qemu.git
4 years agoxen-disk: use g_new0 to fix build
Olaf Hering [Fri, 28 Jul 2017 13:11:51 +0000 (15:11 +0200)] 
xen-disk: use g_new0 to fix build

g_malloc0_n is available since glib-2.24. To allow build with older glib
versions use the generic g_new0, which is already used in many other
places in the code.

Fixes commit 3284fad728 ("xen-disk: add support for multi-page shared rings")

Signed-off-by: Olaf Hering <olaf@aepfle.de>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Signed-off-by: Stefano Stabellini <sstabellini@kernel.org>
4 years agoMerge remote-tracking branch 'remotes/gkurz/tags/for-upstream' into staging
Peter Maydell [Wed, 20 Sep 2017 19:33:48 +0000 (20:33 +0100)] 
Merge remote-tracking branch 'remotes/gkurz/tags/for-upstream' into staging

These patches fix regressions in 2.10

# gpg: Signature made Wed 20 Sep 2017 07:51:07 BST
# gpg:                using DSA key 0x02FC3AEB0101DBC2
# gpg: Good signature from "Greg Kurz <groug@kaod.org>"
# gpg:                 aka "Greg Kurz <groug@free.fr>"
# gpg:                 aka "Greg Kurz <gkurz@linux.vnet.ibm.com>"
# gpg:                 aka "Gregory Kurz (Groug) <groug@free.fr>"
# gpg:                 aka "[jpeg image of size 3330]"
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg:          There is no indication that the signature belongs to the owner.
# Primary key fingerprint: 2BD4 3B44 535E C0A7 9894  DBA2 02FC 3AEB 0101 DBC2

* remotes/gkurz/tags/for-upstream:
  9pfs: check the size of transport buffer before marshaling
  9pfs: fix name_to_path assertion in v9fs_complete_rename()
  9pfs: fix readdir() for 9p2000.u

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 years agoMerge remote-tracking branch 'remotes/ehabkost/tags/machine-next-pull-request' into...
Peter Maydell [Wed, 20 Sep 2017 16:35:36 +0000 (17:35 +0100)] 
Merge remote-tracking branch 'remotes/ehabkost/tags/machine-next-pull-request' into staging

Machine/CPU/NUMA queue, 2017-09-19

# gpg: Signature made Tue 19 Sep 2017 21:17:01 BST
# gpg:                using RSA key 0x2807936F984DC5A6
# gpg: Good signature from "Eduardo Habkost <ehabkost@redhat.com>"
# Primary key fingerprint: 5A32 2FD5 ABC4 D3DB ACCF  D1AA 2807 936F 984D C5A6

* remotes/ehabkost/tags/machine-next-pull-request:
  MAINTAINERS: Update git URLs for my trees
  hw/acpi-build: Fix SRAT memory building in case of node 0 without RAM
  NUMA: Replace MAX_NODES with nb_numa_nodes in for loop
  numa: cpu: calculate/set default node-ids after all -numa CLI options are parsed
  arm: drop intermediate cpu_model -> cpu type parsing and use cpu type directly
  pc: use generic cpu_model parsing
  vl.c: convert cpu_model to cpu type and set of global properties before machine_init()
  cpu: make cpu_generic_init() abort QEMU on error
  qom: cpus: split cpu_generic_init() on feature parsing and cpu creation parts
  hostmem-file: Add "discard-data" option
  osdep: Define QEMU_MADV_REMOVE
  vl: Clean up user-creatable objects when exiting

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 years ago9pfs: check the size of transport buffer before marshaling
Jan Dakinevich [Wed, 20 Sep 2017 06:48:52 +0000 (08:48 +0200)] 
9pfs: check the size of transport buffer before marshaling

v9fs_do_readdir_with_stat() should check for a maximum buffer size
before an attempt to marshal gathered data. Otherwise, buffers assumed
as misconfigured and the transport would be broken.

The patch brings v9fs_do_readdir_with_stat() in conformity with
v9fs_do_readdir() behavior.

Signed-off-by: Jan Dakinevich <jan.dakinevich@gmail.com>
[groug, regression caused my commit 8d37de41cab1 # 2.10]
Signed-off-by: Greg Kurz <groug@kaod.org>
4 years ago9pfs: fix name_to_path assertion in v9fs_complete_rename()
Jan Dakinevich [Wed, 20 Sep 2017 06:48:52 +0000 (08:48 +0200)] 
9pfs: fix name_to_path assertion in v9fs_complete_rename()

The third parameter of v9fs_co_name_to_path() must not contain `/'
character.

The issue is most likely related to 9p2000.u protocol only.

Signed-off-by: Jan Dakinevich <jan.dakinevich@gmail.com>
[groug, regression caused by commit f57f5878578a # 2.10]
Signed-off-by: Greg Kurz <groug@kaod.org>
4 years ago9pfs: fix readdir() for 9p2000.u
Jan Dakinevich [Wed, 20 Sep 2017 06:48:51 +0000 (08:48 +0200)] 
9pfs: fix readdir() for 9p2000.u

If the client is using 9p2000.u, the following occurs:

$ cd ${virtfs_shared_dir}
$ mkdir -p a/b/c
$ ls a/b
ls: cannot access 'a/b/a': No such file or directory
ls: cannot access 'a/b/b': No such file or directory
a  b  c

instead of the expected:

$ ls a/b
c

This is a regression introduced by commit f57f5878578a;
local_name_to_path() now resolves ".." and "." in paths,
and v9fs_do_readdir_with_stat()->stat_to_v9stat() then
copies the basename of the resulting path to the response.
With the example above, this means that "." and ".." are
turned into "b" and "a" respectively...

stat_to_v9stat() currently assumes it is passed a full
canonicalized path and uses it to do two different things:
1) to pass it to v9fs_co_readlink() in case the file is a symbolic
   link
2) to set the name field of the V9fsStat structure to the basename
   part of the given path

It only has two users: v9fs_stat() and v9fs_do_readdir_with_stat().

v9fs_stat() really needs 1) and 2) to be performed since it starts
with the full canonicalized path stored in the fid. It is different
for v9fs_do_readdir_with_stat() though because the name we want to
put into the V9fsStat structure is the d_name field of the dirent
actually (ie, we want to keep the "." and ".." special names). So,
we only need 1) in this case.

This patch hence adds a basename argument to stat_to_v9stat(), to
be used to set the name field of the V9fsStat structure, and moves
the basename logic to v9fs_stat().

Signed-off-by: Jan Dakinevich <jan.dakinevich@gmail.com>
(groug, renamed old name argument to path and updated changelog)
Signed-off-by: Greg Kurz <groug@kaod.org>
4 years agoMAINTAINERS: Update git URLs for my trees
Eduardo Habkost [Fri, 1 Sep 2017 15:39:28 +0000 (12:39 -0300)] 
MAINTAINERS: Update git URLs for my trees

List the branches where I queue patches for Machine Core, NUMA,
Memory Backends, and X86.  Update the NUMA section to list the
"machine-next" branch instead of "numa".

Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <20170901153928.17058-1-ehabkost@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
4 years agohw/acpi-build: Fix SRAT memory building in case of node 0 without RAM
Eduardo Habkost [Fri, 1 Sep 2017 02:10:02 +0000 (10:10 +0800)] 
hw/acpi-build: Fix SRAT memory building in case of node 0 without RAM

Currently, Using the fisrt node without memory on the machine makes
QEMU unhappy. With this example command line:
  ... \
  -m 1024M,slots=4,maxmem=32G \
  -numa node,nodeid=0 \
  -numa node,mem=1024M,nodeid=1 \
  -numa node,nodeid=2 \
  -numa node,nodeid=3 \
Guest reports "No NUMA configuration found" and the NUMA topology is
wrong.

This is because when QEMU builds ACPI SRAT, it regards node 0 as the
default node to deal with the memory hole(640K-1M). this means the
node0 must have some memory(>1M), but, actually it can have no
memory.

Fix this problem by cut out the 640K hole in the same way the PCI
4G hole does.

Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Dou Liyang <douly.fnst@cn.fujitsu.com>
Message-Id: <1504231805-30957-2-git-send-email-douly.fnst@cn.fujitsu.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
4 years agoNUMA: Replace MAX_NODES with nb_numa_nodes in for loop
Dou Liyang [Tue, 22 Aug 2017 07:45:36 +0000 (15:45 +0800)] 
NUMA: Replace MAX_NODES with nb_numa_nodes in for loop

In QEMU, the number of the NUMA nodes is determined by parse_numa_opts().
Then, QEMU uses it for iteration, for example:
  for (i = 0; i < nb_numa_nodes; i++)

However, in memory_region_allocate_system_memory(), it uses MAX_NODES
not nb_numa_nodes.

So, replace MAX_NODES with nb_numa_nodes to keep code consistency and
reduce the loop times.

