On 6/4/21 8:52 AM, Alex Bennée wrote:
+void kvm_sve_get_supported_lens(ARMCPU *cpu, unsigned long *kvm_supported)
+{
+ /* Collect the set of vector lengths supported by KVM. */
+ bitmap_zero(kvm_supported, ARM_MAX_VQ);
+
+ if (kvm_arm_sve_supported()) {
+ kvm_arm_sve_get_vls(CPU(cpu), kvm_supported);
+ } else {
+ assert(!cpu_isar_feature(aa64_sve, cpu));
+ }
+}
+
+void kvm_sve_enable_lens(unsigned long *sve_vq_map,
+ unsigned long *sve_vq_init, uint32_t max_vq,
+ unsigned long *kvm_supported)
+{
+ /*
+ * For KVM we have to automatically enable all supported unitialized
+ * lengths, even when the smaller lengths are not all powers-of-two.
+ */
+ DECLARE_BITMAP(tmp, ARM_MAX_VQ);
+
+ bitmap_andnot(tmp, kvm_supported, sve_vq_init, max_vq);
+ bitmap_or(sve_vq_map, sve_vq_map, tmp, max_vq);
+}
+
+uint32_t kvm_sve_disable_lens(unsigned long *sve_vq_map,
+ unsigned long *sve_vq_init,
+ unsigned long *kvm_supported, Error **errp)
+{
+ uint32_t max_vq, vq;
+
+ /* Disabling a supported length disables all larger lengths. */
+ for (vq = 1; vq <= ARM_MAX_VQ; ++vq) {
+ if (test_bit(vq - 1, sve_vq_init) &&
+ test_bit(vq - 1, kvm_supported)) {
+ break;
+ }
+ }
+
+ max_vq = vq <= ARM_MAX_VQ ? vq - 1 : ARM_MAX_VQ;
+ bitmap_andnot(sve_vq_map, kvm_supported, sve_vq_init, max_vq);
+
+ if (max_vq == 0 || bitmap_empty(sve_vq_map, max_vq)) {
+ error_setg(errp, "cannot disable sve%d", vq * 128);
+ error_append_hint(errp, "Disabling sve%d results in all "
+ "vector lengths being disabled.\n",
+ vq * 128);
+ error_append_hint(errp, "With SVE enabled, at least one "
+ "vector length must be enabled.\n");
+ return 0;
+ }
+
+ return max_vq;
+}
+
+bool kvm_sve_validate_lens(unsigned long *sve_vq_map, uint32_t max_vq,
+ unsigned long *kvm_supported, Error **errp,
+ uint32_t sve_max_vq)
+{
+ /* Ensure the set of lengths matches what KVM supports. */
+ DECLARE_BITMAP(tmp, ARM_MAX_VQ);
+ uint32_t vq;
+
+ bitmap_xor(tmp, sve_vq_map, kvm_supported, max_vq);
+ if (bitmap_empty(tmp, max_vq)) {
+ return true;
+ }
+
+ vq = find_last_bit(tmp, max_vq) + 1;
+ if (test_bit(vq - 1, sve_vq_map)) {
+ if (sve_max_vq) {
+ error_setg(errp, "cannot set sve-max-vq=%d", sve_max_vq);
+ error_append_hint(errp, "This KVM host does not support "
+ "the vector length %d-bits.\n", vq * 128);
+ error_append_hint(errp, "It may not be possible to use "
+ "sve-max-vq with this KVM host. Try "
+ "using only sve<N> properties.\n");
+ } else {
+ error_setg(errp, "cannot enable sve%d", vq * 128);
+ error_append_hint(errp, "This KVM host does not support "
+ "the vector length %d-bits.\n", vq * 128);
+ }
+ } else {
+ error_setg(errp, "cannot disable sve%d", vq * 128);
+ error_append_hint(errp, "The KVM host requires all "
+ "supported vector lengths smaller "
+ "than %d bits to also be enabled.\n", max_vq * 128);
+ }
+ return false;
+}
Also, I think I mentioned this before in review of a previous version, I
strongly suspect that aside from kvm_sve_get_supported_lens, the other
functions will be used without change by all other hw virt systems.
So I think we'll do best, short term, to keep all of this in cpu-sve.c for now.
The tcg_enabled() test should elide all of the code that's not needed when
!CONFIG_TCG.
r~
