On Tue, Mar 15, 2022 at 01:57:20PM +0000, James Morse wrote: > 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@xxxxxxx> > Link: https://lore.kernel.org/all/20220310234858.GB16308@amd/ > Signed-off-by: James Morse <james.morse@xxxxxxx> > --- > 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 > -- > 2.30.2 > Thanks, now queued up! greg k-h