Signed-off-by: Dou Liyang <douly.fnst@cn.fujitsu.com>
Message-Id: <1503387936-3483-1-git-send-email-douly.fnst@cn.fujitsu.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
4 years agonuma: cpu: calculate/set default node-ids after all -numa CLI options are parsed
Igor Mammedov [Thu, 1 Jun 2017 10:53:28 +0000 (12:53 +0200)] 
numa: cpu: calculate/set default node-ids after all -numa CLI options are parsed

Calculating default node-ids for CPUs in possible_cpu_arch_ids()
is rather fragile since defaults calculation uses nb_numa_nodes but
callback might be potentially called early before all -numa CLI
options are parsed, which would lead to cpus assigned only upto
nb_numa_nodes at the time possible_cpu_arch_ids() is called.

Issue was introduced by
(7c88e65 numa: mirror cpu to node mapping in MachineState::possible_cpus)
and for example CLI:
  -smp 4 -numa node,cpus=0 -numa node
would set props.node-id in possible_cpus array for every non
explicitly mapped CPU to the first node.

Issue is not visible to guest nor to mgmt interface due to
  1) implictly mapped cpus are forced to the first node in
     case of partial mapping
  2) in case of default mapping possible_cpu_arch_ids() is
     called after all -numa options are parsed (resulting
     in correct mapping).

However it's fragile to rely on late execution of
possible_cpu_arch_ids(), therefore add machine specific
callback that returns node-id for CPU and use it to calculate/
set defaults at machine_numa_finish_init() time when all -numa
options are parsed.

Reported-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <1496314408-163972-1-git-send-email-imammedo@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
4 years agoMerge remote-tracking branch 'remotes/cohuck/tags/s390x-20170919-v2' into staging
Peter Maydell [Tue, 19 Sep 2017 17:08:48 +0000 (18:08 +0100)] 
Merge remote-tracking branch 'remotes/cohuck/tags/s390x-20170919-v2' into staging

Assorted s390x patches:
- introduce virtio-gpu-ccw, with virtio-gpu endian fixes
- lots of cleanup in the s390x code
- make device_add work for s390x cpus
- enable seccomp on s390x
- an ivshmem endian fix
- set the reserved DHCP client architecture id for netboot
- fixes in the css and pci support

# gpg: Signature made Tue 19 Sep 2017 17:39:45 BST
# gpg:                using RSA key 0xDECF6B93C6F02FAF
# gpg: Good signature from "Cornelia Huck <conny@cornelia-huck.de>"
# gpg:                 aka "Cornelia Huck <huckc@linux.vnet.ibm.com>"
# gpg:                 aka "Cornelia Huck <cornelia.huck@de.ibm.com>"
# gpg:                 aka "Cornelia Huck <cohuck@kernel.org>"
# gpg:                 aka "Cornelia Huck <cohuck@redhat.com>"
# Primary key fingerprint: C3D0 D66D C362 4FF6 A8C0  18CE DECF 6B93 C6F0 2FAF

* remotes/cohuck/tags/s390x-20170919-v2: (38 commits)
  MAINTAINERS/s390x: add terminal3270.c
  virtio-ccw: Create a virtio gpu device for the ccw bus
  virtio-gpu: Handle endian conversion
  s390x/ccw: create s390 phb for compat reasons as well
  configure: Allow --enable-seccomp on s390x, too
  virtio-ccw: remove stale comments on endianness
  s390x: allow CPU hotplug in random core-id order
  s390x: generate sclp cpu information from possible_cpus
  s390x: get rid of cpu_s390x_create()
  s390x: get rid of cpu_states and use possible_cpus instead
  s390x: implement query-hotpluggable-cpus
  s390x: CPU hot unplug via device_del cannot work for now
  s390x: allow cpu hotplug via device_add
  s390x: print CPU definitions in sorted order
  target/s390x: rename next_cpu_id to next_core_id
  target/s390x: use "core-id" for cpu number/address/id handling
  target/s390x: set cpu->id for linux user when realizing
  s390x: allow only 1 CPU with TCG
  target/s390x: use program_interrupt() in per_check_exception()
  target/s390x: use trigger_pgm_exception() in s390_cpu_handle_mmu_fault()
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 years agoMAINTAINERS/s390x: add terminal3270.c
Christian Borntraeger [Mon, 18 Sep 2017 13:04:55 +0000 (15:04 +0200)] 
MAINTAINERS/s390x: add terminal3270.c

Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Acked-by: Halil Pasic <pasic@linux.vnet.ibm.com>
Message-Id: <20170918130455.144262-1-borntraeger@de.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agovirtio-ccw: Create a virtio gpu device for the ccw bus
Farhan Ali [Fri, 15 Sep 2017 14:40:32 +0000 (10:40 -0400)] 
virtio-ccw: Create a virtio gpu device for the ccw bus

Wire up the virtio-gpu device for the CCW bus. The virtio-gpu
is a virtio-1 device, so disable revision 0.

Signed-off-by: Farhan Ali <alifm@linux.vnet.ibm.com>
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Reviewed-by: Halil Pasic <pasic@linux.vnet.ibm.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-Id: <6c53f939cf2d64b66d2a6878b29c9bf3820f3d5b.1505485574.git.alifm@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agovirtio-gpu: Handle endian conversion
Farhan Ali [Fri, 15 Sep 2017 14:40:31 +0000 (10:40 -0400)] 
virtio-gpu: Handle endian conversion

Virtio GPU code currently only supports litte endian format,
and so using the Virtio GPU device on a big endian machine
does not work.

Let's fix it by supporting the correct host cpu byte order.

Signed-off-by: Farhan Ali <alifm@linux.vnet.ibm.com>
Message-Id: <dc748e15f36db808f90b4f2393bc29ba7556a9f6.1505485574.git.alifm@linux.vnet.ibm.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390x/ccw: create s390 phb for compat reasons as well
Cornelia Huck [Fri, 15 Sep 2017 09:45:23 +0000 (11:45 +0200)] 
s390x/ccw: create s390 phb for compat reasons as well

d32bd032d8 ("s390x/ccw: create s390 phb conditionally") made
registering the s390 pci host bridge conditional on presense
of the zpci facility bit. Sadly, that breaks migration from
machines that did not use the cpu model (2.7 and previous).

Create the s390 phb for pre-cpu model machines as well: We can
tweak s390_has_feat() to always indicate the zpci facility bit
when no cpu model is available (on 2.7 and previous compat machines).

Fixes: d32bd032d8 ("s390x/ccw: create s390 phb conditionally")
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agoconfigure: Allow --enable-seccomp on s390x, too
Thomas Huth [Thu, 14 Sep 2017 10:36:03 +0000 (12:36 +0200)] 
configure: Allow --enable-seccomp on s390x, too

libseccomp supports s390x since version 2.3.0, and I was able to start
a VM with "-sandbox on" without any obvious problems by using this patch,
so it should be safe to allow --enable-seccomp on s390x nowadays, too.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <1505385363-27717-1-git-send-email-thuth@redhat.com>
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Acked-by: Eduardo Otubo <otubo@redhat.com>
Acked-by: Halil Pasic <pasic@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agovirtio-ccw: remove stale comments on endianness
Halil Pasic [Thu, 14 Sep 2017 10:55:35 +0000 (12:55 +0200)] 
virtio-ccw: remove stale comments on endianness

We have two stale comments suggesting one should think about virtio
config space endianness a bit longer. We have just done that, and came to
the conclusion we are fine as is: it's the responsibility of the virtio
device and not of the transport (and that is how it works now). Putting
the responsibility into the transport isn't even possible, because the
transport would have to know about the config space layout of each
device.

Let us remove the stale comments.

Signed-off-by: Halil Pasic <pasic@linux.vnet.ibm.com>
Suggested-by: Cornelia Huck <cohuck@redhat.com>
Message-Id: <20170914105535.47941-1-pasic@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390x: allow CPU hotplug in random core-id order
David Hildenbrand [Wed, 13 Sep 2017 13:24:17 +0000 (15:24 +0200)] 
s390x: allow CPU hotplug in random core-id order

SCLP correctly indicates the core-id aka. CPU address for each available
CPU.

As the core-id corresponds to cpu_index, also a newly created kvm vcpu
gets assigned this core-id as vcpu id. So SIGP in the kernel works
correctly (it uses the vcpu id to lookup the correct CPU).

So there should be nothing hindering us from hotplugging CPUs in random
core-id order.

This now makes sure that the output from "query-hotpluggable-cpus"
is completely true. Until now, a specific order is implicit. Performance
vice, hotplugging CPUs in non-sequential order might not be the best thing
to do, as VCPU lookup inside KVM might be a little slower. But that
doesn't hinder us from supporting it.

next_core_id is now used by linux user only.

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170913132417.24384-23-david@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390x: generate sclp cpu information from possible_cpus
David Hildenbrand [Wed, 13 Sep 2017 13:24:16 +0000 (15:24 +0200)] 
s390x: generate sclp cpu information from possible_cpus

This is the first step to allow hot plugging of CPUs in a non-sequential
order. If a cpu is available ("plugged") can directly be decided by
looking at the cpu state pointer.

This makes sure, that really only cpus attached to the machine are
reported.

