Re: [PATCH] x86/asm/entry/64: better check for canonical address

[Denys Vlasenko <vda.linux@googlemail.com>]

On Sun, Mar 29, 2015 at 11:12 PM, Andy Lutomirski <luto@amacapital.net> wrote:
> On Sun, Mar 29, 2015 at 12:36 PM, Denys Vlasenko
> <vda.linux@googlemail.com> wrote:
>> The instruction would need a differentiator whether returned-to code
>> is 64-bit or 32-bit.
>> Then it probably can use the same approach SYSRET{O,L} uses:
>> with REX.W, return is to 64-bit; without it, return is to 32-bit.
>>
>> Interrupt return then can check pt_regs->cs and use
>> IRETL_FAST if it is USER32_CS; use IRETQ_FAST if it is USER_CS
>> or KERNEL_CS; otherwise, fall back to slow but "universal" IRETQ.

Hmm. In fact since we'd need such checks, then instructions
can be even simpler: they don't even need to check CPL,
it can be hardcoded too. We'd need four instructions then:
return to 64 and to 32 bits, to CPL0 and to CPL3.


>> Do we have contacts at Intel to petition for this? :D
>
> Some of us do and have petitioned :)

And what did Intel say?

If there's any interest in doing this, Intel better *do* talk to us
before they commit to implementing it. Their track record
in implementing "fast syscalls" is nothing to write home about.
SYSENTER is a design disaster; SYSRET is buggy.