Re: [Qemu-devel] [Qemu-stable] Patch Round-up for stable 2.1.1, freeze on 2014-09-03

[Michael Roth <mdroth@linux.vnet.ibm.com>]

Quoting Andrey Korolyov (2014-09-02 12:33:13)
> On Tue, Sep 2, 2014 at 7:27 PM, Michael S. Tsirkin <mst@redhat.com> wrote:
> > On Tue, Sep 02, 2014 at 06:25:46PM +0300, Michael S. Tsirkin wrote:
> >> On Tue, Sep 02, 2014 at 10:20:50AM -0500, Michael Roth wrote:
> >> > Quoting Michael Roth (2014-08-27 12:35:57)
> >> > > Hi everyone,
> >> > >
> >> > > The following new patches are queued for QEMU stable v2.1.1:
> >> > >
> >> > >   https://github.com/mdroth/qemu/commits/stable-2.1-staging
> >> >
> >> > As of now the following additional patches have been applied to
> >> > the 2.1.1 staging tree (see stable commit for upstream commit ref):
> >> >
> >> >   a7f9ea2 qxl-render: add more sanity checks
> >> >   1511595 target-arm: Correct Cortex-A57 ISAR5 and AA64ISAR0 ID register values
> >> >   44a1530 target-arm: Fix regression that disabled VFP for ARMv5 CPUs
> >> >   95dcaa4 x86: Clear MTRRs on vCPU reset
> >> >   7fd25d3 x86: kvm: Add MTRR support for kvm_get|put_msrs()
> >> >   0f9c63b x86: Use common variable range MTRR counts
> >> >   0c69595 target-i386: Don't forbid NX bit on PAE PDEs and PTEs
> >> >   89713cb vl: process -object after other backend options
> >> >   93dcbd3 spapr_pci: map the MSI window in each PHB
> >> >
> >> > The following patches have been Cc'd to qemu-stable and are still
> >> > pending upstream commit/merge:
> >> >
> >> >   [PATCH] virtio-net: don't run bh on vm stopped (Michael S. Tsirkin)
> >> >     * pull just sent by Michael Tsirkin
> >> >   [PATCH] net: prevent sending packets while guest is stopped (Stefan Hajnoczi)
> >> >   [PATCH v2 2/4] pci: Avoid losing config updates to MSI/MSIX cap regs (Knut Omang)
> >>
> >>       this is also in that pull
> >>
> >> >   [PATCH V4] net: Forbid dealing with packets when VM is not running (zhanghailiang)
> >> >     * looks like Stefan has this queued for net
> >> >   [PATCH V2] acpi-build: Set FORCE_APIC_CLUSTER_MODEL bit for FADT flags (zhanghailiang)
> >>
> >>       hmm I forgot to include that one. I'll redo the pull request.
> >
> > or maybe not
> > I recalled why I didn't include it - I wanted someone else to review it
> > first
> >
> > if it's ready, I'll send it separately tomorrow
> >
> >
> >> >   [RESEND v2 0/3] Fix some memory leaks about query memdev (Chen Fan)
> >> >     * looks like Michael Tokarev has this queued for trivial
> >> >   [PATCH] target-i386: Support migratable=no properly (Eduardo Habkost)
> >> >     * pinged
> >> >   [PATCH 1/3] pc: Fix disabling of vapic for compat PC models (Jan Kiszka)
> >> >     * pinged
> >> >   [PATCH] pty: Fix byte loss bug when connecting to pty (Sebastian Tanase)
> >> >     * pinged, presumably going through Gerd's tree
> >> >
> >> > Let me know if anything is missing.
> >> >
> >> > >
> >> > > The release is planned for 2014-09-08:
> >> > >
> >> > >   http://wiki.qemu.org/Planning/2.1
> >> > >
> >> > > Please respond here or CC qemu-stable@nongnu.org on any patches you
> >> > > think should be included in the release.
> >> > >
> >> > > Testing/feedback is greatly appreciated.
> >> > >
> >> > > Thanks!
> >> > >
> >> > > ----------------------------------------------------------------
> >> > > Alex Williamson (1):
> >> > >       vfio: Fix MSI-X vector expansion
> >> > >
> >> > > Ben Draper (1):
> >> > >       vmxnet3: Pad short frames to minimum size (60 bytes)
> >> > >
> >> > > Christoffer Dall (2):
> >> > >       target-arm: Rename QEMU PSCI v0.1 definitions
> >> > >       arm/virt: Use PSCI v0.2 function IDs in the DT when KVM uses PSCI v0.2
> >> > >
> >> > > Fam Zheng (1):
> >> > >       blkdebug: Delete BH in bdrv_aio_cancel
> >> > >
> >> > > Gonglei (1):
> >> > >       pcihp: fix possible array out of bounds