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170913132417.24384-22-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390x: get rid of cpu_s390x_create()
David Hildenbrand [Wed, 13 Sep 2017 13:24:15 +0000 (15:24 +0200)] 
s390x: get rid of cpu_s390x_create()

Now that there is only one user of cpu_s390x_create() left, make cpu
creation look like on x86.
- Perform the model/properties split and checks in s390_init_cpus()
- Parse features only once without having to remember if already parsed
- Pass only the typename to s390x_new_cpu()
- Use the typename of an existing CPU for hotplug via cpu-add

Acked-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170913132417.24384-21-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390x: get rid of cpu_states and use possible_cpus instead
David Hildenbrand [Wed, 13 Sep 2017 13:24:14 +0000 (15:24 +0200)] 
s390x: get rid of cpu_states and use possible_cpus instead

Now that we have possible_cpus, we can get rid of the global variable
and rewrite s390_cpu_addr2state() to use it.

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170913132417.24384-20-david@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390x: implement query-hotpluggable-cpus
David Hildenbrand [Wed, 13 Sep 2017 13:24:13 +0000 (15:24 +0200)] 
s390x: implement query-hotpluggable-cpus

CPU hotplug is only possible on a per core basis on s390x. So let's
add possible_cpus and wire everything up properly.

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170913132417.24384-19-david@redhat.com>
Acked-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390x: CPU hot unplug via device_del cannot work for now
David Hildenbrand [Wed, 13 Sep 2017 13:24:12 +0000 (15:24 +0200)] 
s390x: CPU hot unplug via device_del cannot work for now

device_del on a CPU will currently do nothing. Let's emit an error
telling that this is will currently not work (there is no architecture
support on s390x). Error message copied from ppc.

(qemu) device_del cpu1
device_del cpu1
CPU hot unplug not supported on this machine

Reviewed-by: Matthew Rosato <mjrosato@linux.vnet.ibm.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170913132417.24384-18-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390x: allow cpu hotplug via device_add
David Hildenbrand [Wed, 13 Sep 2017 13:24:11 +0000 (15:24 +0200)] 
s390x: allow cpu hotplug via device_add

E.g. the following now works:
    device_add host-s390-cpu,id=cpu1,core-id=1

The system will perform the same checks as when using cpu_add:
- If the core_id is already in use
- If the next sequential core_id isn't used
- If core-id >= max_cpu is specified

In addition, mixed CPU models are checked. E.g. if starting with
-cpu host and trying to hotplug "qemu-s390-cpu":
    "Mixed CPU models are not supported on s390x."

Reviewed-by: Matthew Rosato <mjrosato@linux.vnet.ibm.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170913132417.24384-17-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390x: print CPU definitions in sorted order
David Hildenbrand [Wed, 13 Sep 2017 13:24:10 +0000 (15:24 +0200)] 
s390x: print CPU definitions in sorted order

Other architectures provide nicely sorted lists, let's do it similarly on
s390x.

While at it, clean up the code we have to touch either way.

Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170913132417.24384-16-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agotarget/s390x: rename next_cpu_id to next_core_id
David Hildenbrand [Wed, 13 Sep 2017 13:24:09 +0000 (15:24 +0200)] 
target/s390x: rename next_cpu_id to next_core_id

Adapt to the new term "core_id". While at it, fix the type and drop the
initialization to 0 (which is superfluous).

Reviewed-by: Matthew Rosato <mjrosato@linux.vnet.ibm.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170913132417.24384-15-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agotarget/s390x: use "core-id" for cpu number/address/id handling
David Hildenbrand [Wed, 13 Sep 2017 13:24:08 +0000 (15:24 +0200)] 
target/s390x: use "core-id" for cpu number/address/id handling

Some time ago we discussed that using "id" as property name is not the
right thing to do, as it is a reserved property for other devices and
will not work with device_add.

Switch to the term "core-id" instead, and use it as an equivalent to
"CPU address" mentioned in the PoP. There is no such thing as cpu number,
so rename env.cpu_num to env.core_id. We use "core-id" as this is the
common term to use for device_add later on (x86 and ppc).

We can get rid of cpu->id now. Keep cpu_index and env->core_id in sync.
cpu_index was already implicitly used by e.g. cpu_exists(), so keeping
both in sync seems to be the right thing to do.

cpu_index will now no longer automatically get set via
cpu_exec_realizefn(). For now, we were lucky that both implicitly stayed
in sync.

Our new cpu property "core-id" can be a static property. Range checks can
be avoided by using the correct type and the "setting after realized"
check is done implicitly.

device_add will later need the reserved "id" property. Hotplugging a CPU
on s390x will then be: "device_add host-s390-cpu,id=cpu2,core-id=2".

Reviewed-by: Matthew Rosato <mjrosato@linux.vnet.ibm.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170913132417.24384-14-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agotarget/s390x: set cpu->id for linux user when realizing
David Hildenbrand [Wed, 13 Sep 2017 13:24:07 +0000 (15:24 +0200)] 
target/s390x: set cpu->id for linux user when realizing

scc->next_cpu_id is updated when realizing. Setting it just before that
point looks cleaner.

Reviewed-by: Matthew Rosato <mjrosato@linux.vnet.ibm.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170913132417.24384-13-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390x: allow only 1 CPU with TCG
David Hildenbrand [Wed, 13 Sep 2017 13:24:06 +0000 (15:24 +0200)] 
s390x: allow only 1 CPU with TCG

Specifying more than 1 CPU (e.g. -smp 5) leads to SIGP errors (the
guest tries to bring these CPUs up but fails), because we don't support
multiple CPUs on s390x under TCG.

Let's bail out if more than 1 is specified, so we don't raise people's
hope.

Tested-by: Matthew Rosato <mjrosato@linux.vnet.ibm.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170913132417.24384-12-david@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agotarget/s390x: use program_interrupt() in per_check_exception()
David Hildenbrand [Wed, 13 Sep 2017 13:24:05 +0000 (15:24 +0200)] 
target/s390x: use program_interrupt() in per_check_exception()

Clean it up by reusing program_interrupt(). Add a concern regarding
ilen.

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170913132417.24384-11-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agotarget/s390x: use trigger_pgm_exception() in s390_cpu_handle_mmu_fault()
David Hildenbrand [Wed, 13 Sep 2017 13:24:04 +0000 (15:24 +0200)] 
target/s390x: use trigger_pgm_exception() in s390_cpu_handle_mmu_fault()

This looks cleaner. linux-user will not use the ilen field, so setting
it doesn't do any harm.

Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170913132417.24384-10-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390x: move sclp_service_call() to sclp.h
David Hildenbrand [Wed, 13 Sep 2017 13:24:03 +0000 (15:24 +0200)] 
s390x: move sclp_service_call() to sclp.h

Implemented in sclp.c, so let's move it to the right include file.
Also adjust some includes.

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170913132417.24384-9-david@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agotarget/s390x: move some s390x typedefs to cpu-qom.h
David Hildenbrand [Wed, 13 Sep 2017 13:24:02 +0000 (15:24 +0200)] 
target/s390x: move some s390x typedefs to cpu-qom.h

This allows us to drop inclusion of cpu_models.h in cpu-qom.h, and
prepares for using cpu-qom.h as a s390 specific version of typedefs.h

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170913132417.24384-8-david@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390x: move subsystem_reset() to s390-virtio-ccw.h
David Hildenbrand [Wed, 13 Sep 2017 13:24:01 +0000 (15:24 +0200)] 
s390x: move subsystem_reset() to s390-virtio-ccw.h

Implemented in s390-virtio-ccw.c, so move it to the right header.
We can also drop the extern. Fix up one include.

Reviewed-by: Thomas Huth <thuth@redhat.com>
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170913132417.24384-7-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390x: move s390_virtio_hypercall() to s390-virtio-hcall.h
David Hildenbrand [Wed, 13 Sep 2017 13:24:00 +0000 (15:24 +0200)] 
s390x: move s390_virtio_hypercall() to s390-virtio-hcall.h

Implemented in hw/s390x/s390-virtio-hcall.c, so let's move it to the
right header file.

Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170913132417.24384-6-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390x: rename s390-virtio.h to s390-virtio-hcall.h
David Hildenbrand [Wed, 13 Sep 2017 13:23:59 +0000 (15:23 +0200)] 
s390x: rename s390-virtio.h to s390-virtio-hcall.h

The only interface left, so let's properly rename it.

Reviewed-by: Thomas Huth <thuth@redhat.com>
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170913132417.24384-5-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390x: get rid of s390-virtio.c
David Hildenbrand [Wed, 13 Sep 2017 13:23:58 +0000 (15:23 +0200)] 
s390x: get rid of s390-virtio.c

It is a leftover from the days where we had still the !ccw virtio
machine. As this one is long gone, let's move everything to
s390-virtio-ccw.c.

Suggested-by: Cornelia Huck <cohuck@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170913132417.24384-4-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agocpu: drop old comments describing members
David Hildenbrand [Wed, 13 Sep 2017 13:23:57 +0000 (15:23 +0200)] 
cpu: drop old comments describing members

These comments are obviously stale.

Acked-by: Cornelia Huck <cohuck@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170913132417.24384-3-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agoexec,dump,i386,ppc,s390x: don't include exec/cpu-all.h explicitly
David Hildenbrand [Wed, 13 Sep 2017 13:23:56 +0000 (15:23 +0200)] 
exec,dump,i386,ppc,s390x: don't include exec/cpu-all.h explicitly

All but a handful of files include exec/cpu-all.h via cpu.h only.
As these files already include cpu.h, let's just drop the additional
include.

Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170913132417.24384-2-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390: set DHCP client architecure id for netboot
Viktor Mihajlovski [Mon, 11 Sep 2017 10:33:47 +0000 (12:33 +0200)] 
s390: set DHCP client architecure id for netboot

Setting the client architecture DHCP option to 0x001f (s390 Basic) [1]
allows the DHCP server to return a s390-specific bootfile if wanted.
DHCP servers not configured for the option (or not yet recognizing the
option value) will continue to work as they have done before.

[1] https://www.iana.org/assignments/dhcpv6-parameters

Signed-off-by: Viktor Mihajlovski <mihajlov@linux.vnet.ibm.com>
Message-Id: <1505126027-1704-1-git-send-email-mihajlov@linux.vnet.ibm.com>
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390x/css: remove unused error handling branch
Halil Pasic [Fri, 8 Sep 2017 15:24:45 +0000 (17:24 +0200)] 
s390x/css: remove unused error handling branch

The case in question actually never happens. Let us get rid of the dead
code.

Signed-off-by: Halil Pasic <pasic@linux.vnet.ibm.com>
Message-Id: <20170908152446.14606-4-pasic@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390x/css: fix NULL handling for CCW addresses
Halil Pasic [Fri, 8 Sep 2017 15:24:44 +0000 (17:24 +0200)] 
s390x/css: fix NULL handling for CCW addresses

Back then in the time of df1fe5bb49 ("s390: Virtual channel subsystem
support.", 2013-01-24) -EIO used to map to a channel-program check (via
the default label of the switch statement).  Then 2dc95b4cac
("s390x/3270: 3270 data stream handling", 2016-04-01) came along
and that changed dramatically.

Let us roll back this undesired side effect, and go back to
channel-program check.

Signed-off-by: Halil Pasic <pasic@linux.vnet.ibm.com>
Fixes: 2dc95b4cac "s390x/3270: 3270 data stream handling"
Message-Id: <20170908152446.14606-3-pasic@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390x/css: drop data-check in interpretation
Halil Pasic [Fri, 8 Sep 2017 15:24:43 +0000 (17:24 +0200)] 
s390x/css: drop data-check in interpretation

The architecture says that channel-data check is indicating that
an uncorrected storage (memory) error has been detected in regard
to the data residing in main storage (memory) that is currently
used for an I/O operation. The described detection is done using
the CBC technology.

The ccw interpretation code is however generating a channel-data check
effectively when the (device specific) ccw_cb returns -EFAULT.  In case
of virtio-ccw devices this happens when mapping memory fails, or when a
NULL pointer is encountered. So this behavior is not architecture
conform.

Furthermore the best fit for these situations (null pointer, mapping a
piece of guest memory fails) from architectural perspective the condition
described as the channel subsystem refers to a location that is not
available, which when encountered shall result in a channel-program
check.

To fix this, all we have to do is to get rid of the switch case matching
-EFAULT: the default is generating a channel-program check.

Signed-off-by: Halil Pasic <pasic@linux.vnet.ibm.com>
Message-Id: <20170908152446.14606-2-pasic@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agohw/misc/ivshmem: Fix ivshmem_recv_msg() to also work on big endian systems
Thomas Huth [Wed, 30 Aug 2017 13:39:03 +0000 (15:39 +0200)] 
hw/misc/ivshmem: Fix ivshmem_recv_msg() to also work on big endian systems

The "slow" ivshmem-tests currently fail when they are running on a
big endian host:

$ uname -m
ppc64
$ V=1 QTEST_QEMU_BINARY=x86_64-softmmu/qemu-system-x86_64 tests/ivshmem-test -m slow
/x86_64/ivshmem/single: OK
/x86_64/ivshmem/hotplug: OK
/x86_64/ivshmem/memdev: OK
/x86_64/ivshmem/pair: OK
/x86_64/ivshmem/server-msi: qemu-system-x86_64:
 -device ivshmem-doorbell,chardev=chr0,vectors=2: server sent invalid ID message
Broken pipe

The problem is that the server side code in ivshmem_server_send_one_msg()
correctly translates all messages IDs into little endian 64-bit values,
but the client side code in the ivshmem_recv_msg() function does not swap
the byte order back. Fix it by passing the value through le64_to_cpu().

Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <1504100343-26607-1-git-send-email-thuth@redhat.com>
Tested-by: Cornelia Huck <cohuck@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390x/pci: add iommu replay callback
Yi Min Zhao [Tue, 5 Sep 2017 10:13:00 +0000 (12:13 +0200)] 
s390x/pci: add iommu replay callback

Let's introduce iommu replay callback for s390 pci iommu memory region.
Currently we don't need any dma mapping replay. So let it return
directly. This implementation will avoid meaningless loops calling
translation callback.

Reviewed-by: Pierre Morel <pmorel@linux.vnet.ibm.com>
Reviewed-by: Halil Pasic <pasic@linux.vnet.ibm.com>
Signed-off-by: Yi Min Zhao <zyimin@linux.vnet.ibm.com>
Message-Id: <1504606380-49341-4-git-send-email-zyimin@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390x/pci: fixup ind_offset of msix routing entry
Yi Min Zhao [Tue, 5 Sep 2017 10:12:59 +0000 (12:12 +0200)] 
s390x/pci: fixup ind_offset of msix routing entry

The guest uses the mpcifc instruction to register the aibvo of a zpci
device, which is the starting offset of indicators in the indicator
area and thus remains constant. Each msix vector is an offset from the
aibvo. When we map a msix route to an adapter route, we should not
modify the starting offset, but instead add the vector to the starting
offset to get the absolute offset in the specific route.

Signed-off-by: Yi Min Zhao <zyimin@linux.vnet.ibm.com>
Message-Id: <1504606380-49341-3-git-send-email-zyimin@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390x/pci: remove idx from msix msg data
Yi Min Zhao [Tue, 5 Sep 2017 10:12:58 +0000 (12:12 +0200)] 
s390x/pci: remove idx from msix msg data

PCIDevice pointer has been a parameter of kvm_arch_fixup_msi_route().
So we don't need to store zpci idx in msix message data to find out the
specific zpci device. Instead, we could use pci device id to find its
corresponding zpci device.

Signed-off-by: Yi Min Zhao <zyimin@linux.vnet.ibm.com>
Message-Id: <1504606380-49341-2-git-send-email-zyimin@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agotests: Enable the drive_del test also on s390x
Thomas Huth [Thu, 31 Aug 2017 14:40:08 +0000 (16:40 +0200)] 
tests: Enable the drive_del test also on s390x

We can use the drive_del test on s390x, too, to check that adding and
deleting also works fine with the virtio-ccw bus. But we have to make
sure that we use the devices with the "-ccw" suffix instead of the
"-pci" suffix for the virtio-ccw transport on s390x. Introduce a helper
function called qvirtio_get_dev_type() that returns the correct string
for the current architecture.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <1504190408-11143-1-git-send-email-thuth@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agos390x/css: fix cc handling for XSCH
Halil Pasic [Thu, 31 Aug 2017 12:18:28 +0000 (14:18 +0200)] 
s390x/css: fix cc handling for XSCH

The function ioinst_handle_xsch is presenting cc 2 when it's supposed to
present cc 1 and the other way around, because css_do_xsch has the error
codes mixed up. Because cc 1 has precedence over cc 2 we also have to
swap the two checks.

Let us fix this.

Signed-off-by: Halil Pasic <pasic@linux.vnet.ibm.com>
Reported-by: Pierre Morel <pmorel@linux.vnet.ibm.com>
Message-Id: <20170831121828.85885-1-pasic@linux.vnet.ibm.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
4 years agoMerge remote-tracking branch 'remotes/bonzini/tags/for-upstream' into staging
Peter Maydell [Tue, 19 Sep 2017 14:44:07 +0000 (15:44 +0100)] 
Merge remote-tracking branch 'remotes/bonzini/tags/for-upstream' into staging

* warning improvements (Alistair)
* KVM code cleanup (David)
* scsi-block support for rerror/werror (Fam)
* support for >64 vCPUs in Windows (Gonglei)
* SCSI fix (Hannes)
* SSE bugfixes (Joseph)
* SmartOS compilation fixes (Kamil)
* Hyper-V frequency MSR support (Ladi)
* move more files to accel/tcg (Philippe, Thomas)
* multiboot validation (PJP)
* virtqueue size configuration for virtio-scsi (Richard)
* Hyper-V header cleanup (Roman)
* Maintainer email update (Guangrong)
* checkpatch.pl --branch (Daniel), fixes (Greg)
* introducing scsi/ (me)

# gpg: Signature made Tue 19 Sep 2017 15:21:26 BST
# gpg:                using RSA key 0xBFFBD25F78C7AE83
# gpg: Good signature from "Paolo Bonzini <bonzini@gnu.org>"
# gpg:                 aka "Paolo Bonzini <pbonzini@redhat.com>"
# Primary key fingerprint: 46F5 9FBD 57D6 12E7 BFD4  E2F7 7E15 100C CD36 69B1
#      Subkey fingerprint: F133 3857 4B66 2389 866C  7682 BFFB D25F 78C7 AE83

