The changes to run_on_cpu and friends mean that all helpers are passed
the CPUState of vCPU they are running on. The conversion missed the
field in commit e0eeb4a21a which
introduced bugs.
Reported-by: Claudio Imbrenda <imbrenda@linux.vnet.ibm.com>
Tested-by: Claudio Imbrenda <imbrenda@linux.vnet.ibm.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20161010154625.14881-1-alex.bennee@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Introduce global kvm_msi_use_devid flag plus associated
kvm_msi_devid_required() macro. Passes the device ID,
if needed, while building the MSI route entry. Device IDs are
required by the ARM GICv3 ITS (IRQ remapping function is based on
this information).
Signed-off-by: Pavel Fedin <p.fedin@samsung.com>
Signed-off-by: Eric Auger <eric.auger@redhat.com>
Message-id: 1474616617-366-5-git-send-email-eric.auger@redhat.com
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
CPUState is a fairly common pointer to pass to these helpers. This means
if you need other arguments for the async_run_on_cpu case you end up
having to do a g_malloc to stuff additional data into the routine. For
the current users this isn't a massive deal but for MTTCG this gets
cumbersome when the only other parameter is often an address.
This adds the typedef run_on_cpu_func for helper functions which has an
explicit CPUState * passed as the first parameter. All the users of
run_on_cpu and async_run_on_cpu have had their helpers updated to use
CPUState where available.
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
[Sergey Fedorov:
- eliminate more CPUState in user data;
- remove unnecessary user data passing;
- fix target-s390x/kvm.c and target-s390x/misc_helper.c]
Signed-off-by: Sergey Fedorov <sergey.fedorov@linaro.org>
Acked-by: David Gibson <david@gibson.dropbear.id.au> (ppc parts)
Reviewed-by: Christian Borntraeger <borntraeger@de.ibm.com> (s390 parts)
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <1470158864-17651-3-git-send-email-alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
kvm_setup_guest_memory only does "madvise to QEMU_MADV_DONTFORK" and
is only called by ram_block_add, which actually is duplicate code.
Bonus: add simple comment for kvm_has_sync_mmu to make life easier.
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Cao jin <caoj.fnst@cn.fujitsu.com>
Message-Id: <1473662096-32598-1-git-send-email-caoj.fnst@cn.fujitsu.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <1470224274-31522-5-git-send-email-armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reported from Alexey Kardashevskiy:
3f1fea0fb5 "kvm-irqchip: do explicit commit when update irq" produces
a crash on pseries guest running with VFIO on POWER8 machine as it does
not support KVM_CAP_IRQCHIP (KVM_CAP_IRQ_XICS is there instead). At the
result, KVMState::irq_routes is NULL when VFIO calls
kvm_irqchip_commit_routes.
This makes the routing update conditional.
Reported-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Tested-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
These will help us monitoring irqchip route activities more easily.
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
In the past, we are doing gsi route commit for each irqchip route
update. This is not efficient if we are updating lots of routes in the
same time. This patch removes the committing phase in
kvm_irqchip_update_msi_route(). Instead, we do explicit commit after all
routes updated.
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
One more IEC notifier is added to let msi routes know about the IEC
changes. When interrupt invalidation happens, all registered msi routes
will be updated for all PCI devices.
Since both vfio and vhost are possible gsi route consumers, this patch
will go one step further to keep them safe in split irqchip mode and
when irqfd is enabled.
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
[move trace-events lines into target-i386/trace-events]
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Adding two hooks to be notified when adding/removing msi routes. There
are two kinds of MSI routes:
- in kvm_irqchip_add_irq_route(): before assigning IRQFD. Used by
vhost, vfio, etc.
- in kvm_irqchip_send_msi(): when sending direct MSI message, if
direct MSI not allowed, we will first create one MSI route entry
in the kernel, then trigger it.
This patch only hooks the first one (irqfd case). We do not need to
take care for the 2nd one, since it's only used by QEMU userspace
(kvm-apic) and the messages will always do in-time translation when
triggered. While we need to note them down for the 1st one, so that we
can notify the kernel when cache invalidation happens.
