Decimal Floating Point has been available on POWER7 and later (server)
cpus. However, it can be disabled on the hypervisor, meaning that it's
not available to guests.
We currently handle this by conditionally advertising DFP support in the
device tree depending on whether the guest CPU model supports it - which
can also depend on what's allowed in the host for -cpu host. That can lead
to confusion on migration, since host properties are silently affecting
guest visible properties.
This patch handles it by treating it as an optional capability for the
pseries machine type.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Greg Kurz <groug@kaod.org>
We currently have some conditionals in the spapr device tree code to decide
whether or not to advertise the availability of the VMX (aka Altivec) and
VSX vector extensions to the guest, based on whether the guest cpu has
those features.
This can lead to confusion and subtle failures on migration, since it makes
a guest visible change based only on host capabilities. We now have a
better mechanism for this, in spapr capabilities flags, which explicitly
depend on user options rather than host capabilities.
Rework the advertisement of VSX and VMX based on a new VSX capability. We
no longer bother with a conditional for VMX support, because every CPU
that's ever been supported by the pseries machine type supports VMX.
NOTE: Some userspace distributions (e.g. RHEL7.4) already rely on
availability of VSX in libc, so using cap-vsx=off may lead to a fatal
SIGILL in init.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Greg Kurz <groug@kaod.org>
Now that the "pseries" machine type implements optional capabilities (well,
one so far) there's the possibility of having different capabilities
available at either end of a migration. Although arguably a user error,
it would be nice to catch this situation and fail as gracefully as we can.
This adds code to migrate the capabilities flags. These aren't pulled
directly into the destination's configuration since what the user has
specified on the destination command line should take precedence. However,
they are checked against the destination capabilities.
If the source was using a capability which is absent on the destination,
we fail the migration, since that could easily cause a guest crash or other
bad behaviour. If the source lacked a capability which is present on the
destination we warn, but allow the migration to proceed.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Greg Kurz <groug@kaod.org>
This adds an spapr capability bit for Hardware Transactional Memory. It is
enabled by default for pseries-2.11 and earlier machine types. with POWER8
or later CPUs (as it must be, since earlier qemu versions would implicitly
allow it). However it is disabled by default for the latest pseries-2.12
machine type.
This means that with the latest machine type, HTM will not be available,
regardless of CPU, unless it is explicitly enabled on the command line.
That change is made on the basis that:
* This way running with -M pseries,accel=tcg will start with whatever cpu
and will provide the same guest visible model as with accel=kvm.
- More specifically, this means existing make check tests don't have
to be modified to use cap-htm=off in order to run with TCG
* We hope to add a new "HTM without suspend" feature in the not too
distant future which could work on both POWER8 and POWER9 cpus, and
could be enabled by default.
* Best guesses suggest that future POWER cpus may well only support the
HTM-without-suspend model, not the (frankly, horribly overcomplicated)
POWER8 style HTM with suspend.
* Anecdotal evidence suggests problems with HTM being enabled when it
wasn't wanted are more common than being missing when it was.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Greg Kurz <groug@kaod.org>
Because PAPR is a paravirtual environment access to certain CPU (or other)
facilities can be blocked by the hypervisor. PAPR provides ways to
advertise in the device tree whether or not those features are available to
the guest.
In some places we automatically determine whether to make a feature
available based on whether our host can support it, in most cases this is
based on limitations in the available KVM implementation.
Although we correctly advertise this to the guest, it means that host
factors might make changes to the guest visible environment which is bad:
as well as generaly reducing reproducibility, it means that a migration
between different host environments can easily go bad.
We've mostly gotten away with it because the environments considered mature
enough to be well supported (basically, KVM on POWER8) have had consistent
feature availability. But, it's still not right and some limitations on
POWER9 is going to make it more of an issue in future.
This introduces an infrastructure for defining "sPAPR capabilities". These
are set by default based on the machine version, masked by the capabilities
of the chosen cpu, but can be overriden with machine properties.
The intention is at reset time we verify that the requested capabilities
can be supported on the host (considering TCG, KVM and/or host cpu
limitations). If not we simply fail, rather than silently modifying the
advertised featureset to the guest.
This does mean that certain configurations that "worked" may now fail, but
such configurations were already more subtly broken.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Greg Kurz <groug@kaod.org>
Currently the pseries machine sets the compatibility mode for the
guest's cpus in two places: 1) at machine reset and 2) after CAS
negotiation.
