31 Dec, 2008
40 commits
-
There is no point in doing the ready_for_nmi_injection/
request_nmi_window dance with user space. First, we don't do this for
in-kernel irqchip anyway, while the code path is the same as for user
space irqchip mode. And second, there is nothing to loose if a pending
NMI is overwritten by another one (in contrast to IRQs where we have to
save the number). Actually, there is even the risk of raising spurious
NMIs this way because the reason for the held-back NMI might already be
handled while processing the first one.Therefore this patch creates a simplified user space NMI injection
interface, exporting it under KVM_CAP_USER_NMI and dropping the old
KVM_CAP_NMI capability. And this time we also take care to provide the
interface only on archs supporting NMIs via KVM (right now only x86).Signed-off-by: Jan Kiszka
Signed-off-by: Avi Kivity -
As with the kernel irqchip, don't allow an NMI to stomp over an already
injected IRQ; instead wait for the IRQ injection to be completed.Signed-off-by: Jan Kiszka
Signed-off-by: Avi Kivity -
If an assigned device shares a guest irq with an emulated
device then we currently interpret an ack generated by the
emulated device as originating from the assigned device
leading to e.g. "Unbalanced enable for IRQ 4347" from the
enable_irq() in kvm_assigned_dev_ack_irq().The fix is fairly simple - don't enable the physical device
irq unless it was previously disabled.Of course, this can still lead to a situation where a
non-assigned device ACK can cause the physical device irq to
be reenabled before the device was serviced. However, being
level sensitive, the interrupt will merely be regenerated.Signed-off-by: Mark McLoughlin
Signed-off-by: Avi Kivity -
walk_shadow assumes the caller verified validity of the pdptr pointer in
question, which is not the case for the invlpg handler.Fixes oops during Solaris 10 install.
Signed-off-by: Marcelo Tosatti
Signed-off-by: Avi Kivity -
Signed-off-by: Avi Kivity
-
Userspace might need to act differently.
Signed-off-by: Avi Kivity
-
This changes cpus_hardware_enabled from a cpumask_t to a cpumask_var_t:
equivalent for CONFIG_CPUMASKS_OFFSTACK=n, otherwise dynamically allocated.Signed-off-by: Rusty Russell
Signed-off-by: Avi Kivity -
We're getting rid on on-stack cpumasks for large NR_CPUS.
1) Use cpumask_var_t/alloc_cpumask_var.
2) smp_call_function_mask -> smp_call_function_many
3) cpus_clear, cpus_empty, cpu_set -> cpumask_clear, cpumask_empty,
cpumask_set_cpu.This actually generates slightly smaller code than the old one with
CONFIG_CPUMASKS_OFFSTACK=n. (gcc knows that cpus cannot be NULL in
that case, where cpumask_var_t is cpumask_t[1]).Signed-off-by: Rusty Russell
Signed-off-by: Avi Kivity -
Avi said:
> Wow, code duplication from Rusty. Things must be bad.Something about glass houses comes to mind. But instead, a patch.
Signed-off-by: Rusty Russell
Signed-off-by: Avi Kivity -
There is a race between a "close of the file descriptors" and module
unload in the kvm module.You can easily trigger this problem by applying this debug patch:
>--- kvm.orig/virt/kvm/kvm_main.c
>+++ kvm/virt/kvm/kvm_main.c
>@@ -648,10 +648,14 @@ void kvm_free_physmem(struct kvm *kvm)
> kvm_free_physmem_slot(&kvm->memslots[i], NULL);
> }
>
>+#include
> static void kvm_destroy_vm(struct kvm *kvm)
> {
> struct mm_struct *mm = kvm->mm;
>
>+ printk("off1\n");
>+ msleep(5000);
>+ printk("off2\n");
> spin_lock(&kvm_lock);
> list_del(&kvm->vm_list);
> spin_unlock(&kvm_lock);and killing the userspace, followed by an rmmod.
The problem is that kvm_destroy_vm can run while the module count
is 0. That means, you can remove the module while kvm_destroy_vm
is running. But kvm_destroy_vm is part of the module text. This
causes a kerneloops. The race exists without the msleep but is much
harder to trigger.This patch requires the fix for anon_inodes (anon_inodes: use fops->owner
for module refcount).
With this patch, we can set the owner of all anonymous KVM inodes file
operations. The VFS will then control the KVM module refcount as long as there
is an open file. kvm_destroy_vm will be called by the release function of the
last closed file - before the VFS drops the module refcount.Signed-off-by: Christian Borntraeger
Signed-off-by: Avi Kivity -
There is an imbalance for anonymous inodes. If the fops->owner field is set,
the module reference count of owner is decreases on release.
