From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:38589) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1dv0Ej-0000y5-MX for qemu-devel@nongnu.org; Thu, 21 Sep 2017 08:05:49 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1dv0Ed-0007ev-Hq for qemu-devel@nongnu.org; Thu, 21 Sep 2017 08:05:45 -0400 Received: from mx1.redhat.com ([209.132.183.28]:50414) by eggs.gnu.org with esmtps (TLS1.0:DHE_RSA_AES_256_CBC_SHA1:32) (Exim 4.71) (envelope-from ) id 1dv0Ed-0007eY-3P for qemu-devel@nongnu.org; Thu, 21 Sep 2017 08:05:39 -0400 Message-Id: <20170921120415.241151389@redhat.com> Date: Thu, 21 Sep 2017 09:03:43 -0300 From: Marcelo Tosatti References: <20170921120341.590850215@redhat.com> Content-Disposition: inline; filename=qemu-toggle-kvmrt-prio Subject: [Qemu-devel] [patch 2/2] qemu: kvm: support RT_PRIO_HC hypercall List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: kvm@vger.kernel.org, qemu-devel@nongnu.org Cc: Marcelo Tosatti Add the following CPU options: allow-rt-prio-hc: allow guest to execute hypercall to change vcpu thread priority. rt-prio: SCHED_FIFO priority to be used when that hypercall is invoked. See kernel patchset for details about this hypercall. Signed-off-by: Marcelo Tosatti --- target/i386/cpu.c | 2 ++ target/i386/cpu.h | 4 ++++ target/i386/kvm.c | 23 +++++++++++++++++++++++ 3 files changed, 29 insertions(+) Index: qemu-fifoprio/target/i386/cpu.c =================================================================== --- qemu-fifoprio.orig/target/i386/cpu.c +++ qemu-fifoprio/target/i386/cpu.c @@ -4106,6 +4106,8 @@ static Property x86_cpu_properties[] = { false), DEFINE_PROP_BOOL("vmware-cpuid-freq", X86CPU, vmware_cpuid_freq, true), DEFINE_PROP_BOOL("tcg-cpuid", X86CPU, expose_tcg, true), + DEFINE_PROP_BOOL("allow-rt-prio-hc", X86CPU, allow_rt_prio_hc, false), + DEFINE_PROP_UINT32("rt-prio", X86CPU, rt_prio_hc, 0), DEFINE_PROP_END_OF_LIST() }; Index: qemu-fifoprio/target/i386/cpu.h =================================================================== --- qemu-fifoprio.orig/target/i386/cpu.h +++ qemu-fifoprio/target/i386/cpu.h @@ -1215,6 +1215,7 @@ struct X86CPU { bool hyperv_runtime; bool hyperv_synic; bool hyperv_stimer; + bool allow_rt_prio_hc; bool check_cpuid; bool enforce_cpuid; bool expose_kvm; @@ -1270,6 +1271,9 @@ struct X86CPU { /* Number of physical address bits supported */ uint32_t phys_bits; + /* RT priority of VCPU thread, when hypercall is invoked by guest */ + uint32_t rt_prio_hc; + /* in order to simplify APIC support, we leave this pointer to the user */ struct DeviceState *apic_state; Index: qemu-fifoprio/target/i386/kvm.c =================================================================== --- qemu-fifoprio.orig/target/i386/kvm.c +++ qemu-fifoprio/target/i386/kvm.c @@ -673,6 +673,28 @@ static int hyperv_handle_properties(CPUS static Error *invtsc_mig_blocker; +static int enable_allow_rt_prio_hc(CPUState *cs) +{ + int ret; + struct kvm_vcpu_rt_prio rt_prio; + X86CPU *cpu = X86_CPU(cs); + + if (!kvm_check_extension(kvm_state, KVM_CAP_VCPU_RT_PRIO_HC)) { + fprintf(stderr, "RT prio hypercall not supported by kernel\n"); + return -ENOSYS; + } + + rt_prio.sched_priority = cpu->rt_prio_hc; + rt_prio.enabled = 1; + + ret = kvm_vcpu_ioctl(cs, KVM_ENABLE_VCPU_RT_PRIO_HC, &rt_prio); + if (ret) { + fprintf(stderr, "RT prio hypercall failed, ret=%d\n", ret); + } + + return ret; +} + #define KVM_MAX_CPUID_ENTRIES 100 int kvm_arch_init_vcpu(CPUState *cs) @@ -758,6 +780,12 @@ int kvm_arch_init_vcpu(CPUState *cs) has_msr_hv_hypercall = true; } + if (cpu->allow_rt_prio_hc) { + if (enable_allow_rt_prio_hc(cs)) { + abort(); + } + } + if (cpu->expose_kvm) { memcpy(signature, "KVMKVMKVM\0\0\0", 12); c = &cpuid_data.entries[cpuid_i++];