Re: [PATCH 1/2] KVM: VMX: Move IRQ invocation to assembly subroutine

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Mon, Sep 14, 2020 at 02:07:19PM -0700, Sean Christopherson wrote:
> 16-byte aligned to correctly mimic CPU behavior when vectoring an IRQ/NMI.
> When not changing stack, the CPU aligns RSP before pushing the frame.

16 byte alignment is not needed for the internal kernel ABI because it doesn't use
SSE.

-Andi

> 
> The above shenanigans work because the x86-64 ABI also requires RSP to be
> 16-byte aligned prior to CALL.  RSP is thus 8-byte aligned due to CALL
> pushing the return IP, and so creating the stack frame by pushing RBP makes
> it 16-byte aliagned again.
> 
> > > +
> > > +#ifdef CONFIG_X86_64
> > > +	push $__KERNEL_DS
> > > +	push %_ASM_BP
> > > +#endif
> > > +	pushf
> > > +	push $__KERNEL_CS
> > > +	CALL_NOSPEC _ASM_ARG1
> > > +
> > > +	/*
> > > +	 * "Restore" RSP from RBP, even though IRET has already unwound RSP to
> > > +	 * the correct value.  objtool doesn't know the target will IRET and so
> > > +	 * thinks the stack is getting walloped (without the explicit restore).
> > > +	 */
> > > +	mov %_ASM_BP, %rsp
> > > +	pop %_ASM_BP
> > > +	ret
> > 
> > BTW, there *is* actually an unwind hint for this situation:
> > UNWIND_HINT_RET_OFFSET.
> 
> I played with that one, but for the life of me couldn't figure out how to
> satisfy both the "stack size" and "cfa.offset" checks.  In the code below,
> cfa.offset will be 8, stack_size will be 40 and initial_func_cfi.cfa.offset
> will be 8.  But rereading this, I assume I missed something that would allow
> maniuplating cfa.offset?  Or maybe I botched my debugging?
> 
> static bool has_modified_stack_frame(struct instruction *insn, struct insn_state *state)
> {
> 	...
> 
>         if (cfi->cfa.offset != initial_func_cfi.cfa.offset + ret_offset)
>                 return true;
> 
>         if (cfi->stack_size != initial_func_cfi.cfa.offset + ret_offset)
>                 return true;
> 
> 	...
> }
>  
> > So you might be able to do something like the following (depending on
> > what your alignment requirements actually are):
> > 
> > SYM_FUNC_START(vmx_do_interrupt_nmi_irqoff)
> > #ifdef CONFIG_X86_64
> > 	push $__KERNEL_DS
> > 	push %_ASM_BP
> > #endif
> > 	pushf
> > 	push $__KERNEL_CS
> > 	CALL_NOSPEC _ASM_ARG1
> > 
> > 	/* The call popped the pushes */
> > 	UNWIND_HINT_RET_OFFSET sp_offset=32
> > 
> > 	ret
> > SYM_FUNC_END(vmx_do_interrupt_nmi_irqoff)
> > 
> > -- 
> > Josh
> > 



[Index of Archives]     [KVM ARM]     [KVM ia64]     [KVM ppc]     [Virtualization Tools]     [Spice Development]     [Libvirt]     [Libvirt Users]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite Questions]     [Linux Kernel]     [Linux SCSI]     [XFree86]

  Powered by Linux