Don't create the s390 pci host bridge if we do not provide the zpci
facility.
Reviewed-by: Thomas Huth <thuth@redhat.com>
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
Only set the zpci feature bit on builds that actually support pci.
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
The nt2 event class is pci-only - don't look for events if pci is
not in the active cpu model.
Reviewed-by: Thomas Huth <thuth@redhat.com>
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
Some non-pci code calls into zpci code. Provide some stubs for builds
without pci.
Reviewed-by: Thomas Huth <thuth@redhat.com>
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Acked-by: Halil Pasic <pasic@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
The msi routing code in kvm calls some pci functions: provide
some stubs to enable builds without pci.
Also, to make this more obvious, guard them via a pci_available boolean
(which also can be reused in other places).
Fixes: e1d4fb2de ("kvm-irqchip: x86: add msi route notify fn")
Fixes: 767a554a0 ("kvm-all: Pass requester ID to MSI routing functions")
Reviewed-by: Pierre Morel <pmorel@linux.vnet.ibm.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
Nothing in fsdev/ or hw/9pfs/ depends on pci; it should rather depend
on CONFIG_VIRTFS and CONFIG_VIRTIO/CONFIG_XEN only.
Acked-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
KVM guests on s390 need a different page table layout than normal
processes (2kb page table + 2kb page status extensions vs 2kb page table
only). As of today this has to be enabled via the vm.allocate_pgste
sysctl.
Newer kernels (>= 4.12) on s390 check for an S390_PGSTE program header
and enable the pgste page table extensions in that case. This makes the
vm.allocate_pgste sysctl unnecessary. We enable this program header for
the s390 system emulation (qemu-system-s390x) if we build on s390
- for s390 system emulation
- the linker supports --s390-pgste (binutils >= 2.29)
- KVM is enabled
This will allow distributions to disable the global vm.allocate_pgste
sysctl, which will improve the page table allocation for non KVM
processes as only 2kb chunks are necessary.
Cc: Christian Ehrhardt <christian.ehrhardt@canonical.com>
Cc: Alexander Graf <agraf@suse.de>
Cc: Dan Horak <dhorak@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Acked-by: Janosch Frank <frankja@linux.vnet.ibm.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-Id: <1503483383-199649-1-git-send-email-borntraeger@de.ibm.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
QEMU currently aborts when the user tries to hot-unplug a diag288
device:
$ qemu-system-s390x -nographic -nodefaults -S -monitor stdio
QEMU 2.9.92 monitor - type 'help' for more information
(qemu) device_add diag288,id=x
(qemu) device_del x
**
ERROR:qemu/qdev-monitor.c:872:qdev_unplug: assertion failed: (hotplug_ctrl)
Aborted (core dumped)
The device is not designed as hot-pluggable (it should only be used
via the "-watchdog" parameter), so let's simply remove the possibility
to hotplug it to prevent that users can run into this ugly situation.
Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <1502892528-22618-1-git-send-email-thuth@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
While the PoP is silent on the issue, z/VM documentation states
that unknown diagnose codes trigger a specification exception.
We already do that when running with kvm, so change tcg to do so
as well.
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
With some small modifications, we can also use the the netfilter,
the filter-mirror and the filter-redirector tests on s390x.
Signed-off-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Message-Id: <1502951113-4246-3-git-send-email-thuth@redhat.com>
Reviewed-by: Zhang Chen <zhangchen.fnst@cn.fujitsu.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
This way we can get rid of the ugly #ifdefs in the code which makes
it easier to extend later.
Signed-off-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Message-Id: <1502951113-4246-2-git-send-email-thuth@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Zhang Chen <zhangchen.fnst@cn.fujitsu.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
Now that we've got a firmware that can do TFTP booting on s390x (i.e.
the pc-bios/s390-netboot.img), we can enable the PXE tester for this
architecture, too.
Signed-off-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Message-Id: <1502431076-22849-3-git-send-email-thuth@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
Re-using the boot_sector code buffer from x86 for other architectures
is not very nice, especially if we add more architectures later. It's
also ugly that the test uses a huge pre-initialized array at all - the
size of the executable is very huge due to this array. So let's use a
separate buffer for each architecture instead, allocated from the heap,
so that we really just use the memory that we need.
Suggested-by: Michael Tsirkin <mst@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Message-Id: <1502431076-22849-2-git-send-email-thuth@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
The s390-ipl device can not be created by the user, since it is meant only
to be instantiated once internally to load the ROMs and kernel. If the user
tries to do a "device_add s390-ipl" via the monitor later, QEMU aborts with
a "ROM images must be loaded at startup" error message.
Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <1502861458-30270-1-git-send-email-thuth@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
A successful completion of rchp should signal a solicited channel path
initialized CRW (channel report word), while the current implementation
always generates an un-solicited one. Let's fix this.
Reported-by: Halil Pasic <pasic@linux.vnet.ibm.com>
Signed-off-by: Dong Jia Shi <bjsdjshi@linux.vnet.ibm.com>
Reviewed-by: Halil Pasic <pasic@linux.vnet.ibm.com>
Message-Id: <20170803003527.86979-3-bjsdjshi@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
Let's use a macro for the ERC (error recover code) when generating a
Channel Subsystem Event-information pending CRW (channel report word).
While we are at it, let's also add all other ERCs.
Signed-off-by: Dong Jia Shi <bjsdjshi@linux.vnet.ibm.com>
Reviewed-by: Halil Pasic <pasic@linux.vnet.ibm.com>
Message-Id: <20170803003527.86979-2-bjsdjshi@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
Add the scripts/ directory to sys.path so Python 2.6 will be able to
import argparse.
Cc: Daniel P. Berrange <berrange@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Acked-by: John Snow <jsnow@redhat.com>
Acked-by: Fam Zheng <famz@redhat.com>
Message-id: 20170825155732.15665-4-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Add the scripts/ directory to sys.path so Python 2.6 will be able to
import argparse.
Cc: Fam Zheng <famz@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Acked-by: John Snow <jsnow@redhat.com>
Acked-by: Fam Zheng <famz@redhat.com>
Message-id: 20170825155732.15665-3-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The minimum Python version supported by QEMU is 2.6. The argparse
standard library module was only added in Python 2.7. Many scripts
would like to use argparse because it supports command-line
sub-commands.
This patch adds argparse. See the top of argparse.py for details.
Suggested-by: Daniel P. Berrange <berrange@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Acked-by: John Snow <jsnow@redhat.com>
Message-id: 20170825155732.15665-2-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
There's a few cases which we're passing an Error pointer to a function
only to discard it immediately afterwards without checking it. In
these cases we can simply remove the variable and pass NULL instead.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20170829120836.16091-1-berto@igalia.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
If QEMU is running on a system that's out of memory and mmap()
fails, QEMU aborts with no error message at all, making it hard
to debug the reason for the failure.
Add perror() calls that will print error information before
aborting.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20170829212053.6003-1-ehabkost@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
LeakyBucket.burst_length is defined as an unsigned integer but the
code never checks for overflows and it only makes sure that the value
is not 0.
In practice this means that the user can set something like
throttling.iops-total-max-length=4294967300 despite being larger than
UINT_MAX and the final value after casting to unsigned int will be 4.
This patch changes the data type to uint64_t. This does not increase
the storage size of LeakyBucket, and allows us to assign the value
directly from qemu_opt_get_number() or BlockIOThrottle and then do the
checks directly in throttle_is_valid().
The value of burst_length does not have a specific upper limit,
but since the bucket size is defined by max * burst_length we have
to prevent overflows. Instead of going for UINT64_MAX or something
similar this patch reuses THROTTLE_VALUE_MAX, which allows I/O bursts
of 1 GiB/s for 10 days in a row.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-id: 1b2e3049803f71cafb2e1fa1be4fb47147a0d398.1503580370.git.berto@igalia.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Both the throttling limits set with the throttling.iops-* and
throttling.bps-* options and their QMP equivalents defined in the
BlockIOThrottle struct are integer values.
Those limits are also reported in the BlockDeviceInfo struct and they
are integers there as well.
Therefore there's no reason to store them internally as double and do
the conversion everytime we're setting or querying them, so this patch
uses uint64_t for those types. Let's also use an unsigned type because
we don't allow negative values anyway.
LeakyBucket.level and LeakyBucket.burst_level do however remain double
because their value changes depending on the fraction of time elapsed
since the previous I/O operation.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-id: f29b840422767b5be2c41c2dfdbbbf6c5f8fedf8.1503580370.git.berto@igalia.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The throttling code can change internally the value of bkt->max if it
hasn't been set by the user. The problem with this is that if we want
to retrieve the original value we have to undo this change first. This
is ugly and unnecessary: this patch removes the throttle_fix_bucket()
and throttle_unfix_bucket() functions completely and moves the logic
to throttle_compute_wait().
