Until now, KVM/arm didn't care much for page aging (who was swapping
anyway?), and simply provided empty hooks to the core KVM code. With
server-type systems now being available, things are quite different.
This patch implements very simple support for page aging, by clearing
the Access flag in the Stage-2 page tables. On access fault, the current
fault handling will write the PTE or PMD again, putting the Access flag
back on.
It should be possible to implement a much faster handling for Access
faults, but that's left for a later patch.
With this in place, performance in VMs is degraded much more gracefully.
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Acked-by: Christoffer Dall <christoffer.dall@linaro.org>
Signed-off-by: Christoffer Dall <christoffer.dall@linaro.org>
In order to be able to detect the point where the guest enables
its MMU and caches, trap all the VM related system registers.
Once we see the guest enabling both the MMU and the caches, we
can go back to a saner mode of operation, which is to leave these
registers in complete control of the guest.
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Reviewed-by: Christoffer Dall <christoffer.dall@linaro.org>
So far, KVM/ARM used a fixed HCR configuration per guest, except for
the VI/VF/VA bits to control the interrupt in absence of VGIC.
With the upcoming need to dynamically reconfigure trapping, it becomes
necessary to allow the HCR to be changed on a per-vcpu basis.
The fix here is to mimic what KVM/arm64 already does: a per vcpu HCR
field, initialized at setup time.
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Reviewed-by: Christoffer Dall <christoffer.dall@linaro.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Update comments to reflect what is really going on and add the TWE bit
to the comments in kvm_arm.h.
Also renames the function to kvm_handle_wfx like is done on arm64 for
consistency and uber-correctness.
Signed-off-by: Christoffer Dall <christoffer.dall@linaro.org>
On an (even slightly) oversubscribed system, spinlocks are quickly
becoming a bottleneck, as some vcpus are spinning, waiting for a
lock to be released, while the vcpu holding the lock may not be
running at all.
This creates contention, and the observed slowdown is 40x for
hackbench. No, this isn't a typo.
The solution is to trap blocking WFEs and tell KVM that we're
now spinning. This ensures that other vpus will get a scheduling
boost, allowing the lock to be released more quickly. Also, using
CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT slightly improves the performance
when the VM is severely overcommited.
Quick test to estimate the performance: hackbench 1 process 1000
2xA15 host (baseline): 1.843s
2xA15 guest w/o patch: 2.083s
4xA15 guest w/o patch: 80.212s
8xA15 guest w/o patch: Could not be bothered to find out
2xA15 guest w/ patch: 2.102s
4xA15 guest w/ patch: 3.205s
8xA15 guest w/ patch: 6.887s
So we go from a 40x degradation to 1.5x in the 2x overcommit case,
which is vaguely more acceptable.
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <christoffer.dall@linaro.org>
The T{0,1}SZ fields of TTBCR are 3 bits wide when using the long descriptor
format. Likewise, the T0SZ field of the HTCR is 3-bits. KVM currently
defines TTBCR_T{0,1}SZ as 3, not 7.
The T0SZ mask is used to calculate the value for the HTCR, both to pick out
TTBCR.T0SZ and mask off the equivalent field in the HTCR during
read-modify-write. The incorrect mask size causes the (UNKNOWN) reset value
of HTCR.T0SZ to leak in to the calculated HTCR value. Linux will hang when
initializing KVM if HTCR's reset value has bit 2 set (sometimes the case on
A7/TC2)
Fixing T0SZ allows A7 cores to boot and T1SZ is also fixed for completeness.
Signed-off-by: Jonathan Austin <jonathan.austin@arm.com>
Acked-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <christoffer.dall@linaro.org>
S2_PGD_SIZE defines the number of pages used by a stage-2 PGD
and is unused, except for a VM_BUG_ON check that missuses the
define.
As the check is very unlikely to ever triggered except in
circumstances where KVM is the least of our worries, just kill
both the define and the VM_BUG_ON check.
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <cdall@cs.columbia.edu>
Bit 8 is cache maintenance, bit 9 is external abort.
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <cdall@cs.columbia.edu>
When the guest accesses I/O memory this will create data abort
exceptions and they are handled by decoding the HSR information
(physical address, read/write, length, register) and forwarding reads
and writes to QEMU which performs the device emulation.
Certain classes of load/store operations do not support the syndrome
information provided in the HSR. We don't support decoding these (patches
are available elsewhere), so we report an error to user space in this case.
This requires changing the general flow somewhat since new calls to run
the VCPU must check if there's a pending MMIO load and perform the write
after userspace has made the data available.
Reviewed-by: Will Deacon <will.deacon@arm.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <c.dall@virtualopensystems.com>
Adds a new important function in the main KVM/ARM code called
handle_exit() which is called from kvm_arch_vcpu_ioctl_run() on returns
from guest execution. This function examines the Hyp-Syndrome-Register
(HSR), which contains information telling KVM what caused the exit from
the guest.
Some of the reasons for an exit are CP15 accesses, which are
not allowed from the guest and this commit handles these exits by
emulating the intended operation in software and skipping the guest
instruction.
Minor notes about the coproc register reset:
1) We reserve a value of 0 as an invalid cp15 offset, to catch bugs in our
table, at cost of 4 bytes per vcpu.
2) Added comments on the table indicating how we handle each register, for
simplicity of understanding.
Reviewed-by: Will Deacon <will.deacon@arm.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Christoffer Dall <c.dall@virtualopensystems.com>
Provides complete world-switch implementation to switch to other guests
running in non-secure modes. Includes Hyp exception handlers that
capture necessary exception information and stores the information on
the VCPU and KVM structures.