* remotes/bonzini/tags/for-upstream: (51 commits)
  docker: fix creation of archives
  default-configs: Replace $(and ...) with $(call land, ...)
  osdep.h: Prohibit disabling assert() in supported builds
  checkpatch: add hwaddr to @typeList
  accel/hax: move hax-stub.c to accel/stubs/
  target/i386: fix "info mem" for LA57 mode
  scripts: let checkpatch.pl process an entire GIT branch
  update-linux-headers: prepare for hyperv.h removal
  hyperv: add header with protocol definitions
  i386/cpu/hyperv: support over 64 vcpus for windows guests
  Convert remaining single line fprintf() to warn_report()
  Makefile: Remove libqemustub.a
  ptimer-test: do not link to libqemustub.a/libqemuutil.a
  target/mips: Convert VM clock update prints to warn_report
  General warn report fixups
  Convert multi-line fprintf() to warn_report()
  Convert single line fprintf(.../n) to warn_report()
  Convert remaining error_report() to warn_report()
  hw/i386: Improve some of the warning messages
  test-qga: add missing qemu-ga tool dependency
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 years agodocker: fix creation of archives
Paolo Bonzini [Tue, 19 Sep 2017 12:08:31 +0000 (14:08 +0200)] 
docker: fix creation of archives

The pixman submodule does not exist anymore, and its removal broke
docker-based tests.  Fix it.

Cc: Fam Zheng <famz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agodefault-configs: Replace $(and ...) with $(call land, ...)
Thomas Huth [Mon, 18 Sep 2017 18:32:18 +0000 (20:32 +0200)] 
default-configs: Replace $(and ...) with $(call land, ...)

Using $(and ...) is dangerous here: It only works as long as the first
argument is set to 'y' or completely unset. It does not work if the
first argument is set to 'n' for example. Let's use the "land" make
function instead which has been written explicitely for this purpose.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <1505759538-15365-1-git-send-email-thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoosdep.h: Prohibit disabling assert() in supported builds
Eric Blake [Mon, 11 Sep 2017 21:13:20 +0000 (16:13 -0500)] 
osdep.h: Prohibit disabling assert() in supported builds

We already have several files that knowingly require assert()
to work, sometimes because refactoring the code for proper
error handling has not been tackled yet; there are probably
other files that have a similar situation but with no comments
documenting the same.  In fact, we have places in migration
that handle untrusted input with assertions, where disabling
the assertions risks a worse security hole than the current
behavior of losing the guest to SIGABRT when migration fails
because of the assertion.  Promote our current per-file
safety-valve to instead be project-wide, and expand it to also
cover glib's g_assert().

Note that we do NOT want to encourage 'assert(side-effects);'
(that is a bad practice that prevents copy-and-paste of code to
other projects that CAN disable assertions; plus it costs
unnecessary reviewer mental cycles to remember whether a project
special-cases the crippling of asserts); and we would LIKE to
fix migration to not rely on asserts (but that takes a big code
audit).  But in the meantime, we DO want to send a message
that anyone that disables assertions has to tweak code in order
to compile, making it obvious that they are taking on additional
risk that we are not going to support.  At the same time, leave
comments mentioning NDEBUG in files that we know still need to
be scrubbed, so there is at least something to grep for.

It would be possible to come up with some other mechanism for
doing runtime checking by default, but which does not abort
the program on failure, while leaving side effects in place
(unlike how crippling assert() avoids even the side effects),
perhaps under the name q_verify(); but it was not deemed worth
the effort (developers should not have to learn a replacement
when the standard C macro works just fine, and it would be a lot
of churn for little gain).  The patch specifically uses #error
rather than #warn so that a user is forced to tweak the header
to acknowledge the issue, even when not using a -Werror
compilation.

Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20170911211320.25385-1-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agocheckpatch: add hwaddr to @typeList
Greg Kurz [Thu, 14 Sep 2017 09:12:07 +0000 (11:12 +0200)] 
checkpatch: add hwaddr to @typeList

The script doesn't know about all possible types and learn them as
it parses the code. If it reaches a line with a type cast but the
type isn't known yet, it is misinterpreted as an identifier.

For example the following line:

    foo = (hwaddr) -1;

results in the following false-positive to be reported:

ERROR: spaces required around that '-' (ctx:VxV)

Let's add this standard QEMU type to the list of pre-known types.

Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <150538015789.8149.10902725348939486674.stgit@bahia.lan>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoaccel/hax: move hax-stub.c to accel/stubs/
Philippe Mathieu-Daudé [Wed, 13 Sep 2017 22:11:49 +0000 (19:11 -0300)] 
accel/hax: move hax-stub.c to accel/stubs/

Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20170913221149.30382-1-f4bug@amsat.org>
Reviewed-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
4 years agotarget/i386: fix "info mem" for LA57 mode
Paolo Bonzini [Tue, 12 Sep 2017 10:24:03 +0000 (12:24 +0200)] 
target/i386: fix "info mem" for LA57 mode

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoscripts: let checkpatch.pl process an entire GIT branch
Daniel P. Berrange [Wed, 13 Sep 2017 09:10:00 +0000 (10:10 +0100)] 
scripts: let checkpatch.pl process an entire GIT branch

Currently before submitting a series, devs should run checkpatch.pl
across each patch to be submitted. This can be automated using a
command such as:

  git rebase -i master -x 'git show | ./scripts/checkpatch.pl -'

This is rather long winded to type, so this patch introduces a way
to tell checkpatch.pl to validate a series of GIT revisions.

There are now three modes it can operate in 1) check a patch 2) check a source
file, or 3) check a git branch.

If no flags are given, the mode is determined by checking the args passed to
the command. If the args contain a literal ".." it is treated as a GIT revision
list. If the args end in ".patch" or equal "-" it is treated as a patch file.
Otherwise it is treated as a source file.

This automatic guessing can be overridden using --[no-]patch --[no-]file or
--[no-]branch

For example to check a GIT revision list:

    $ ./scripts/checkpatch.pl master..
    total: 0 errors, 0 warnings, 297 lines checked

    b886d352a2bf58f0996471fb3991a138373a2957 has no obvious style problems and is ready for submission.
    total: 0 errors, 0 warnings, 182 lines checked

    2a731f9a9ce145e0e0df6d42dd2a3ce4dfc543fa has no obvious style problems and is ready for submission.
    total: 0 errors, 0 warnings, 102 lines checked

    11844169bcc0c8ed4449eb3744a69877ed329dd7 has no obvious style problems and is ready for submission.

If a genuine patch filename contains the characters '..' it is
possible to force interpretation of the arg as a patch

  $ ./scripts/checkpatch.pl --patch master..

will force it to load a patch file called "master..", or equivalently

  $ ./scripts/checkpatch.pl --no-branch master..

will simply turn off guessing of GIT revision lists.

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Message-Id: <20170913091000.9005-1-berrange@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoupdate-linux-headers: prepare for hyperv.h removal
Roman Kagan [Thu, 13 Jul 2017 20:15:22 +0000 (23:15 +0300)] 
update-linux-headers: prepare for hyperv.h removal

All definitions related to Hyper-V emulation are now taken from the QEMU
own header, so the one imported from the kernel is no longer needed.

Unfortunately it's included by kvm_para.h.

So, until this is fixed in the kernel, teach the header harvesting
script to substitute kernel's hyperv.h with a dummy.

Signed-off-by: Roman Kagan <rkagan@virtuozzo.com>
Message-Id: <20170713201522.13765-3-rkagan@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agohyperv: add header with protocol definitions
Roman Kagan [Thu, 13 Jul 2017 20:15:21 +0000 (23:15 +0300)] 
hyperv: add header with protocol definitions

The definitions for Hyper-V emulation are currently taken from a header
imported from the Linux kernel.

However, as these describe a third-party protocol rather than a kernel
API, it probably wasn't a good idea to publish it in the kernel uapi.

This patch introduces a header that provides all the necessary
definitions, superseding the one coming from the kernel.

The new header supports (temporary) coexistence with the kernel one.
The constants explicitly named in the Hyper-V specification (e.g. msr
numbers) are defined in a non-conflicting way.  Other constants and
types have got new names.

While at this, the protocol data structures are defined in a more
conventional way, without bitfields, enums, and excessive unions.

The code using this stuff is adjusted, too; it can now be built both
with and without the kernel header in the tree.

Signed-off-by: Roman Kagan <rkagan@virtuozzo.com>
Message-Id: <20170713201522.13765-2-rkagan@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoi386/cpu/hyperv: support over 64 vcpus for windows guests
Gonglei [Mon, 11 Sep 2017 15:20:27 +0000 (23:20 +0800)] 
i386/cpu/hyperv: support over 64 vcpus for windows guests

Starting with Windows Server 2012 and Windows 8, if
CPUID.40000005.EAX contains a value of -1, Windows assumes specific
limit to the number of VPs. In this case, Windows Server 2012
guest VMs may use more than 64 VPs, up to the maximum supported
number of processors applicable to the specific Windows
version being used.

https://docs.microsoft.com/en-us/virtualization/hyper-v-on-windows/reference/tlfs

For compatibility, Let's introduce a new property for X86CPU,
named "x-hv-max-vps" as Eduardo's suggestion, and set it
to 0x40 before machine 2.10.