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Manos Pitsidianakis <el13635@mail.ntua.gr>
Message-id: 5b0b9e1ac6eb208d709eddc7b09e7669a523bff3.1503580370.git.berto@igalia.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Use a pointer to the bucket instead of repeating cfg->buckets[i] all
the time. This makes the code more concise and will help us expand the
checks later and save a few line breaks.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-id: 763ffc40a26b17d54cf93f5a999e4656049fcf0c.1503580370.git.berto@igalia.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The way the throttling algorithm works is that requests start being
throttled once the bucket level exceeds the burst limit. When we get
there the bucket leaks at the level set by the user (bkt->avg), and
that leak rate is what prevents guest I/O from exceeding the desired
limit.
If we don't allow bursts (i.e. bkt->max == 0) then we can start
throttling requests immediately. The problem with keeping the
threshold at 0 is that it only allows one request at a time, and as
soon as there's a bit of I/O from the guest every other request will
be throttled and performance will suffer considerably. That can even
make the guest unable to reach the throttle limit if that limit is
high enough, and that happens regardless of the block scheduler used
by the guest.
Increasing that threshold gives flexibility to the guest, allowing it
to perform short bursts of I/O before being throttled. Increasing the
threshold too much does not make a difference in the long run (because
it's the leak rate what defines the actual throughput) but it does
allow the guest to perform longer initial bursts and exceed the
throttle limit for a short while.
A burst value of bkt->avg / 10 allows the guest to perform 100ms'
worth of I/O at the target rate without being throttled.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-id: 31aae6645f0d1fbf3860fb2b528b757236f0c0a7.1503580370.git.berto@igalia.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The level of the burst bucket is stored in bkt.burst_level, not
bkt.burst_length.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Manos Pitsidianakis <el13635@mail.ntua.gr>
Message-id: 49aab2711d02f285567f3b3b13a113847af33812.1503580370.git.berto@igalia.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The number of queues that should be return by the admin command should:
1) Only mention the number of non-admin queues.
2) It is zero-based, meaning that '0 == one non-admin queue',
'1 == two non-admin queues', and so forth.
Because our `num_queues` means the number of queues _plus_ the admin
queue, then the right calculation for the number returned from the admin
command is `num_queues - 2`, combining the two requirements mentioned.
The issue was discovered by reducing num_queues from 64 to 8 and running
a Linux VM with an SMP parameter larger than that (e.g. 22). It tries to
utilize all queues, and therefore fails with an invalid queue number
when trying to queue I/Os on the last queue.
Signed-off-by: Dan Aloni <dan@kernelim.com>
CC: Alex Friedman <alex@e8storage.com>
CC: Keith Busch <keith.busch@intel.com>
CC: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Keith Busch <keith.busch@intel.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The following scenario leads to an assertion failure in
qio_channel_yield():
1. Request coroutine calls qio_channel_yield() successfully when sending
would block on the socket. It is now yielded.
2. nbd_read_reply_entry() calls nbd_recv_coroutines_enter_all() because
nbd_receive_reply() failed.
3. Request coroutine is entered and returns from qio_channel_yield().
Note that the socket fd handler has not fired yet so
ioc->write_coroutine is still set.
4. Request coroutine attempts to send the request body with nbd_rwv()
but the socket would still block. qio_channel_yield() is called
again and assert(!ioc->write_coroutine) is hit.
The problem is that nbd_read_reply_entry() does not distinguish between
request coroutines that are waiting to receive a reply and those that
are not.
This patch adds a per-request bool receiving flag so
nbd_read_reply_entry() can avoid spurious aio_wake() calls.
Reported-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20170822125113.5025-1-stefanha@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Tested-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
Non-shared storage migration with NBD and drive-mirror is currently not
tested by qemu-iotests. This test case covers the basic migration
scenario.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Based-on: <20170823134242.12080-1-famz@redhat.com>
Message-Id: <20170823140506.28723-1-stefanha@redhat.com>
Tested-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
In the ->inactivate() callbacks, permissions are updated, which
typically involves a recursive check of the whole graph. Setting
BDRV_O_INACTIVE right before doing that creates a state that
bdrv_is_writable() returns false, which causes permission update
failure.
Reorder them so the flag is updated after calling the function. Note
that this doesn't break the assert in bdrv_child_cb_inactivate() because
for any specific BDS, we still update its flags first before calling
->inactivate() on it one level deeper in the recursion.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>
Message-Id: <20170823134242.12080-5-famz@redhat.com>
Tested-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
These two conditions corresponds to mirror job's source and target,
which need to be allowed as they are part of the non-shared storage
migration workflow: failing to inactivate either will result in a
failure during migration completion.
Signed-off-by: Fam Zheng <famz@redhat.com>
Message-Id: <20170823134242.12080-3-famz@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
[eblake: improve comment grammar]
Signed-off-by: Eric Blake <eblake@redhat.com>
The logic will be fixed (extended), move it to a separate function.
