[PATCH bpf-next v5 2/3] arm64: patching: Add aarch64_insn_set()

[Mark Rutland]

On Fri, Sep 08, 2023 at 02:43:19PM +0000, Puranjay Mohan wrote:
> The BPF JIT needs to write invalid instructions to RX regions of memory
> to invalidate removed BPF programs. This needs a function like memset()
> that can work with RX memory.
> 
> Implement aarch64_insn_set() which is similar to text_poke_set() of x86.
> 
> Signed-off-by: Puranjay Mohan <puranjay12 at gmail.com>
> ---
>  arch/arm64/include/asm/patching.h |  1 +
>  arch/arm64/kernel/patching.c      | 40 +++++++++++++++++++++++++++++++
>  2 files changed, 41 insertions(+)
> 
> diff --git a/arch/arm64/include/asm/patching.h b/arch/arm64/include/asm/patching.h
> index f78a0409cbdb..551933338739 100644
> --- a/arch/arm64/include/asm/patching.h
> +++ b/arch/arm64/include/asm/patching.h
> @@ -8,6 +8,7 @@ int aarch64_insn_read(void *addr, u32 *insnp);
>  int aarch64_insn_write(void *addr, u32 insn);
>  
>  int aarch64_insn_write_literal_u64(void *addr, u64 val);
> +int aarch64_insn_set(void *dst, const u32 insn, size_t len);
>  void *aarch64_insn_copy(void *dst, const void *src, size_t len);
>  
>  int aarch64_insn_patch_text_nosync(void *addr, u32 insn);
> diff --git a/arch/arm64/kernel/patching.c b/arch/arm64/kernel/patching.c
> index 243d6ae8d2d8..63d9e0e77806 100644
> --- a/arch/arm64/kernel/patching.c
> +++ b/arch/arm64/kernel/patching.c
> @@ -146,6 +146,46 @@ noinstr void *aarch64_insn_copy(void *dst, const void *src, size_t len)
>  	return dst;
>  }
>  
> +/**
> + * aarch64_insn_set - memset for RX memory regions.
> + * @dst: address to modify
> + * @c: value to set
> + * @len: length of memory region.
> + *
> + * Useful for JITs to fill regions of RX memory with illegal instructions.
> + */
> +noinstr int aarch64_insn_set(void *dst, const u32 insn, size_t len)
> +{
> +	unsigned long flags;
> +	size_t patched = 0;
> +	size_t size;
> +	void *waddr;
> +	void *ptr;
> +
> +	/* A64 instructions must be word aligned */
> +	if ((uintptr_t)dst & 0x3)
> +		return -EINVAL;
> +
> +	raw_spin_lock_irqsave(&patch_lock, flags);
> +
> +	while (patched < len) {
> +		ptr = dst + patched;
> +		size = min_t(size_t, PAGE_SIZE - offset_in_page(ptr),
> +			     len - patched);
> +
> +		waddr = patch_map(ptr, FIX_TEXT_POKE0);
> +		memset32(waddr, insn, size / 4);

Do we need to use a specific instruction passed by the caller, or can we
hard-code a trapping instruction here?

If we don't care about the specific instruction, it'd be best to memset this to
0, as 0x00000000 is UDF #0 (which will trap), and that way memset can use DC
ZVA to zero the memory, which is faster than 4 bytes at a time.

If we did that, we can rename this to something like:

	aarch64_insn_clear(void *dst, size_t len)

> +		patch_unmap(FIX_TEXT_POKE0);
> +
> +		patched += size;
> +	}
> +	raw_spin_unlock_irqrestore(&patch_lock, flags);
> +
> +	caches_clean_inval_pou((uintptr_t)dst, (uintptr_t)dst + len);

Assuming the point of this is to trap if/when we accidentally execute the freed
instructions, we need an IPI here, and so this should use flush_icache_range()
or make it the caller's responsibility to do so.

Mark.

> +
> +	return 0;
> +}
> +
>  int __kprobes aarch64_insn_patch_text_nosync(void *addr, u32 insn)
>  {
>  	u32 *tp = addr;
> -- 
> 2.40.1
> 
>