Watchdog is taken at critical exception level. So this patch
is tested with host watchdog exception happening when guest
is running.
Signed-off-by: NBharat Bhushan <bharat.bhushan@freescale.com>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Signed-off-by: NBharat Bhushan <bharat.bhushan@freescale.com>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Added the decrementer auto-reload support. DECAR is readable
on e500v2/e500mc and later cpus.
Signed-off-by: NBharat Bhushan <bharat.bhushan@freescale.com>
Signed-off-by: NAlexander Graf <agraf@suse.de>

This adds a new ioctl to enable userspace to control the size of the guest
hashed page table (HPT) and to clear it out when resetting the guest.
The KVM_PPC_ALLOCATE_HTAB ioctl is a VM ioctl and takes as its parameter
a pointer to a u32 containing the desired order of the HPT (log base 2
of the size in bytes), which is updated on successful return to the
actual order of the HPT which was allocated.

There must be no vcpus running at the time of this ioctl.  To enforce
this, we now keep a count of the number of vcpus running in
kvm->arch.vcpus_running.

If the ioctl is called when a HPT has already been allocated, we don't
reallocate the HPT but just clear it out.  We first clear the
kvm->arch.rma_setup_done flag, which has two effects: (a) since we hold
the kvm->lock mutex, it will prevent any vcpus from starting to run until
we're done, and (b) it means that the first vcpu to run after we're done
will re-establish the VRMA if necessary.

If userspace doesn't call this ioctl before running the first vcpu, the
kernel will allocate a default-sized HPT at that point.  We do it then
rather than when creating the VM, as the code did previously, so that
userspace has a chance to do the ioctl if it wants.

When allocating the HPT, we can allocate either from the kernel page
allocator, or from the preallocated pool.  If userspace is asking for
a different size from the preallocated HPTs, we first try to allocate
using the kernel page allocator.  Then we try to allocate from the
preallocated pool, and then if that fails, we try allocating decreasing
sizes from the kernel page allocator, down to the minimum size allowed
(256kB).  Note that the kernel page allocator limits allocations to
1 << CONFIG_FORCE_MAX_ZONEORDER pages, which by default corresponds to
16MB (on 64-bit powerpc, at least).
Signed-off-by: NPaul Mackerras <paulus@samba.org>
[agraf: fix module compilation]
Signed-off-by: NAlexander Graf <agraf@suse.de>

When handling the H_BULK_REMOVE hypercall, we were forgetting to
invalidate and unlock the hashed page table entry (HPTE) in the case
where the page had been paged out.  This fixes it by clearing the
first doubleword of the HPTE in that case.

This fixes a regression introduced in commit a92bce95 ("KVM: PPC:
Book3S HV: Keep HPTE locked when invalidating").  The effect of the
regression is that the host kernel will sometimes hang when under
memory pressure.
Signed-off-by: NPaul Mackerras <paulus@samba.org>
Signed-off-by: NAlexander Graf <agraf@suse.de>

The code forgot to scramble the VSIDs the way we normally do
and was basically using the "proto VSID" directly with the MMU.

This means that in practice, KVM used random VSIDs that could
collide with segments used by other user space programs.
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>
[agraf: simplify ppc32 case]
Signed-off-by: NAlexander Graf <agraf@suse.de>

When jumping back into the kernel to code that knows that it would be
using HSRR registers instead of SRR registers, we need to make sure we
pass it all information on where to jump to in HSRR registers.

Unfortunately, we used r10 to store the information to distinguish between
the HSRR and SRR case. That register got clobbered in between though,
rendering the later comparison invalid.

Instead, let's use cr1 to store this information. That way we don't
need yet another register and everyone's happy.

This fixes PR KVM on POWER7 bare metal for me.
Signed-off-by: NAlexander Graf <agraf@suse.de>

When running on a system that is HV capable, some interrupts use HSRR
SPRs instead of the normal SRR SPRs. These are also used in the Linux
handlers to jump back to code after an interrupt got processed.

Unfortunately, in our "jump back to the real host handler after we've
done the context switch" code, we were only setting the SRR SPRs,
rendering Linux to jump back to some invalid IP after it's processed
the interrupt.

This fixes random crashes on p7 opal mode with PR KVM for me.
Signed-off-by: NAlexander Graf <agraf@suse.de>

In addition to normal "priviledged instruction" traps, we can also receive
"emulation assist" traps on newer hardware that has the HV bit set.

Handle that one the same way as a privileged instruction, including the
instruction fetching. That way we don't execute old instructions that we
happen to still leave in that field when an emul assist trap comes.

