linux/virt/kvm, branch v4.3

arm/arm64: KVM: Fix disabled distributor operation

2015-10-20T16:09:13Z

We currently do a single update of the vgic state when the distributor enable/disable control register is accessed and then bypass updating the state for as long as the distributor remains disabled. This is incorrect, because updating the state does not consider the distributor enable bit, and this you can end up in a situation where an interrupt is marked as pending on the CPU interface, but not pending on the distributor, which is an impossible state to be in, and triggers a warning. Consider for example the following sequence of events: 1. An interrupt is marked as pending on the distributor - the interrupt is also forwarded to the CPU interface 2. The guest turns off the distributor (it's about to do a reboot) - we stop updating the CPU interface state from now on 3. The guest disables the pending interrupt - we remove the pending state from the distributor, but don't touch the CPU interface, see point 2. Since the distributor disable bit really means that no interrupts should be forwarded to the CPU interface, we modify the code to keep updating the internal VGIC state, but always set the CPU interface pending bits to zero when the distributor is disabled. Signed-off-by: Christoffer Dall

arm/arm64: KVM: Clear map->active on pend/active clear

2015-10-20T16:06:34Z

When a guest reboots or offlines/onlines CPUs, it is not uncommon for it to clear the pending and active states of an interrupt through the emulated VGIC distributor. However, since the architected timers are defined by the architecture to be level triggered and the guest rightfully expects them to be that, but we emulate them as edge-triggered, we have to mimic level-triggered behavior for an edge-triggered virtual implementation. We currently do not signal the VGIC when the map->active field is true, because it indicates that the guest has already been signalled of the interrupt as required. Normally this field is set to false when the guest deactivates the virtual interrupt through the sync path. We also need to catch the case where the guest deactivates the interrupt through the emulated distributor, again allowing guests to boot even if the original virtual timer signal hit before the guest's GIC initialization sequence is run. Reviewed-by: Eric Auger Signed-off-by: Christoffer Dall

arm/arm64: KVM: Fix arch timer behavior for disabled interrupts

2015-10-20T16:04:54Z

We have an interesting issue when the guest disables the timer interrupt on the VGIC, which happens when turning VCPUs off using PSCI, for example. The problem is that because the guest disables the virtual interrupt at the VGIC level, we never inject interrupts to the guest and therefore never mark the interrupt as active on the physical distributor. The host also never takes the timer interrupt (we only use the timer device to trigger a guest exit and everything else is done in software), so the interrupt does not become active through normal means. The result is that we keep entering the guest with a programmed timer that will always fire as soon as we context switch the hardware timer state and run the guest, preventing forward progress for the VCPU. Since the active state on the physical distributor is really part of the timer logic, it is the job of our virtual arch timer driver to manage this state. The timer->map->active boolean field indicates whether we have signalled this interrupt to the vgic and if that interrupt is still pending or active. As long as that is the case, the hardware doesn't have to generate physical interrupts and therefore we mark the interrupt as active on the physical distributor. We also have to restore the pending state of an interrupt that was queued to an LR but was retired from the LR for some reason, while remaining pending in the LR. Cc: Marc Zyngier Reported-by: Lorenzo Pieralisi Signed-off-by: Christoffer Dall

KVM: arm/arm64: Do not inject spurious interrupts

2015-10-20T16:04:43Z

When lowering a level-triggered line from userspace, we forgot to lower the pending bit on the emulated CPU interface and we also did not re-compute the pending_on_cpu bitmap for the CPU affected by the change. Update vgic_update_irq_pending() to fix the two issues above and also raise a warning in vgic_quue_irq_to_lr if we encounter an interrupt pending on a CPU which is neither marked active nor pending. [ Commit text reworked completely - Christoffer ] Signed-off-by: Pavel Fedin Signed-off-by: Christoffer Dall

KVM: disable halt_poll_ns as default for s390x

2015-09-25T08:31:30Z

We observed some performance degradation on s390x with dynamic halt polling. Until we can provide a proper fix, let's enable halt_poll_ns as default only for supported architectures. Architectures are now free to set their own halt_poll_ns default value. Signed-off-by: David Hildenbrand Signed-off-by: Paolo Bonzini

Merge tag 'kvm-arm-for-4.3-rc2-2' of git://git.kernel.org/pub/scm/linux/kernel/git/kvmarm/kvmarm into kvm-master

2015-09-17T14:51:59Z

Second set of KVM/ARM changes for 4.3-rc2 - Workaround for a Cortex-A57 erratum - Bug fix for the debugging infrastructure - Fix for 32bit guests with more than 4GB of address space on a 32bit host - A number of fixes for the (unusual) case when we don't use the in-kernel GIC emulation - Removal of ThumbEE handling on arm64, since these have been dropped from the architecture before anyone actually ever built a CPU - Remove the KVM_ARM_MAX_VCPUS limitation which has become fairly pointless

arm/arm64: KVM: Remove 'config KVM_ARM_MAX_VCPUS'