Also, we do not hook IOAPIC msi routes (we have explicit notifier for
IOAPIC to keep its cache updated). We only need to care about irqfd
users.
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Changing the original MSIMessage parameter in kvm_irqchip_add_msi_route
into the vector number. Vector index provides more information than the
MSIMessage, we can retrieve the MSIMessage using the vector easily. This
will avoid fetching MSIMessage every time before adding MSI routes.
Meanwhile, the vector info will be used in the coming patches to further
enable gsi route update notifications.
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
As stated in linux/Documentation/virtual/kvm/api.txt:
The maximum possible value for max_vcpu_id can be retrieved using the
KVM_CAP_MAX_VCPU_ID of the KVM_CHECK_EXTENSION ioctl() at run-time.
If the KVM_CAP_MAX_VCPU_ID does not exist, you should assume that
max_vcpu_id is the same as the value returned from KVM_CAP_MAX_VCPUS.
Signed-off-by: Greg Kurz <gkurz@linux.vnet.ibm.com>
Message-Id: <146424974323.5666.5471538288045048119.stgit@bahia.huguette.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
qemu/osdep.h checks whether MAP_ANONYMOUS is defined, but this check
is bogus without a previous inclusion of sys/mman.h. Include it in
sysemu/os-posix.h and remove it from everywhere else.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Introduce kvm_get_max_memslots() API that can be used to obtain the
maximum number of memslots supported by KVM.
Signed-off-by: Bharata B Rao <bharata@linux.vnet.ibm.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
In order to deal well with the kvm vcpus (which can not be removed without any
protection), we do not close KVM vcpu fd, just record and mark it as stopped
into a list, so that we can reuse it for the appending cpu hot-add request if
possible. It is also the approach that kvm guys suggested:
https://www.mail-archive.com/kvm@vger.kernel.org/msg102839.html
Signed-off-by: Chen Fan <chen.fan.fnst@cn.fujitsu.com>
Signed-off-by: Gu Zheng <guz.fnst@cn.fujitsu.com>
Signed-off-by: Zhu Guihua <zhugh.fnst@cn.fujitsu.com>
Signed-off-by: Bharata B Rao <bharata@linux.vnet.ibm.com>
[- Explicit CPU_REMOVE() from qemu_kvm/tcg_destroy_vcpu()
isn't needed as it is done from cpu_exec_exit()
- Use iothread mutex instead of global mutex during
destroy
- Don't cleanup vCPU object from vCPU thread context
but leave it to the callers (device_add/device_del)]
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The KVM API restricts vcpu ids to be < KVM_CAP_MAX_VCPUS. On PowerPC
targets, depending on the number of threads per core in the host and
in the guest, some topologies do generate higher vcpu ids actually.
When this happens, QEMU bails out with the following error:
kvm_init_vcpu failed: Invalid argument
The KVM_CREATE_VCPU ioctl has several EINVAL return paths, so it is
not possible to fully disambiguate.
This patch adds a check in the code that computes vcpu ids, so that
we can detect the error earlier, and print a friendlier message instead
of calling KVM_CREATE_VCPU with an obviously bogus vcpu id.
Signed-off-by: Greg Kurz <gkurz@linux.vnet.ibm.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
This can be used when probing whether KVM support specific device. Here,
a raw vmfd is used.
Signed-off-by: Peter Xu <peterx@redhat.com>
Acked-by: Sergey Fedorov <serge.fdrv@gmail.com>
Message-id: 1458788142-17509-4-git-send-email-peterx@redhat.com
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
By using utilities in bitops and bitmap, this patch tries to make it more
friendly to audience. No functional change.
Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Message-Id: <1457229445-25954-1-git-send-email-richard.weiyang@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
All references to mr->ram_addr are replaced by
memory_region_get_ram_addr(mr) (except for a few assertions that are
replaced with mr->ram_block).