This fixes -M mac99 / -M g3beige on p7 bare metal for me.
Signed-off-by: NAlexander Graf <agraf@suse.de>

The H_REGISTER_VPA hcall implementation in HV Power KVM needs to pin some
guest memory pages into host memory so that they can be safely accessed
from usermode.  It does this used get_user_pages_fast().  When the VPA is
unregistered, or the VCPUs are cleaned up, these pages are released using
put_page().

However, the get_user_pages() is invoked on the specific memory are of the
VPA which could lie within hugepages.  In case the pinned page is huge,
we explicitly find the head page of the compound page before calling
put_page() on it.

At least with the latest kernel, this is not correct.  put_page() already
handles finding the correct head page of a compound, and also deals with
various counts on the individual tail page which are important for
transparent huge pages.  We don't support transparent hugepages on Power,
but even so, bypassing this count maintenance can lead (when the VM ends)
to a hugepage being released back to the pool with a non-zero mapcount on
one of the tail pages.  This can then lead to a bad_page() when the page
is released from the hugepage pool.

This removes the explicit compound_head() call to correct this bug.
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>
Signed-off-by: NPaul Mackerras <paulus@samba.org>
Acked-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAvi Kivity <avi@redhat.com>

When reading and writing SPRs, every SPR emulation piece had to read
or write the respective GPR the value was read from or stored in itself.

This approach is pretty prone to failure. What if we accidentally
implement mfspr emulation where we just do "break" and nothing else?
Suddenly we would get a random value in the return register - which is
always a bad idea.

So let's consolidate the generic code paths and only give the core
specific SPR handling code readily made variables to read/write from/to.

Functionally, this patch doesn't change anything, but it increases the
readability of the code and makes is less prone to bugs.
Signed-off-by: NAlexander Graf <agraf@suse.de>

Instructions on PPC are pretty similarly encoded. So instead of
every instruction emulation code decoding the instruction fields
itself, we can move that code to more generic places and rely on
the compiler to optimize the unused bits away.

This has 2 advantages. It makes the code smaller and it makes the
code less error prone, as the instruction fields are always
available, so accidental misusage is reduced.

Functionally, this patch doesn't change anything.
Signed-off-by: NAlexander Graf <agraf@suse.de>

This is necessary for qemu to be able to pass the right information
to the guest, such as the supported page sizes and corresponding
encodings in the SLB and hash table, which can vary depending
on the processor type, the type of KVM used (PR vs HV) and the
version of KVM
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>
[agraf: fix compilation on hv, adjust for newer ioctl numbers]
Signed-off-by: NAlexander Graf <agraf@suse.de>

There is nothing in the code for emulating TCE tables in the kernel
that prevents it from working on "PR" KVM... other than ifdef's and
location of the code.

This and moves the bulk of the code there to a new file called
book3s_64_vio.c.

This speeds things up a bit on my G5.
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>
[agraf: fix for hv kvm, 32bit, whitespace]
Signed-off-by: NAlexander Graf <agraf@suse.de>

Guest r8 register is held in the scratch register and stored correctly,
so remove the instruction that clobbers it. Guest r13 was missing from vcpu,
store it there.
Signed-off-by: NMihai Caraman <mihai.caraman@freescale.com>
Signed-off-by: NAlexander Graf <agraf@suse.de>

While handling an exit, we should listen for interrupts and make sure to
receive them when they arrive, to keep our latencies low.
Signed-off-by: NAlexander Graf <agraf@suse.de>

When running on a system that is HV capable, some interrupts use HSRR
SPRs instead of the normal SRR SPRs. These are also used in the Linux
handlers to jump back to code after an interrupt got processed.

Unfortunately, in our "jump back to the real host handler after we've
done the context switch" code, we were only setting the SRR SPRs,
rendering Linux to jump back to some invalid IP after it's processed
the interrupt.

This fixes random crashes on p7 opal mode with PR KVM for me.
Signed-off-by: NAlexander Graf <agraf@suse.de>

Stbux writes the address it's operating on to the register specified in ra,
not into the data source register.
Signed-off-by: NAlexander Graf <agraf@suse.de>

Interrupt code used PPC_LL/PPC_STL macros to load/store some of u32 fields
which led to memory overflow on 64-bit. Use lwz/stw instead.
Signed-off-by: NMihai Caraman <mihai.caraman@freescale.com>
Signed-off-by: NAlexander Graf <agraf@suse.de>

While messing around with the SLBs we're running in real mode. The
entry to guest space goes through rfid, which is context synchronizing,
so there's no need to manually synchronize anything through isync.

With this patch and a simple priviledged SPR access loop guest, I get
a speed bump from 2035607 to 2181301 exits per second.
Signed-off-by: NAlexander Graf <agraf@suse.de>

By shuffling a few instructions around we can execute more memory
loads in parallel, giving us a small performance boost.