This means that if we set or negotiate a compatiblity mode, then
hotplug a cpu, the hotplugged cpu doesn't get the right mode set and
will incorrectly have the full native features.
To correct this, we set the compatibility mode on a cpu when it is
brought online with the 'start-cpu' RTAS call. Given that we no
longer need to set the compatibility mode on all CPUs at machine
reset, so we change that to only set the mode for the boot cpu.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reported-by: Satheesh Rajendran <sathnaga@linux.vnet.ibm.com>
Tested-by: Satheesh Rajendran <sathnaga@linux.vnet.ibm.com>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
if KVM is enabled and KVM capabilities MMU radix is available,
the partition table entry (patb_entry) for the radix mode is
initialized by default in ppc_spapr_reset().
It's a problem if we want to migrate the guest to a POWER8 host
while the kernel is not started to set the value to the one
expected for a POWER8 CPU.
The "-machine max-cpu-compat=power8" should allow to migrate
a POWER9 KVM host to a POWER8 KVM host, but because patb_entry
is set, the destination QEMU tries to enable radix mode on the
POWER8 host. This fails and cancels the migration:
Process table config unsupported by the host
error while loading state for instance 0x0 of device 'spapr'
load of migration failed: Invalid argument
This patch doesn't set the PATB entry if the user provides
a CPU compatibility mode that doesn't support radix mode.
Signed-off-by: Laurent Vivier <lvivier@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
We conditionally adjust part of the guest device tree based on the
global msi_nonbroken flag. However, the main machine type code
initializes msi_nonbroken to true and there's nothing that would set
it to false again.
So replace the test with an assert().
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Machine objects have two init functions - the generic QOM level
instance_init which should only do static object initialization, and
the Machine specific MachineClass::init which does the actual
construction of the machine.
In spapr the functions implementing these two have names -
ppc_machine_initfn() and ppc_spapr_init() - which don't correspond closely
to either of those. To prevent people (read, me) from confusing which is
which, rename them spapr_instance_init() and spapr_machine_init() to
make it clearer which is which.
While we're there rename ppc_spapr_reset() to spapr_machine_reset() to
match.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com>
SPAPR is the last user of numa_get_node() and a bunch of
supporting code to maintain numa_info[x].addr list.
Get LMB node id from pc-dimm list, which allows to
remove ~80LOC maintaining dynamic address range
lookup list.
It also removes pc-dimm dependency on numa_[un]set_mem_node_id()
and makes pc-dimms a sole source of information about which
node it belongs to and removes duplicate data from global
numa_info.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
xics_get_qirq() is only used by the sPAPR machine. Let's move it there
and change its name to reflect its scope. It will be useful for XIVE
support which will use its own set of qirqs.
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
It will make synchronisation easier with the XIVE interrupt mode when
available. The 'irq' parameter refers to the global IRQ number space.
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Also change the prototype to use a sPAPRMachineState and prefix them
with spapr_irq_. It will let us synchronise the IRQ allocation with
the XIVE interrupt mode when available.
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The current code assumes that only the CPU core object holds a
reference on each individual CPU object, and happily frees their
allocated memory when the core is unrealized. This is dangerous
as some other code can legitimely keep a pointer to a CPU if it
calls object_ref(), but it would end up with a dangling pointer.
Let's allocate all CPUs with object_new() and let QOM free them
when their reference count reaches zero. This greatly simplify the
code as we don't have to fiddle with the instance size anymore.
Signed-off-by: Greg Kurz <groug@kaod.org>
Acked-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
While we're at it fix a couple of small errors in the 2.11 and 2.10 models
(they didn't have any real effect, but don't quite match the template).
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
At guest reset time, we allocate a hash page table (HPT) for the guest
based on the guest's RAM size. If dynamic HPT resizing is not available we
use the maximum RAM size, if it is we use the current RAM size.
But the "current RAM size" calculation is incorrect - we just use the
"base" ram_size from the machine structure. This doesn't include any
pluggable DIMMs that are already plugged at reset time.
This means that if you try to start a 'pseries' machine with a DIMM
specified on the command line that's much larger than the "base" RAM size,
then the guest will get a woefully inadequate HPT. This can lead to a
guest freeze during boot as it runs out of HPT space during initial MMU
setup.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Greg Kurz <groug@kaod.org>
Tested-by: Greg Kurz <groug@kaod.org>
Migration of pseries is broken with TCG because
QEMU tries to restore KVM MMU state unconditionally.
