- Linaro-open-discussions - op-lists.linaro.org

by Salil Mehta

[+] Adding back LOD mailing list Hi Leo, > From: Leonardo Augusto Guimarães Garcia <leonardo.garcia(a)linaro.org> > Sent: Wednesday, August 30, 2023 1:59 PM > To: Marcin Juszkiewicz <marcin.juszkiewicz(a)linaro.org>; Jonathan Cameron > <jonathan.cameron(a)huawei.com>; Salil Mehta <salil.mehta(a)huawei.com> > Cc: Lorenzo Pieralisi <lorenzo.pieralisi(a)linaro.org>; Joyce Qi > <joyce.qi(a)linaro.org>; James Morse <james.morse(a)arm.com>; Salil Mehta > <salil.mehta(a)opnsrc.net>; Radoslaw Biernacki <rad(a)semihalf.com>; Leif > Lindholm <quic_llindhol(a)quicinc.com>; Peter Maydell <peter.maydell(a)linaro.org> > Subject: Re: [Linaro-open-discussions] Re: About the LOD next week (Aug 22th) > > > On 2023/08/30 09:11, Marcin Juszkiewicz wrote: > > W dniu 30.08.2023 o 13:59, Jonathan Cameron pisze: > >> Salil Mehta <salil.mehta(a)huawei.com> wrote: > > > >>> But I am not sure about qemu-sbsa. Maybe someone can help throw > >>> light on the goals of qemu-sbsa and who wants it? > >> > >> Architecture verification - not use as a VM except for that purpose. > >> It's not a replacement for virt - purpose is completely different. > >> > >> Of the current machines in QEMU the only one that I can see vCPU HP > >> mattering for is virt as that's the one targeting cloud VMs etc. > > > > Clouds use 'virt' as it was made for that purpose. And vCPU hotplug > > may have sense there. > > > > SBSA Reference Platform (sbsa-ref in QEMU) is emulation of physical > > computer (similar to 'q35' in x86-64 world). It does not use any of > > virtio expansions but emulate expansion cards so you get > > 'bochs-display' instead of 'virtio-gpu', 'ahci' instead of > > 'virtio-blk/scsi' etc. > > > > The goal is to have virtual system which can be used to check how > > operating systems behave on SBSA hardware. > > > > We are going for SBSA level 3 compliance and we are getting closer. > > > > Does vCPU hotplug make sense for sbsa-ref? Probably not as you do not > > hotplug cpus into your physical box, right? > > > This is correct. Today we don't have Arm physical machines that support > CPUs hot plug/unplug. > > Given that, I agree with Jonathan that the cVPU hot plug/unplug code > don't need to worry about SBSA QEMU implementation. There shouldn't be > any use case for that in the foreseeable future. Agreed for the reasons mentioned above. Thanks again for the confirmation. Thanks Salil. > > Cheers, > > Leo

1 year, 9 months

1
0
0 0

Re: About the LOD next week (Aug 22th)

by Salil Mehta

[+] Adding the LOD list for the obvious reasons Hi Marcin, > From: Marcin Juszkiewicz <marcin.juszkiewicz(a)linaro.org> > Sent: Wednesday, August 30, 2023 1:12 PM > To: Jonathan Cameron <jonathan.cameron(a)huawei.com>; Salil Mehta > <salil.mehta(a)huawei.com> > Cc: Leonardo Augusto Guimarães Garcia <leonardo.garcia(a)linaro.org>; Lorenzo > Pieralisi <lorenzo.pieralisi(a)linaro.org>; Joyce Qi <joyce.qi(a)linaro.org>; > James Morse <james.morse(a)arm.com>; Salil Mehta <salil.mehta(a)opnsrc.net>; > Radoslaw Biernacki <rad(a)semihalf.com>; Leif Lindholm > <quic_llindhol(a)quicinc.com>; Peter Maydell <peter.maydell(a)linaro.org> > Subject: Re: [Linaro-open-discussions] Re: About the LOD next week (Aug > 22th) > > W dniu 30.08.2023 o 13:59, Jonathan Cameron pisze: > > Salil Mehta <salil.mehta(a)huawei.com> wrote: > > >> But I am not sure about qemu-sbsa. Maybe someone can help throw > >> light on the goals of qemu-sbsa and who wants it? > > > > Architecture verification - not use as a VM except for that purpose. > > It's not a replacement for virt - purpose is completely different. > > > > Of the current machines in QEMU the only one that I can see vCPU HP > > mattering for is virt as that's the one targeting cloud VMs etc. > > Clouds use 'virt' as it was made for that purpose. And vCPU hotplug may > have sense there. > > SBSA Reference Platform (sbsa-ref in QEMU) is emulation of physical > computer (similar to 'q35' in x86-64 world). It does not use any of > virtio expansions but emulate expansion cards so you get 'bochs-display' > instead of 'virtio-gpu', 'ahci' instead of 'virtio-blk/scsi' etc. > > The goal is to have virtual system which can be used to check how > operating systems behave on SBSA hardware. > > We are going for SBSA level 3 compliance and we are getting closer. Thanks for confirming this and sharing this useful information. > > Does vCPU hotplug make sense for sbsa-ref? Probably not as you do not > hotplug cpus into your physical box, right? Agreed, for the reference verification platform it does not. Thanks Salil.

1 year, 9 months

1
0
0 0

[Question] Bugzilla Id for the vCPU Hotplug related _OSC capability Bits

by Salil Mehta

Hi James, Do we have Bugzilla Id for the vCPU Hotplug related _OSC capability Bits (Present and enabled) change proposed for ACPI Specification? Thanks Salil

1 year, 10 months

1
0
0 0

Linaro OP-TEE Contribution, LOC, monthly meeting August 22

by Jens Wiklander

Hi, Time flies, on Tuesday, August 22 it's for another LOC monthly meeting. For time and connection details see the calendar at https://www.trustedfirmware.org/meetings/ I'm happy to report that the Xen patches needed to run OP-TEE with FF-A have just been merged [1] and will be included in the next Xen release. With this, we may need to focus more on for how long we may hog the CPU with non-secure interrupts masked. [1] https://patchew.org/Xen/20230731121536.934239-1-jens.wiklander@linaro.org/#… Any other topics? Thanks, Jens