> >> > >
> >> > > Hu Tao (3):
> >> > >       hw:i386: typo fix: MEMORY_HOPTLUG_DEVICE -> MEMORY_HOTPLUG_DEVICE
> >> > >       pc-dimm: validate node property
> >> > >       numa: show hex number in error message for consistency and prefix them with 0x
> >> > >
> >> > > Jan Kiszka (1):
> >> > >       pci: Use bus master address space for delivering MSI/MSI-X messages
> >> > >
> >> > > Michael S. Tsirkin (4):
> >> > >       pc-dimm: fix up error message
> >> > >       acpi: align RSDP
> >> > >       hostmem: set MPOL_MF_MOVE
> >> > >       pc: reserve more memory for ACPI for new machine types
> >> > >
> >> > > Michael Tokarev (2):
> >> > >       l2tpv3 (configure): it is linux-specific
> >> > >       ide: only constrain read/write requests to drive size, not other types
> >> > >
> >> > > Peter Lieven (1):
> >> > >       block/iscsi: fix memory corruption on iscsi resize
> >> > >
> >> > > Peter Maydell (1):
> >> > >       target-arm: Fix return address for A64 BRK instructions
> >> > >
> >> > > Stefan Hajnoczi (6):
> >> > >       qmp: hide "hotplugged" device property from device-list-properties
> >> > >       qdev-monitor: include QOM properties in -device FOO, help output
> >> > >       raw-posix: fix O_DIRECT short reads
> >> > >       qemu-iotests: add test case 101 for short file I/O
> >> > >       thread-pool: avoid per-thread-pool EventNotifier
> >> > >       thread-pool: avoid deadlock in nested aio_poll() calls
> >> > >
> >> > > zhanghailiang (1):
> >> > >       virtio-blk: fix reference a pointer which might be freed
> >> > >
> >> > >  backends/hostmem.c              |  2 +-
> >> > >  block/blkdebug.c                |  4 +++
> >> > >  block/iscsi.c                   |  3 ++-
> >> > >  block/raw-posix.c               |  9 +++++++
> >> > >  configure                       |  1 +
> >> > >  hw/acpi/pcihp.c                 |  2 +-
> >> > >  hw/arm/virt.c                   | 31 ++++++++++++++++++----
> >> > >  hw/block/virtio-blk.c           |  3 ++-
> >> > >  hw/i386/acpi-build.c            |  2 +-
> >> > >  hw/i386/acpi-dsdt.dsl           |  4 +--
> >> > >  hw/i386/acpi-dsdt.hex.generated |  8 +++---
> >> > >  hw/i386/pc.c                    | 12 ++++++---
> >> > >  hw/i386/pc_piix.c               |  1 +
> >> > >  hw/i386/pc_q35.c                |  1 +
> >> > >  hw/i386/q35-acpi-dsdt.dsl       |  4 +--
> >> > >  hw/i386/ssdt-mem.dsl            | 16 ++++++------
> >> > >  hw/i386/ssdt-misc.dsl           |  2 +-
> >> > >  hw/ide/core.c                   |  3 ++-
> >> > >  hw/mem/pc-dimm.c                |  6 +++++
> >> > >  hw/misc/vfio.c                  | 38 ++++++++++++++++++++-------
> >> > >  hw/net/vmxnet3.c                | 10 +++++++
> >> > >  hw/pci/msi.c                    |  2 +-
> >> > >  hw/pci/msix.c                   |  2 +-
> >> > >  include/hw/acpi/pc-hotplug.h    |  2 +-
> >> > >  include/hw/i386/pc.h            |  2 ++
> >> > >  numa.c                          |  4 +--
> >> > >  qdev-monitor.c                  | 40 ++++++++++++----------------
> >> > >  qmp.c                           |  1 +
> >> > >  target-arm/kvm-consts.h         | 49 ++++++++++++++++++++++++++--------
> >> > >  target-arm/translate-a64.c      |  2 +-
> >> > >  tests/qemu-iotests/101          | 58 +++++++++++++++++++++++++++++++++++++++++
> >> > >  tests/qemu-iotests/101.out      | 10 +++++++
> >> > >  tests/qemu-iotests/group        |  1 +
> >> > >  thread-pool.c                   | 27 ++++++++++---------
> >> > >  34 files changed, 269 insertions(+), 93 deletions(-)
> >> > >  create mode 100755 tests/qemu-iotests/101
> >> > >  create mode 100644 tests/qemu-iotests/101.out
> >
> 
> 
> Hi, I`ve got assertion failure while testing upcoming tree with net
> patches applied after start - boot - reset cycle of an i686 VM.