(The "x-" prefix indicates that the property is not supposed to
be a stable user interface.)

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Message-Id: <1505143227-14324-1-git-send-email-arei.gonglei@huawei.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoConvert remaining single line fprintf() to warn_report()
Alistair Francis [Fri, 18 Aug 2017 18:47:35 +0000 (11:47 -0700)] 
Convert remaining single line fprintf() to warn_report()

Convert any remaining uses of fprintf(stderr, "warning:"...
to use warn_report() instead. This helps standardise on a single
method of printing warnings to the user.

All of the warnings were changed using this command:
  find ./* -type f -exec sed -i 's|fprintf(.*".*warning[,:] |warn_report("|Ig' {} +

The #include lines and chagnes to the test Makefile were manually
updated to allow the code to compile.

Signed-off-by: Alistair Francis <alistair.francis@xilinx.com>
Message-Id: <2c94ac3bb116cc6b8ebbcd66a254920a69665515.1503077821.git.alistair.francis@xilinx.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoMakefile: Remove libqemustub.a
Paolo Bonzini [Tue, 19 Sep 2017 14:20:31 +0000 (16:20 +0200)] 
Makefile: Remove libqemustub.a

Using two libraries (libqemuutil.a and libqemustub.a) would sometimes
result in circular dependencies. To avoid these issues let's just
combine both into a single library that functions as both.

Signed-off-by: Alistair Francis <alistair.francis@xilinx.com>
Message-Id: <54e6458745493d10901964624479a7d9a872f481.1503077821.git.alistair.francis@xilinx.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoptimer-test: do not link to libqemustub.a/libqemuutil.a
Paolo Bonzini [Tue, 19 Sep 2017 14:18:42 +0000 (16:18 +0200)] 
ptimer-test: do not link to libqemustub.a/libqemuutil.a

This test provides its own mocks, so do not use the "standard"
stubs in libqemustub.a or the event loop implementation in
libqemuutil.a.

This is required on OS X, which otherwise brings in qemu-timer.o,
async.o and main-loop.o from libqemuutil.a.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agotarget/mips: Convert VM clock update prints to warn_report
Alistair Francis [Mon, 11 Sep 2017 19:52:59 +0000 (12:52 -0700)] 
target/mips: Convert VM clock update prints to warn_report

Convert the fprintf() messages in kvm_mips_update_state() to use
warn_report() as they aren't errors, but are just warnings.

Signed-off-by: Alistair Francis <alistair.francis@xilinx.com>
Cc: James Hogan <james.hogan@imgtec.com>
Message-Id: <e6acff8db6d264f913a18c86858b9aa600554e51.1505158760.git.alistair.francis@xilinx.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoGeneral warn report fixups
Alistair Francis [Mon, 11 Sep 2017 19:52:56 +0000 (12:52 -0700)] 
General warn report fixups

Tidy up some of the warn_report() messages after having converted them
to use warn_report().

Signed-off-by: Alistair Francis <alistair.francis@xilinx.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <9cb1d23551898c9c9a5f84da6773e99871285120.1505158760.git.alistair.francis@xilinx.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoConvert multi-line fprintf() to warn_report()
Alistair Francis [Mon, 11 Sep 2017 19:52:53 +0000 (12:52 -0700)] 
Convert multi-line fprintf() to warn_report()

Convert all the multi-line uses of fprintf(stderr, "warning:"..."\n"...
to use warn_report() instead. This helps standardise on a single
method of printing warnings to the user.

All of the warnings were changed using these commands:
  find ./* -type f -exec sed -i \
    'N; {s|fprintf(.*".*warning[,:] \(.*\)\\n"\(.*\));|warn_report("\1"\2);|Ig}' \
    {} +
  find ./* -type f -exec sed -i \
    'N;N; {s|fprintf(.*".*warning[,:] \(.*\)\\n"\(.*\));|warn_report("\1"\2);|Ig}' \
    {} +
  find ./* -type f -exec sed -i \
    'N;N;N; {s|fprintf(.*".*warning[,:] \(.*\)\\n"\(.*\));|warn_report("\1"\2);|Ig}' \
    {} +
  find ./* -type f -exec sed -i \
    'N;N;N;N {s|fprintf(.*".*warning[,:] \(.*\)\\n"\(.*\));|warn_report("\1"\2);|Ig}' \
    {} +
  find ./* -type f -exec sed -i \
    'N;N;N;N;N {s|fprintf(.*".*warning[,:] \(.*\)\\n"\(.*\));|warn_report("\1"\2);|Ig}' \
    {} +
  find ./* -type f -exec sed -i \
    'N;N;N;N;N;N {s|fprintf(.*".*warning[,:] \(.*\)\\n"\(.*\));|warn_report("\1"\2);|Ig}' \
    {} +
  find ./* -type f -exec sed -i \
    'N;N;N;N;N;N;N; {s|fprintf(.*".*warning[,:] \(.*\)\\n"\(.*\));|warn_report("\1"\2);|Ig}' \
    {} +

Indentation fixed up manually afterwards.

Some of the lines were manually edited to reduce the line length to below
80 charecters. Some of the lines with newlines in the middle of the
string were also manually edit to avoid checkpatch errrors.

The #include lines were manually updated to allow the code to compile.

Several of the warning messages can be improved after this patch, to
keep this patch mechanical this has been moved into a later patch.

Signed-off-by: Alistair Francis <alistair.francis@xilinx.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Kevin Wolf <kwolf@redhat.com>
Cc: Max Reitz <mreitz@redhat.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Peter Maydell <peter.maydell@linaro.org>
Cc: Stefano Stabellini <sstabellini@kernel.org>
Cc: Anthony Perard <anthony.perard@citrix.com>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Eduardo Habkost <ehabkost@redhat.com>
Cc: Aurelien Jarno <aurelien@aurel32.net>
Cc: Yongbok Kim <yongbok.kim@imgtec.com>
Cc: Cornelia Huck <cohuck@redhat.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Alexander Graf <agraf@suse.de>
Cc: Jason Wang <jasowang@redhat.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Acked-by: Cornelia Huck <cohuck@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <5def63849ca8f551630c6f2b45bcb1c482f765a6.1505158760.git.alistair.francis@xilinx.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoConvert single line fprintf(.../n) to warn_report()
Alistair Francis [Mon, 11 Sep 2017 19:52:50 +0000 (12:52 -0700)] 
Convert single line fprintf(.../n) to warn_report()

Convert all the single line uses of fprintf(stderr, "warning:"..."\n"...
to use warn_report() instead. This helps standardise on a single
method of printing warnings to the user.

All of the warnings were changed using this command:
  find ./* -type f -exec sed -i \
    's|fprintf(.*".*warning[,:] \(.*\)\\n"\(.*\));|warn_report("\1"\2);|Ig' \
    {} +

Some of the lines were manually edited to reduce the line length to below
80 charecters.

The #include lines were manually updated to allow the code to compile.

Signed-off-by: Alistair Francis <alistair.francis@xilinx.com>
Cc: Kevin Wolf <kwolf@redhat.com>
Cc: Max Reitz <mreitz@redhat.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Eduardo Habkost <ehabkost@redhat.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: Michael Roth <mdroth@linux.vnet.ibm.com>
Cc: James Hogan <james.hogan@imgtec.com>
Cc: Aurelien Jarno <aurelien@aurel32.net>
Cc: Yongbok Kim <yongbok.kim@imgtec.com>
Cc: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: James Hogan <james.hogan@imgtec.com> [mips]
Message-Id: <ae8f8a7f0a88ded61743dff2adade21f8122a9e7.1505158760.git.alistair.francis@xilinx.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoConvert remaining error_report() to warn_report()
Alistair Francis [Mon, 11 Sep 2017 19:52:46 +0000 (12:52 -0700)] 
Convert remaining error_report() to warn_report()

In a previous patch (3dc6f8693694a649a9c83f1e2746565b47683923) we
converted uses of error_report("warning:"... to use warn_report()
instead. This was to help standardise on a single method of printing
warnings to the user.

There appears to have been some cases that slipped through in patch sets
applied around the same time, this patch catches the few remaining
cases.

All of the warnings were changed using this command:
  find ./* -type f -exec sed -i \
    's|error_report(".*warning[,:] |warn_report("|Ig' {} +

Indentation fixed up manually afterwards.

Two messages were manually fixed up as well.