("filp_close" --> "__fput" ---> "fops_put")On the other hand, anon_inode_getfd does not increase the module reference
count of owner. This causes two problems:- if owner is set, the module refcount goes negative
- if owner is not set, the module can be unloaded while code is runningThis patch changes anon_inode_getfd to be symmetric regarding fops->owner
handling.I have checked all existing users of anon_inode_getfd. Noone sets fops->owner,
thats why nobody has seen the module refcount negative. The refcounting was
tested with a patched and unpatched KVM module.(see patch 2/2) I also did an
epoll_open/close test.Signed-off-by: Christian Borntraeger
Reviewed-by: Davide Libenzi
Signed-off-by: Avi Kivity -
kvm_get_tsc_khz() currently returns the previously-calculated preset_lpj
value, but it is in loops-per-jiffy, not kHz. The current code works
correctly only when HZ=1000.Signed-off-by: Eduardo Habkost
Signed-off-by: Avi Kivity -
If the guest executes invlpg, peek into the pagetable and attempt to
prepopulate the shadow entry.Also stop dirty fault updates from interfering with the fork detector.
2% improvement on RHEL3/AIM7.
Signed-off-by: Marcelo Tosatti
Signed-off-by: Avi Kivity -
Skip syncing global pages on cr3 switch (but not on cr4/cr0). This is
important for Linux 32-bit guests with PAE, where the kmap page is
marked as global.Signed-off-by: Marcelo Tosatti
Signed-off-by: Avi Kivity -
Collapse remote TLB flushes on root sync.
kernbench is 2.7% faster on 4-way guest. Improvements have been seen
with other loads such as AIM7.Signed-off-by: Marcelo Tosatti
Signed-off-by: Avi Kivity -
Instead of invoking the handler directly collect pages into
an array so the caller can work with it.Simplifies TLB flush collapsing.
Signed-off-by: Marcelo Tosatti
Signed-off-by: Avi Kivity -
The VMMCALL instruction doesn't get recognised and isn't processed
by the emulator.This is seen on an Intel host that tries to execute the VMMCALL
instruction after a guest live migrates from an AMD host.Signed-off-by: Amit Shah
Signed-off-by: Avi Kivity -
Add emulation of shld and shrd instructions
Signed-off-by: Guillaume Thouvenin
Signed-off-by: Avi Kivity -
Add the assembler code for instruction with three operands and one
operand is stored in ECX registerSigned-off-by: Guillaume Thouvenin
Signed-off-by: Avi Kivity -
Add SrcOne operand type when we need to decode an implied '1' like with
regular shift instructionSigned-off-by: Guillaume Thouvenin
Signed-off-by: Avi Kivity -
Instruction like shld has three operands, so we need to add a Src2
decode set. We start with Src2None, Src2CL, and Src2ImmByte, Src2One to
support shld/shrd and we will expand it later.Signed-off-by: Guillaume Thouvenin
Signed-off-by: Avi Kivity -
Extend the opcode descriptor to 32 bits. This is needed by the
introduction of a new Src2 operand type.Signed-off-by: Guillaume Thouvenin
Signed-off-by: Avi Kivity -
Right now, KVM does not remove a slot when we do a
register ioctl for size 0 (would be the expected behaviour).Instead, we only mark it as empty, but keep all bitmaps
and allocated data structures present. It completely
nullifies our chances of reusing that same slot again
for mapping a different piece of memory.In this patch, we destroy rmaps, and vfree() the
pointers that used to hold the dirty bitmap, rmap
and lpage_info structures.Signed-off-by: Glauber Costa
Signed-off-by: Avi Kivity -
The only significant changes were to kvmppc_exit_timing_write() and
kvmppc_exit_timing_show(), both of which were dramatically simplified.Signed-off-by: Hollis Blanchard
Signed-off-by: Avi Kivity -
Existing KVM statistics are either just counters (kvm_stat) reported for
KVM generally or trace based aproaches like kvm_trace.
For KVM on powerpc we had the need to track the timings of the different exit
types. While this could be achieved parsing data created with a kvm_trace
extension this adds too much overhead (at least on embedded PowerPC) slowing
down the workloads we wanted to measure.Therefore this patch adds a in-kernel exit timing statistic to the powerpc kvm
code. These statistic is available per vm&vcpu under the kvm debugfs directory.
