When starting lots of dataplane devices the bootup takes very long on
Christian's s390 with irqfd patches. With larger setups he is even
able to trigger some timeouts in some components. Turns out that the
KVM_SET_GSI_ROUTING ioctl takes very long (strace claims up to 0.1 sec)
when having multiple CPUs. This is caused by the synchronize_rcu and
the HZ=100 of s390. By changing the code to use a private srcu we can
speed things up. This patch reduces the boot time till mounting root
from 8 to 2 seconds on my s390 guest with 100 disks.
Uses of hlist_for_each_entry_rcu, hlist_add_head_rcu, hlist_del_init_rcu
are fine because they do not have lockdep checks (hlist_for_each_entry_rcu
uses rcu_dereference_raw rather than rcu_dereference, and write-sides
do not do rcu lockdep at all).
Note that we're hardly relying on the "sleepable" part of srcu. We just
want SRCU's faster detection of grace periods.
Testing was done by Andrew Theurer using netperf tests STREAM, MAERTS
and RR. The difference between results "before" and "after" the patch
has mean -0.2% and standard deviation 0.6%. Using a paired t-test on the
data points says that there is a 2.5% probability that the patch is the
cause of the performance difference (rather than a random fluctuation).
(Restricting the t-test to RR, which is the most likely to be affected,
changes the numbers to respectively -0.3% mean, 0.7% stdev, and 8%
probability that the numbers actually say something about the patch.
The probability increases mostly because there are fewer data points).
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Christian Borntraeger <borntraeger@de.ibm.com> # s390
Reviewed-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
With KVM, MMIO is much slower than PIO, due to the need to
do page walk and emulation. But with EPT, it does not have to be: we
know the address from the VMCS so if the address is unique, we can look
up the eventfd directly, bypassing emulation.
Unfortunately, this only works if userspace does not need to match on
access length and data. The implementation adds a separate FAST_MMIO
bus internally. This serves two purposes:
- minimize overhead for old userspace that does not use eventfd with lengtth = 0
- minimize disruption in other code (since we don't know the length,
devices on the MMIO bus only get a valid address in write, this
way we don't need to touch all devices to teach them to handle
an invalid length)
At the moment, this optimization only has effect for EPT on x86.
It will be possible to speed up MMIO for NPT and MMU using the same
idea in the future.
With this patch applied, on VMX MMIO EVENTFD is essentially as fast as PIO.
I was unable to detect any measureable slowdown to non-eventfd MMIO.
Making MMIO faster is important for the upcoming virtio 1.0 which
includes an MMIO signalling capability.
The idea was suggested by Peter Anvin. Lots of thanks to Gleb for
pre-review and suggestions.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
It is sometimes benefitial to ignore IO size, and only match on address.
In hindsight this would have been a better default than matching length
when KVM_IOEVENTFD_FLAG_DATAMATCH is not set, In particular, this kind
of access can be optimized on VMX: there no need to do page lookups.
This can currently be done with many ioeventfds but in a suboptimal way.
However we can't change kernel/userspace ABI without risk of breaking
some applications.
Use len = 0 to mean "ignore length for matching" in a more optimal way.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
When registering a new irqfd, we call its ->poll method to collect any
event that might have previously been pending so that we can trigger it.
This is done under the kvm->irqfds.lock, which means the eventfd's ctx
lock is taken under it.
However, if we get a POLLHUP in irqfd_wakeup, we will be called with the
ctx lock held before getting the irqfds.lock to deactivate the irqfd,
causing lockdep to complain.
Calling the ->poll method does not really need the irqfds.lock, so let's
just move it after we've given up the irqfds.lock in kvm_irqfd_assign().
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
We can easily reach the 1000 limit by start VM with a couple
hundred I/O devices (multifunction=on). The hardcode limit
already been adjusted 3 times (6 ~ 200 ~ 300 ~ 1000).
In userspace, we already have maximum file descriptor to
limit ioeventfd count. But kvm_io_bus devices also are used
for pit, pic, ioapic, coalesced_mmio. They couldn't be limited
by maximum file descriptor.
Currently only ioeventfds take too much kvm_io_bus devices,
so just exclude it from counting kvm_io_range limit.
Also fixed one indent issue in kvm_host.h
Signed-off-by: Amos Kong <akong@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Quite a bit of code in KVM has been conditionalized on availability of
IOAPIC emulation. However, most of it is generically applicable to
platforms that don't have an IOPIC, but a different type of irq chip.
Make code that only relies on IRQ routing, not an APIC itself, on
CONFIG_HAVE_KVM_IRQ_ROUTING, so that we can reuse it later.