Reviewed-by: Gonglei <arei.gonglei@huawei.com>
Signed-off-by: Fam Zheng <famz@redhat.com>
Message-Id: <1456813104-25902-5-git-send-email-famz@redhat.com>
Acked-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Andrew Jones <drjones@redhat.com>
Message-Id: <1454355464-14999-1-git-send-email-drjones@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Clean up includes so that osdep.h is included first and headers
which it implies are not included manually.
This commit was created with scripts/clean-includes.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1454089805-5470-16-git-send-email-peter.maydell@linaro.org
The arguments of error_report() should yield a short error string
without newlines.
A few places try to print additional help after the error message by
embedding newlines in the error string. That's nice, but let's do it
the right way. Commit 474c213 cleaned up some, but they keep coming
back. Offenders tracked down with the Coccinelle semantic patch from
commit 312fd5f.
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Pavel Fedin <p.fedin@samsung.com>
Signed-off-by: Markus Armbruster <armbru@pond.sub.org>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
This patch adds support for split IRQ chip mode. When
KVM_CAP_SPLIT_IRQCHIP is enabled:
1.) The PIC, PIT, and IOAPIC are implemented in userspace while
the LAPIC is implemented by KVM.
2.) The software IOAPIC delivers interrupts to the KVM LAPIC via
kvm_set_irq. Interrupt delivery is configured via the MSI routing
table, for which routes are reserved in target-i386/kvm.c then
configured in hw/intc/ioapic.c
3.) KVM delivers IOAPIC EOIs via a new exit KVM_EXIT_IOAPIC_EOI,
which is handled in target-i386/kvm.c and relayed to the software
IOAPIC via ioapic_eoi_broadcast.
Signed-off-by: Matt Gingell <gingell@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The ioeventfd mechanism is used by vhost, dataplane, and virtio-pci to
turn guest MMIO/PIO writes into eventfd file descriptor events. This
allows arbitrary threads to be notified when the guest writes to a
specific MMIO/PIO address.
qtest and TCG do not support ioeventfd because memory writes are not
checked against registered ioeventfds in QEMU. This patch implements
this in memory_region_dispatch_write() so qtest can use ioeventfd.
Also this patch fixes vhost aborting on some misconfigured old kernels
like 3.18.0 on ARM. It is possible to explicitly enable CONFIG_EVENTFD
in expert settings, while MMIO binding support in KVM will still be
missing.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Pavel Fedin <p.fedin@samsung.com>
Message-Id: <006e01d12377$0b9c2d40$22d487c0$@samsung.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Just noticed this while grepping TARGET_PAGE_SIZE for an unrelated
reason. I didn't use qemu_real_host_page_size as kvm_set_phys_mem()
does, because we'd need to make sure page_size_init() has run first.
Signed-off-by: Andrew Jones <drjones@redhat.com>
Message-Id: <1447115022-4142-1-git-send-email-drjones@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
The HOST_PAGE_ALIGN macros don't work until the page size variables
have been set up; later in postcopy I use those macros in the RAM
code, and it can be triggered using -object.
Fix this by initialising page_size_init() earlier - it's currently
initialised inside the accelerators, move it up into vl.c.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
This reverts commit de9d61e83d.
Now 'cpu_clean_all_dirty' is useless, we can revert the related code.
Conflicts:
include/sysemu/kvm.h
Signed-off-by: Liang Li <liang.z.li@intel.com>
Message-Id: <1446695464-27116-3-git-send-email-liang.z.li@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The functions for checking xcrs, xsave and pit_state2 are
only used on x86, so they should reside in target-i386/kvm.c.
Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <1444933820-6968-1-git-send-email-thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
In-kernel ITS emulation on ARM64 will require to supply requester IDs.
These IDs can now be retrieved from the device pointer using new
pci_requester_id() function.
This patch adds pci_dev pointer to KVM GSI routing functions and makes
callers passing it.
x86 architecture does not use requester IDs, but hw/i386/kvm/pci-assign.c
also made passing PCI device pointer instead of NULL for consistency with
the rest of the code.