2015-09-17T12:13:27Z

This patch removes config option of KVM_ARM_MAX_VCPUS, and like other ARCHs, just choose the maximum allowed value from hardware, and follows the reasons: 1) from distribution view, the option has to be defined as the max allowed value because it need to meet all kinds of virtulization applications and need to support most of SoCs; 2) using a bigger value doesn't introduce extra memory consumption, and the help text in Kconfig isn't accurate because kvm_vpu structure isn't allocated until request of creating VCPU is sent from QEMU; 3) the main effect is that the field of vcpus[] in 'struct kvm' becomes a bit bigger(sizeof(void *) per vcpu) and need more cache lines to hold the structure, but 'struct kvm' is one generic struct, and it has worked well on other ARCHs already in this way. Also, the world switch frequecy is often low, for example, it is ~2000 when running kernel building load in VM from APM xgene KVM host, so the effect is very small, and the difference can't be observed in my test at all. Cc: Dann Frazier Signed-off-by: Ming Lei Reviewed-by: Christoffer Dall Signed-off-by: Marc Zyngier

KVM: add halt_attempted_poll to VCPU stats

2015-09-16T10:17:00Z

This new statistic can help diagnosing VCPUs that, for any reason, trigger bad behavior of halt_poll_ns autotuning. For example, say halt_poll_ns = 480000, and wakeups are spaced exactly like 479us, 481us, 479us, 481us. Then KVM always fails polling and wastes 10+20+40+80+160+320+480 = 1110 microseconds out of every 479+481+479+481+479+481+479 = 3359 microseconds. The VCPU then is consuming about 30% more CPU than it would use without polling. This would show as an abnormally high number of attempted polling compared to the successful polls. Acked-by: Christian Borntraeger Signed-off-by: Paolo Bonzini

kvm: fix zero length mmio searching

2015-09-15T14:59:46Z

Currently, if we had a zero length mmio eventfd assigned on KVM_MMIO_BUS. It will never be found by kvm_io_bus_cmp() since it always compares the kvm_io_range() with the length that guest wrote. This will cause e.g for vhost, kick will be trapped by qemu userspace instead of vhost. Fixing this by using zero length if an iodevice is zero length. Cc: stable@vger.kernel.org Cc: Gleb Natapov Cc: Paolo Bonzini Signed-off-by: Jason Wang Reviewed-by: Cornelia Huck Signed-off-by: Paolo Bonzini

kvm: fix double free for fast mmio eventfd

2015-09-15T14:59:31Z

We register wildcard mmio eventfd on two buses, once for KVM_MMIO_BUS and once on KVM_FAST_MMIO_BUS but with a single iodev instance. This will lead to an issue: kvm_io_bus_destroy() knows nothing about the devices on two buses pointing to a single dev. Which will lead to double free[1] during exit. Fix this by allocating two instances of iodevs then registering one on KVM_MMIO_BUS and another on KVM_FAST_MMIO_BUS. CPU: 1 PID: 2894 Comm: qemu-system-x86 Not tainted 3.19.0-26-generic #28-Ubuntu Hardware name: LENOVO 2356BG6/2356BG6, BIOS G7ET96WW (2.56 ) 09/12/2013 task: ffff88009ae0c4b0 ti: ffff88020e7f0000 task.ti: ffff88020e7f0000 RIP: 0010:[] [] ioeventfd_release+0x28/0x60 [kvm] RSP: 0018:ffff88020e7f3bc8 EFLAGS: 00010292 RAX: dead000000200200 RBX: ffff8801ec19c900 RCX: 000000018200016d RDX: ffff8801ec19cf80 RSI: ffffea0008bf1d40 RDI: ffff8801ec19c900 RBP: ffff88020e7f3bd8 R08: 000000002fc75a01 R09: 000000018200016d R10: ffffffffc07df6ae R11: ffff88022fc75a98 R12: ffff88021e7cc000 R13: ffff88021e7cca48 R14: ffff88021e7cca50 R15: ffff8801ec19c880 FS: 00007fc1ee3e6700(0000) GS:ffff88023e240000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007f8f389d8000 CR3: 000000023dc13000 CR4: 00000000001427e0 Stack: ffff88021e7cc000 0000000000000000 ffff88020e7f3be8 ffffffffc07e2622 ffff88020e7f3c38 ffffffffc07df69a ffff880232524160 ffff88020e792d80 0000000000000000 ffff880219b78c00 0000000000000008 ffff8802321686a8 Call Trace: [] ioeventfd_destructor+0x12/0x20 [kvm] [] kvm_put_kvm+0xca/0x210 [kvm] [] kvm_vcpu_release+0x18/0x20 [kvm] [] __fput+0xe7/0x250 [] ____fput+0xe/0x10 [] task_work_run+0xd4/0xf0 [] do_exit+0x368/0xa50 [] ? recalc_sigpending+0x1f/0x60 [] do_group_exit+0x45/0xb0 [] get_signal+0x291/0x750 [] do_signal+0x28/0xab0 [] ? do_futex+0xdb/0x5d0 [] ? __wake_up_locked_key+0x18/0x20 [] ? SyS_futex+0x76/0x170 [] do_notify_resume+0x69/0xb0 [] int_signal+0x12/0x17 Code: 5d c3 90 0f 1f 44 00 00 55 48 89 e5 53 48 89 fb 48 83 ec 08 48 8b 7f 20 e8 06 d6 a5 c0 48 8b 43 08 48 8b 13 48 89 df 48 89 42 08 <48> 89 10 48 b8 00 01 10 00 00 RIP [] ioeventfd_release+0x28/0x60 [kvm] RSP Cc: stable@vger.kernel.org Cc: Gleb Natapov Cc: Paolo Bonzini Signed-off-by: Jason Wang Reviewed-by: Cornelia Huck Signed-off-by: Paolo Bonzini