1 year, 10 months

1
0
0 0

Re: [External] : RE: vCPU hotplug/unplug for the ARMv8 processors

by Salil Mehta

Hi Vishnu, I got tied down with the refactoring of the patches in preparation of the patch-set and hence could not follow this bug last week. Thanks for your testing efforts and also suggesting a probable fix. > From: Vishnu Pajjuri <vishnu(a)amperemail.onmicrosoft.com> > Sent: Tuesday, August 8, 2023 12:53 PM > To: Salil Mehta <salil.mehta(a)huawei.com>; Vishnu Pajjuri OS > <vishnu(a)os.amperecomputing.com>; Salil Mehta <salil.mehta(a)opnsrc.net>; > Miguel Luis <miguel.luis(a)oracle.com> > Cc: linaro-open-discussions(a)op-lists.linaro.org > Subject: Re: [External] : RE: vCPU hotplug/unplug for the ARMv8 processors > > Hi Salil, > Below patch resolves VM live migration CPU stall issues. Good catch. > diff --git a/hw/arm/virt.c b/hw/arm/virt.c > index e9ced9f172..23b06d44cc 100644 > --- a/hw/arm/virt.c > +++ b/hw/arm/virt.c > @@ -3224,6 +3224,11 @@ static void virt_cpu_plug(HotplugHandler > *hotplug_dev, DeviceState *dev, > fw_cfg_modify_i16(vms->fw_cfg, FW_CFG_NB_CPUS, vms->boot_cpus); > > cs->disabled = false; > + if (qemu_enabled_cpu(cs)) { > + cs->vmcse = > qemu_add_vm_change_state_handler(kvm_arm_vm_state_change, cs); > + } > + This is not the correct location of the fix. > return; > fail: > error_propagate(errp, local_err); > diff --git a/target/arm/kvm64.c b/target/arm/kvm64.c > index 80a456ef16..510f829e4b 100644 > --- a/target/arm/kvm64.c > +++ b/target/arm/kvm64.c > @@ -883,9 +883,6 @@ int kvm_arch_init_vcpu(CPUState *cs) > return -EINVAL; > } > > - if (qemu_enabled_cpu(cs)) > - cs->vmcse = > qemu_add_vm_change_state_handler(kvm_arm_vm_state_change, > - cs); We cannot remove this. Its corresponding destruction leg also has similar problem and needs to be fixed as well. Installation and un-installation must happen when realization of vCPUs is happening. > > /* Determine init features for this CPU */ > memset(cpu->kvm_init_features, 0, sizeof(cpu->kvm_init_features)); > > Root Cause: > It's too early to attach kvm_arm_vm_state_change handler in > kvm_arch_init_vcpu > since vcpus were not yet enabled at this stage Yes, the problem is related to updating of virtual time but it is not so much about the time when these handlers must be installed. VM State change handler must be installed during realization phase of the vCPU thread and when vCPU exits, these handlers must be uninstalled as well. And even now, this is getting done at the right place and time. The problem is due to the wrong qemu_enabled_cpu() check, none of the VM state change handlers are getting installed. Hence, the jump in the virtual time becomes large during migration as the adjustment of the virtual time with Host KVM is not happening. > > Because of this, QEMU is not fetching virtual time from KVM before saving > VM state > Also not able to restore virtual time back to KVM after restoring VM state > It is causing misconfiguration of vtimer offset and leading to CPU stall > issues. This is correct. But the reason why this is happening is due to absence of the VM state change handlers at the first place. And what you just mentioned is the visible side affect of the real cause mentioned earlier above. > Attaching kvm_arm_vm_state_change handler in virt_cpu_plug after enabling > each vCPU > makes more reasonable to me. Any suggestions are welcome here. As mentioned above, we should not move installation of handler to any context other than the vCPU thread. This is because when vCPU exits, its symmetric destruction legs ensures uninstallation of the VM state change handler. Having these at 2 different places is an invitation to many problems and actually not required at all. > This helps to properly configure vtimer offset. And didn't see any timing > issues Sure, I have pushed the fix in the below commit. Please pull the change and check if everything works for you. This will ensure that handlers are only installed when vCPU threads are being spawned and vCPU is being realized. https://github.com/salil-mehta/qemu/commit/a220a9e22f8d6452e8c9759e05a12b75… diff --git a/target/arm/kvm64.c b/target/arm/kvm64.c index 80a456ef16..96ceebc7cc 100644 --- a/target/arm/kvm64.c +++ b/target/arm/kvm64.c @@ -883,7 +883,7 @@ int kvm_arch_init_vcpu(CPUState *cs) return -EINVAL; } - if (qemu_enabled_cpu(cs)) + if (cs->thread_id) cs->vmcse = qemu_add_vm_change_state_handler(kvm_arm_vm_state_change, cs); @@ -962,7 +962,7 @@ int kvm_arch_init_vcpu(CPUState *cs) int kvm_arch_destroy_vcpu(CPUState *cs) { - if (qemu_enabled_cpu(cs)) + if (cs->thread_id) qemu_del_vm_change_state_handler(cs->vmcse); return 0; Meny thanks Salil. > > Regards, > -Vishnu. > On 26-07-2023 15:15, Salil Mehta wrote: > HI Vishnu, > Got it. I think I am now able to recreate the condition successfully with > just 2 vCPUs and even when all of them are present and online. > I am analyzing the issue and will get back to you soon. > > Thanks > Salil > From: Vishnu Pajjuri OS mailto:vishnu@os.amperecomputing.com > Sent: Friday, July 21, 2023 10:39 AM > To: Salil Mehta mailto:salil.mehta@opnsrc.net; Salil Mehta > mailto:salil.mehta@huawei.com > Cc: Vishnu Pajjuri OS mailto:vishnu@os.amperecomputing.com; Miguel Luis > mailto:miguel.luis@oracle.com > Subject: Re: [External] : RE: vCPU hotplug/unplug for the ARMv8 processors > > Hi Salil, > On 20-07-2023 22:56, Salil Mehta wrote: > [+] For your reference, I am attaching the scripts which I am using. Note > the subtle difference between both > Thanks for sharing your scripts for VM live migration. > > > > Thanks! > On Thu, Jul 20, 2023 at 12:57 PM Salil Mehta > <mailto:salil.mehta@huawei.com> wrote: > Hi Vishnu, > Aah, I see looks like you have 2 physical machine? I am using same host > setup and use below commands to migrate VMs > > Src VM (Same Host) > Qemu> migrate tcp:0:4444 > > Dst VM (Same Host) > $ qemu-arm64-nosmmu […] incoming tcp:0:4444 > I have prepared setup like below on the same Host with the help of qemu > commands only > [VM1 (P4:C2:H0:U2:O2)] --> [VM2 (P4:C2:H0:U2:O2)] --> [VM3 > (P4:C2:H0:U2:O2)] > From VM1 to VM2 works fine. But when I tried to do live migration from > VM2(same VM migrated from VM1) to VM3 > I observed CPU stall issues in this scenario also. > One more observation is that time is jumping in the VM2 to VM3 iteration > always > [root@fedora ~]# date > Wed Nov 21 06:34:54 AM EST 2114 > VM1.sh > TOPDIR=/home/vishnu/vcpu_hp > BASECODEPATH=$TOPDIR/pre_v2 > QEMUPATH=$BASECODEPATH/qemu/ > QEMUBIN=$QEMUPATH/build/aarch64-softmmu/qemu-system-aarch64 > ROOTFS_IMG=/home/vishnu/vms/images/vm11.raw > BIOS=/usr/share/AAVMF/AAVMF_CODE.fd > LOG_FILE=/tmp/serial1.log > > $QEMUBIN -M virt,accel=kvm,gic-version=3 \ > -smp cpus=2,maxcpus=4 \ > -m 4g \ > -cpu host \ > -nodefaults \ > -nographic \ > -drive if=pflash,format=raw,readonly,file=$BIOS \ > -boot c -hda $ROOTFS_IMG \ > -chardev stdio,id=char0,logfile=$LOG_FILE,signal=off \ > -serial chardev:char0 \ > -monitor tcp:127.0.0.1:55555,server,nowait > VM2.sh > TOPDIR=/home/vishnu/vcpu_hp > BASECODEPATH=$TOPDIR/pre_v2 > QEMUPATH=$BASECODEPATH/qemu/ > QEMUBIN=$QEMUPATH/build/aarch64-softmmu/qemu-system-aarch64 > ROOTFS_IMG=/home/vishnu/vms/images/vm11.raw > BIOS=/usr/share/AAVMF/AAVMF_CODE.fd > LOG_FILE=/tmp/serial1.log > > $QEMUBIN -M virt,accel=kvm,gic-version=3 \ > -smp cpus=2,maxcpus=4 \ > -m 4g \ > -cpu host \ > -nodefaults \ > -nographic \ > -drive if=pflash,format=raw,readonly,file=$BIOS \ > -boot c -hda $ROOTFS_IMG \ > -chardev stdio,id=char0,logfile=$LOG_FILE,signal=off \ > -serial chardev:char0 \ > -monitor tcp:127.0.0.1:55556,server,nowait \ > -incoming tcp:0:4444 > VM3.sh > TOPDIR=/home/vishnu/vcpu_hp > BASECODEPATH=$TOPDIR/pre_v2 > QEMUPATH=$BASECODEPATH/qemu/ > QEMUBIN=$QEMUPATH/build/aarch64-softmmu/qemu-system-aarch64 > ROOTFS_IMG=/home/vishnu/vms/images/vm11.raw > BIOS=/usr/share/AAVMF/AAVMF_CODE.fd > LOG_FILE=/tmp/serial1.log > > $QEMUBIN -M virt,accel=kvm,gic-version=3 \ > -smp cpus=2,maxcpus=4 \ > -m 4g \ > -cpu host \ > -nodefaults \ > -nographic \ > -drive if=pflash,format=raw,readonly,file=$BIOS \ > -boot c -hda $ROOTFS_IMG \ > -chardev stdio,id=char0,logfile=$LOG_FILE,signal=off \ > -serial chardev:char0 \ > -monitor tcp:127.0.0.1:55557,server,nowait \ > -incoming tcp:0:4445 > Log:- > [root@fedora ~]# [ 189.884992] rcu: INFO: rcu_sched detected stalls on > CPUs/tasks: > [ 189.887474] rcu: 1-...!: (0 ticks this GP) idle=c268/0/0x0 > softirq=13837/13837 fqs=0 (false positive?) > [ 189.891022] rcu: (detected by 0, t=6002 jiffies, g=7521, q=764 > ncpus=2) > [ 189.893494] Task dump for CPU 1: > [ 189.894875] task:swapper/1 state:R running > task stack:0 pid:0 ppid:1 flags:0x00000008 > [ 189.898597] Call trace: > [ 189.899505] __switch_to+0xc8/0xf8 > [ 189.900825] 0xffff8000080ebde0 > [ 189.902236] rcu: rcu_sched kthread timer wakeup didn't happen for 6001 > jiffies! g7521 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x402 > [ 189.906499] rcu: Possible timer handling issue on cpu=1 timer- > softirq=3803 > [ 189.909125] rcu: rcu_sched kthread starved for 6002 jiffies! g7521 f0x0 > RCU_GP_WAIT_FQS(5) ->state=0x402 ->cpu=1 > [ 189.912934] rcu: Unless rcu_sched kthread gets sufficient CPU time, > OOM is now expected behavior. > [ 189.916226] rcu: RCU grace-period kthread stack dump: > [ 189.918122] task:rcu_sched state:I > stack:0 pid:15 ppid:2 flags:0x00000008 > [ 189.921241] Call trace: > [ 189.922165] __switch_to+0xc8/0xf8 > [ 189.923453] __schedule+0x248/0x6c0 > [ 189.924821] schedule+0x58/0xf0 > [ 189.926010] schedule_timeout+0x90/0x178 > [ 189.927483] rcu_gp_fqs_loop+0x124/0x438 > [ 189.928969] rcu_gp_kthread+0x164/0x1c8 > [ 189.930401] kthread+0xe0/0xf0 > [ 189.931544] ret_from_fork+0x10/0x20 > [ 189.932874] rcu: Stack dump where RCU GP kthread last ran: > [ 189.935207] Task dump for CPU 1: > [ 189.936338] task:swapper/1 state:R running > task stack:0 pid:0 ppid:1 flags:0x00000008 > [ 189.939811] Call trace: > [ 189.940673] __switch_to+0xc8/0xf8 > [ 189.941888] 0xffff8000080ebde0 > Lets first use Qemu prompt to migrate instead of web interface if you can? > This will also help us converge whether problem is an interface problem or > we are really missing something. > Could you please share your Linux and Qemu source repos as well? > I'll try to figure out if I'm using any wrong source repositories. > > Regards, > -Vishnu. > Thanks > Salil > From: Vishnu Pajjuri OS <mailto:vishnu@os.amperecomputing.com> > Sent: Thursday, July 20, 2023 12:49 PM > To: Salil Mehta <mailto:salil.mehta@huawei.com>; Vishnu Pajjuri OS > <mailto:vishnu@os.amperecomputing.com>; Miguel Luis > <mailto:miguel.luis@oracle.com> > Cc: Salil Mehta <mailto:salil.mehta@opnsrc.net> > Subject: Re: [External] : RE: vCPU hotplug/unplug for the ARMv8 processors > > Hi Salil, > On 20-07-2023 16:58, Salil Mehta wrote: > Hi Vishnu, > Can I request you to reduce the number of vCPUs to 6. > > Test Case: > [VM1 (P6:C4:H0:U2:O4)] --> [VM2 (P6:C4:H0:U2:O4)] > <Exit from qemu prompt. Restart VM1 and then execute below> > With virsh(libvirt) commands, > After Migrating VM1 to DstHost, there is nothing left on SrcHost. Even ps > command shows nothing on SrcHost regarding qemu command. > So operating VM1 on SrcHost after migrating VM1 to DstHost is not possible. > On DstHost side VM1 is live, there I can able to access VM1's console. > I can able to manipulate VM1 on DstHost after migration. > Regards, > -Vishnu. > [VM1 (P6:C4:H0:U2:O4)] <-- [VM2 (P6:C4:H0:U2:O4)] > > [...‘n’ times...] > > [VM1 (P6:C4:H0:U2:O4)] --> [VM2 (P6:C4:H0:U2:O4)] > <Exit from qemu prompt. Restart VM1 and then execute below> > [VM1 (P6:C4:H0:U2:O4)] <-- [VM2 (P6:C4:H0:U2:O4)] > > Thanks > Salil. > > From: Vishnu Pajjuri OS mailto:vishnu@os.amperecomputing.com > Sent: Thursday, July 20, 2023 12:16 PM > To: Salil Mehta mailto:salil.mehta@huawei.com; Vishnu Pajjuri OS > mailto:vishnu@os.amperecomputing.com; Miguel Luis > mailto:miguel.luis@oracle.com > Cc: Salil Mehta mailto:salil.mehta@opnsrc.net > Subject: Re: [External] : RE: vCPU hotplug/unplug for the ARMv8 processors > > Hi Salil, > On 20-07-2023 15:52, Salil Mehta wrote: > Hi Vishnu, > > From: Vishnu Pajjuri OS mailto:vishnu@os.amperecomputing.com > Sent: Tuesday, July 18, 2023 12:39 PM > > [...] > > @Miguel/Vishnu BTW, I was able to reproduce the CPU stall issue and it > looks > to me it only happens when number of online vCPUs being migrated at source > VM > do not match number of vCPUs being online'd during boot at the destination > VM. > This means "maxcpus=N" kernel parameter at the destination VM should have > value equal to number of online'd vCPUs being migrated. > > I still need to check if this matches the x86_64 behavior. If possible, can > you guys confirm below: > > [1] always keep the online vcpus same at source and destination and then > wait > and check for the CPU stall messages? > Hopefully below is the test scenario that you are looking at, but still I > observed CPU stall issues > > P: Possible vCPUs > C: Cold Plugged vCPUs > H: Hot Plugged vCPUs > U: Unplugged vCPUs > O: Onlined vCPUs > sVM: Source VM > dVM: Destination VM > > > Test Case: > [sVM (P80:C72:H0:U8:O72)] --> [dVM (P80:C72:H0:U8:O72)] - PASS > > Migrate the same VM back > > [sVM (P80:C72:H0:U8:O72)] <-- [dVM (P80:C72:H0:U8:O72)] - Observed CPU > stalls on VM > > > I have now repeated above scenario at-least 5 times and cannot reproduce > stalls > in the case you have mentioned. > Good hear that you didn't observed any CPU stall issues with your setup. > I'm using virsh commands to create VM live setup. If possible I'll adapt > your qemu commands to libvirt, > Otherwise I would like to reproduce your setup, > for that could you please provide your kernel and qemu repo's > Also your complete Qemu command of VM launch and migration > Commands you have been using both at source and destination side? > > Few questions: > Q1: In above case, there is no Hotplug happening at all - right? > Ans: Yes > > Q2: Do you restart the source VM after it has been migrated once to the > destination? > Ans: No. > > > > Q3: Can you please paste your complete Qemu command of VM launch and > migration > Commands you have been using both at source and destination side? > Ans: I'm using virsh commands to migrate VMs > > Commands on SrcHost: > qemu command after running with virsh command: > virsh start vm1 --console > /usr/local/bin/qemu-system-aarch64 -name guest=vm1,debug-threads=on -S - > object {"qom- > type":"secret","id":"masterKey0","format":"raw","file":"/var/lib/libvirt/qe > mu/domain-4-vm1/master-key.aes"} -blockdev > {"driver":"file","filename":"/usr/share/edk2/aarch64/QEMU_EFI-silent- > pflash.raw","node-name":"libvirt-pflash0-storage","auto-read- > only":true,"discard":"unmap"} -blockdev {"node-name":"libvirt-pflash0- > format","read-only":true,"driver":"raw","file":"libvirt-pflash0-storage"} - > blockdev > {"driver":"file","filename":"/var/lib/libvirt/qemu/nvram/vm1_VARS.fd","node > -name":"libvirt-pflash1-storage","auto-read-only":true,"discard":"unmap"} - > blockdev {"node-name":"libvirt-pflash1-format","read- > only":false,"driver":"raw","file":"libvirt-pflash1-storage"} -machine virt- > 7.2,usb=off,dump-guest-core=off,gic-version=3,pflash0=libvirt-pflash0- > format,pflash1=libvirt-pflash1-format,memory-backend=mach-virt.ram -accel > kvm -cpu host -m 32768 -object {"qom-type":"memory-backend-ram","id":"mach- > virt.ram","size":34359738368} -overcommit mem-lock=off -smp > 2,maxcpus=4,sockets=1,dies=1,cores=4,threads=1 -uuid ea78094e-de32-4caa- > 827d-7628afff6524 -display none -no-user-config -nodefaults -chardev > socket,id=charmonitor,fd=23,server=on,wait=off -mon > chardev=charmonitor,id=monitor,mode=control -rtc base=utc -no-shutdown - > boot strict=on -device {"driver":"pcie-root- > port","port":8,"chassis":1,"id":"pci.1","bus":"pcie.0","multifunction":true > ,"addr":"0x1"} -device {"driver":"pcie-root- > port","port":9,"chassis":2,"id":"pci.2","bus":"pcie.0","addr":"0x1.0x1"} - > device {"driver":"pcie-root- > port","port":10,"chassis":3,"id":"pci.3","bus":"pcie.0","addr":"0x1.0x2"} - > device {"driver":"pcie-root- > port","port":11,"chassis":4,"id":"pci.4","bus":"pcie.0","addr":"0x1.0x3"} - > device {"driver":"pcie-root- > port","port":12,"chassis":5,"id":"pci.5","bus":"pcie.0","addr":"0x1.0x4"} - > device {"driver":"pcie-root- > port","port":13,"chassis":6,"id":"pci.6","bus":"pcie.0","addr":"0x1.0x5"} - > device {"driver":"pcie-root- > port","port":14,"chassis":7,"id":"pci.7","bus":"pcie.0","addr":"0x1.0x6"} - > device {"driver":"pcie-root- > port","port":15,"chassis":8,"id":"pci.8","bus":"pcie.0","addr":"0x1.0x7"} - > device {"driver":"pcie-root- > port","port":16,"chassis":9,"id":"pci.9","bus":"pcie.0","multifunction":tru > e,"addr":"0x2"} -device {"driver":"pcie-root- > port","port":17,"chassis":10,"id":"pci.10","bus":"pcie.0","addr":"0x2.0x1"} > -device {"driver":"pcie-root- > port","port":18,"chassis":11,"id":"pci.11","bus":"pcie.0","addr":"0x2.0x2"} > -device {"driver":"pcie-root- > port","port":19,"chassis":12,"id":"pci.12","bus":"pcie.0","addr":"0x2.0x3"} > -device {"driver":"pcie-root- > port","port":20,"chassis":13,"id":"pci.13","bus":"pcie.0","addr":"0x2.0x4"} > -device {"driver":"pcie-root- > port","port":21,"chassis":14,"id":"pci.14","bus":"pcie.0","addr":"0x2.0x5"} > -device {"driver":"qemu- > xhci","p2":15,"p3":15,"id":"usb","bus":"pci.2","addr":"0x0"} -device > {"driver":"virtio-scsi-pci","id":"scsi0","bus":"pci.3","addr":"0x0"} - > device {"driver":"virtio-serial-pci","id":"virtio- > serial0","bus":"pci.4","addr":"0x0"} -blockdev > {"driver":"file","filename":"/mnt/nfs/vm11.raw","node-name":"libvirt-2- > storage","auto-read-only":true,"discard":"unmap"} -blockdev {"node- > name":"libvirt-2-format","read- > only":false,"discard":"unmap","driver":"raw","file":"libvirt-2-storage"} - > device {"driver":"virtio-blk- > pci","bus":"pci.5","addr":"0x0","drive":"libvirt-2-format","id":"virtio- > disk0","bootindex":1} -blockdev > {"driver":"file","filename":"/home/vishnu/vms/Fedora-Server-dvd-aarch64-36- > 1.5.iso","node-name":"libvirt-1-storage","auto-read- > only":true,"discard":"unmap"} -blockdev {"node-name":"libvirt-1- > format","read-only":true,"driver":"raw","file":"libvirt-1-storage"} -device > {"driver":"scsi-cd","bus":"scsi0.0","channel":0,"scsi- > id":0,"lun":0,"device_id":"drive-scsi0-0-0-0","drive":"libvirt-1- > format","id":"scsi0-0-0-0"} -netdev > tap,fd=24,vhost=on,vhostfd=26,id=hostnet0 -device {"driver":"virtio-net- > pci","netdev":"hostnet0","id":"net0","mac":"52:54:00:4d:8d:9e","bus":"pci.1 > ","addr":"0x0"} -chardev pty,id=charserial0 -serial chardev:charserial0 - > chardev socket,id=charchannel0,fd=22,server=on,wait=off -device > {"driver":"virtserialport","bus":"virtio- > serial0.0","nr":1,"chardev":"charchannel0","id":"channel0","name":"org.qemu > .guest_agent.0"} -chardev socket,id=chrtpm,path=/run/libvirt/qemu/swtpm/4- > vm1-swtpm.sock -tpmdev emulator,id=tpm-tpm0,chardev=chrtpm -device > {"driver":"tpm-tis-device","tpmdev":"tpm-tpm0","id":"tpm0"} -audiodev > {"id":"audio1","driver":"none"} -device {"driver":"virtio-balloon- > pci","id":"balloon0","bus":"pci.6","addr":"0x0"} -object {"qom-type":"rng- > random","id":"objrng0","filename":"/dev/urandom"} -device > {"driver":"virtio-rng- > pci","rng":"objrng0","id":"rng0","bus":"pci.7","addr":"0x0"} -msg > timestamp=on > virsh command to migrate vm1 from SrcHost to DstHost > virsh migrate --live vm1 qemu+ssh://vCPU-HP-Host2/system --unsafe -- > verbose > Commands on DstHost: > qemu command after vm1 migrated to DstHost: > /usr/local/bin/qemu-system-aarch64 -name guest=vm1,debug-threads=on -S - > object {"qom- > type":"secret","id":"masterKey0","format":"raw","file":"/var/lib/libvirt/qe > mu/domain-1-vm1/master-key.aes"} -blockdev > {"driver":"file","filename":"/usr/share/edk2/aarch64/QEMU_EFI-silent- > pflash.raw","node-name":"libvirt-pflash0-storage","auto-read- > only":true,"discard":"unmap"} -blockdev {"node-name":"libvirt-pflash0- > format","read-only":true,"driver":"raw","file":"libvirt-pflash0-storage"} - > blockdev > {"driver":"file","filename":"/var/lib/libvirt/qemu/nvram/vm1_VARS.fd","node > -name":"libvirt-pflash1-storage","auto-read-only":true,"discard":"unmap"} - > blockdev {"node-name":"libvirt-pflash1-format","read- > only":false,"driver":"raw","file":"libvirt-pflash1-storage"} -machine virt- > 7.2,usb=off,gic-version=3,dump-guest-core=off,memory-backend=mach- > virt.ram,pflash0=libvirt-pflash0-format,pflash1=libvirt-pflash1-format - > accel kvm -cpu host -m 32768 -object {"qom-type":"memory-backend- > ram","id":"mach-virt.ram","size":34359738368} -overcommit mem-lock=off -smp > 2,maxcpus=4,sockets=1,dies=1,cores=4,threads=1 -uuid ea78094e-de32-4caa- > 827d-7628afff6524 -display none -no-user-config -nodefaults -chardev > socket,id=charmonitor,fd=23,server=on,wait=off -mon > chardev=charmonitor,id=monitor,mode=control -rtc base=utc -no-shutdown - > boot strict=on -device {"driver":"pcie-root- > port","port":8,"chassis":1,"id":"pci.1","bus":"pcie.0","multifunction":true > ,"addr":"0x1"} -device {"driver":"pcie-root- > port","port":9,"chassis":2,"id":"pci.2","bus":"pcie.0","addr":"0x1.0x1"} - > device {"driver":"pcie-root- > port","port":10,"chassis":3,"id":"pci.3","bus":"pcie.0","addr":"0x1.0x2"} - > device {"driver":"pcie-root- > port","port":11,"chassis":4,"id":"pci.4","bus":"pcie.0","addr":"0x1.0x3"} - > device {"driver":"pcie-root- > port","port":12,"chassis":5,"id":"pci.5","bus":"pcie.0","addr":"0x1.0x4"} - > device {"driver":"pcie-root- > port","port":13,"chassis":6,"id":"pci.6","bus":"pcie.0","addr":"0x1.0x5"} - > device {"driver":"pcie-root- > port","port":14,"chassis":7,"id":"pci.7","bus":"pcie.0","addr":"0x1.0x6"} - > device {"driver":"pcie-root- > port","port":15,"chassis":8,"id":"pci.8","bus":"pcie.0","addr":"0x1.0x7"} - > device {"driver":"pcie-root- > port","port":16,"chassis":9,"id":"pci.9","bus":"pcie.0","multifunction":tru > e,"addr":"0x2"} -device {"driver":"pcie-root- > port","port":17,"chassis":10,"id":"pci.10","bus":"pcie.0","addr":"0x2.0x1"} > -device {"driver":"pcie-root- > port","port":18,"chassis":11,"id":"pci.11","bus":"pcie.0","addr":"0x2.0x2"} > -device {"driver":"pcie-root- > port","port":19,"chassis":12,"id":"pci.12","bus":"pcie.0","addr":"0x2.0x3"} > -device {"driver":"pcie-root- > port","port":20,"chassis":13,"id":"pci.13","bus":"pcie.0","addr":"0x2.0x4"} > -device {"driver":"pcie-root- > port","port":21,"chassis":14,"id":"pci.14","bus":"pcie.0","addr":"0x2.0x5"} > -device {"driver":"qemu- > xhci","p2":15,"p3":15,"id":"usb","bus":"pci.2","addr":"0x0"} -device > {"driver":"virtio-scsi-pci","id":"scsi0","bus":"pci.3","addr":"0x0"} - > device {"driver":"virtio-serial-pci","id":"virtio- > serial0","bus":"pci.4","addr":"0x0"} -blockdev > {"driver":"file","filename":"/mnt/nfs/vm11.raw","node-name":"libvirt-2- > storage","auto-read-only":true,"discard":"unmap"} -blockdev {"node- > name":"libvirt-2-format","read- > only":false,"discard":"unmap","driver":"raw","file":"libvirt-2-storage"} - > device {"driver":"virtio-blk- > pci","bus":"pci.5","addr":"0x0","drive":"libvirt-2-format","id":"virtio- > disk0","bootindex":1} -blockdev > {"driver":"file","filename":"/home/vishnu/vms/Fedora-Server-dvd-aarch64-36- > 1.5.iso","node-name":"libvirt-1-storage","auto-read- > only":true,"discard":"unmap"} -blockdev {"node-name":"libvirt-1- > format","read-only":true,"driver":"raw","file":"libvirt-1-storage"} -device > {"driver":"scsi-cd","bus":"scsi0.0","channel":0,"scsi- > id":0,"lun":0,"device_id":"drive-scsi0-0-0-0","drive":"libvirt-1- > format","id":"scsi0-0-0-0"} -netdev > {"type":"tap","fd":"24","vhost":true,"vhostfd":"26","id":"hostnet0"} - > device {"driver":"virtio-net- > pci","netdev":"hostnet0","id":"net0","mac":"52:54:00:4d:8d:9e","bus":"pci.1 > ","addr":"0x0"} -chardev pty,id=charserial0 -serial chardev:charserial0 - > chardev socket,id=charchannel0,fd=22,server=on,wait=off -device > {"driver":"virtserialport","bus":"virtio- > serial0.0","nr":1,"chardev":"charchannel0","id":"channel0","name":"org.qemu > .guest_agent.0"} -chardev socket,id=chrtpm,path=/run/libvirt/qemu/swtpm/1- > vm1-swtpm.sock -tpmdev emulator,id=tpm-tpm0,chardev=chrtpm -device > {"driver":"tpm-tis-device","tpmdev":"tpm-tpm0","id":"tpm0"} -audiodev > {"id":"audio1","driver":"none"} -incoming defer -device {"driver":"virtio- > balloon-pci","id":"balloon0","bus":"pci.6","addr":"0x0"} -object {"qom- > type":"rng-random","id":"objrng0","filename":"/dev/urandom"} -device > {"driver":"virtio-rng- > pci","rng":"objrng0","id":"rng0","bus":"pci.7","addr":"0x0"} -msg > timestamp=on > virsh command to migrate vm1 back to SrcHost: > # virsh migrate --live vm1 qemu+ssh://vCPU-HP-Host1/system --unsafe -- > verbose > > > > Thanks > Salil. > > > >