Signed-off-by: Fam Zheng <famz@redhat.com>
Message-Id: <20170823134242.12080-2-famz@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
The 'm->numa_auto_assign_ram = numa_legacy_auto_assign_ram;' line
was supposed to be in pc_i440fx_2_9_machine_options() (see commit
3bfe5716 "numa: equally distribute memory on nodes"), but the
merge commit adb354dd ("Merge remote-tracking branch
'mst/tags/for_upstream' into staging") moved it to the
pc_i440fx_2_10_machine_options().
Move the line back to pc_i440fx_2_9_machine_options().
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Message-id: 20170818190943.23858-1-ehabkost@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
travis builds fail at HEAD at rc3 master with
block/nbd-client.c: In function ‘nbd_read_reply_entry’:
block/nbd-client.c:110:8: error: ‘ret’ may be used uninitialized in this function [-Werror=uninitialized]
fix it by initializing 'ret' to 0
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
This is identical to the pull request from yesterday (20180822),
except that a bug in one patch is fixed so that it doesn't break TCG
on a ppc host.
Last minute ppc related fixes for qemu-2.10. I'm not sure if these
are critical enough to prompt another rc, but I'm submitting them for
consideration.
First, is Cornelia's fix for 480bc11e6 which meant "make check" would
always fail on a ppc host. Tracking that down delayed submission of
the rest of these patches, sorry.
The rest are all fairly important bugfixes for qemu crashes or guest
behaviour regression on ppc. Patches 2-4 specifically are fixes for
regressions from qemu-2.9, caused by the compatibility mode and
hotplug handling cleanups for the pseries machine type.
-----BEGIN PGP SIGNATURE-----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=xE8H
-----END PGP SIGNATURE-----
Merge remote-tracking branch 'remotes/dgibson/tags/ppc-for-2.10-20170823' into staging
ppc patch queue 2017-08-23
This is identical to the pull request from yesterday (20180822),
except that a bug in one patch is fixed so that it doesn't break TCG
on a ppc host.
Last minute ppc related fixes for qemu-2.10. I'm not sure if these
are critical enough to prompt another rc, but I'm submitting them for
consideration.
First, is Cornelia's fix for 480bc11e6 which meant "make check" would
always fail on a ppc host. Tracking that down delayed submission of
the rest of these patches, sorry.
The rest are all fairly important bugfixes for qemu crashes or guest
behaviour regression on ppc. Patches 2-4 specifically are fixes for
regressions from qemu-2.9, caused by the compatibility mode and
hotplug handling cleanups for the pseries machine type.
# gpg: Signature made Wed 23 Aug 2017 01:31:47 BST
# gpg: using RSA key 0x6C38CACA20D9B392
# gpg: Good signature from "David Gibson <david@gibson.dropbear.id.au>"
# gpg: aka "David Gibson (Red Hat) <dgibson@redhat.com>"
# gpg: aka "David Gibson (ozlabs.org) <dgibson@ozlabs.org>"
# gpg: aka "David Gibson (kernel.org) <dwg@kernel.org>"
# Primary key fingerprint: 75F4 6586 AE61 A66C C44E 87DC 6C38 CACA 20D9 B392
* remotes/dgibson/tags/ppc-for-2.10-20170823:
hw/ppc/spapr_iommu: Fix crash when removing the "spapr-tce-table" device
hw/ppc/spapr_rtc: Mark the RTC device with user_creatable = false
hw/ppc/spapr: Fix segfault when instantiating a 'pc-dimm' without 'memdev'
spapr: Allow configure-connector to be called multiple times
ppc: fix ppc_set_compat() with KVM PR
target/ppc: 'PVR != host PVR' in KVM_SET_SREGS workaround
boot-serial-test: prefer tcg accelerator
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
QEMU currently aborts unexpectedly when the user tries to add and
remove a "spapr-tce-table" device:
$ qemu-system-ppc64 -nographic -S -nodefaults -monitor stdio
QEMU 2.9.92 monitor - type 'help' for more information
(qemu) device_add spapr-tce-table,id=x
(qemu) device_del x
**
ERROR:qemu/qdev-monitor.c:872:qdev_unplug: assertion failed: (hotplug_ctrl)
Aborted (core dumped)
The device should not be accessable for the users at all, it's just
used internally, so mark it with user_creatable = false.
