[RFC PATCH 03/36] arm64: hibernate: mask DAIF before restoring hibernated kernel

[Jinjie Ruan]

On 7/9/2026 8:13 PM, Vladimir Murzin wrote:
> From: Ada Couprie Diaz <ada.coupriediaz at arm.com>
> 
> The arm64 hibernate code manages the exception masking in an unsound
> way, leading to potential crashes and/or warnings during resume.
> 
> When a hibernation image is saved in `swsusp_arch_suspend()`, all DAIF
> exceptions are masked (by virtue of `local_daif_save()`), and the
> suspended image is saved assuming that all DAIF exceptions will remain
> masked when the image is restored.
> 
> When a hibernation image is resumed by `swsusp_arch_resume()`, only
> interrupts are masked (by virtue of `local_irq_save()` in
> `resume_target_kernel()`). When pseudo-NMI is enabled the DAIF.IF bits
> will be clear, and regardless of pseudo-NMI the DAIF.DA bits will be
> clear.
> 
> This means that there are two problems:
> 
> (1) It is possible to take Debug, SError, or pseudo-NMI exceptions
>     during the resume process. This is unsafe, as during the resume
>     process both the old ane new kernels will tranisently be in an
>     inconsistent state, and swsusp_arch_suspend_exit() won't retain
>     an executable mapping of any exception vectors.
> 
>     Any exception taken here will be fatal and silent.
> 
> (2) When re-entering the resumed kernel, some DAIF bits will be clear
>     unexpectedly. This permits Debug, SError, or pseudo-NMI exceptions
>     to be taken for a short period while the resumed kernel is not yet
>     in a consistent state.
> 
>     This is detected by CONFIG_ARM64_DEBUG_PRIORITY_MASKING.
> 
> Avoid these issues by masking all DAIF exceptions during resume.
> 
> Cc: stable at vger.kernel.org
> Signed-off-by: Ada Couprie Diaz <ada.coupriediaz at arm.com>
> Signed-off-by: Vladimir Murzin <vladimir.murzin at arm.com>
> ---
>  arch/arm64/kernel/hibernate.c | 13 +++++++++++++
>  1 file changed, 13 insertions(+)
> 
> diff --git a/arch/arm64/kernel/hibernate.c b/arch/arm64/kernel/hibernate.c
> index 9717568518ba..d0d9bd91e639 100644
> --- a/arch/arm64/kernel/hibernate.c
> +++ b/arch/arm64/kernel/hibernate.c
> @@ -405,6 +405,7 @@ int swsusp_arch_suspend(void)
>  int __nocfi swsusp_arch_resume(void)
>  {
>  	int rc;
> +	unsigned long flags;
>  	void *zero_page;
>  	size_t exit_size;
>  	pgd_t *tmp_pg_dir;
> @@ -465,9 +466,21 @@ int __nocfi swsusp_arch_resume(void)
>  	if (el2_reset_needed())
>  		__hyp_set_vectors(el2_vectors);
>  
> +	/*
> +	 * It is necessary to mask all DAIF exceptions here as:
> +	 *
> +	 * - The copy of swsusp_arch_suspend_exit() in the hibernation
> +	 *   text cannot handle taking any exceptions.
> +	 *
> +	 * - The suspended kernel masked all DAIF exceptions in
> +	 *   swsusp_arch_resume(), and expects to be re-entered in the
> +	 *   same state : with all DAIF exceptions masked.
> +	 */
> +	flags = local_daif_save();
>  	hibernate_exit(virt_to_phys(tmp_pg_dir), resume_hdr.ttbr1_el1,
>  		       resume_hdr.reenter_kernel, restore_pblist,
>  		       resume_hdr.__hyp_stub_vectors, virt_to_phys(zero_page));
> +	local_daif_restore(flags);

I believe that the local_daif_save() here is also unnecessary because
hibernate_exit() returns from the "return 0 branch" of
__cpu_suspend_enter() in swsusp_arch_suspend(), and before that,
local_daif_save() has already been called to mask all exceptions.

swsusp_arch_suspend(void)
    -> flags = local_daif_save();
    -> if (__cpu_suspend_enter(&state)) {
           ...
        } else {                // _cpu_resume <- hibernate_exit()

           ...
           in_suspend = 0;
           ...
        }
     -> local_daif_restore(flags);

>  
>  	return 0;
>  }