1 year, 10 months

1
0
0 0

Re: [External] : RE: vCPU hotplug/unplug for the ARMv8 processors

by Vishnu Pajjuri

Hi Salil, Below patch resolves VM live migration CPU stall issues. diff --git a/hw/arm/virt.c b/hw/arm/virt.c index e9ced9f172..23b06d44cc 100644 --- a/hw/arm/virt.c +++ b/hw/arm/virt.c @@ -3224,6 +3224,11 @@ static void virt_cpu_plug(HotplugHandler *hotplug_dev, DeviceState *dev, fw_cfg_modify_i16(vms->fw_cfg, FW_CFG_NB_CPUS, vms->boot_cpus); cs->disabled = false; + if (qemu_enabled_cpu(cs)) { + cs->vmcse = qemu_add_vm_change_state_handler(kvm_arm_vm_state_change, cs); + } + return; fail: error_propagate(errp, local_err); diff --git a/target/arm/kvm64.c b/target/arm/kvm64.c index 80a456ef16..510f829e4b 100644 --- a/target/arm/kvm64.c +++ b/target/arm/kvm64.c @@ -883,9 +883,6 @@ int kvm_arch_init_vcpu(CPUState *cs) return -EINVAL; } - if (qemu_enabled_cpu(cs)) - cs->vmcse = qemu_add_vm_change_state_handler(kvm_arm_vm_state_change, - cs); /* Determine init features for this CPU */ memset(cpu->kvm_init_features, 0, sizeof(cpu->kvm_init_features)); Root Cause: It's too early to attach kvm_arm_vm_state_change handler in kvm_arch_init_vcpu since vcpus were not yet enabled at this stage Because of this, QEMU is not fetching virtual time from KVM before saving VM state Also not able to restore virtual time back to KVM after restoring VM state It is causing misconfiguration of vtimer offset and leading to CPU stall issues. Attaching kvm_arm_vm_state_change handler in virt_cpu_plug after enabling each vCPU makes more reasonable to me. Any suggestions are welcome here. This helps to properly configure vtimer offset. And didn't see any timing issues _Regards_, -Vishnu. On 26-07-2023 15:15, Salil Mehta wrote: > > HI Vishnu, > > Got it. I think I am now able to recreate the condition successfully > with just 2 vCPUs and even when all of them are present and online. > > I am analyzing the issue and will get back to you soon. > > Thanks > > Salil > > *From:* Vishnu Pajjuri OS <vishnu(a)os.amperecomputing.com> > *Sent:* Friday, July 21, 2023 10:39 AM > *To:* Salil Mehta <salil.mehta(a)opnsrc.net>; Salil Mehta > <salil.mehta(a)huawei.com> > *Cc:* Vishnu Pajjuri OS <vishnu(a)os.amperecomputing.com>; Miguel Luis > <miguel.luis(a)oracle.com> > *Subject:* Re: [External] : RE: vCPU hotplug/unplug for the ARMv8 > processors > > Hi Salil, > > On 20-07-2023 22:56, Salil Mehta wrote: > > [+] For your reference, I am attaching the scripts which I am > using. Note the subtle difference between both > > Thanks for sharing your scripts for VM live migration. > > > > Thanks! > > On Thu, Jul 20, 2023 at 12:57 PM Salil Mehta > <salil.mehta(a)huawei.com> wrote: > > Hi Vishnu, > > Aah, I see looks like you have 2 physical machine? I am using > same host setup and use below commands to migrate VMs > > *Src VM (Same Host)* > > Qemu> migrate tcp:0:4444 > > *Dst VM (Same Host)* > > $ qemu-arm64-nosmmu[…] incoming tcp:0:4444 > > I have prepared setup like below on the same Host with the help of > qemu commands only > > [VM1 (P4:C2:H0:U2:O2)] --> [VM2 (P4:C2:H0:U2:O2)] --> [VM3 > (P4:C2:H0:U2:O2)] > > From VM1 to VM2 works fine. But when I tried to do live migration from > VM2(same VM migrated from VM1) to VM3 > > I observed CPU stall issues in this scenario also. > > One more observation is that time is jumping in the VM2 to VM3 > iteration always > > [root@fedora ~]# date > Wed Nov 21 06:34:54 AM EST 2114 > > VM1.sh > > TOPDIR=/home/vishnu/vcpu_hp > BASECODEPATH=$TOPDIR/pre_v2 > QEMUPATH=$BASECODEPATH/qemu/ > QEMUBIN=$QEMUPATH/build/aarch64-softmmu/qemu-system-aarch64 > ROOTFS_IMG=/home/vishnu/vms/images/vm11.raw > BIOS=/usr/share/AAVMF/AAVMF_CODE.fd > LOG_FILE=/tmp/serial1.log > > $QEMUBIN -M virt,accel=kvm,gic-version=3 \ > -smp cpus=2,maxcpus=4 \ > -m 4g \ > -cpu host \ > -nodefaults \ > -nographic \ > -drive if=pflash,format=raw,readonly,file=$BIOS \ > -boot c -hda $ROOTFS_IMG \ > -chardev stdio,id=char0,logfile=$LOG_FILE,signal=off \ > -serial chardev:char0 \ > -monitor tcp:127.0.0.1:55555,server,nowait > > VM2.sh > > TOPDIR=/home/vishnu/vcpu_hp > BASECODEPATH=$TOPDIR/pre_v2 > QEMUPATH=$BASECODEPATH/qemu/ > QEMUBIN=$QEMUPATH/build/aarch64-softmmu/qemu-system-aarch64 > ROOTFS_IMG=/home/vishnu/vms/images/vm11.raw > BIOS=/usr/share/AAVMF/AAVMF_CODE.fd > LOG_FILE=/tmp/serial1.log > > $QEMUBIN -M virt,accel=kvm,gic-version=3 \ > -smp cpus=2,maxcpus=4 \ > -m 4g \ > -cpu host \ > -nodefaults \ > -nographic \ > -drive if=pflash,format=raw,readonly,file=$BIOS \ > -boot c -hda $ROOTFS_IMG \ > -chardev stdio,id=char0,logfile=$LOG_FILE,signal=off \ > -serial chardev:char0 \ > -monitor tcp:127.0.0.1:55556,server,nowait \ > -incoming tcp:0:4444 > > VM3.sh > > TOPDIR=/home/vishnu/vcpu_hp > BASECODEPATH=$TOPDIR/pre_v2 > QEMUPATH=$BASECODEPATH/qemu/ > QEMUBIN=$QEMUPATH/build/aarch64-softmmu/qemu-system-aarch64 > ROOTFS_IMG=/home/vishnu/vms/images/vm11.raw > BIOS=/usr/share/AAVMF/AAVMF_CODE.fd > LOG_FILE=/tmp/serial1.log > > $QEMUBIN -M virt,accel=kvm,gic-version=3 \ > -smp cpus=2,maxcpus=4 \ > -m 4g \ > -cpu host \ > -nodefaults \ > -nographic \ > -drive if=pflash,format=raw,readonly,file=$BIOS \ > -boot c -hda $ROOTFS_IMG \ > -chardev stdio,id=char0,logfile=$LOG_FILE,signal=off \ > -serial chardev:char0 \ > -monitor tcp:127.0.0.1:55557,server,nowait \ > -incoming tcp:0:4445 > > Log:- > [root@fedora ~]# [ 189.884992] rcu: INFO: rcu_sched detected stalls > on CPUs/tasks: > [ 189.887474] rcu: 1-...!: (0 ticks this GP) idle=c268/0/0x0 > softirq=13837/13837 fqs=0 (false positive?) > [ 189.891022] rcu: (detected by 0, t=6002 jiffies, g=7521, q=764 > ncpus=2) > [ 189.893494] Task dump for CPU 1: > [ 189.894875] task:swapper/1 state:R running task > stack:0 pid:0 ppid:1 flags:0x00000008 > [ 189.898597] Call trace: > [ 189.899505] __switch_to+0xc8/0xf8 > [ 189.900825] 0xffff8000080ebde0 > [ 189.902236] rcu: rcu_sched kthread timer wakeup didn't happen for > 6001 jiffies! g7521 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x402 > [ 189.906499] rcu: Possible timer handling issue on cpu=1 > timer-softirq=3803 > [ 189.909125] rcu: rcu_sched kthread starved for 6002 jiffies! g7521 > f0x0 RCU_GP_WAIT_FQS(5) ->state=0x402 ->cpu=1 > [ 189.912934] rcu: Unless rcu_sched kthread gets sufficient CPU > time, OOM is now expected behavior. > [ 189.916226] rcu: RCU grace-period kthread stack dump: > [ 189.918122] task:rcu_sched state:I stack:0 pid:15 > ppid:2 flags:0x00000008 > [ 189.921241] Call trace: > [ 189.922165] __switch_to+0xc8/0xf8 > [ 189.923453] __schedule+0x248/0x6c0 > [ 189.924821] schedule+0x58/0xf0 > [ 189.926010] schedule_timeout+0x90/0x178 > [ 189.927483] rcu_gp_fqs_loop+0x124/0x438 > [ 189.928969] rcu_gp_kthread+0x164/0x1c8 > [ 189.930401] kthread+0xe0/0xf0 > [ 189.931544] ret_from_fork+0x10/0x20 > [ 189.932874] rcu: Stack dump where RCU GP kthread last ran: > [ 189.935207] Task dump for CPU 1: > [ 189.936338] task:swapper/1 state:R running task > stack:0 pid:0 ppid:1 flags:0x00000008 > [ 189.939811] Call trace: > [ 189.940673] __switch_to+0xc8/0xf8 > [ 189.941888] 0xffff8000080ebde0 > > Lets first use Qemu prompt to migrate instead of web interface > if you can? This will also help us converge whether problem is > an interface problem or we are really missing something. > > Could you please share your Linux and Qemu source repos as well? > > I'll try to figure out if I'm using