With this patch and a simple priviledged SPR access loop guest, I get
a speed bump from 2013052 to 2035607 exits per second.
Signed-off-by: NAlexander Graf <agraf@suse.de>

For Guest accessible SPRGs 4-7, save/restore must be handled differently for 64bit and
non-64 bit case. Use the PPC_STD/PPC_LD macros for saving/restoring to/from these registers.
Signed-off-by: NVarun Sethi <Varun.Sethi@freescale.com>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Introduced PPC_STD/PPC_LD macros for saving/restoring guest registers to/from their 64 bit copies.
Signed-off-by: NVarun Sethi <Varun.Sethi@freescale.com>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Time for which the hrtimer is started for decrementer emulation is calculated
using tb_ticks_per_usec. While hrtimer uses the clockevent for DEC
reprogramming (if needed) and which calculate timebase ticks using the
multiplier and shifter mechanism implemented within clockevent layer.

It was observed that this conversion (timebase->time->timebase) are not
correct because the mechanism are not consistent.
In our setup it adds 2% jitter.

With this patch clockevent multiplier and shifter mechanism are used when
starting hrtimer for decrementer emulation. Now the jitter is < 0.5%.
Signed-off-by: NBharat Bhushan <bharat.bhushan@freescale.com>
Signed-off-by: NAlexander Graf <agraf@suse.de>

Keep track of minimum and maximum address mapped by tlb1.
This helps in TLBMISS handling in KVM to quick check whether the address lies in mapped range.
If address does not lies in this range then no need to look in each tlb1 entry of tlb1 array.
Signed-off-by: NBharat Bhushan <bharat.bhushan@freescale.com>
Signed-off-by: NAlexander Graf <agraf@suse.de>

When the kernel calls into RTAS, it switches to 32-bit mode. The
magic page was is longer accessible in that case, causing the
patched instructions in the RTAS call wrapper to crash.

This fixes it by making available a 32-bit mapping of the magic
page in that case. This mapping is flushed whenever we switch
the kernel back to 64-bit mode.
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>
[agraf: add a check if the magic page is mapped]
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAvi Kivity <avi@redhat.com>

When running kvm_vcpu_block and it realizes that the CPU is actually good
to run, we get a request bit set for KVM_REQ_UNHALT. Right now, there's
nothing we can do with that bit, so let's unset it right after the call
again so we don't get confused in our later checks for pending work.
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAvi Kivity <avi@redhat.com>

When running PR KVM on a p7 system in bare metal, we get HV exits instead
of normal supervisor traps. Semantically they are identical though and the
HSRR vs SRR difference is already taken care of in the exit code.

So all we need to do is handle them in addition to our normal exits.
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAvi Kivity <avi@redhat.com>

There are 4 conditional trapping instructions: tw, twi, td, tdi. The
ones with an i take an immediate comparison, the others compare two
registers. All of them arrive in the emulator when the condition to
trap was successfully fulfilled.

Unfortunately, we were only implementing the i versions so far, so
let's also add support for the other two.

This fixes kernel booting with recents book3s_32 guest kernels.
Reported-by: NJörg Sommer <joerg@alea.gnuu.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAvi Kivity <avi@redhat.com>

When emulating updating load/store instructions (lwzu, stwu, ...) we need to
write the effective address of the load/store into a register.

Currently, we write the physical address in there, which is very wrong. So
instead let's save off where the virtual fault was on MMIO and use that
information as value to put into the register.

While at it, also move the XOP variants of the above instructions to the new
scheme of using the already known vaddr instead of calculating it themselves.
Reported-by: NJörg Sommer <joerg@alea.gnuu.de>
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAvi Kivity <avi@redhat.com>

It turns out that on POWER7, writing to the DABR can cause a corrupted
value to be written if the PMU is active and updating SDAR in continuous
sampling mode.  To work around this, we make sure that the PMU is inactive
and SDAR updates are disabled (via MMCRA) when we are context-switching
DABR.

When the guest sets DABR via the H_SET_DABR hypercall, we use a slightly
different workaround, which is to read back the DABR and write it again
if it got corrupted.

While we are at it, make it consistent that the saving and restoring
of the guest's non-volatile GPRs and the FPRs are done with the guest
setup of the PMU active.
Signed-off-by: NPaul Mackerras <paulus@samba.org>
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAvi Kivity <avi@redhat.com>

No instruction which can change Condition Register (CR) should be executed after
Guest CR is loaded. So the guest CR is restored after the Exit Timing in
lightweight_exit executes cmpw, which can clobber CR.
Signed-off-by: NBharat Bhushan <bharat.bhushan@freescale.com>
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAvi Kivity <avi@redhat.com>

This adds code to measure "stolen" time per virtual core in units of
timebase ticks, and to report the stolen time to the guest using the
dispatch trace log (DTL).  The guest can register an area of memory
for the DTL for a given vcpu.  The DTL is a ring buffer where KVM
fills in one entry every time it enters the guest for that vcpu.