Signed-off-by: Alistair Francis <alistair.francis@xilinx.com>
Cc: Kevin Wolf <kwolf@redhat.com>
Cc: Max Reitz <mreitz@redhat.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Cornelia Huck <cohuck@redhat.com>
Cc: Alexander Graf <agraf@suse.de>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Stefan Hajnoczi <stefanha@redhat.com>
Acked-by: Cornelia Huck <cohuck@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <eec8cba0d5434bd828639e5e45f12182490ff47d.1505158760.git.alistair.francis@xilinx.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agohw/i386: Improve some of the warning messages
Alistair Francis [Mon, 11 Sep 2017 19:52:43 +0000 (12:52 -0700)] 
hw/i386: Improve some of the warning messages

Signed-off-by: Alistair Francis <alistair.francis@xilinx.com>
Suggested-by: Eduardo Habkost <ehabkost@redhat.com>
Cc: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <1d6ef2ccd9667878ed5820fcf17eef35957ea5d8.1505158760.git.alistair.francis@xilinx.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agotest-qga: add missing qemu-ga tool dependency
Philippe Mathieu-Daudé [Mon, 11 Sep 2017 21:01:29 +0000 (18:01 -0300)] 
test-qga: add missing qemu-ga tool dependency

this fixes running 'make check-unit' without running 'make all' beforehand:

$ make check-unit
  ...
  GTESTER tests/test-qga
**
ERROR:tests/test-qga.c:73:fixture_setup: assertion failed (error == NULL): Failed to execute child process "/build/qemu/qemu-ga" (No such file or directory) (g-exec-error-quark, 8)
make: *** [check-tests/test-qga] Error 1

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20170911210129.5874-1-f4bug@amsat.org>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agokvm: kvm_log_sync() is only called with known memory sections
David Hildenbrand [Mon, 11 Sep 2017 17:49:33 +0000 (19:49 +0200)] 
kvm: kvm_log_sync() is only called with known memory sections

Flatview will make sure that we can only end up in this function with
memory sections that correspond to exactly one slot. So we don't
have to iterate multiple times. There won't be overlapping slots but
only matching slots.

Properly align the section and look up the corresponding slot. This
heavily simplifies this function.

We can now get rid of kvm_lookup_overlapping_slot().

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170911174933.20789-7-david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agokvm: kvm_log_start/stop are only called with known sections
David Hildenbrand [Mon, 11 Sep 2017 17:49:32 +0000 (19:49 +0200)] 
kvm: kvm_log_start/stop are only called with known sections

Let's properly align the sections first and bail out if we would ever
get called with a memory section we don't know yet.

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170911174933.20789-6-david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agokvm: we never have overlapping slots in kvm_set_phys_mem()
David Hildenbrand [Mon, 11 Sep 2017 17:49:31 +0000 (19:49 +0200)] 
kvm: we never have overlapping slots in kvm_set_phys_mem()

The way flatview handles memory sections, we will never have overlapping
memory sections in kvm.

address_space_update_topology_pass() will make sure that we will only
get called for

a) an existing memory section for which we only update parameters
(log_start, log_stop).
b) an existing memory section we want to delete (region_del)
c) a brand new memory section we want to add (region_add)

We cannot have overlapping memory sections in kvm as we will first remove
the overlapping sections and then add the ones without conflicts.

Therefore we can remove the complexity for handling prefix and suffix
slots.

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170911174933.20789-5-david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agokvm: use start + size for memory ranges
David Hildenbrand [Mon, 11 Sep 2017 17:49:30 +0000 (19:49 +0200)] 
kvm: use start + size for memory ranges

Convert kvm_lookup_matching_slot().

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170911174933.20789-4-david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agokvm: factor out alignment of memory section
David Hildenbrand [Mon, 11 Sep 2017 17:49:29 +0000 (19:49 +0200)] 
kvm: factor out alignment of memory section

Factor it out, so we can reuse it later.

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170911174933.20789-3-david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agokvm: require JOIN_MEMORY_REGIONS_WORKS
David Hildenbrand [Mon, 11 Sep 2017 17:49:28 +0000 (19:49 +0200)] 
kvm: require JOIN_MEMORY_REGIONS_WORKS

We already require DESTROY_MEMORY_REGION_WORKS, JOIN_MEMORY_REGIONS_WORKS
was added just half a year later.

In addition, with flatview overlapping memory regions are first
removed before adding the changed one. So we can't really detect joining
memory regions this way.

Let's just get rid of this special handling.

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170911174933.20789-2-david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agomultiboot: validate multiboot header address values
Prasad J Pandit [Thu, 7 Sep 2017 06:32:56 +0000 (12:02 +0530)] 
multiboot: validate multiboot header address values

While loading kernel via multiboot-v1 image, (flags & 0x00010000)
indicates that multiboot header contains valid addresses to load
the kernel image. These addresses are used to compute kernel
size and kernel text offset in the OS image. Validate these
address values to avoid an OOB access issue.

This is CVE-2017-14167.

Reported-by: Thomas Garnier <thgarnie@google.com>
Signed-off-by: Prasad J Pandit <pjp@fedoraproject.org>
Message-Id: <20170907063256.7418-1-ppandit@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoscsi/esp: Rename the ESP macro to ESP_STATE
Kamil Rytarowski [Sat, 9 Sep 2017 14:21:16 +0000 (16:21 +0200)] 
scsi/esp: Rename the ESP macro to ESP_STATE

SunOS defines ESP (x86 register) in <sys/regset.h> as 7.

This fixes build on SmartOS (Joyent).

Signed-off-by: Kamil Rytarowski <n54@gmx.com>
Message-Id: <20170909142116.26816-1-n54@gmx.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agomemory: Rename queue to mrqueue (memory region queue)
Kamil Rytarowski [Sun, 3 Sep 2017 16:33:04 +0000 (18:33 +0200)] 
memory: Rename queue to mrqueue (memory region queue)

SunOS declares struct queue in <netinet/in.h>.

This fixes build on SmartOS (Joyent).

Patch cherry-picked from pkgsrc by jperkin (Joyent).

Signed-off-by: Kamil Rytarowski <n54@gmx.com>
Message-Id: <20170903163304.17919-1-n54@gmx.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoMAINTAINERS: update email, add missing test entry for megasas
Philippe Mathieu-Daudé [Sun, 10 Sep 2017 17:15:57 +0000 (14:15 -0300)] 
MAINTAINERS: update email, add missing test entry for megasas

and update maintainer email address

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20170910171557.12689-1-f4bug@amsat.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoi386/kvm: advertise Hyper-V frequency MSRs
Ladi Prosek [Mon, 7 Aug 2017 08:57:03 +0000 (10:57 +0200)] 
i386/kvm: advertise Hyper-V frequency MSRs

As of kernel commit eb82feea59d6 ("KVM: hyperv: support HV_X64_MSR_TSC_FREQUENCY
and HV_X64_MSR_APIC_FREQUENCY"), KVM supports two new MSRs which are required
for nested Hyper-V to read timestamps with RDTSC + TSC page.

This commit makes QEMU advertise the MSRs with CPUID.40000003H:EAX[11] and
CPUID.40000003H:EDX[8] as specified in the Hyper-V TLFS and experimentally
verified on a Hyper-V host. The feature is enabled with the existing hv-time CPU
flag, and only if the TSC frequency is stable across migrations and known.

Signed-off-by: Ladi Prosek <lprosek@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170807085703.32267-5-lprosek@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoi386/kvm: introduce tsc_is_stable_and_known()
Ladi Prosek [Mon, 7 Aug 2017 08:57:02 +0000 (10:57 +0200)] 
i386/kvm: introduce tsc_is_stable_and_known()

Move the "is TSC stable and known" condition to a reusable helper.

Signed-off-by: Ladi Prosek <lprosek@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170807085703.32267-4-lprosek@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoi386/kvm: set tsc_khz before configuring Hyper-V CPUID
Ladi Prosek [Mon, 7 Aug 2017 08:57:01 +0000 (10:57 +0200)] 
i386/kvm: set tsc_khz before configuring Hyper-V CPUID

Timing-related Hyper-V enlightenments will benefit from knowing the final
tsc_khz value. This commit just moves the code in preparation for further
changes.

Signed-off-by: Ladi Prosek <lprosek@redhat.com>
Message-Id: <20170807085703.32267-3-lprosek@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoi386/kvm: use a switch statement for MSR detection
Ladi Prosek [Mon, 7 Aug 2017 08:57:00 +0000 (10:57 +0200)] 
i386/kvm: use a switch statement for MSR detection

Switch is easier on the eye and might lead to better codegen.

Signed-off-by: Ladi Prosek <lprosek@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Message-Id: <20170807085703.32267-2-lprosek@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoMAINTAINERS: update mail address for NVDIMM
Xiao Guangrong [Tue, 21 Mar 2017 05:33:57 +0000 (13:33 +0800)] 
MAINTAINERS: update mail address for NVDIMM

My Intel mail account will be disabled soon, update the mail info
to my private mail

Signed-off-by: Xiao Guangrong <xiaoguangrong.eric@gmail.com>
Message-Id: <1490074437-17059-1-git-send-email-guangrong.xiao@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoarm: drop intermediate cpu_model -> cpu type parsing and use cpu type directly
Igor Mammedov [Wed, 13 Sep 2017 16:04:57 +0000 (18:04 +0200)] 
arm: drop intermediate cpu_model -> cpu type parsing and use cpu type directly

there are 2 use cases to deal with:
  1: fixed CPU models per board/soc
  2: boards with user configurable cpu_model and fallback to
     default cpu_model if user hasn't specified one explicitly

For the 1st
  drop intermediate cpu_model parsing and use const cpu type
  directly, which replaces:
     typename = object_class_get_name(
           cpu_class_by_name(TYPE_ARM_CPU, cpu_model))
     object_new(typename)
  with
     object_new(FOO_CPU_TYPE_NAME)
  or
     cpu_generic_init(BASE_CPU_TYPE, "my cpu model")
  with
     cpu_create(FOO_CPU_TYPE_NAME)

as result 1st use case doesn't have to invoke not necessary
translation and not needed code is removed.