Signed-off-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
QEMU currently aborts unexpectedly when a user tries to do something
like this:
$ qemu-system-ppc64 -nographic -S -nodefaults -monitor stdio
QEMU 2.9.92 monitor - type 'help' for more information
(qemu) device_add spapr-rtc,id=spapr-rtc
(qemu) device_del spapr-rtc
**
ERROR:qemu/qdev-monitor.c:872:qdev_unplug: assertion failed: (hotplug_ctrl)
Aborted (core dumped)
The RTC device is not meant to be hot-pluggable - it's an internal
device only and it even should not be possible to create it a
second time with the "-device" parameter, so let's mark this
with "user_creatable = false".
Signed-off-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
QEMU currently crashes when trying to use a 'pc-dimm' on the pseries
machine without specifying its 'memdev' property. This happens because
pc_dimm_get_memory_region() does not check whether the 'memdev' property
has properly been set by the user. Looking closer at this function, it's
also obvious that it is using &error_abort to call another function - and
this is bad in a function that is used in the hot-plugging calling chain
since this can also cause QEMU to exit unexpectedly.
So let's fix these issues in a proper way now: Add a "Error **errp"
parameter to pc_dimm_get_memory_region() which we use in case the 'memdev'
property has not been set by the user, and which we can use instead of
the &error_abort, and change the callers of get_memory_region() to make
use of this "errp" parameter for proper error checking.
Signed-off-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
In case of in-kernel memory hot unplug, when the guest is not able
to remove all the LMBs that are requested for removal, it will add back
any LMBs that have been successfully removed. The DR Connectors of
these LMBs wouldn't have been unconfigured and hence the addition of
these LMBs will result in configure-connector call being issued on
LMB DR connectors that are already in configured state. Such
configure-connector calls will fail resulting in a DIMM which is
partially unplugged.
This however worked till recently before we overhauled the DRC
implementation in QEMU. Commit 9d4c0f4f0a71e: "spapr: Consolidate
DRC state variables" is the first commit where this problem shows up
as per git bisect.
Ideally guest shouldn't be issuing configure-connector call on an
already configured DR connector. However for now, work around this in
QEMU by allowing configure-connector to be called multiple times for
all types of DR connectors.
Signed-off-by: Bharata B Rao <bharata@linux.vnet.ibm.com>
[dwg: Corrected buglet that would have initialized fdt pointers ready
for reading on a device not present at reset]
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
When running in KVM PR mode, kvmppc_set_compat() always fail because the
current PR implementation doesn't handle KVM_REG_PPC_ARCH_COMPAT. Now that
the machine code inconditionally calls ppc_set_compat_all() at reset time
to restore the compat mode default value (commit 66d5c492dd), it is
impossible to start a guest with PR:
qemu-system-ppc64: Unable to set CPU compatibility mode in KVM:
Invalid argument
A tentative patch [1] was recently sent by Suraj to address the issue, but
it would prevent the compat mode to be turned off on reset. And we really
don't want to explicitely check for KVM PR. During the patch's review,
David suggested that we should only call the KVM ioctl() if the compat
PVR changes. This allows at least to run with KVM PR, provided no compat
mode is requested from the command line (which should be the case when
running PR nested). This is what this patch does.
While here, we also fix the side effect where KVM would fail but we would
change the CPU state in QEMU anyway.
[1] http://patchwork.ozlabs.org/patch/782039/
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Commit d5fc133eed ("ppc: Rework CPU compatibility testing
across migration") changed the way cpu_post_load behaves with
the PVR setting, causing an unexpected bug in KVM-HV migrations
between hosts that are compatible (POWER8 and POWER8E, for example).
Even with pvr_match() returning true, the guest freezes right after
cpu_post_load. The reason is that the guest kernel can't handle a
different PVR value other that the running host in KVM_SET_SREGS.
In [1] it was discussed the possibility of a new KVM capability
that would indicate that the guest kernel can handle a different
PVR in KVM_SET_SREGS. Even if such feature is implemented, there is
still the problem with older kernels that will not have this capability
and will fail to migrate.
This patch implements a workaround for that scenario. If running
with KVM, check if the guest kernel does not have the capability
(named here as 'cap_ppc_pvr_compat'). If it doesn't, calls
kvmppc_is_pr() to see if the guest is running in KVM-HV. If all this
happens, set env->spr[SPR_PVR] to the same value as the current
host PVR. This ensures that we allow migrations with 'close enough'
PVRs to still work in KVM-HV but also makes the code ready for
this new KVM capability when it is done.
A new function called 'kvmppc_pvr_workaround_required' was created
to encapsulate the conditions said above and to avoid calling too
many kvm.c internals inside cpu_post_load.
[1] https://lists.gnu.org/archive/html/qemu-ppc/2017-06/msg00503.html
Signed-off-by: Daniel Henrique Barboza <danielhb@linux.vnet.ibm.com>
[dwg: Fix for the case of using TCG on a PPC host]
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>