any wrong source repositories. > > _Regards,_ > > -Vishnu. > > Thanks > > Salil > > *From:* Vishnu Pajjuri OS <vishnu(a)os.amperecomputing.com> > *Sent:* Thursday, July 20, 2023 12:49 PM > *To:* Salil Mehta <salil.mehta(a)huawei.com>; Vishnu Pajjuri OS > <vishnu(a)os.amperecomputing.com>; Miguel Luis > <miguel.luis(a)oracle.com> > *Cc:* Salil Mehta <salil.mehta(a)opnsrc.net> > *Subject:* Re: [External] : RE: vCPU hotplug/unplug for the > ARMv8 processors > > Hi Salil, > > On 20-07-2023 16:58, Salil Mehta wrote: > > Hi Vishnu, > > Can I request you to reduce the number of vCPUs to 6. > > Test Case: > > [VM1 (P6:C4:H0:U2:O4)] --> [VM2 (P6:C4:H0:U2:O4)] > > <Exit from qemu prompt. Restart VM1 and then execute below> > > With virsh(libvirt) commands, > > After Migrating VM1 to DstHost, there is nothing left on > SrcHost. Even ps command shows nothing on SrcHost regarding > qemu command. > > So operating VM1 on SrcHost after migrating VM1 to DstHost is > not possible. > > On DstHost side VM1 is live, there I can able to access VM1's > console. > > I can able to manipulate VM1 on DstHost after migration. > > _Regards_, > > -Vishnu. > > [VM1 (P6:C4:H0:U2:O4)] <-- [VM2 (P6:C4:H0:U2:O4)] > > [...‘n’ times...] > > [VM1 (P6:C4:H0:U2:O4)] --> [VM2 (P6:C4:H0:U2:O4)] > > <Exit from qemu prompt. Restart VM1 and then execute below> > > [VM1 (P6:C4:H0:U2:O4)] <-- [VM2 (P6:C4:H0:U2:O4)] > > Thanks > > Salil. > > *From:* Vishnu Pajjuri OS <vishnu(a)os.amperecomputing.com> > <mailto:vishnu@os.amperecomputing.com> > *Sent:* Thursday, July 20, 2023 12:16 PM > *To:* Salil Mehta <salil.mehta(a)huawei.com> > <mailto:salil.mehta@huawei.com>; Vishnu Pajjuri OS > <vishnu(a)os.amperecomputing.com> > <mailto:vishnu@os.amperecomputing.com>; Miguel Luis > <miguel.luis(a)oracle.com> <mailto:miguel.luis@oracle.com> > *Cc:* Salil Mehta <salil.mehta(a)opnsrc.net> > <mailto:salil.mehta@opnsrc.net> > *Subject:* Re: [External] : RE: vCPU hotplug/unplug for > the ARMv8 processors > > Hi Salil, > > On 20-07-2023 15:52, Salil Mehta wrote: > > Hi Vishnu, > > > > From: Vishnu Pajjuri OS<vishnu(a)os.amperecomputing.com> <mailto:vishnu@os.amperecomputing.com> > > Sent: Tuesday, July 18, 2023 12:39 PM > > > > [...] > > > > @Miguel/Vishnu BTW, I was able to reproduce the CPU stall issue and it looks > > to me it only happens when number of online vCPUs being migrated at source VM > > do not match number of vCPUs being online'd during boot at the destination VM. > > This means "maxcpus=N" kernel parameter at the destination VM should have > > value equal to number of online'd vCPUs being migrated. > > > > I still need to check if this matches the x86_64 behavior. If possible, can > > you guys confirm below: > > > > [1] always keep the online vcpus same at source and destination and then wait > > and check for the CPU stall messages? > > Hopefully below is the test scenario that you are looking at, but still I observed CPU stall issues > > > > P: Possible vCPUs > > C: Cold Plugged vCPUs > > H: Hot Plugged vCPUs > > U: Unplugged vCPUs > > O: Onlined vCPUs > > sVM: Source VM > > dVM: Destination VM > > > > > > Test Case: > > [sVM (P80:C72:H0:U8:O72)] --> [dVM (P80:C72:H0:U8:O72)] - PASS > > > > Migrate the same VM back > > > > [sVM (P80:C72:H0:U8:O72)] <-- [dVM (P80:C72:H0:U8:O72)] - Observed CPU stalls on VM > > > > > > I have now repeated above scenario at-least 5 times and cannot reproduce stalls > > in the case you have mentioned. > > Good hear that you didn't observed any CPU stall issues > with your setup. > > I'm using virsh commands to create VM live setup. If > possible I'll adapt your qemu commands to libvirt, > > Otherwise I would like to reproduce your setup, > > for that could you please provide your kernel and qemu repo's > > Also your complete Qemu command of VM launch and migration > > Commands you have been using both at source and > destination side? > > > > Few questions: > > Q1: In above case, there is no Hotplug happening at all - right? > > Ans: Yes > > > > Q2: Do you restart the source VM after it has been migrated once to the destination? > > Ans: No. > > > > Q3: Can you please paste your complete Qemu command of VM launch and migration > > Commands you have been using both at source and destination side? > > Ans: I'm using virsh commands to migrate VMs > > *Commands on SrcHost:* > > *qemu command after running with virsh command:* > > virsh start vm1 --console > > /usr/local/bin/qemu-system-aarch64 -name > guest=vm1,debug-threads=on -S -object > {"qom-type":"secret","id":"masterKey0","format":"raw","file":"/var/lib/libvirt/qemu/domain-4-vm1/master-key.aes"} > -blockdev > {"driver":"file","filename":"/usr/share/edk2/aarch64/QEMU_EFI-silent-pflash.raw","node-name":"libvirt-pflash0-storage","auto-read-only":true,"discard":"unmap"} > -blockdev > {"node-name":"libvirt-pflash0-format","read-only":true,"driver":"raw","file":"libvirt-pflash0-storage"} > -blockdev > {"driver":"file","filename":"/var/lib/libvirt/qemu/nvram/vm1_VARS.fd","node-name":"libvirt-pflash1-storage","auto-read-only":true,"discard":"unmap"} > -blockdev > {"node-name":"libvirt-pflash1-format","read-only":false,"driver":"raw","file":"libvirt-pflash1-storage"} > -machine > virt-7.2,usb=off,dump-guest-core=off,gic-version=3,pflash0=libvirt-pflash0-format,pflash1=libvirt-pflash1-format,memory-backend=mach-virt.ram > -accel kvm -cpu host -m 32768 -object > {"qom-type":"memory-backend-ram","id":"mach-virt.ram","size":34359738368} > -overcommit mem-lock=off -smp > 2,maxcpus=4,sockets=1,dies=1,cores=4,threads=1 -uuid > ea78094e-de32-4caa-827d-7628afff6524 -display none > -no-user-config -nodefaults -chardev > socket,id=charmonitor,fd=23,server=on,wait=off -mon > chardev=charmonitor,id=monitor,mode=control -rtc base=utc > -no-shutdown -boot strict=on -device > {"driver":"pcie-root-port","port":8,"chassis":1,"id":"pci.1","bus":"pcie.0","multifunction":true,"addr":"0x1"} > -device > {"driver":"pcie-root-port","port":9,"chassis":2,"id":"pci.2","bus":"pcie.0","addr":"0x1.0x1"} > -device > {"driver":"pcie-root-port","port":10,"chassis":3,"id":"pci.3","bus":"pcie.0","addr":"0x1.0x2"} > -device > {"driver":"pcie-root-port","port":11,"chassis":4,"id":"pci.4","bus":"pcie.0","addr":"0x1.0x3"} > -device > {"driver":"pcie-root-port","port":12,"chassis":5,"id":"pci.5","bus":"pcie.0","addr":"0x1.0x4"} > -device > {"driver":"pcie-root-port","port":13,"chassis":6,"id":"pci.6","bus":"pcie.0","addr":"0x1.0x5"} > -device > {"driver":"pcie-root-port","port":14,"chassis":7,"id":"pci.7","bus":"pcie.0","addr":"0x1.0x6"} > -device > {"driver":"pcie-root-port","port":15,"chassis":8,"id":"pci.8","bus":"pcie.0","addr":"0x1.0x7"} > -device > {"driver":"pcie-root-port","port":16,"chassis":9,"id":"pci.9","bus":"pcie.0","multifunction":true,"addr":"0x2"} > -device > {"driver":"pcie-root-port","port":17,"chassis":10,"id":"pci.10","bus":"pcie.0","addr":"0x2.0x1"} > -device > {"driver":"pcie-root-port","port":18,"chassis":11,"id":"pci.11","bus":"pcie.0","addr":"0x2.0x2"} > -device > {"driver":"pcie-root-port","port":19,"chassis":12,"id":"pci.12","bus":"pcie.0","addr":"0x2.0x3"} > -device > {"driver":"pcie-root-port","port":20,"chassis":13,"id":"pci.13","bus":"pcie.0","addr":"0x2.0x4"} > -device > {"driver":"pcie-root-port","port":21,"chassis":14,"id":"pci.14","bus":"pcie.0","addr":"0x2.0x5"} > -device > {"driver":"qemu-xhci","p2":15,"p3":15,"id":"usb","bus":"pci.2","addr":"0x0"} > -device > {"driver":"virtio-scsi-pci","id":"scsi0","bus":"pci.3","addr":"0x0"} > -device > {"driver":"virtio-serial-pci","id":"virtio-serial0","bus":"pci.4","addr":"0x0"} > -blockdev > {"driver":"file","filename":"/mnt/nfs/vm11.raw","node-name":"libvirt-2-storage","auto-read-only":true,"discard":"unmap"} > -blockdev > {"node-name":"libvirt-2-format","read-only":false,"discard":"unmap","driver":"raw","file":"libvirt-2-storage"} > -device > {"driver":"virtio-blk-pci","bus":"pci.5","addr":"0x0","drive":"libvirt-2-format","id":"virtio-disk0","bootindex":1} > -blockdev > {"driver":"file","filename":"/home/vishnu/vms/Fedora-Server-dvd-aarch64-36-1.5.iso","node-name":"libvirt-1-storage","auto-read-only":true,"discard":"unmap"} > -blockdev > {"node-name":"libvirt-1-format","read-only":true,"driver":"raw","file":"libvirt-1-storage"} > -device > {"driver":"scsi-cd","bus":"scsi0.0","channel":0,"scsi-id":0,"lun":0,"device_id":"drive-scsi0-0-0-0","drive":"libvirt-1-format","id":"scsi0-0-0-0"} > -netdev tap,fd=24,vhost=on,vhostfd=26,id=hostnet0 -device > {"driver":"virtio-net-pci","netdev":"hostnet0","id":"net0","mac":"52:54:00:4d:8d:9e","bus":"pci.1","addr":"0x0"} > -chardev pty,id=charserial0 -serial chardev:charserial0 > -chardev socket,id=charchannel0,fd=22,server=on,wait=off > -device > {"driver":"virtserialport","bus":"virtio-serial0.0","nr":1,"chardev":"charchannel0","id":"channel0","name":"org.qemu.guest_agent.0"} > -chardev > socket,id=chrtpm,path=/run/libvirt/qemu/swtpm/4-vm1-swtpm.sock > -tpmdev emulator,id=tpm-tpm0,chardev=chrtpm -device > {"driver":"tpm-tis-device","tpmdev":"tpm-tpm0","id":"tpm0"} > -audiodev {"id":"audio1","driver":"none"} -device > {"driver":"virtio-balloon-pci","id":"balloon0","bus":"pci.6","addr":"0x0"} > -object > {"qom-type":"rng-random","id":"objrng0","filename":"/dev/urandom"} > -device > {"driver":"virtio-rng-pci","rng":"objrng0","id":"rng0","bus":"pci.7","addr":"0x0"} > -msg timestamp=on > > *virsh command to migrate vm1 from SrcHost to DstHost * > > virsh migrate --live vm1 qemu+ssh://vCPU-HP-Host2/system > --unsafe --verbose > > *Commands on DstHost:* > > *qemu command after vm1 migrated to DstHost:* > > /usr/local/bin/qemu-system-aarch64 -name > guest=vm1,debug-threads=on -S -object > {"qom-type":"secret","id":"masterKey0","format":"raw","file":"/var/lib/libvirt/qemu/domain-1-vm1/master-key.aes"} > -blockdev > {"driver":"file","filename":"/usr/share/edk2/aarch64/QEMU_EFI-silent-pflash.raw","node-name":"libvirt-pflash0-storage","auto-read-only":true,"discard":"unmap"} > -blockdev > {"node-name":"libvirt-pflash0-format","read-only":true,"driver":"raw","file":"libvirt-pflash0-storage"} > -blockdev > {"driver":"file","filename":"/var/lib/libvirt/qemu/nvram/vm1_VARS.fd","node-name":"libvirt-pflash1-storage","auto-read-only":true,"discard":"unmap"} > -blockdev > {"node-name":"libvirt-pflash1-format","read-only":false,"driver":"raw","file":"libvirt-pflash1-storage"} > -machine > virt-7.2,usb=off,gic-version=3,dump-guest-core=off,memory-backend=mach-virt.ram,pflash0=libvirt-pflash0-format,pflash1=libvirt-pflash1-format > -accel kvm -cpu host -m 32768 -object > {"qom-type":"memory-backend-ram","id":"mach-virt.ram","size":34359738368} > -overcommit mem-lock=off -smp > 2,maxcpus=4,sockets=1,dies=1,cores=4,threads=1 -uuid > ea78094e-de32-4caa-827d-7628afff6524 -display none > -no-user-config -nodefaults -chardev > socket,id=charmonitor,fd=23,server=on,wait=off -mon > chardev=charmonitor,id=monitor,mode=control -rtc base=utc > -no-shutdown -boot strict=on -device > {"driver":"pcie-root-port","port":8,"chassis":1,"id":"pci.1","bus":"pcie.0","multifunction":true,"addr":"0x1"} > -device > {"driver":"pcie-root-port","port":9,"chassis":2,"id":"pci.2","bus":"pcie.0","addr":"0x1.0x1"} > -device > {"driver":"pcie-root-port","port":10,"chassis":3,"id":"pci.3","bus":"pcie.0","addr":"0x1.0x2"} > -device > {"driver":"pcie-root-port","port":11,"chassis":4,"id":"pci.4","bus":"pcie.0","addr":"0x1.0x3"} > -device > {"driver":"pcie-root-port","port":12,"chassis":5,"id":"pci.5","bus":"pcie.0","addr":"0x1.0x4"} > -device > {"driver":"pcie-root-port","port":13,"chassis":6,"id":"pci.6","bus":"pcie.0","addr":"0x1.0x5"} > -device > {"driver":"pcie-root-port","port":14,"chassis":7,"id":"pci.7","bus":"pcie.0","addr":"0x1.0x6"} > -device > {"driver":"pcie-root-port","port":15,"chassis":8,"id":"pci.8","bus":"pcie.0","addr":"0x1.0x7"} > -device > {"driver":"pcie-root-port","port":16,"chassis":9,"id":"pci.9","bus":"pcie.0","multifunction":true,"addr":"0x2"} > -device > {"driver":"pcie-root-port","port":17,"chassis":10,"id":"pci.10","bus":"pcie.0","addr":"0x2.0x1"} > -device > {"driver":"pcie-root-port","port":18,"chassis":11,"id":"pci.11","bus":"pcie.0","addr":"0x2.0x2"} > -device > {"driver":"pcie-root-port","port":19,"chassis":12,"id":"pci.12","bus":"pcie.0","addr":"0x2.0x3"} > -device > {"driver":"pcie-root-port","port":20,"chassis":13,"id":"pci.13","bus":"pcie.0","addr":"0x2.0x4"} > -device > {"driver":"pcie-root-port","port":21,"chassis":14,"id":"pci.14","bus":"pcie.0","addr":"0x2.0x5"} > -device > {"driver":"qemu-xhci","p2":15,"p3":15,"id":"usb","bus":"pci.2","addr":"0x0"} > -device > {"driver":"virtio-scsi-pci","id":"scsi0","bus":"pci.3","addr":"0x0"} > -device > {"driver":"virtio-serial-pci","id":"virtio-serial0","bus":"pci.4","addr":"0x0"} > -blockdev > {"driver":"file","filename":"/mnt/nfs/vm11.raw","node-name":"libvirt-2-storage","auto-read-only":true,"discard":"unmap"} > -blockdev > {"node-name":"libvirt-2-format","read-only":false,"discard":"unmap","driver":"raw","file":"libvirt-2-storage"} > -device > {"driver":"virtio-blk-pci","bus":"pci.5","addr":"0x0","drive":"libvirt-2-format","id":"virtio-disk0","bootindex":1} > -blockdev > {"driver":"file","filename":"/home/vishnu/vms/Fedora-Server-dvd-aarch64-36-1.5.iso","node-name":"libvirt-1-storage","auto-read-only":true,"discard":"unmap"} > -blockdev > {"node-name":"libvirt-1-format","read-only":true,"driver":"raw","file":"libvirt-1-storage"} > -device > {"driver":"scsi-cd","bus":"scsi0.0","channel":0,"scsi-id":0,"lun":0,"device_id":"drive-scsi0-0-0-0","drive":"libvirt-1-format","id":"scsi0-0-0-0"} > -netdev > {"type":"tap","fd":"24","vhost":true,"vhostfd":"26","id":"hostnet0"} > -device > {"driver":"virtio-net-pci","netdev":"hostnet0","id":"net0","mac":"52:54:00:4d:8d:9e","bus":"pci.1","addr":"0x0"} > -chardev pty,id=charserial0 -serial chardev:charserial0 > -chardev socket,id=charchannel0,fd=22,server=on,wait=off > -device > {"driver":"virtserialport","bus":"virtio-serial0.0","nr":1,"chardev":"charchannel0","id":"channel0","name":"org.qemu.guest_agent.0"} > -chardev > socket,id=chrtpm,path=/run/libvirt/qemu/swtpm/1-vm1-swtpm.sock > -tpmdev emulator,id=tpm-tpm0,chardev=chrtpm -device > {"driver":"tpm-tis-device","tpmdev":"tpm-tpm0","id":"tpm0"} > -audiodev {"id":"audio1","driver":"none"} -incoming defer > -device > {"driver":"virtio-balloon-pci","id":"balloon0","bus":"pci.6","addr":"0x0"} > -object > {"qom-type":"rng-random","id":"objrng0","filename":"/dev/urandom"} > -device > {"driver":"virtio-rng-pci","rng":"objrng0","id":"rng0","bus":"pci.7","addr":"0x0"} > -msg timestamp=on > > *virsh command to migrate vm1 back to SrcHost: * > > # virsh migrate --live vm1 qemu+ssh://vCPU-HP-Host1/system > --unsafe --verbose > > > > > > Thanks > > Salil. > > > > > > > > >