The result is a SIGSEGV in kvm_vm_ioctl():
#0 kvm_vm_ioctl (s=0x0, type=-2146390353)
at qemu/accel/kvm/kvm-all.c:2032
#1 0x00000001003e3e2c in kvmppc_configure_v3_mmu (cpu=<optimized out>,
radix=<optimized out>, gtse=<optimized out>, proc_tbl=<optimized out>)
at qemu/target/ppc/kvm.c:396
#2 0x00000001002f8b88 in spapr_post_load (opaque=0x1019103c0,
version_id=<optimized out>) at qemu/hw/ppc/spapr.c:1578
#3 0x000000010059e4cc in vmstate_load_state (f=0x106230000,
vmsd=0x1009479e0 <vmstate_spapr>, opaque=0x1019103c0,
version_id=<optimized out>) at qemu/migration/vmstate.c:165
#4 0x00000001005987e0 in vmstate_load (f=<optimized out>, se=<optimized out>)
at qemu/migration/savevm.c:748
This patch fixes the problem by not calling the KVM function with the
TCG mode.
Fixes: d39c90f5f3 ("spapr: Fix migration of Radix guests")
Signed-off-by: Laurent Vivier <lvivier@redhat.com>
Reviewed-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The patb_entry is used to store the location of the process table in
guest memory. The msb is also used to indicate the mmu mode of the
guest, that is patb_entry & 1 << 63 ? radix_mode : hash_mode.
Currently we set this to zero in spapr_setup_hpt_and_vrma() since if
this function gets called then we know we're hash. However some code
paths, such as setting up the hpt on incoming migration of a hash guest,
call spapr_reallocate_hpt() directly bypassing this higher level
function. Since we assume radix if the host is capable this results in
the msb in patb_entry being left set so in spapr_post_load() we call
kvmppc_configure_v3_mmu() and tell the host we're radix which as
expected means addresses cannot be translated once we actually run the cpu.
To fix this move the zeroing of patb_entry into spapr_reallocate_hpt().
Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
LUNs >= 256 have to be encoded with the so-called "flat space
addressing method" for virtio-scsi, where an additional bit has to
be set. SLOF already took care of this with the following commit:
https://git.qemu.org/?p=SLOF.git;a=commitdiff;h=f72a37713fea47da
(see https://bugzilla.redhat.com/show_bug.cgi?id=1431584 for details)
But QEMU does not use this encoding yet for device tree paths
that have to be handed over to SLOF to deal with the "bootindex"
property, so SLOF currently fails to boot from virtio-scsi devices
with LUNs >= 256 in the right boot order. Fix it by using the bit
to indicate the "flat space addressing method" for LUNs >= 256.
Signed-off-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
A DRC with a pending unplug request releases its associated device at
machine reset time.
In the case of LMB, when all DRCs for a DIMM device have been reset,
the DIMM gets unplugged, causing guest memory to disappear. This may
be very confusing for anything still using this memory.
This is exactly what happens with vhost backends, and QEMU aborts
with:
qemu-system-ppc64: used ring relocated for ring 2
qemu-system-ppc64: qemu/hw/virtio/vhost.c:649: vhost_commit: Assertion
`r >= 0' failed.
The issue is that each DRC registers a QEMU reset handler, and we
don't control the order in which these handlers are called (ie,
a LMB DRC will unplug a DIMM before the virtio device using the
memory on this DIMM could stop its vhost backend).
To avoid such situations, let's reset DRCs after all devices
have been reset.
Reported-by: Mallesh N. Koti <mallesh@linux.vnet.ibm.com>
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Daniel Henrique Barboza <danielhb@linux.vnet.ibm.com>
Reviewed-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The device tree nodes ibm,arch-vec-5-platform-support and ibm,pa-features
are used to communicate features of the cpu to the guest operating
system. The properties of each of these are determined based on the
selected cpu model and the availability of hypervisor features.
Currently the compatibility mode of the cpu is not taken into account.
The ibm,arch-vec-5-platform-support node is used to communicate the
level of support for various ISAv3 processor features to the guest
before CAS to inform the guests' request. The available mmu mode should
only be hash unless the cpu is a POWER9 which is not in a prePOWER9
compat mode, in which case the available modes depend on the
accelerator and the hypervisor capabilities.