Signed-off-by: Alexander Graf <agraf@suse.de>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Userspace may deliver RTC interrupt without query the status. So we
want to track RTC EOI for this case.
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Reviewed-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
PIO and MMIO are separate address spaces, but
ioeventfd registration code mistakenly detected
two eventfds as duplicate if they use the same address,
even if one is PIO and another one MMIO.
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Enhance KVM_IOEVENTFD with a new flag that allows to attach to virtio-ccw
devices on s390 via the KVM_VIRTIO_CCW_NOTIFY_BUS.
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Currently, eventfd introduces module_init/module_exit functions
to initialize/cleanup the irqfd workqueue. This only works, however,
if no other module_init/module_exit functions are built into the
same module.
Let's just move the initialization and cleanup to kvm_init and kvm_exit.
This way, it is also clearer where kvm startup may fail.
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
I'm not sure why, but the hlist for each entry iterators were conceived
list_for_each_entry(pos, head, member)
The hlist ones were greedy and wanted an extra parameter:
hlist_for_each_entry(tpos, pos, head, member)
Why did they need an extra pos parameter? I'm not quite sure. Not only
they don't really need it, it also prevents the iterator from looking
exactly like the list iterator, which is unfortunate.
Besides the semantic patch, there was some manual work required:
- Fix up the actual hlist iterators in linux/list.h
- Fix up the declaration of other iterators based on the hlist ones.
- A very small amount of places were using the 'node' parameter, this
was modified to use 'obj->member' instead.
- Coccinelle didn't handle the hlist_for_each_entry_safe iterator
properly, so those had to be fixed up manually.
The semantic patch which is mostly the work of Peter Senna Tschudin is here:
@@
iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host;
type T;
expression a,c,d,e;
identifier b;
statement S;
@@
-T b;
<+... when != b
(
hlist_for_each_entry(a,
- b,
c, d) S
|
hlist_for_each_entry_continue(a,
- b,
c) S
|
hlist_for_each_entry_from(a,
- b,
c) S
|
hlist_for_each_entry_rcu(a,
- b,
c, d) S
|
hlist_for_each_entry_rcu_bh(a,
- b,
c, d) S
|
hlist_for_each_entry_continue_rcu_bh(a,
- b,
c) S
|
for_each_busy_worker(a, c,
- b,
d) S
|
ax25_uid_for_each(a,
- b,
c) S
|
ax25_for_each(a,
- b,
c) S
|
inet_bind_bucket_for_each(a,
- b,
c) S
|
sctp_for_each_hentry(a,
- b,
c) S
|
sk_for_each(a,
- b,
c) S
|
sk_for_each_rcu(a,
- b,
c) S
|
sk_for_each_from
-(a, b)
+(a)
S
+ sk_for_each_from(a) S
|
sk_for_each_safe(a,
- b,
c, d) S
|
sk_for_each_bound(a,
- b,
c) S
|
hlist_for_each_entry_safe(a,
- b,
c, d, e) S
|
hlist_for_each_entry_continue_rcu(a,
- b,
c) S
|
nr_neigh_for_each(a,
- b,
c) S
|
nr_neigh_for_each_safe(a,
- b,
c, d) S
|
nr_node_for_each(a,
- b,
c) S
|
nr_node_for_each_safe(a,
- b,
c, d) S
|
- for_each_gfn_sp(a, c, d, b) S
+ for_each_gfn_sp(a, c, d) S
|
- for_each_gfn_indirect_valid_sp(a, c, d, b) S
+ for_each_gfn_indirect_valid_sp(a, c, d) S
|
for_each_host(a,
- b,
c) S
|
for_each_host_safe(a,
- b,
c, d) S
|
for_each_mesh_entry(a,
- b,
c, d) S
)
...+>
[akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c]
[akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c]
[akpm@linux-foundation.org: checkpatch fixes]
[akpm@linux-foundation.org: fix warnings]
[akpm@linux-foudnation.org: redo intrusive kvm changes]
Tested-by: Peter Senna Tschudin <peter.senna@gmail.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The current eventfd code assumes that when we have eventfd, we also have
irqfd for in-kernel interrupt delivery. This is not necessarily true. On
PPC we don't have an in-kernel irqchip yet, but we can still support easily
support eventfd.
Signed-off-by: Alexander Graf <agraf@suse.de>
Pull KVM updates from Avi Kivity:
"Highlights of the changes for this release include support for vfio
level triggered interrupts, improved big real mode support on older
Intels, a streamlines guest page table walker, guest APIC speedups,
PIO optimizations, better overcommit handling, and read-only memory."