1 year, 10 months

1
0
0 0

MPAM buglet..

by Jonathan Cameron

Given you mentioned rebasing MPAM :) One trivial bug that might still be hanging around. When adding resources parsed from the MPAM ACPI table I think you are adding one byte too much for IORESOURCE_MEM https://git.kernel.org/pub/scm/linux/kernel/git/morse/linux.git/tree/driver… should be a -1 similar to IORT handling. https://elixir.bootlin.com/linux/latest/source/drivers/acpi/arm64/iort.c#L1… As a side note, I'm emulating MPAM on QEMU so I can poke the corners of the code beyond the real implementations we have access to and in a fashion I can talk about more freely. Other than the above, mostly seems to be working fine beyond a bug in the upstream PPTT parsing code that doesn't affect what you have implemented so far. There is an infinite loop / deadlock hiding somewhere though in one of the error paths though that I haven't chased down yet. Maybe I'll be lucky and you've fixed it already :) It's holiday season so might take a while to get the QEMU implementation finished / through internal review - but I'll post any bugs I find in the meantime. Jonathan

1 year, 10 months

3
6
0 0

ARM64 vCPU Hotplug Kernel patches for discussion in community?

by Salil Mehta

Hi James, Are you planning to push the next version (-V2) of vCPU Hotplug kernel changes to the community? Is it possible to figure out any blockers on your end which are restricting you to move ahead. I think kernel can move even when Qemu patches have not been pushed to community and are only shared as repository. Though I shall been posting them soon. It would be better if we can get kernel patches in discussion and then hope for their acceptance as soon as possible. This is very important for acceptance of Qemu patches and to be frank for anybody to take Qemu patches seriously even for discussion. Many thanks Salil.

1 year, 10 months

2
2
0 0

[Question] Regarding the new _OSC capability to negotiate type of CPU Hotplug support (i.e. present/enable)

by Salil Mehta

Hi James, Few questions related to new _OSC capabilities to negotiate the type of CPU hotplug(if any): Q1: ACPI Standard Update: Do we have any Bugzilla-id for the change related to these new _OSC capabilities for the type of CPU hotplug support (i.e. Enabled/Present cpu)? Q2: Have you intentionally omitted below function in the _OSC related kernel change shared in the repository [1]? bool acpi_processor_hotplug_enabled_supported(void) Q3: I am sure you would want to extend usage of these APIs in the acpi_processor.c as well to decide what kind of hotplug (if any) is enabled? Is code related to _OSC changes complete? [1] https://gitlab.arm.com/linux-arm/linux-jm/-/commit/220b0d8b0261d7467c8705e6… Thanks Salil

1 year, 10 months

3
13
0 0

[Question - ARM CCA] Regarding vCPU Hotplug Support in ARM Realm world

by Salil Mehta

Hi James, I recently got a chance to look into the ARM CCA specifications. And if I understood correctly then in Realm world PSCI is managed by RMM but the Realm vCPUs are taken care by the host. As per our recent changes we handle the SMC/HVC hyper call exits in user-space. If the VMs are to run in Realm world and vCPU Hotplug needs to be supported then do we need to consider any change there? Thanks Salil

1 year, 11 months

2
2
0 0

2025

2024

2023

2022

2021

2020

Linaro-open-discussions