Re: [PATCH v3 29/39] arm64: mte: Switch GCR_EL1 in kernel entry and exit

[Catalin Marinas]

On Fri, Sep 25, 2020 at 12:50:36AM +0200, Andrey Konovalov wrote:
> diff --git a/arch/arm64/kernel/entry.S b/arch/arm64/kernel/entry.S
> index ff34461524d4..c7cc1fdfbd1a 100644
> --- a/arch/arm64/kernel/entry.S
> +++ b/arch/arm64/kernel/entry.S
> @@ -175,6 +175,49 @@ alternative_else_nop_endif
>  #endif
>  	.endm
>  
> +	.macro mte_set_gcr, tmp, tmp2
> +#ifdef CONFIG_ARM64_MTE
> +alternative_if_not ARM64_MTE
> +	b	1f
> +alternative_else_nop_endif

You don't need the alternative here. The macro is only invoked in an
alternative path already (I'd be surprised if it even works, we don't
handle nested alternatives well).

> +	/*
> +	 * Calculate and set the exclude mask preserving
> +	 * the RRND (bit[16]) setting.
> +	 */
> +	mrs_s	\tmp2, SYS_GCR_EL1
> +	bfi	\tmp2, \tmp, #0, #16
> +	msr_s	SYS_GCR_EL1, \tmp2
> +	isb
> +1:
> +#endif
> +	.endm
> +
> +	.macro mte_set_kernel_gcr, tsk, tmp, tmp2

What's the point of a 'tsk' argument here?

> +#ifdef CONFIG_KASAN_HW_TAGS
> +#ifdef CONFIG_ARM64_MTE

Does KASAN_HW_TAGS depend on ARM64_MTE already? Just to avoid too may
ifdefs. Otherwise, you can always write it as:

#if defined(CONFIG_KASAN_HW_TAGS) && defined(CONFIG_ARM64_MTE)

to save two lines (and its easier to read).

> +alternative_if_not ARM64_MTE
> +	b	1f
> +alternative_else_nop_endif
> +	ldr_l	\tmp, gcr_kernel_excl
> +
> +	mte_set_gcr \tmp, \tmp2
> +1:
> +#endif
> +#endif
> +	.endm
> +
> +	.macro mte_set_user_gcr, tsk, tmp, tmp2
> +#ifdef CONFIG_ARM64_MTE
> +alternative_if_not ARM64_MTE
> +	b	1f
> +alternative_else_nop_endif
> +	ldr	\tmp, [\tsk, #THREAD_GCR_EL1_USER]
> +
> +	mte_set_gcr \tmp, \tmp2
> +1:
> +#endif
> +	.endm
> +
>  	.macro	kernel_entry, el, regsize = 64
>  	.if	\regsize == 32
>  	mov	w0, w0				// zero upper 32 bits of x0
> @@ -214,6 +257,8 @@ alternative_else_nop_endif
>  
>  	ptrauth_keys_install_kernel tsk, x20, x22, x23
>  
> +	mte_set_kernel_gcr tsk, x22, x23
> +
>  	scs_load tsk, x20
>  	.else
>  	add	x21, sp, #S_FRAME_SIZE
> @@ -332,6 +377,8 @@ alternative_else_nop_endif
>  	/* No kernel C function calls after this as user keys are set. */
>  	ptrauth_keys_install_user tsk, x0, x1, x2
>  
> +	mte_set_user_gcr tsk, x0, x1
> +
>  	apply_ssbd 0, x0, x1
>  	.endif
>  
> diff --git a/arch/arm64/kernel/mte.c b/arch/arm64/kernel/mte.c
> index 393d0c794be4..c3b4f056fc54 100644
> --- a/arch/arm64/kernel/mte.c
> +++ b/arch/arm64/kernel/mte.c
> @@ -22,6 +22,8 @@
>  #include <asm/ptrace.h>
>  #include <asm/sysreg.h>
>  
> +u64 gcr_kernel_excl __ro_after_init;
> +
>  static void mte_sync_page_tags(struct page *page, pte_t *ptep, bool check_swap)
>  {
>  	pte_t old_pte = READ_ONCE(*ptep);
> @@ -116,6 +118,13 @@ void *mte_set_mem_tag_range(void *addr, size_t size, u8 tag)
>  	return ptr;
>  }
>  
> +void mte_init_tags(u64 max_tag)
> +{
> +	u64 incl = GENMASK(max_tag & MTE_TAG_MAX, 0);
> +
> +	gcr_kernel_excl = ~incl & SYS_GCR_EL1_EXCL_MASK;
> +}
> +
>  static void update_sctlr_el1_tcf0(u64 tcf0)
>  {
>  	/* ISB required for the kernel uaccess routines */
> @@ -151,7 +160,11 @@ static void update_gcr_el1_excl(u64 excl)
>  static void set_gcr_el1_excl(u64 excl)
>  {
>  	current->thread.gcr_user_excl = excl;
> -	update_gcr_el1_excl(excl);
> +
> +	/*
> +	 * SYS_GCR_EL1 will be set to current->thread.gcr_user_incl value
                                                      ^^^^^^^^^^^^^
That's gcr_user_excl now.

> +	 * by mte_restore_gcr() in kernel_exit,

I don't think mte_restore_gcr is still around in this patch.

-- 
Catalin