The ibm,pa-featues node is used to communicate the level of cpu support
for various features to the guest os. This should only contain features
relevant to the operating mode of the processor, that is the selected
cpu model taking into account any compat mode. This means that the
compat mode should be taken into account when choosing the properties of
ibm,pa-features and they should match the compat mode selected, or the
cpu model selected if no compat mode.
Update the setting of these cpu features in the device tree as described
above to properly take into account any compat mode. We use the
ppc_check_compat function which takes into account the current processor
model and the cpu compat mode.
Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
use generic cpu_model parsing introduced by
(6063d4c0f vl.c: convert cpu_model to cpu type and set of global properties before machine_init())
it allows to:
* replace sPAPRMachineClass::tcg_default_cpu with
MachineClass::default_cpu_type
* drop cpu_parse_cpu_model() from hw/ppc/spapr.c and reuse
one in vl.c
* simplify spapr_get_cpu_core_type() by removing
not needed anymore recurrsion since alias look up
happens earlier at vl.c and spapr_get_cpu_core_type()
works only with resulted from that cpu type.
* spapr no more needs to parse/depend on being phased out
MachineState::cpu_model, all tha parsing done by generic
code and target specific callback.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
[dwg: Correct minor compile error]
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Acked-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
replace sPAPRCPUCoreClass::cpu_class with cpu type name
since it were needed just to get that at points it were
accessed.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Acked-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
there is a dedicated callback CPUClass::parse_features
which purpose is to convert -cpu features into a set of
global properties AND deal with compat/legacy features
that couldn't be directly translated into CPU's properties.
Create ppc variant of it (ppc_cpu_parse_featurestr) and
move 'compat=val' handling from spapr_cpu_core.c into it.
That removes a dependency of board/core code on cpu_model
parsing and would let to reuse common -cpu parsing
introduced by 6063d4c0
Set "max-cpu-compat" property only if it exists, in practice
it should limit 'compat' hack to spapr machine and allow
to avoid including machine/spapr headers in target/ppc/cpu.c
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
LMB removal is completed only when the spapr_lmb_release callback
is called after all DRCs of the dimm are detached. During this
time, it is possible that a unplug request for the same dimm
arrives, trying to detach DRCs that were detached by the guest
in the first unplug_request.
BQL doesn't help in this case - the lock will prevent any concurrent
removal from happening until the end of spapr_memory_unplug_request
only. What happens is that the second unplug_request ends up calling
spapr_drc_detach in a DRC that were detached already, causing an
assert error in spapr_drc_detach (e.g
https://bugs.launchpad.net/qemu/+bug/1718118).
spapr_lmb_release uses a structure called sPAPRDIMMState, stored in the
spapr->pending_dimm_unplugs QTAIL, to track how many LMB DRCs are left
to be detached by the guest. When there are no more DRCs left, this
structure is deleted and the pc-dimm unplug handler is called to
finish the process.
This patch reuses the sPAPRDIMMState to allow unplug_request to know
if there is an ongoing unplug process for a given dimm, aborting the
unplug request in this case, by doing the following changes:
- in spapr_lmb_release callback, move the dimm state removal to the
end, after pc-dimm unplug handler. With this change we can check for
the existence of the dimm state to see if the unplug process is
done.
- use spapr_pending_dimm_unplugs_find in spapr_memory_unplug_request
to check if the dimm state exists. If positive, there is an unplug
operation already in progress for this dimm, meaning that we should
abort it and warn the user about it.
Fixes: https://bugs.launchpad.net/qemu/+bug/1718118
Signed-off-by: Daniel Henrique Barboza <danielhb@linux.vnet.ibm.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The CAS buffer is provided by SLOF. A broken SLOF could pass a silly
size: either smaller than the diff header, in which case the current
code will try to allocate 16 Exabytes of memory and g_malloc0() will
abort, or bigger than the maximum memory provisioned for SLOF (ie,
40 Megabytes), which doesn't make sense. Both cases indicate that
SLOF has a bug.
Let's print out an explicit error message and exit since rebooting as
we do with other errors would only result in a reset loop.
Signed-off-by: Greg Kurz <groug@kaod.org>
[dwg: Fix format specifier that broke 32-bit builds]
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The offset of the root node is guaranteed to be 0.
This doesn't fix anything, it's just trivial cleanup of the two
remaining places where this was done under hw/ppc.
