[PATCH 21/35] arm64/gcs: Implement shadow stack prctl() interface

Edgecombe, Rick P rick.p.edgecombe at intel.com
Tue Jul 18 10:51:54 PDT 2023


On Sun, 2023-07-16 at 22:51 +0100, Mark Brown wrote:
> +int arch_set_shadow_stack_status(struct task_struct *task, unsigned
> long arg)
> +{
> +       unsigned long gcs, size;
> +
> +       if (!system_supports_gcs())
> +               return -EINVAL;
> +
> +       if (is_compat_thread(task_thread_info(task)))
> +               return -EINVAL;
> +
> +       /* Reject unknown flags */
> +       if (arg & ~PR_SHADOW_STACK_SUPPORTED_STATUS_MASK)
> +               return -EINVAL;
> +
> +       /* If the task has been locked block any attempted changes */
> +       if (task->thread.gcs_el0_mode & PR_SHADOW_STACK_LOCK)
> +               return -EBUSY;
> +
> +       /* Drop flags other than lock if disabling */
> +       if (!(arg & PR_SHADOW_STACK_ENABLE))
> +               arg &= ~PR_SHADOW_STACK_LOCK;
> +
> +       /* If we are enabling GCS then make sure we have a stack */
> +       if (arg & PR_SHADOW_STACK_ENABLE) {
> +               if (!task_gcs_el0_enabled(task)) {
> +                       /* Do not allow GCS to be reenabled */
> +                       if (task->thread.gcs_base)
> +                               return -EINVAL;
> +
> +                       size = gcs_size(0);
> +                       gcs = alloc_gcs(task->thread.gcspr_el0, size,
> +                                       0, 0);
> +                       if (!gcs)
> +                               return -ENOMEM;
> +
> +                       task->thread.gcspr_el0 = gcs + size -
> sizeof(u64);
> +                       task->thread.gcs_base = gcs;
> +                       task->thread.gcs_size = size;
> +                       if (task == current)
> +                               write_sysreg_s(task-
> >thread.gcspr_el0,
> +                                              SYS_GCSPR_EL0);
> +
> +               }
> +       }
> +
> +       task->thread.gcs_el0_mode = arg;
> +       if (task == current)
> +               gcs_set_el0_mode(task);
> +

Ah! So does this task == current part mean this can be called from
another task via ptrace?

If so, then is the alloc_gcs() part on the wrong mm?


More information about the linux-arm-kernel mailing list