Stolen time is measured as time when the virtual core is not running,
either because the vcore is not runnable (e.g. some of its vcpus are
executing elsewhere in the kernel or in userspace), or when the vcpu
thread that is running the vcore is preempted.  This includes time
when all the vcpus are idle (i.e. have executed the H_CEDE hypercall),
which is OK because the guest accounts stolen time while idle as idle
time.

Each vcpu keeps a record of how much stolen time has been reported to
the guest for that vcpu so far.  When we are about to enter the guest,
we create a new DTL entry (if the guest vcpu has a DTL) and report the
difference between total stolen time for the vcore and stolen time
reported so far for the vcpu as the "enqueue to dispatch" time in the
DTL entry.
Signed-off-by: NPaul Mackerras <paulus@samba.org>
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAvi Kivity <avi@redhat.com>

The PAPR API allows three sorts of per-virtual-processor areas to be
registered (VPA, SLB shadow buffer, and dispatch trace log), and
furthermore, these can be registered and unregistered for another
virtual CPU.  Currently we just update the vcpu fields pointing to
these areas at the time of registration or unregistration.  If this
is done on another vcpu, there is the possibility that the target vcpu
is using those fields at the time and could end up using a bogus
pointer and corrupting memory.

This fixes the race by making the target cpu itself do the update, so
we can be sure that the update happens at a time when the fields
aren't being used.  Each area now has a struct kvmppc_vpa which is
used to manage these updates.  There is also a spinlock which protects
access to all of the kvmppc_vpa structs, other than to the pinned_addr
fields.  (We could have just taken the spinlock when using the vpa,
slb_shadow or dtl fields, but that would mean taking the spinlock on
every guest entry and exit.)

This also changes 'struct dtl' (which was undefined) to 'struct dtl_entry',
which is what the rest of the kernel uses.

Thanks to Michael Ellerman <michael@ellerman.id.au> for pointing out
the need to initialize vcpu->arch.vpa_update_lock.
Signed-off-by: NPaul Mackerras <paulus@samba.org>
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAvi Kivity <avi@redhat.com>

Currently on POWER7, if we are running the guest on a core and we don't
need all the hardware threads, we do nothing to ensure that the unused
threads aren't executing in the kernel (other than checking that they
are offline).  We just assume they're napping and we don't do anything
to stop them trying to enter the kernel while the guest is running.
This means that a stray IPI can wake up the hardware thread and it will
then try to enter the kernel, but since the core is in guest context,
it will execute code from the guest in hypervisor mode once it turns the
MMU on, which tends to lead to crashes or hangs in the host.

This fixes the problem by adding two new one-byte flags in the
kvmppc_host_state structure in the PACA which are used to interlock
between the primary thread and the unused secondary threads when entering
the guest.  With these flags, the primary thread can ensure that the
unused secondaries are not already in kernel mode (i.e. handling a stray
IPI) and then indicate that they should not try to enter the kernel
if they do get woken for any reason.  Instead they will go into KVM code,
find that there is no vcpu to run, acknowledge and clear the IPI and go
back to nap mode.
Signed-off-by: NPaul Mackerras <paulus@samba.org>
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAvi Kivity <avi@redhat.com>

On PPC, CR2-CR4 are nonvolatile, thus have to be saved across function calls.
We didn't respect that for any architecture until Paul spotted it in his
patch for Book3S-HV. This patch saves/restores CR for all KVM capable PPC hosts.
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAvi Kivity <avi@redhat.com>

SPAPR support includes various in-kernel hypercalls, improving performance
by cutting out the exit to userspace.  H_BULK_REMOVE is implemented in this
patch.
Signed-off-by: NMatt Evans <matt@ozlabs.org>
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAvi Kivity <avi@redhat.com>

So far, we've always called prepare_to_enter even when all we did was return
to the host. This patch changes that semantic to only call prepare_to_enter
when we actually want to get back into the guest.
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAvi Kivity <avi@redhat.com>

When we get a performance monitor interrupt, we need to make sure that
the host receives it. So reinject it like we reinject the other host
destined interrupts.
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAvi Kivity <avi@redhat.com>

When reinjecting an interrupt into the host interrupt handler after we're
back in host kernel land, we need to tell the kernel where the interrupt
happened. We can't tell it that we were in guest state, because that might
lead to random code walking host addresses. So instead, we tell it that
we came from the interrupt reinject code.

This helps getting reasonable numbers out of perf.
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAvi Kivity <avi@redhat.com>