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Daniel Henrique Barboza <danielhb@linux.vnet.ibm.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
When running with KVM PR, if a new HPT is allocated we need to inform
KVM about the HPT address and size. This is currently done by hacking
the value of SDR1 and pushing it to KVM in several places.
Also, migration breaks the guest since it is very unlikely the HPT has
the same address in source and destination, but we push the incoming
value of SDR1 to KVM anyway.
This patch introduces a new virtual hypervisor hook so that the spapr
code can provide the correct value of SDR1 to be pushed to KVM each
time kvmppc_put_books_sregs() is called.
It allows to get rid of all the hacking in the spapr/kvmppc code and
it fixes migration of nested KVM PR.
Suggested-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
This consolidates some duplicated code in a dedicated helpers.
Signed-off-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The use of KVM_PPC_GET_HTAB_FD is open-coded in kvmppc_read_hptes()
and kvmppc_write_hpte().
This patch modifies kvmppc_get_htab_fd() so that it can be used
everywhere we need to access the in-kernel htab:
- add an index argument
=> only kvmppc_read_hptes() passes an actual index, all other users
pass 0
- add an errp argument to propagate error messages to the caller.
=> spapr migration code prints the error
=> hpte helpers pass &error_abort to keep the current behavior
of hw_error()
While here, this also fixes a bug in kvmppc_write_hpte() so that it
opens the htab fd for writing instead of reading as it currently does.
This never broke anything because we currently never call this code,
as explained in the changelog of commit c138593380:
"This support updating htab managed by the hypervisor. Currently
we don't have any user for this feature. This actually bring the
store_hpte interface in-line with the load_hpte one. We may want
to use this when we want to emulate henter hcall in qemu for HV
kvm."
The above is still true today.
Signed-off-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
When kvmppc_get_htab_fd() fails, its return value is propagated up to
qemu_savevm_state_iterate() or to qemu_savevm_state_complete_precopy().
All savevm handlers expect to receive a negative errno on error.
Let's patch kvmppc_get_htab_fd() accordingly.
While here, let's change htab_load() in the spapr code to also
propagate the error, since it doesn't make sense to abort() if
we couldn't get the htab fd from KVM.
Signed-off-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
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>
On POWER9, the Client Architecture Support (CAS) negotiation process
determines whether the guest operates in XIVE Legacy compatibility or
in XIVE exploitation mode. Now that we have initial guest support for
the XIVE interrupt controller, let's fix the bits definition which have
evolved in the latest specs.
The platform advertises the XIVE Exploitation Mode support using the
property "ibm,arch-vec-5-platform-support-vec-5", byte 23 bits 0-1 :
- 0b00 XIVE legacy mode Only
- 0b01 XIVE exploitation mode Only
- 0b10 XIVE legacy or exploitation mode
The OS asks for XIVE Exploitation Mode support using the property
"ibm,architecture-vec-5", byte 23 bits 0-1:
- 0b00 XIVE legacy mode Only
- 0b01 XIVE exploitation mode Only
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
This patch removes the qdev_get_machine() calls that are made in
spapr.c in situations where we can get an existing pointer for
the MachineState by either passing it as an argument to the function
or by using other already available pointers.
The following changes were made:
- spapr_node0_size: static function that is called two times:
at spapr_setup_hpt_and_vrma and ppc_spapr_init. In both cases we can
pass an existing MachineState pointer to it.
- spapr_build_fdt: MachineState pointer can be retrieved from
the existing sPAPRMachineState pointer.
- spapr_boot_set: the opaque in the first arg is a sPAPRMachineState
pointer as we can see inside ppc_spapr_init:
qemu_register_boot_set(spapr_boot_set, spapr);
We can get a MachineState pointer from it.
- spapr_machine_device_plug and spapr_machine_device_unplug_request: the
MachineState, sPAPRMachineState, MachineClass and sPAPRMachineClass pointers
can all be retrieved from the HotplugHandler pointer.
Signed-off-by: Daniel Henrique Barboza <danielhb@linux.vnet.ibm.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
KVM now allows writing to KVM_CAP_PPC_SMT which has previously been
read only. Doing so causes KVM to act, for that VM, as if the host's
SMT mode was the given value. This is particularly important on Power
9 systems because their default value is 1, but they are able to
support values up to 8.
This patch introduces a way to control this capability via a new
machine property called VSMT ("Virtual SMT"). If the value is not set
on the command line a default is chosen that is, when possible,
compatible with legacy systems.
