KVM's infrastructure for spectre mitigations in the vectors in v5.10 and earlier is different, it uses templates which are used to build a set of vectors at runtime.
There are two copy-and-paste errors in the templates: __spectre_bhb_loop_k24 should loop 24 times and __spectre_bhb_loop_k32 32.
Fix these.
Reported-by: Pavel Machek pavel@denx.de Link: https://lore.kernel.org/all/20220310234858.GB16308@amd/ Signed-off-by: James Morse james.morse@arm.com --- arch/arm64/kvm/hyp/smccc_wa.S | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/arch/arm64/kvm/hyp/smccc_wa.S b/arch/arm64/kvm/hyp/smccc_wa.S index 24b281912463..533b0aa73256 100644 --- a/arch/arm64/kvm/hyp/smccc_wa.S +++ b/arch/arm64/kvm/hyp/smccc_wa.S @@ -68,7 +68,7 @@ SYM_DATA_START(__spectre_bhb_loop_k24) esb sub sp, sp, #(8 * 2) stp x0, x1, [sp, #(8 * 0)] - mov x0, #8 + mov x0, #24 2: b . + 4 subs x0, x0, #1 b.ne 2b @@ -85,7 +85,7 @@ SYM_DATA_START(__spectre_bhb_loop_k32) esb sub sp, sp, #(8 * 2) stp x0, x1, [sp, #(8 * 0)] - mov x0, #8 + mov x0, #32 2: b . + 4 subs x0, x0, #1 b.ne 2b