Thanks for testing! Is this a consistent failure? I'm unable to reproduce with
current 2.1.1 tree with these applied:

  0d33fef vhost_net: start/stop guest notifiers properly
  27066b6 pci: avoid losing config updates to MSI/MSIX cap regs
  64a953a virtio-net: don't run bh on vm stopped
  1dcb588 vhost_net: cleanup start/stop condition

I haven't tested with these but I don't see them coming into play in
this scenario:

  383004 New          [V4] net: Forbid dealing with packets when VM is not  zhiang
  385148 New          net: prevent sending packets while guest is stopped

I'm running a 32-bit fedora 15 guest via qemu x86_64, and confirmed vhost is
being started/stopped on each boot/reset. My command-line is:

./x86_64-softmmu/qemu-system-x86_64 -drive file=/home/mdroth/vm/fedora15_32.raw,if=virtio -smp 1 -m 512M -name vm3 -device virtio-net-pci,mac=52:54:00:12:34:03,netdev=netdev0,id=net0 -netdev tap,script=/etc/qemu-ifup,vhost=on,id=netdev0 -vnc :3 -device virtio-serial -balloon virtio -mon chardev=hmp0 -chardev socket,path=/tmp/vm3-hmp0.sock,server,nowait,id=hmp0 -mon chardev=qmp0,mode=control -chardev socket,path=/tmp/vm3-qmp0.sock,server,nowait,id=qmp0 -device virtserialport,chardev=vs0,name=vs0 -chardev socket,path=/tmp/vm3-vs0.sock,server,nowait,id=vs0 -device virtserialport,chardev=vs1,name=vs1 -chardev socket,path=/tmp/vm3-vs1.sock,server,nowait,id=vs1 -device virtserialport,chardev=qga,name=org.qemu.guest_agent.0 -chardev socket,path=/tmp/vm3-qga.sock,server,nowait,id=qga -device isa-serial,chardev=serial0 -chardev socket,path=/tmp/vm3-serial0.sock,server,nowait,id=serial0 -L ./pc-bios --enable-kvm

I've tried rebooting from within the guest as well as system_reset via
hmp and it doesn't seem to trigger on my end.

> 
> qemu-system-x86_64:
> /tmp/buildd/qemu-2.1.0+f1/hw/net/virtio-net.c:1543:
> virtio_net_guest_notifier_mask: Assertion `n->vhost_started' failed.
> 
> net patches applied:
> 383004 New          [V4] net: Forbid dealing with packets when VM is not  zhiang
> 385148 New          net: prevent sending packets while guest is stopped
> 385177 New          [PULL,09/13] vhost_net: cleanup start/stop condition
> 385180 New          [PULL,11/13] virtio-net: don't run bh on vm stopped
> 385179 New          [PULL,12/13] pci: avoid losing config updates to
> MSI/MSIX cap regs
> 385181 New          [PULL,13/13] vhost_net: start/stop guest notifiers properly
> 
> bad one is the
> 
> Author: Jason Wang <jasowang@redhat.com>
> Date:   Tue Sep 2 18:07:46 2014 +0300
> 
>     vhost_net: start/stop guest notifiers properly