As this statistic is low, but still some overhead it can be enabled via a
.config entry and should be off by default.Since this patch touched all powerpc kvm_stat code anyway this code is now
merged and simplified together with the exit timing statistic code (still
working with exit timing disabled in .config).Signed-off-by: Christian Ehrhardt
Signed-off-by: Hollis Blanchard
Signed-off-by: Avi Kivity -
Store shadow TLB entries in memory, but only use it on host context switch
(instead of every guest entry). This improves performance for most workloads on
440 by reducing the guest TLB miss rate.Signed-off-by: Hollis Blanchard
Signed-off-by: Avi Kivity -
Formerly, we used to maintain a per-vcpu shadow TLB and on every entry to the
guest would load this array into the hardware TLB. This consumed 1280 bytes of
memory (64 entries of 16 bytes plus a struct page pointer each), and also
required some assembly to loop over the array on every entry.Instead of saving a copy in memory, we can just store shadow mappings directly
into the hardware TLB, accepting that the host kernel will clobber these as
part of the normal 440 TLB round robin. When we do that we need less than half
the memory, and we have decreased the exit handling time for all guest exits,
at the cost of increased number of TLB misses because the host overwrites some
guest entries.These savings will be increased on processors with larger TLBs or which
implement intelligent flush instructions like tlbivax (which will avoid the
need to walk arrays in software).In addition to that and to the code simplification, we have a greater chance of
leaving other host userspace mappings in the TLB, instead of forcing all
subsequent tasks to re-fault all their mappings.Signed-off-by: Hollis Blanchard
Signed-off-by: Avi Kivity -
KVM currently ignores the host's round robin TLB eviction selection, instead
maintaining its own TLB state and its own round robin index. However, by
participating in the normal 44x TLB selection, we can drop the alternate TLB
processing in KVM. This results in a significant performance improvement,
since that processing currently must be done on *every* guest exit.Accordingly, KVM needs to be able to access and increment tlb_44x_index.
(KVM on 440 cannot be a module, so there is no need to export this symbol.)Signed-off-by: Hollis Blanchard
Acked-by: Josh Boyer
Signed-off-by: Avi Kivity -
KVM on 440 has always been able to handle large guest mappings with 4K host
pages -- we must, since the guest kernel uses 256MB mappings.This patch makes KVM work when the host has large pages too (tested with 64K).
Signed-off-by: Hollis Blanchard
Signed-off-by: Avi Kivity -
Split out the logic corresponding to undoing assign_irq() and
clean it up a bit.Signed-off-by: Mark McLoughlin
Signed-off-by: Avi Kivity -
Make sure kvm_request_irq_source_id() never returns
KVM_USERSPACE_IRQ_SOURCE_ID.Likewise, check that kvm_free_irq_source_id() never accepts
KVM_USERSPACE_IRQ_SOURCE_ID.Signed-off-by: Mark McLoughlin
Signed-off-by: Avi Kivity -
Set assigned_dev->irq_source_id to -1 so that we can avoid freeing
a source ID which we never allocated.Signed-off-by: Mark McLoughlin
Signed-off-by: Avi Kivity -
We never pass a NULL notifier pointer here, but we may well
pass a notifier struct which hasn't previously been
registered.Guard against this by using hlist_del_init() which will
not do anything if the node hasn't been added to the list
and, when removing the node, will ensure that a subsequent
call to hlist_del_init() will be fine too.Fixes an oops seen when an assigned device is freed before
and IRQ is assigned to it.Signed-off-by: Mark McLoughlin
Signed-off-by: Avi Kivity -
We will obviously never pass a NULL struct kvm_irq_ack_notifier* to
this functions. They are always embedded in the assigned device
structure, so the assertion add nothing.The irqchip_in_kernel() assertion is very out of place - clearly
this little abstraction needs to know nothing about the upper
layer details.Signed-off-by: Mark McLoughlin
Signed-off-by: Avi Kivity -
Impact: make global function static
arch/x86/kvm/vmx.c:134:3: warning: symbol 'vmx_capability' was not declared. Should it be static?
Signed-off-by: Hannes Eder
Signed-off-by: Avi Kivity -
Impact: make global function static
virt/kvm/kvm_main.c:85:6: warning: symbol 'kvm_rebooting' was not declared. Should it be static?
Signed-off-by: Hannes Eder
Signed-off-by: Avi Kivity -
Notices by Guillaume Thouvenin.
Signed-off-by: Avi Kivity
-
Set operand type and size to get correct writeback behavior.
Signed-off-by: Avi Kivity
-
'ret' did not set the operand type or size for the destination, so
writeback ignored it.Signed-off-by: Avi Kivity
-
Signed-off-by: Avi Kivity