* tag 'kvm-3.7-1' of git://git.kernel.org/pub/scm/virt/kvm/kvm: (138 commits)
KVM: s390: Fix vcpu_load handling in interrupt code
KVM: x86: Fix guest debug across vcpu INIT reset
KVM: Add resampling irqfds for level triggered interrupts
KVM: optimize apic interrupt delivery
KVM: MMU: Eliminate pointless temporary 'ac'
KVM: MMU: Avoid access/dirty update loop if all is well
KVM: MMU: Eliminate eperm temporary
KVM: MMU: Optimize is_last_gpte()
KVM: MMU: Simplify walk_addr_generic() loop
KVM: MMU: Optimize pte permission checks
KVM: MMU: Update accessed and dirty bits after guest pagetable walk
KVM: MMU: Move gpte_access() out of paging_tmpl.h
KVM: MMU: Optimize gpte_access() slightly
KVM: MMU: Push clean gpte write protection out of gpte_access()
KVM: clarify kvmclock documentation
KVM: make processes waiting on vcpu mutex killable
KVM: SVM: Make use of asm.h
KVM: VMX: Make use of asm.h
KVM: VMX: Make lto-friendly
KVM: x86: lapic: Clean up find_highest_vector() and count_vectors()
...
Conflicts:
arch/s390/include/asm/processor.h
arch/x86/kvm/i8259.c
To emulate level triggered interrupts, add a resample option to
KVM_IRQFD. When specified, a new resamplefd is provided that notifies
the user when the irqchip has been resampled by the VM. This may, for
instance, indicate an EOI. Also in this mode, posting of an interrupt
through an irqfd only asserts the interrupt. On resampling, the
interrupt is automatically de-asserted prior to user notification.
This enables level triggered interrupts to be posted and re-enabled
from vfio with no userspace intervention.
All resampling irqfds can make use of a single irq source ID, so we
reserve a new one for this interface.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Prune this down to just the struct kvm_irqfd so we can avoid
changing function definition for every flag or field we use.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Acked-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Currently the method of dealing with an IO operation on a bus (PIO/MMIO)
is to call the read or write callback for each device registered
on the bus until we find a device which handles it.
Since the number of devices on a bus can be significant due to ioeventfds
and coalesced MMIO zones, this leads to a lot of overhead on each IO
operation.
Instead of registering devices, we now register ranges which points to
a device. Lookup is done using an efficient bsearch instead of a linear
search.
Performance test was conducted by comparing exit count per second with
200 ioeventfds created on one byte and the guest is trying to access a
different byte continuously (triggering usermode exits).
Before the patch the guest has achieved 259k exits per second, after the
patch the guest does 274k exits per second.
Cc: Avi Kivity <avi@redhat.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Sasha Levin <levinsasha928@gmail.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
irqfd in kvm used flush_work incorrectly: it assumed that work scheduled
previously can't run after flush_work, but since kvm uses a non-reentrant
workqueue (by means of schedule_work) we need flush_work_sync to get that
guarantee.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reported-by: Jean-Philippe Menil <jean-philippe.menil@univ-nantes.fr>
Tested-by: Jean-Philippe Menil <jean-philippe.menil@univ-nantes.fr>
Signed-off-by: Avi Kivity <avi@redhat.com>
The RCU use in kvm_irqfd_deassign is tricky: we have rcu_assign_pointer
but no synchronize_rcu: synchronize_rcu is done by kvm_irq_routing_update
which we share a spinlock with.
Fix up a comment in an attempt to make this clearer.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Store irq routing table pointer in the irqfd object,
and use that to inject MSI directly without bouncing out to
a kernel thread.
While we touch this structure, rearrange irqfd fields to make fastpath
better packed for better cache utilization.
This also adds some comments about locking rules and rcu usage in code.
Some notes on the design:
- Use pointer into the rt instead of copying an entry,
to make it possible to use rcu, thus side-stepping
locking complexities. We also save some memory this way.
- Old workqueue code is still used for level irqs.
I don't think we DTRT with level anyway, however,
it seems easier to keep the code around as
it has been thought through and debugged, and fix level later than
rip out and re-instate it later.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Marcelo Tosatti <mtosatti@redhat.com>
Acked-by: Gregory Haskins <ghaskins@novell.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Christian Borntraeger
Michael S. Tsirkin
Cornelia Huck
Al Viro
Amos Kong
Alexander Graf
Yang Zhang
Sasha Levin
Alex Williamson
Linus Torvalds
Tejun Heo
Sasha Levin
Lucas De Marchi