For the 2nd
 1: set default cpu type with MachineClass::default_cpu_type and
 2: use generic cpu_model parsing that done before machine_init()
    is run and:
    2.1: drop custom cpu_model parsing where pattern is:
       typename = object_class_get_name(
           cpu_class_by_name(TYPE_ARM_CPU, cpu_model))
       [parse_features(typename, cpu_model, &err) ]

    2.2: or replace cpu_generic_init() which does what
         2.1 does + create_cpu(typename) with just
         create_cpu(machine->cpu_type)
as result cpu_name -> cpu_type translation is done using
generic machine code one including parsing optional features
if supported/present (removes a bunch of duplicated cpu_model
parsing code) and default cpu type is defined in an uniform way
within machine_class_init callbacks instead of adhoc places
in boadr's machine_init code.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <1505318697-77161-6-git-send-email-imammedo@redhat.com>
Reviewed-by: Alistair Francis <alistair.francis@xilinx.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
4 years agopc: use generic cpu_model parsing
Igor Mammedov [Wed, 13 Sep 2017 16:04:56 +0000 (18:04 +0200)] 
pc: use generic cpu_model parsing

define default CPU type in generic way in pc_machine_class_init()
and let common machine code to handle cpu_model parsing

Patch also introduces TARGET_DEFAULT_CPU_TYPE define for 2 purposes:
  * make foo_machine_class_init() look uniform on every target
  * use define in [bsd|linux]-user targets to pick default
    cpu type

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <1505318697-77161-5-git-send-email-imammedo@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
4 years agovl.c: convert cpu_model to cpu type and set of global properties before machine_init()
Igor Mammedov [Wed, 13 Sep 2017 16:04:55 +0000 (18:04 +0200)] 
vl.c: convert cpu_model to cpu type and set of global properties before machine_init()

All machines that support user specified cpu_model either call
cpu_generic_init() or cpu_class_by_name()/CPUClass::parse_features
to parse feature string and to get CPU type to create.

Which leads to code duplication and hard-codding default CPU model
within machine_foo_init() code. Which makes it impossible to
get CPU type before machine_init() is run.

So instead of setting default CPUs models and doing parsing in
target specific machine_foo_init() in various ways, provide
a generic data driven cpu_model parsing before machine_init()
is called.

in follow up per target patches, it will allow to:
  * define default CPU type in consistent/generic manner
    per machine type and drop custom code that fallbacks
    to default if cpu_model is NULL
  * drop custom features parsing in targets and do it
    in centralized way.
  * for cases of
      cpu_generic_init(TYPE_BASE/DEFAULT_CPU, "some_cpu")
    replace it with
      cpu_create(machine->cpu_type) || cpu_create(TYPE_FOO)
    depending if CPU type is user settable or not.
    not doing useless parsing and clearly documenting where
    CPU model is user settable or fixed one.

Patch allows machine subclasses to define default CPU type
per machine class at class_init() time and if that is set
generic code will parse cpu_model into a MachineState::cpu_type
which will be used to create CPUs for that machine instance
and allows gradual per board conversion.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <1505318697-77161-4-git-send-email-imammedo@redhat.com>
Acked-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
4 years agocpu: make cpu_generic_init() abort QEMU on error
Igor Mammedov [Wed, 13 Sep 2017 16:04:54 +0000 (18:04 +0200)] 
cpu: make cpu_generic_init() abort QEMU on error

Almost every user of cpu_generic_init() checks for
returned NULL and then reports failure in a custom way
and aborts process.
Some users assume that call can't fail and don't check
for failure, though they should have checked for it.

In either cases cpu_generic_init() failure is fatal,
so instead of checking for failure and reporting
it various ways, make cpu_generic_init() report
errors in consistent way and terminate QEMU on failure.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <1505318697-77161-3-git-send-email-imammedo@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
4 years agoqom: cpus: split cpu_generic_init() on feature parsing and cpu creation parts
Igor Mammedov [Wed, 13 Sep 2017 16:04:53 +0000 (18:04 +0200)] 
qom: cpus: split cpu_generic_init() on feature parsing and cpu creation parts

it would allow to reuse feature parsing part in various machines
that have CPU features instead of re-implementing the same feature
parsing each time.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <1505318697-77161-2-git-send-email-imammedo@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
4 years agoscsi: move block/scsi.h to include/scsi/constants.h
Paolo Bonzini [Tue, 22 Aug 2017 07:23:55 +0000 (09:23 +0200)] 
scsi: move block/scsi.h to include/scsi/constants.h

Complete the transition by renaming this header, which was
shared by block/iscsi.c and the SCSI emulation code.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agohostmem-file: Add "discard-data" option
Eduardo Habkost [Thu, 24 Aug 2017 19:23:15 +0000 (16:23 -0300)] 
hostmem-file: Add "discard-data" option

The new option can be used to indicate that the file contents can
be destroyed and don't need to be flushed to disk when QEMU exits
or when the memory backend object is removed.

Internally, it will trigger a madvise(MADV_REMOVE) call when the
memory backend is removed.

Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <20170824192315.5897-4-ehabkost@redhat.com>
[ehabkost: fixup: improved documentation]
Reviewed-by: Daniel P. Berrange <berrange@redhat.com>
Tested-by: Zack Cornelius <zack.cornelius@kove.net>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
4 years agoosdep: Define QEMU_MADV_REMOVE
Eduardo Habkost [Thu, 24 Aug 2017 19:23:14 +0000 (16:23 -0300)] 
osdep: Define QEMU_MADV_REMOVE

Define QEMU_MADV_REMOVE, so we can use it with qemu_madvise().

Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <20170824192315.5897-3-ehabkost@redhat.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Tested-by: Zack Cornelius <zack.cornelius@kove.net>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
4 years agovl: Clean up user-creatable objects when exiting
Eduardo Habkost [Thu, 24 Aug 2017 19:23:13 +0000 (16:23 -0300)] 
vl: Clean up user-creatable objects when exiting

Delete all user-creatable objects in /objects when exiting QEMU, so they
can perform cleanup actions.

Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <20170824192315.5897-2-ehabkost@redhat.com>
Acked-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: Zack Cornelius <zack.cornelius@kove.net>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
4 years agoscsi: introduce sg_io_sense_from_errno
Paolo Bonzini [Tue, 22 Aug 2017 07:43:14 +0000 (09:43 +0200)] 
scsi: introduce sg_io_sense_from_errno

Move more knowledge of SG_IO out of hw/scsi/scsi-generic.c, for
reusability.

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoscsi: introduce scsi_build_sense
Paolo Bonzini [Tue, 22 Aug 2017 07:42:59 +0000 (09:42 +0200)] 
scsi: introduce scsi_build_sense

Move more knowledge of sense data format out of hw/scsi/scsi-bus.c
for reusability.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoscsi: move non-emulation specific code to scsi/
Paolo Bonzini [Tue, 22 Aug 2017 05:08:27 +0000 (07:08 +0200)] 
scsi: move non-emulation specific code to scsi/

util/scsi.c includes some SCSI code that is shared by block/iscsi.c and
hw/scsi, but the introduction of the persistent reservation helper
will add many more instances of this.  There is also include/block/scsi.h,
which actually is not part of the core block layer.

The persistent reservation manager will also need a home.  A scsi/
directory provides one for both the aforementioned shared code and
the PR manager code.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoscsi: rename scsi_build_sense to scsi_convert_sense
Paolo Bonzini [Tue, 22 Aug 2017 07:31:36 +0000 (09:31 +0200)] 
scsi: rename scsi_build_sense to scsi_convert_sense

After introducing the scsi/ subdirectory, there will be a scsi_build_sense
function that is the same as scsi_req_build_sense but without needing
a SCSIRequest.  The existing scsi_build_sense function gets in the way,
remove it.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoscsi-block: Support rerror/werror
Fam Zheng [Mon, 21 Aug 2017 14:10:08 +0000 (22:10 +0800)] 
scsi-block: Support rerror/werror

This makes the werror/rerror options available on the scsi-block device,
to allow user specify error handling policy similar to scsi-hd.

Signed-off-by: Fam Zheng <famz@redhat.com>
Message-Id: <20170821141008.19383-5-famz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 years agoscsi: Introduce scsi_sense_buf_to_errno
Fam Zheng [Mon, 21 Aug 2017 14:10:07 +0000 (22:10 +0800)] 
scsi: Introduce scsi_sense_buf_to_errno

This recognizes the "fixed" and "descriptor" format sense data, extracts
the sense key/asc/ascq fields then converts them to an errno.

Signed-off-by: Fam Zheng <famz@redhat.com>
Message-Id: <20170821141008.19383-4-famz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>