KVM: arm64: pkvm: Simplify BTI handling on CPU boot

In order to perform an indirect branch to kvm_host_psci_cpu_entry()
on a BTI-aware system, we first branch to a 'BTI j' landing pad,
and from there branch again to the target.

While this works, this is really not required:

- BLR works with 'BTI c' and 'PACIASP' as the landing pad

- Even if LR gets clobbered by BLR, we are going to restore the
  host's registers, so it is pointless to try and avoid touching
  LR

Given the above, drop the veneer and directly call into C code.
If we were to come back from it, we'd directly enter the error
handler.

Reviewed-by: Fuad Tabba <tabba@google.com>
Tested-by: Fuad Tabba <tabba@google.com>
Link: https://patch.msgid.link/20260321212419.2803972-3-maz@kernel.org
Signed-off-by: Marc Zyngier <maz@kernel.org>
This commit is contained in:
Marc Zyngier 2026-03-21 21:24:16 +00:00
parent 4ebfa3230b
commit 1536a0b138
2 changed files with 5 additions and 14 deletions

View File

@ -291,13 +291,3 @@ SYM_CODE_START(__kvm_hyp_host_forward_smc)
ret
SYM_CODE_END(__kvm_hyp_host_forward_smc)
/*
* kvm_host_psci_cpu_entry is called through br instruction, which requires
* bti j instruction as compilers (gcc and llvm) doesn't insert bti j for external
* functions, but bti c instead.
*/
SYM_CODE_START(kvm_host_psci_cpu_entry)
bti j
b __kvm_host_psci_cpu_entry
SYM_CODE_END(kvm_host_psci_cpu_entry)

View File

@ -213,12 +213,13 @@ SYM_CODE_START_LOCAL(__kvm_hyp_init_cpu)
mov x0, x28
bl ___kvm_hyp_init // Clobbers x0..x2
/* Leave idmap. */
/* Leave idmap -- using BLR is OK, LR is restored from host context */
mov x0, x29
ldr x1, =kvm_host_psci_cpu_entry
br x1
ldr x1, =__kvm_host_psci_cpu_entry
blr x1
// The core booted in EL1. KVM cannot be initialized on it.
// The core booted in EL1, or the C code unexpectedly returned.
// Either way, KVM cannot be initialized on it.
1: wfe
wfi
b 1b