Note that the intialization of KVM_CAP_PPC_SMT has changed slightly
because it has changed (in KVM) from a global capability to a
VM-specific one. This won't cause a problem on older KVMs because VM
capabilities fall back to global ones.
Signed-off-by: Sam Bobroff <sam.bobroff@au1.ibm.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The concept of a VCPU ID that differs from the CPU's index
(cpu->cpu_index) exists only within SPAPR machines so, move the
functions ppc_get_vcpu_id() and ppc_get_cpu_by_vcpu_id() into spapr.c
and rename them appropriately.
Signed-off-by: Sam Bobroff <sam.bobroff@au1.ibm.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
This field actually records the VCPU ID used by KVM and, although the
value is also used in the device tree it is primarily the VCPU ID so
rename it as such.
Signed-off-by: Sam Bobroff <sam.bobroff@au1.ibm.com>
[dwg: Updated comment missed in cpu.h]
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
This patch is a follow up on the discussions made in patch
"hw/ppc: disable hotplug before CAS is completed" that can be
found at [1].
At this moment, we do not support CPU/memory hotplug in early
boot stages, before CAS. When a hotplug occurs, the event is logged
in an internal RTAS event log queue and an IRQ pulse is fired. In
regular conditions, the guest handles the interrupt by executing
check_exception, fetching the generated hotplug event and enabling
the device for use.
In early boot, this IRQ isn't caught (SLOF does not handle hotplug
events), leaving the event in the rtas event log queue. If the guest
executes check_exception due to another hotplug event, the re-assertion
of the IRQ ends up de-queuing the first hotplug event as well. In short,
a device hotplugged before CAS is considered coldplugged by SLOF.
This leads to device misbehavior and, in some cases, guest kernel
Ooops when trying to unplug the device.
A proper fix would be to turn every device hotplugged before CAS
as a colplugged device. This is not trivial to do with the current
code base though - the FDT is written in the guest memory at
ppc_spapr_reset and can't be retrieved without adding extra state
(fdt_size for example) that will need to managed and migrated. Adding
the hotplugged DT in the middle of CAS negotiation via the updated DT
tree works with CPU devs, but panics the guest kernel at boot. Additional
analysis would be necessary for LMBs and PCI devices. There are
questions to be made in QEMU/SLOF/kernel level about how we can make
this change in a sustainable way.
With Linux guests, a fix would be the kernel executing check_exception
at boot time, de-queueing the events that happened in early boot and
processing them. However, even if/when the newer kernels start
fetching these events at boot time, we need to take care of older
kernels that won't be doing that.
This patch works around the situation by issuing a CAS reset if a hotplugged
device is detected during CAS:
- the DRC conditions that warrant a CAS reset is the same as those that
triggers a DRC migration - the DRC must have a device attached and
the DRC state is not equal to its ready_state. With that in mind, this
patch makes use of 'spapr_drc_needed' to determine if a CAS reset
is needed.
- In the middle of CAS negotiations, the function
'spapr_hotplugged_dev_before_cas' goes through all the DRCs to see
if there are any DRC that requires a reset, using spapr_drc_needed. If
that happens, returns '1' in 'spapr_h_cas_compose_response' which will set
spapr->cas_reboot to true, causing the machine to reboot.
No changes are made for coldplug devices.
[1] http://lists.nongnu.org/archive/html/qemu-devel/2017-08/msg02855.html
Signed-off-by: Daniel Henrique Barboza <danielhb@linux.vnet.ibm.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The sPAPR machine isn't clearing up the pending events QTAILQ on
machine reboot. This allows for unprocessed hotplug/epow events
to persist in the queue after reset and, when reasserting the IRQs in
check_exception later on, these will be being processed by the OS.
This patch implements a new function called 'spapr_clear_pending_events'
that clears up the pending_events QTAILQ. This helper is then called
inside ppc_spapr_reset to clear up the events queue, preventing
old/deprecated events from persisting after a reset.
Signed-off-by: Daniel Henrique Barboza <danielhb@linux.vnet.ibm.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>
This reverts commit b87680427e.