The following Hyp-ABI is also documented in the code:
Hyp-ABI: Calling HYP-mode functions from host (in SVC mode):
Switching to Hyp mode is done through a simple HVC #0 instruction. The
exception vector code will check that the HVC comes from VMID==0 and if
so will push the necessary state (SPSR, lr_usr) on the Hyp stack.
- r0 contains a pointer to a HYP function
- r1, r2, and r3 contain arguments to the above function.
- The HYP function will be called with its arguments in r0, r1 and r2.
On HYP function return, we return directly to SVC.
A call to a function executing in Hyp mode is performed like the following:
<svc code>
ldr r0, =BSYM(my_hyp_fn)
ldr r1, =my_param
hvc #0 ; Call my_hyp_fn(my_param) from HYP mode
<svc code>
Otherwise, the world-switch is pretty straight-forward. All state that
can be modified by the guest is first backed up on the Hyp stack and the
VCPU values is loaded onto the hardware. State, which is not loaded, but
theoretically modifiable by the guest is protected through the
virtualiation features to generate a trap and cause software emulation.
Upon guest returns, all state is restored from hardware onto the VCPU
struct and the original state is restored from the Hyp-stack onto the
hardware.
SMP support using the VMPIDR calculated on the basis of the host MPIDR
and overriding the low bits with KVM vcpu_id contributed by Marc Zyngier.
Reuse of VMIDs has been implemented by Antonios Motakis and adapated from
a separate patch into the appropriate patches introducing the
functionality. Note that the VMIDs are stored per VM as required by the ARM
architecture reference manual.
To support VFP/NEON we trap those instructions using the HPCTR. When
we trap, we switch the FPU. After a guest exit, the VFP state is
returned to the host. When disabling access to floating point
instructions, we also mask FPEXC_EN in order to avoid the guest
receiving Undefined instruction exceptions before we have a chance to
switch back the floating point state. We are reusing vfp_hard_struct,
so we depend on VFPv3 being enabled in the host kernel, if not, we still
trap cp10 and cp11 in order to inject an undefined instruction exception
whenever the guest tries to use VFP/NEON. VFP/NEON developed by
Antionios Motakis and Rusty Russell.
Aborts that are permission faults, and not stage-1 page table walk, do
not report the faulting address in the HPFAR. We have to resolve the
IPA, and store it just like the HPFAR register on the VCPU struct. If
the IPA cannot be resolved, it means another CPU is playing with the
page tables, and we simply restart the guest. This quirk was fixed by
Marc Zyngier.
Reviewed-by: Will Deacon <will.deacon@arm.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Antonios Motakis <a.motakis@virtualopensystems.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <c.dall@virtualopensystems.com>
All interrupt injection is now based on the VM ioctl KVM_IRQ_LINE. This
works semantically well for the GIC as we in fact raise/lower a line on
a machine component (the gic). The IOCTL uses the follwing struct.
struct kvm_irq_level {
union {
__u32 irq; /* GSI */
__s32 status; /* not used for KVM_IRQ_LEVEL */
};
__u32 level; /* 0 or 1 */
};
ARM can signal an interrupt either at the CPU level, or at the in-kernel irqchip
(GIC), and for in-kernel irqchip can tell the GIC to use PPIs designated for
specific cpus. The irq field is interpreted like this:
bits: | 31 ... 24 | 23 ... 16 | 15 ... 0 |
field: | irq_type | vcpu_index | irq_number |
The irq_type field has the following values:
- irq_type[0]: out-of-kernel GIC: irq_number 0 is IRQ, irq_number 1 is FIQ
- irq_type[1]: in-kernel GIC: SPI, irq_number between 32 and 1019 (incl.)
(the vcpu_index field is ignored)
- irq_type[2]: in-kernel GIC: PPI, irq_number between 16 and 31 (incl.)
The irq_number thus corresponds to the irq ID in as in the GICv2 specs.
This is documented in Documentation/kvm/api.txt.
Reviewed-by: Will Deacon <will.deacon@arm.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Christoffer Dall <c.dall@virtualopensystems.com>
Sets up KVM code to handle all exceptions taken to Hyp mode.
When the kernel is booted in Hyp mode, calling an hvc instruction with r0
pointing to the new vectors, the HVBAR is changed to the the vector pointers.
This allows subsystems (like KVM here) to execute code in Hyp-mode with the
MMU disabled.
We initialize other Hyp-mode registers and enables the MMU for Hyp-mode from
the id-mapped hyp initialization code. Afterwards, the HVBAR is changed to
point to KVM Hyp vectors used to catch guest faults and to switch to Hyp mode
to perform a world-switch into a KVM guest.
Also provides memory mapping code to map required code pages, data structures,
and I/O regions accessed in Hyp mode at the same virtual address as the host
kernel virtual addresses, but which conforms to the architectural requirements
for translations in Hyp mode. This interface is added in arch/arm/kvm/arm_mmu.c
and comprises:
- create_hyp_mappings(from, to);
- create_hyp_io_mappings(from, to, phys_addr);
- free_hyp_pmds();
Reviewed-by: Will Deacon <will.deacon@arm.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <c.dall@virtualopensystems.com>
Targets KVM support for Cortex A-15 processors.
Contains all the framework components, make files, header files, some
tracing functionality, and basic user space API.
Only supported core is Cortex-A15 for now.
Most functionality is in arch/arm/kvm/* or arch/arm/include/asm/kvm_*.h.
Reviewed-by: Will Deacon <will.deacon@arm.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <c.dall@virtualopensystems.com>
