On 06.02.2018 12:21, Christian Borntraeger wrote:
> instead of having huge jump tables for function selection,
> lets use normal switch/case statements for the instruction
> handlers in priv.c
>
> bloat-o-meter shows that the saving are even bigger than
> just the removed jump tables.
>
> add/remove: 0/11 grow/shrink: 8/0 up/down: 1934/-10246 (-8312)
> Function old new delta
> kvm_s390_handle_b2 42 958 +916
> handle_iske 178 558 +380
> handle_rrbe 178 546 +368
> kvm_s390_handle_b9 42 222 +180
> kvm_s390_handle_01 42 74 +32
> kvm_s390_handle_eb 42 70 +28
> handle_sckpf 176 204 +28
> handle_lctlg 628 630 +2
> handle_ptff 36 - -36
> handle_sckpf.part 78 - -78
> handle_epsw 154 - -154
> handle_stfl 316 - -316
> handle_rrbe.part 470 - -470
> handle_iske.part 482 - -482
> handle_io_inst 518 - -518
> x01_handlers 2048 - -2048
> eb_handlers 2048 - -2048
> b9_handlers 2048 - -2048
> b2_handlers 2048 - -2048
>
> Signed-off-by: Christian Borntraeger <borntraeger@xxxxxxxxxx>
Reviewed-by: Janosch Frank <frankja@xxxxxxxxxxxxxxxxxx>
> -
> int kvm_s390_handle_b2(struct kvm_vcpu *vcpu)
> {
> - intercept_handler_t handler;
> -
> - /*
> - * A lot of B2 instructions are priviledged. Here we check for
> - * the privileged ones, that we can handle in the kernel.
> - * Anything else goes to userspace.
> - */
Shall we move that to kvm_handle_sie_intercept in a more general way?
Something like:
Here we handle all privileged and performance critical
instructions/interceptions everything else gets handled by userspace via
EOPNOTSUPP.
Attachment:
signature.asc
Description: OpenPGP digital signature
