Re: linux-next: manual merge of the tip tree with Linus' tree

[Mathieu Desnoyers <mathieu.desnoyers@efficios.com>]

----- On Feb 5, 2018, at 7:40 PM, Stephen Rothwell sfr@canb.auug.org.au wrote:

> Hi all,
> 
> Today's linux-next merge of the tip tree got a conflict in:
> 
>  arch/arm64/kernel/entry.S
> 
> between commit:
> 
>  4bf3286d29f3 ("arm64: entry: Hook up entry trampoline to exception vectors")
> 
> from Linus' tree and commit:
> 
>  f1e3a12b6543 ("membarrier/arm64: Provide core serializing command")
> 
> from the tip tree.
> 
> I fixed it up (probably not completely - see below) and can carry the
> fix as necessary. This is now fixed as far as linux-next is concerned,
> but any non trivial conflicts should be mentioned to your upstream
> maintainer when your tree is submitted for merging.  You may also want
> to consider cooperating with the maintainer of the conflicting tree to
> minimise any particularly complex conflicts.

The change introduced by 4bf3286d29 "arm64: entry: Hook up entry trampoline
to exception vectors" adds a trampoline as mechanism for return:

-       eret                                    // return to kernel
+
+#ifndef CONFIG_UNMAP_KERNEL_AT_EL0
+       eret
+#else
+       .if     \el == 0
+       bne     4f
+       msr     far_el1, x30
+       tramp_alias     x30, tramp_exit_native
+       br      x30
+4:
+       tramp_alias     x30, tramp_exit_compat
+       br      x30
+       .else
+       eret
+       .endif
+#endif

Which means that with CONFIG_UNMAP_KERNEL_AT_EL0, for return to EL0,
the eret is in the trampoline:

        .macro tramp_exit, regsize = 64
        adr     x30, tramp_vectors
        msr     vbar_el1, x30
        tramp_unmap_kernel      x30
        .if     \regsize == 64
        mrs     x30, far_el1
        .endif
        eret
        .endm

ENTRY(tramp_exit_native)
        tramp_exit
END(tramp_exit_native)

ENTRY(tramp_exit_compat)
        tramp_exit      32
END(tramp_exit_compat)


One approach I would consider for this is to duplicate this
comment and add it just above the "eret" instruction within the
macro:

	/*
	 * ARCH_HAS_MEMBARRIER_SYNC_CORE rely on eret context synchronization
	 * when returning from IPI handler, and when returning to user-space.
	 */

Or perhaps Will has something else in mind ?

Thanks,

Mathieu

> 
> --
> Cheers,
> Stephen Rothwell
> 
> diff --cc arch/arm64/kernel/entry.S
> index b34e717d7597,5edde1c2e93e..000000000000
> --- a/arch/arm64/kernel/entry.S
> +++ b/arch/arm64/kernel/entry.S
> @@@ -324,21 -302,11 +324,25 @@@ alternative_else_nop_endi
>  	ldp	x28, x29, [sp, #16 * 14]
>  	ldr	lr, [sp, #S_LR]
>  	add	sp, sp, #S_FRAME_SIZE		// restore sp
> +
> +	.if	\el == 0
> +alternative_insn eret, nop, ARM64_UNMAP_KERNEL_AT_EL0
> +#ifdef CONFIG_UNMAP_KERNEL_AT_EL0
> +	bne	4f
> +	msr	far_el1, x30
> +	tramp_alias	x30, tramp_exit_native
> +	br	x30
> +4:
> +	tramp_alias	x30, tramp_exit_compat
> +	br	x30
> +#endif
> +	.else
> + 	/*
> + 	 * ARCH_HAS_MEMBARRIER_SYNC_CORE rely on eret context synchronization
> + 	 * when returning from IPI handler, and when returning to user-space.
> + 	 */
> -	eret					// return to kernel
> +	eret
> +	.endif
>  	.endm
>  
>   	.macro	irq_stack_entry

-- 
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com