that early yet and that piece of code looks
like this:

ffffffff81000067: eb 06 jmp ffffffff8100006f <secondary_startup_64+0x1f>
ffffffff81000069: 81 c9 00 00 01 00 or $0x10000,%ecx
ffffffff8100006f: 0f 22 e1 mov %rcx,%cr4

so we'll never set X86_CR4_FSGSBASE during early boot.

Stopping a guest with gdb just before that shows the same thing:

Dump of assembler code from 0x1000069 to 0x100007b:
=> 0x0000000001000069: eb 06 jmp 0x1000071
0x000000000100006b: 81 c9 00 00 01 00 or $0x10000,%ecx
0x0000000001000071: 0f 22 e1 mov %rcx,%cr4
0x0000000001000074: 48 03 05 95 ff 20 01 add 0x120ff95(%rip),%rax # 0x2210010

the unconditional JMP is there and it hasn't been patched out yet.

If you really need to test CPUID flags, you need to do something similar
to what verify_cpu does that early. And looking at that thing:

* verify_cpu, returns the status of longmode and SSE in register %eax.
* 0: Success 1: Failure

you could return the FSGSBASE CPUID bit there too and act accordingly.

Hmm.

--
Regards/Gruss,
Boris.

https://people.kernel.org/tglx/notes-about-netiquette