From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:48902) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1a3Zxw-000745-M6 for qemu-devel@nongnu.org; Mon, 30 Nov 2015 20:42:50 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1a3Zxu-00019l-T2 for qemu-devel@nongnu.org; Mon, 30 Nov 2015 20:42:48 -0500 Date: Tue, 1 Dec 2015 11:55:58 +1100 From: David Gibson Message-ID: <20151201005558.GC31343@voom.redhat.com> References: <1448024079-20808-1-git-send-email-bharata@linux.vnet.ibm.com> <1448024079-20808-5-git-send-email-bharata@linux.vnet.ibm.com> MIME-Version: 1.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="z4+8/lEcDcG5Ke9S" Content-Disposition: inline In-Reply-To: <1448024079-20808-5-git-send-email-bharata@linux.vnet.ibm.com> Subject: Re: [Qemu-devel] [PATCH v5 04/10] cpu: Reclaim vCPU objects List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Bharata B Rao Cc: Zhu Guihua , mdroth@linux.vnet.ibm.com, aik@ozlabs.ru, agraf@suse.de, qemu-devel@nongnu.org, Chen Fan , pbonzini@redhat.com, qemu-ppc@nongnu.org, tyreld@linux.vnet.ibm.com, nfont@linux.vnet.ibm.com, Gu Zheng , imammedo@redhat.com, afaerber@suse.de --z4+8/lEcDcG5Ke9S Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable On Fri, Nov 20, 2015 at 06:24:33PM +0530, Bharata B Rao wrote: > From: Gu Zheng >=20 > In order to deal well with the kvm vcpus (which can not be removed withou= t any > protection), we do not close KVM vcpu fd, just record and mark it as stop= ped > into a list, so that we can reuse it for the appending cpu hot-add reques= t if > possible. It is also the approach that kvm guys suggested: > https://www.mail-archive.com/kvm@vger.kernel.org/msg102839.html >=20 > Signed-off-by: Chen Fan > Signed-off-by: Gu Zheng > Signed-off-by: Zhu Guihua > Signed-off-by: Bharata B Rao > [- Explicit CPU_REMOVE() from qemu_kvm/tcg_destroy_vcpu() > isn't needed as it is done from cpu_exec_exit()] > --- > cpus.c | 41 +++++++++++++++++++++++++++++++++++++ > include/qom/cpu.h | 10 +++++++++ > include/sysemu/kvm.h | 1 + > kvm-all.c | 57 ++++++++++++++++++++++++++++++++++++++++++++++= +++++- > kvm-stub.c | 5 +++++ > 5 files changed, 113 insertions(+), 1 deletion(-) >=20 > diff --git a/cpus.c b/cpus.c > index 877bd70..af2b274 100644 > --- a/cpus.c > +++ b/cpus.c > @@ -953,6 +953,21 @@ void async_run_on_cpu(CPUState *cpu, void (*func)(vo= id *data), void *data) > qemu_cpu_kick(cpu); > } > =20 > +static void qemu_kvm_destroy_vcpu(CPUState *cpu) > +{ > + if (kvm_destroy_vcpu(cpu) < 0) { > + error_report("kvm_destroy_vcpu failed.\n"); > + exit(EXIT_FAILURE); > + } > + > + object_unparent(OBJECT(cpu)); > +} > + > +static void qemu_tcg_destroy_vcpu(CPUState *cpu) > +{ > + object_unparent(OBJECT(cpu)); > +} > + > static void flush_queued_work(CPUState *cpu) > { > struct qemu_work_item *wi; > @@ -1053,6 +1068,11 @@ static void *qemu_kvm_cpu_thread_fn(void *arg) > } > } > qemu_kvm_wait_io_event(cpu); > + if (cpu->exit && !cpu_can_run(cpu)) { > + qemu_kvm_destroy_vcpu(cpu); > + qemu_mutex_unlock(&qemu_global_mutex); This looks like a change to locking semantics, and I can't see the connection to the described purpose of the patch. > + return NULL; > + } > } > =20 > return NULL; > @@ -1108,6 +1128,7 @@ static void tcg_exec_all(void); > static void *qemu_tcg_cpu_thread_fn(void *arg) > { > CPUState *cpu =3D arg; > + CPUState *remove_cpu =3D NULL; > =20 > rcu_register_thread(); > =20 > @@ -1145,6 +1166,16 @@ static void *qemu_tcg_cpu_thread_fn(void *arg) > } > } > qemu_tcg_wait_io_event(QTAILQ_FIRST(&cpus)); > + CPU_FOREACH(cpu) { > + if (cpu->exit && !cpu_can_run(cpu)) { > + remove_cpu =3D cpu; > + break; > + } > + } > + if (remove_cpu) { > + qemu_tcg_destroy_vcpu(remove_cpu); > + remove_cpu =3D NULL; > + } Any particular reason to only cleanup one cpu per iteration? Also, any particular reason this isn't folded into tcg_exec_all with the other cpu->exit logic? > } > =20 > return NULL; > @@ -1301,6 +1332,13 @@ void resume_all_vcpus(void) > } > } > =20 > +void cpu_remove(CPUState *cpu) > +{ > + cpu->stop =3D true; > + cpu->exit =3D true; > + qemu_cpu_kick(cpu); > +} > + > /* For temporary buffers for forming a name */ > #define VCPU_THREAD_NAME_SIZE 16 > =20 > @@ -1506,6 +1544,9 @@ static void tcg_exec_all(void) > break; > } > } else if (cpu->stop || cpu->stopped) { > + if (cpu->exit) { > + next_cpu =3D CPU_NEXT(cpu); > + } > break; > } > } > diff --git a/include/qom/cpu.h b/include/qom/cpu.h > index 51a1323..67e05b0 100644 > --- a/include/qom/cpu.h > +++ b/include/qom/cpu.h > @@ -223,6 +223,7 @@ struct kvm_run; > * @halted: Nonzero if the CPU is in suspended state. > * @stop: Indicates a pending stop request. > * @stopped: Indicates the CPU has been artificially stopped. > + * @exit: Indicates the CPU has exited due to an unplug operation. > * @crash_occurred: Indicates the OS reported a crash (panic) for this C= PU > * @tcg_exit_req: Set to force TCG to stop executing linked TBs for this > * CPU and return to its top level loop. > @@ -274,6 +275,7 @@ struct CPUState { > bool created; > bool stop; > bool stopped; > + bool exit; > bool crash_occurred; > bool exit_request; > uint32_t interrupt_request; > @@ -696,6 +698,14 @@ void cpu_exit(CPUState *cpu); > void cpu_resume(CPUState *cpu); > =20 > /** > + * cpu_remove: > + * @cpu: The CPU to remove. > + * > + * Requests the CPU to be removed. > + */ > +void cpu_remove(CPUState *cpu); > + > +/** > * qemu_init_vcpu: > * @cpu: The vCPU to initialize. > * > diff --git a/include/sysemu/kvm.h b/include/sysemu/kvm.h > index b31f325..dd1b783 100644 > --- a/include/sysemu/kvm.h > +++ b/include/sysemu/kvm.h > @@ -206,6 +206,7 @@ int kvm_has_intx_set_mask(void); > =20 > int kvm_init_vcpu(CPUState *cpu); > int kvm_cpu_exec(CPUState *cpu); > +int kvm_destroy_vcpu(CPUState *cpu); > =20 > #ifdef NEED_CPU_H > =20 > diff --git a/kvm-all.c b/kvm-all.c > index c648b81..3befc59 100644 > --- a/kvm-all.c > +++ b/kvm-all.c > @@ -60,6 +60,12 @@ > =20 > #define KVM_MSI_HASHTAB_SIZE 256 > =20 > +struct KVMParkedVcpu { > + unsigned long vcpu_id; > + int kvm_fd; > + QLIST_ENTRY(KVMParkedVcpu) node; > +}; > + > struct KVMState > { > AccelState parent_obj; > @@ -93,6 +99,7 @@ struct KVMState > QTAILQ_HEAD(msi_hashtab, KVMMSIRoute) msi_hashtab[KVM_MSI_HASHTAB_SI= ZE]; > #endif > KVMMemoryListener memory_listener; > + QLIST_HEAD(, KVMParkedVcpu) kvm_parked_vcpus; > }; > =20 > KVMState *kvm_state; > @@ -235,6 +242,53 @@ static int kvm_set_user_memory_region(KVMMemoryListe= ner *kml, KVMSlot *slot) > return kvm_vm_ioctl(s, KVM_SET_USER_MEMORY_REGION, &mem); > } > =20 > +int kvm_destroy_vcpu(CPUState *cpu) > +{ > + KVMState *s =3D kvm_state; > + long mmap_size; > + struct KVMParkedVcpu *vcpu =3D NULL; > + int ret =3D 0; > + > + DPRINTF("kvm_destroy_vcpu\n"); > + > + mmap_size =3D kvm_ioctl(s, KVM_GET_VCPU_MMAP_SIZE, 0); > + if (mmap_size < 0) { > + ret =3D mmap_size; > + DPRINTF("KVM_GET_VCPU_MMAP_SIZE failed\n"); > + goto err; > + } > + > + ret =3D munmap(cpu->kvm_run, mmap_size); > + if (ret < 0) { > + goto err; > + } > + > + vcpu =3D g_malloc0(sizeof(*vcpu)); > + vcpu->vcpu_id =3D kvm_arch_vcpu_id(cpu); > + vcpu->kvm_fd =3D cpu->kvm_fd; > + QLIST_INSERT_HEAD(&kvm_state->kvm_parked_vcpus, vcpu, node); > +err: > + return ret; > +} > + > +static int kvm_get_vcpu(KVMState *s, unsigned long vcpu_id) > +{ > + struct KVMParkedVcpu *cpu; > + > + QLIST_FOREACH(cpu, &s->kvm_parked_vcpus, node) { > + if (cpu->vcpu_id =3D=3D vcpu_id) { > + int kvm_fd; > + > + QLIST_REMOVE(cpu, node); > + kvm_fd =3D cpu->kvm_fd; > + g_free(cpu); > + return kvm_fd; > + } > + } Hmm.. use of a simple list here does mean that unplugging, then replugging all (except 1) vcpus would be an O(n^2) operation. That's probably still alright, I guess. > + > + return kvm_vm_ioctl(s, KVM_CREATE_VCPU, (void *)vcpu_id); > +} > + > int kvm_init_vcpu(CPUState *cpu) > { > KVMState *s =3D kvm_state; > @@ -243,7 +297,7 @@ int kvm_init_vcpu(CPUState *cpu) > =20 > DPRINTF("kvm_init_vcpu\n"); > =20 > - ret =3D kvm_vm_ioctl(s, KVM_CREATE_VCPU, (void *)kvm_arch_vcpu_id(cp= u)); > + ret =3D kvm_get_vcpu(s, kvm_arch_vcpu_id(cpu)); > if (ret < 0) { > DPRINTF("kvm_create_vcpu failed\n"); > goto err; > @@ -1468,6 +1522,7 @@ static int kvm_init(MachineState *ms) > #ifdef KVM_CAP_SET_GUEST_DEBUG > QTAILQ_INIT(&s->kvm_sw_breakpoints); > #endif > + QLIST_INIT(&s->kvm_parked_vcpus); > s->vmfd =3D -1; > s->fd =3D qemu_open("/dev/kvm", O_RDWR); > if (s->fd =3D=3D -1) { > diff --git a/kvm-stub.c b/kvm-stub.c > index dc97a5e..0b39456 100644 > --- a/kvm-stub.c > +++ b/kvm-stub.c > @@ -32,6 +32,11 @@ bool kvm_allowed; > bool kvm_readonly_mem_allowed; > bool kvm_ioeventfd_any_length_allowed; > =20 > +int kvm_destroy_vcpu(CPUState *cpu) > +{ > + return -ENOSYS; > +} > + > int kvm_init_vcpu(CPUState *cpu) > { > return -ENOSYS; --=20 David Gibson | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_ | _way_ _around_! http://www.ozlabs.org/~dgibson --z4+8/lEcDcG5Ke9S Content-Type: application/pgp-signature; name="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1 iQIcBAEBAgAGBQJWXPAeAAoJEGw4ysog2bOSf1IP/A8V1hEbNnT8ESEkuGfFyMHx Vl6PSvaPvPikc+4hz4vnaHhtw/ebSOaFeaxDQSPVeBJ2tX45CZNbqf8Qd8lq9cSi YvXshDHtSkzDqZuKtFEMaJiCuTyt47HmE40RaxM1nD2m8Az9o/b85r7b5zwGx51i SXke7sLQ1YpQ/zxLUyRQdZmPItOpl3v0APWEZ3UKCSbclit4ESpQSD5Or06TLxrV rslzrfLRwGsVNeiCYhCVHUd6ppsTJF52z3YrsVTOYiaww8Lyvpredx8NgH0+fgpG Qtm3oSSBTFc5Ykzt+axkoGPBrIhXUFej+Bq76gNS4He1ndroH6xo52c8L3S2J9eh 3AdLCQh9A1Ncp+E/3n9uAIYU63Gn515r7rH5llJ96Id2kgOQzLN/NzdgHjRMOwkR MBDqKfsDn1lisgxqbVbq2T0PTTewxZAaf0hcKYVAZwFUR/XLWpbcjTfthRoV5TmS csW9dwwLOcKYIn5WqwqfNSZBBk3O05I4ATjeA4uo2CXgzj1e5iAnXNRnXhC2ai/M czIO8OEK/jrYodKMKW7QppsjfgJhdDDFsjsCDhaxv434pxCi5kROrujk8Om8tpxq 8aSJ+jbwD8LgPUqAA5jLxU+5T45JN3YDU7mVnmiFZWEUCPZijP3hG4l34UFBXWTD BvHQ/YyajN1FFb08QOyU =Istm -----END PGP SIGNATURE----- --z4+8/lEcDcG5Ke9S--