On 05/23/2019 03:52 PM, Ard Biesheuvel wrote:
> In order to avoid transient inconsistencies where freed code pages
> are remapped writable while stale TLB entries still exist on other
> cores, mark the kprobes text pages with the VM_FLUSH_RESET_PERMS
> attribute. This instructs the core vmalloc code not to defer the
> TLB flush when this region is unmapped and returned to the page
> allocator.
Makes sense.
>
> Signed-off-by: Ard Biesheuvel <[email protected]>
> ---
> arch/arm64/kernel/probes/kprobes.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/arch/arm64/kernel/probes/kprobes.c
> b/arch/arm64/kernel/probes/kprobes.c
> index 2509fcb6d404..036cfbf9682a 100644
> --- a/arch/arm64/kernel/probes/kprobes.c
> +++ b/arch/arm64/kernel/probes/kprobes.c
> @@ -131,8 +131,10 @@ void *alloc_insn_page(void)
> void *page;
>
> page = vmalloc_exec(PAGE_SIZE);
> - if (page)
> + if (page) {
> set_memory_ro((unsigned long)page, 1);
> + set_vm_flush_reset_perms(page);
> + }
Looks good. It seems there might be more users who would like to set
VM_FLUSH_RESET_PERMS right after their allocation for the same reason.
Hence would not it help to have a variant like vmalloc_exec_reset() or
such which will tag vm_struct->flags with VM_FLUSH_RESET_PERMS right
after it's allocation without requiring the caller to do the same.