I thought this was a harmless preliminary for XIVE enablement patches
we expect later on. However, due to some subtle interactions between
qemu and SLOF (guest firmware) this breaks some things. Revert it for
now, we'll work out how to fix it when the rest of the XIVE patches
are ready.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Commit 0cffce56 (hw/ppc/spapr.c: adding pending_dimm_unplugs to
sPAPRMachineState) introduced a new way to track pending LMBs of DIMM
device that is marked for removal. Since this commit we can hit the
assert in spapr_pending_dimm_unplugs_add() in the following situation:
- DIMM device removal fails as the guest doesn't allow the removal.
- Subsequent attempt to remove the same DIMM would hit the assert
as the corresponding sPAPRDIMMState is still part of the
pending_dimm_unplugs list.
Fix this by removing the assert and conditionally adding the
sPAPRDIMMState to pending_dimm_unplugs list only when it is not
already present.
Fixes: 0cffce56ae
Signed-off-by: Bharata B Rao <bharata@linux.vnet.ibm.com>
[dwg: Tweaked to avoid returning NULL when spapr_pending_dimm_unplugs_add()
does find an existing entry]
Reviewed-by: Daniel Barboza <danielhb@linux.vnet.ibm.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Commit 3a38429 ("spapr: Add a "no HPT" encoding to HTAB migration stream")
allows to migrate an empty HPT, but doesn't mark correctly the
end of the migration stream.
The end condition (value returned by htab_save_iterate())
should be 1, whereas in 3a38429 it returns 0.
The problem can be reproduced with QEMU monitor command "savevm":
the command never stops and the disk image grows without limit.
Fixes: 3a38429748
Signed-off-by: Laurent Vivier <lvivier@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
In case of error, we must ensure the dynamically allocated base_core_type
is freed, like it is done everywhere else in this function.
This is a regression introduced in QEMU 2.9 by commit 8149e2992f.
Signed-off-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
We've now implemented a PAPR extension allowing PAPR guest to resize
their hash page table (HPT) during runtime.
This patch makes use of that facility to allocate smaller HPTs by default.
Specifically when a guest is aware of the HPT resize facility, qemu sizes
the HPT to the initial memory size, rather than the maximum memory size on
the assumption that the guest will resize its HPT if necessary for hot
plugged memory.
When the initial memory size is much smaller than the maximum memory size
(a common configuration with e.g. oVirt / RHEV) then this can save
significant memory on the HPT.
If the guest does *not* advertise HPT resize awareness when it makes the
ibm,client-architecture-support call, qemu resizes the HPT for maxmimum
memory size (unless it's been configured not to allow such guests at all).
For now we make that reallocation assuming the guest has not yet used the
HPT at all. That's true in practice, but not, strictly, an architectural
or PAPR requirement. If we need to in future we can fix this by having
the client-architecture-support call reboot the guest with the revised
HPT size (the client-architecture-support call is explicitly permitted to
trigger a reboot in this way).
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com>
We've now implemented a PAPR extensions which allows PAPR guests (i.e.
"pseries" machine type) to resize their hash page table during runtime.
However, that extension is only enabled if explicitly chosen on the
command line. This patch enables it by default for spapr-2.10, but leaves
it disabled (by default) for older machine types.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Laurent Vivier <lvivier@redhat.com>
This patch implements hypercalls allowing a PAPR guest to resize its own
hash page table. This will eventually allow for more flexible memory
hotplug.
The implementation is partially asynchronous, handled in a special thread
running the hpt_prepare_thread() function. The state of a pending resize
is stored in SPAPR_MACHINE->pending_hpt.
The H_RESIZE_HPT_PREPARE hypercall will kick off creation of a new HPT, or,
if one is already in progress, monitor it for completion. If there is an
existing HPT resize in progress that doesn't match the size specified in
the call, it will cancel it, replacing it with a new one matching the
given size.
The H_RESIZE_HPT_COMMIT completes transition to a resized HPT, and can only
be called successfully once H_RESIZE_HPT_PREPARE has successfully
completed initialization of a new HPT. The guest must ensure that there
are no concurrent accesses to the existing HPT while this is called (this
effectively means stop_machine() for Linux guests).
For now H_RESIZE_HPT_COMMIT goes through the whole old HPT, rehashing each
HPTE into the new HPT. This can have quite high latency, but it seems to
be of the order of typical migration downtime latencies for HPTs of size
up to ~2GiB (which would be used in a 256GiB guest).
In future we probably want to move more of the rehashing to the "prepare"
phase, by having H_ENTER and other hcalls update both current and
pending HPTs. That's a project for another day, but should be possible
without any changes to the guest interface.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
