Public bug reported:
Latest QEMU master fails to run with too many MMIO devices specified.
After patch 3ac7d43a6fb [1] QEMU just prints an error message and exits.
> kvm_mem_ioeventfd_add: error adding ioeventfd: No space left on device
This is reproducible e.g. with the following setup:
qemu-3.1.50-dirty \
-machine pc-i440fx-2.7,accel=kvm \
-cpu host -m 4096 \
-smp 2,sockets=2,cores=1,threads=1 \
-drive file=freebsd_vm_1.qcow2,format=qcow2,if=none,id=bootdr \
-device ide-hd,drive=bootdr,bootindex=0 \
-device virtio-scsi-pci,id=vc0 \
-device virtio-scsi-pci,id=vc1 \
-device virtio-scsi-pci,id=vc2 \
-device virtio-scsi-pci,id=vc3 \
Running with just 3 Virtio-SCSI controllers seems to work fine, adding
more than that causes the error above. Note that this is not Virtio-SCSI
specific. I've also reproduced this without any Virtio devices
whatsoever.
strace shows the following ioctl chain over and over:
145787 ioctl(11, KVM_UNREGISTER_COALESCED_MMIO, 0x7f60a4985410) = 0
145787 ioctl(11, KVM_UNREGISTER_COALESCED_MMIO, 0x7f60a4985410) = 0
145787 ioctl(11, KVM_REGISTER_COALESCED_MMIO, 0x7f60a49853b0) = 0
145787 ioctl(11, KVM_REGISTER_COALESCED_MMIO, 0x7f60a49853b0) = -1 ENOSPC (No
space left on device)
145787 ioctl(11, KVM_REGISTER_COALESCED_MMIO, 0x7f60a49853b0) = -1 ENOSPC (No
space left on device)
145787 ioctl(11, KVM_REGISTER_COALESCED_MMIO, 0x7f60a49853b0) = -1 ENOSPC (No
space left on device)
145787 ioctl(11, KVM_REGISTER_COALESCED_MMIO, 0x7f60a49853b0) = -1 ENOSPC (No
space left on device)
145787 ioctl(11, KVM_REGISTER_COALESCED_MMIO, 0x7f60a49853b0) = -1 ENOSPC (No
space left on device)
145787 ioctl(11, KVM_REGISTER_COALESCED_MMIO, 0x7f60a49853b0) = -1 ENOSPC (No
space left on device)
145787 ioctl(11, KVM_REGISTER_COALESCED_MMIO, 0x7f60a49853b0) = -1 ENOSPC (No
space left on device)
145787 ioctl(11, KVM_REGISTER_COALESCED_MMIO, 0x7f60a49853b0) = -1 ENOSPC (No
space left on device)
Which suggests there's some kind of MMIO region leak.
[1]
commit 3ac7d43a6fbb5d4a3d01fc9a055c218030af3727
Author: Paolo Bonzini <pbonz...@redhat.com>
AuthorDate: Wed Nov 28 17:28:45 2018 +0100
Commit: Paolo Bonzini <pbonz...@redhat.com>
CommitDate: Fri Jan 11 13:57:24 2019 +0100
memory: update coalesced_range on transaction_commit
** Affects: qemu
Importance: Undecided
Status: New
--
You received this bug notification because you are a member of qemu-
devel-ml, which is subscribed to QEMU.
https://bugs.launchpad.net/bugs/1813940
Title:
kvm_mem_ioeventfd_add: error adding ioeventfd: No space left on device
Status in QEMU:
New
Bug description:
Latest QEMU master fails to run with too many MMIO devices specified.
After patch 3ac7d43a6fb [1] QEMU just prints an error message and exits.
> kvm_mem_ioeventfd_add: error adding ioeventfd: No space left on device
This is reproducible e.g. with the following setup:
qemu-3.1.50-dirty \
-machine pc-i440fx-2.7,accel=kvm \
-cpu host -m 4096 \
-smp 2,sockets=2,cores=1,threads=1 \
-drive file=freebsd_vm_1.qcow2,format=qcow2,if=none,id=bootdr \
-device ide-hd,drive=bootdr,bootindex=0 \
-device virtio-scsi-pci,id=vc0 \
-device virtio-scsi-pci,id=vc1 \
-device virtio-scsi-pci,id=vc2 \
-device virtio-scsi-pci,id=vc3 \
Running with just 3 Virtio-SCSI controllers seems to work fine, adding
more than that causes the error above. Note that this is not Virtio-
SCSI specific. I've also reproduced this without any Virtio devices
whatsoever.
strace shows the following ioctl chain over and over:
145787 ioctl(11, KVM_UNREGISTER_COALESCED_MMIO, 0x7f60a4985410) = 0
145787 ioctl(11, KVM_UNREGISTER_COALESCED_MMIO, 0x7f60a4985410) = 0
145787 ioctl(11, KVM_REGISTER_COALESCED_MMIO, 0x7f60a49853b0) = 0
145787 ioctl(11, KVM_REGISTER_COALESCED_MMIO, 0x7f60a49853b0) = -1 ENOSPC (No
space left on device)
145787 ioctl(11, KVM_REGISTER_COALESCED_MMIO, 0x7f60a49853b0) = -1 ENOSPC (No
space left on device)
145787 ioctl(11, KVM_REGISTER_COALESCED_MMIO, 0x7f60a49853b0) = -1 ENOSPC (No
space left on device)
145787 ioctl(11, KVM_REGISTER_COALESCED_MMIO, 0x7f60a49853b0) = -1 ENOSPC (No
space left on device)
145787 ioctl(11, KVM_REGISTER_COALESCED_MMIO, 0x7f60a49853b0) = -1 ENOSPC (No
space left on device)
145787 ioctl(11, KVM_REGISTER_COALESCED_MMIO, 0x7f60a49853b0) = -1 ENOSPC (No
space left on device)
145787 ioctl(11, KVM_REGISTER_COALESCED_MMIO, 0x7f60a49853b0) = -1 ENOSPC (No
space left on device)
145787 ioctl(11, KVM_REGISTER_COALESCED_MMIO, 0x7f60a49853b0) = -1 ENOSPC (No
space left on device)
Which suggests there's some kind of MMIO region leak.
[1]
commit 3ac7d43a6fbb5d4a3d01fc9a055c218030af3727
Author: Paolo Bonzini <pbonz...@redhat.com>
AuthorDate: Wed Nov 28 17:28:45 2018 +0100
Commit: Paolo Bonzini <pbonz...@redhat.com>
CommitDate: Fri Jan 11 13:57:24 2019 +0100
memory: update coalesced_range on transaction_commit
To manage notifications about this bug go to:
https://bugs.launchpad.net/qemu/+bug/1813940/+subscriptions