Signed-off-by: Pavel Fedin <p.fedin@samsung.com>
Message-Id: <ce081423ba2394a4efc30f30708fca07656bc500.1444916432.git.p.fedin@samsung.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This capability is useful to determine whether we can use KVM ITS
emulation on ARM
Signed-off-by: Pavel Fedin <p.fedin@samsung.com>
Message-Id: <ff4ccb09b837d37defd639b885526949a25276de.1444916432.git.p.fedin@samsung.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
As the comment in kvm_set_phys_mem() says, KVM works in page size chunks.
However it uses hardcoded TARGET_PAGE_SIZE which is 4K on most platforms
while actual host may use different page size, for example, PPC64 hosts
use 64K system pages.
This replaces static TARGET_PAGE_SIZE with run-time calculated
qemu_real_host_page_size.
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Message-Id: <1444102257-17405-1-git-send-email-aik@ozlabs.ru>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Some functions previously used only by vGICv2 are useful also for vGICv3
implementation. Untie them from GICState and make accessible from within
other modules:
- kvm_arm_gic_set_irq()
- kvm_gic_supports_attr() - moved to common code and renamed to
kvm_device_check_attr()
- kvm_gic_access() - turned into GIC-independent kvm_device_access().
Data pointer changed to void * because some GICv3 registers are
64-bit wide
Some of these changes are not used right now, but they will be helpful for
implementing live migration.
Actually kvm_dist_get() and kvm_dist_put() could also be made reusable, but
they would require two extra parameters (s->dev_fd and s->num_cpu) as well as
lots of typecasts of 's' to DeviceState * and back to GICState *. This makes
the code very ugly so i decided to stop at this point. I tried also an
approach with making a base class for all possible GICs, but it would contain
only three variables (dev_fd, cpu_num and irq_num), and accessing them through
the rest of the code would be again tedious (either ugly casts or qemu-style
separate object pointer). So i disliked it too.
Signed-off-by: Pavel Fedin <p.fedin@samsung.com>
Tested-by: Ashok kumar <ashoks@broadcom.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 2ef56d1dd64ffb75ed02a10dcdaf605e5b8ff4f8.1441784344.git.p.fedin@samsung.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
KVM kernel can send guest crash events into userspace.
Appropriate guest crash handler is called when kernel guest
crash event received. Guest crash event recognized by a
KVM_SYSTEM_EVENT_CRASH type of system event.
Signed-off-by: Andrey Smetanin <asmetanin@virtuozzo.com>
Signed-off-by: Denis V. Lunev <den@openvz.org>
CC: Paolo Bonzini <pbonzini@redhat.com>
CC: Andreas Färber <afaerber@suse.de>
Message-Id: <1435924905-8926-11-git-send-email-den@openvz.org>
[Rebase: add lock/unlock iothread around qemu_system_guest_panicked - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
When we add new adapter routes we call kvm_irqchip_add_route() for every
virtqueue and in the same step also do the KVM_SET_GSI_ROUTING ioctl.
This is unnecessary costly as the interface allows us to set multiple
routes in one go. Let's first add all routes to the table stored in the
global kvm_state and then do the ioctl to commit the routes to the
in-kernel irqchip.
This saves us several ioctls to the kernel where for each call a list
is reallocated and populated.
Signed-off-by: Jens Freimann <jfrei@linux.vnet.ibm.com>
Reviewed-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
VFIO platform device needs to setup irqfd but it does not know the
gsi corresponding to the device qemu_irq. This patch proposes to
store a hash table in kvm_state using the qemu_irq as key and the gsi
as a value.
kvm_irqchip_set_qemuirq_gsi allows to insert such a pair. The interrupt
controller is supposed to use it.
kvm_irqchip_[add, remove]_irqfd_notifier allows to setup/tear down
irqfd directly from the qemu_irq.
Signed-off-by: Eric Auger <eric.auger@linaro.org>
Tested-by: Vikram Sethi <vikrams@codeaurora.org>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Anticipating for the introduction of new add/remove functions taking
a qemu_irq parameter, let's rename existing ones with a gsi suffix.
