skip CONFIG_SCHED_STACK_END_CHECK requirement when CONFIG_VMAP_STACK is set about kconfig-hardened-check HOT 2 CLOSED

As I remember, SCHED_STACK_END_CHECK checks the magic value at the end of the kernel thread stack, and VMAP_STACK adds guard pages near it. So they do a bit different things, but VMAP_STACK is more reliable.

Yes, SCHED_STACK_END_CHECK checks a magic value at certain times such as exiting the kernel back to userspace, at which point the exploit can already have succeeded. The attacker may also have been able to clobber the value so that it's not detected. VMAP_STACK directly detects it with memory protection, which combined with making sure no large stack frames or VLAs exist prevents an overflow past the guard.

from kconfig-hardened-check.

Hello @thestinger,

As I remember, SCHED_STACK_END_CHECK checks the magic value at the end of the kernel thread stack, and VMAP_STACK adds guard pages near it. So they do a bit different things, but VMAP_STACK is more reliable.

I agree with your point.

Added the commit c0fc9e8.

from kconfig-hardened-check.