Signed-off-by: Eric Auger <eric.auger@linaro.org>
Tested-by: Vikram Sethi <vikrams@codeaurora.org>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
KVM_CREATE_IRQCHIP should never fail, and so should its userspace
wrapper kvm_irqchip_create. The function does not do anything
if the irqchip capability is not available, as is the case for PPC.
With this patch, kvm_arch_init can allocate memory and it will not
be leaked.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Make kvm_memory_listener_register public, and assign a kernel
address space id to each KVMMemoryListener.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
No semantic change, but s->slots moves into a new struct
KVMMemoryListener. KVM's memory listener becomes a member of struct
KVMState, and becomes of type KVMMemoryListener.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
i386 code will have to define a different KVMMemoryListener. Create
an internal header so that KVMSlot is not exposed outside.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Currently kvm_mem_flags just translates bools to bits, let's
make it also determine the bools first. This avoids its parameter
list growing each time we add a flag.
Signed-off-by: Andrew Jones <drjones@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Do not take the BQL before dispatching MMIO requests of KVM VCPUs.
Instead, address_space_rw will do it if necessary. This enables completely
BQL-free MMIO handling in KVM mode for upcoming devices with fine-grained
locking.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <1434646046-27150-10-git-send-email-pbonzini@redhat.com>
Do not take the BQL before dispatching PIO requests of KVM VCPUs.
Instead, address_space_rw will do it if necessary. This enables
completely BQL-free PIO handling in KVM mode for upcoming devices with
fine-grained locking.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <1434646046-27150-8-git-send-email-pbonzini@redhat.com>
This opens the path to get rid of the iothread lock on vmexits in KVM
mode. On x86, the in-kernel irqchips has to be used because we otherwise
need to synchronize APIC and other per-cpu state accesses that could be
changed concurrently.
Regarding pre/post-run callbacks, s390x and ARM should be fine without
specific locking as the callbacks are empty. MIPS and POWER require
locking for the pre-run callback.
For the handle_exit callback, it is non-empty in x86, POWER and s390.
Some POWER cases could do without the locking, but it is left in
place for now.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <1434646046-27150-7-git-send-email-pbonzini@redhat.com>
Last month, we experienced several guests crash(6cores-8cores), qemu logs
display the following messages:
qemu-system-x86_64: /build/qemu-2.1.2/kvm-all.c:976:
kvm_irqchip_commit_routes: Assertion `ret == 0' failed.
After analysis and verification, we can confirm it's irq-balance
daemon(in guest) leads to the assertion failure. Start a 8 core guest with
two disks, execute the following scripts will reproduce the BUG quickly:
irq_affinity.sh
========================================================================
vda_irq_num=25
vdb_irq_num=27
while [ 1 ]
do
for irq in {1,2,4,8,10,20,40,80}
do
echo $irq > /proc/irq/$vda_irq_num/smp_affinity
echo $irq > /proc/irq/$vdb_irq_num/smp_affinity
dd if=/dev/vda of=/dev/zero bs=4K count=100 iflag=direct
dd if=/dev/vdb of=/dev/zero bs=4K count=100 iflag=direct
done
done
========================================================================
QEMU setup static irq route entries in kvm_pc_setup_irq_routing(), PIC and
IOAPIC share the first 15 GSI numbers, take up 23 GSI numbers, but take up
38 irq route entries. When change irq smp_affinity in guest, a dynamic route
entry may be setup, the current logic is: if allocate GSI number succeeds,
a new route entry can be added. The available dynamic GSI numbers is
1021(KVM_MAX_IRQ_ROUTES-23), but available irq route entries is only
986(KVM_MAX_IRQ_ROUTES-38), GSI numbers greater than route entries.
irq-balance's behavior will eventually leads to total irq route entries
exceed KVM_MAX_IRQ_ROUTES, ioctl(KVM_SET_GSI_ROUTING) fail and
kvm_irqchip_commit_routes() trigger assertion failure.
This patch fix the BUG.
Signed-off-by: Wenshuang Ma <kevinnma@tencent.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
