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>
I think I see the following (theoretical) race:
During irqfd assign, we drop irqfds lock before we
schedule inject work. Therefore, deassign running
on another CPU could cause shutdown and flush to run
before inject, causing user after free in inject.
A simple fix it to schedule inject under the lock.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Gregory Haskins <ghaskins@novell.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
wqh is unused, so we do not need to store it in irqfd anymore
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
kvm didn't clear irqfd counter on deassign, as a result we could get a
spurious interrupt when irqfd is assigned back. this leads to poor
performance and, in theory, guest crash.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Looks like repeatedly binding same fd to multiple gsi's with irqfd can
use up a ton of kernel memory for irqfd structures.
A simple fix is to allow each fd to only trigger one gsi: triggering a
storm of interrupts in guest is likely useless anyway, and we can do it
by binding a single gsi to many interrupts if we really want to.
Cc: stable@kernel.org
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Acked-by: Gregory Haskins <ghaskins@novell.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
The only thing it protects now is interrupt injection into lapic and
this can work lockless. Even now with kvm->irq_lock in place access
to lapic is not entirely serialized since vcpu access doesn't take
kvm->irq_lock.
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
This code is not executed before file has been initialized to the result of
calling eventfd_fget. This function returns an ERR_PTR value in an error
case instead of NULL. Thus the test that file is not NULL is always true.
A simplified version of the semantic match that finds this problem is as
follows: (http://coccinelle.lip6.fr/)
// <smpl>
@match exists@
expression x, E;
statement S1, S2;
@@
x = eventfd_fget(...)
... when != x = E
(
* if (x == NULL || ...) S1 else S2
|
* if (x == NULL && ...) S1 else S2
)
// </smpl>
Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: Avi Kivity <avi@redhat.com>
ioeventfd is a mechanism to register PIO/MMIO regions to trigger an eventfd
signal when written to by a guest. Host userspace can register any
arbitrary IO address with a corresponding eventfd and then pass the eventfd
to a specific end-point of interest for handling.
Normal IO requires a blocking round-trip since the operation may cause
side-effects in the emulated model or may return data to the caller.
Therefore, an IO in KVM traps from the guest to the host, causes a VMX/SVM
"heavy-weight" exit back to userspace, and is ultimately serviced by qemu's
device model synchronously before returning control back to the vcpu.
However, there is a subclass of IO which acts purely as a trigger for
other IO (such as to kick off an out-of-band DMA request, etc). For these
patterns, the synchronous call is particularly expensive since we really
only want to simply get our notification transmitted asychronously and
return as quickly as possible. All the sychronous infrastructure to ensure
proper data-dependencies are met in the normal IO case are just unecessary
overhead for signalling. This adds additional computational load on the
system, as well as latency to the signalling path.
Therefore, we provide a mechanism for registration of an in-kernel trigger
point that allows the VCPU to only require a very brief, lightweight
exit just long enough to signal an eventfd. This also means that any
clients compatible with the eventfd interface (which includes userspace
and kernelspace equally well) can now register to be notified. The end
result should be a more flexible and higher performance notification API
for the backend KVM hypervisor and perhipheral components.
To test this theory, we built a test-harness called "doorbell". This
module has a function called "doorbell_ring()" which simply increments a
counter for each time the doorbell is signaled. It supports signalling
from either an eventfd, or an ioctl().
We then wired up two paths to the doorbell: One via QEMU via a registered
io region and through the doorbell ioctl(). The other is direct via
ioeventfd.
You can download this test harness here:
ftp://ftp.novell.com/dev/ghaskins/doorbell.tar.bz2
The measured results are as follows:
qemu-mmio: 110000 iops, 9.09us rtt
ioeventfd-mmio: 200100 iops, 5.00us rtt
ioeventfd-pio: 367300 iops, 2.72us rtt
I didn't measure qemu-pio, because I have to figure out how to register a
PIO region with qemu's device model, and I got lazy. However, for now we
can extrapolate based on the data from the NULLIO runs of +2.56us for MMIO,
and -350ns for HC, we get:
qemu-pio: 153139 iops, 6.53us rtt
ioeventfd-hc: 412585 iops, 2.37us rtt
these are just for fun, for now, until I can gather more data.
Here is a graph for your convenience:
http://developer.novell.com/wiki/images/7/76/Iofd-chart.png
The conclusion to draw is that we save about 4us by skipping the userspace
hop.
--------------------
Signed-off-by: Gregory Haskins <ghaskins@novell.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Protect irq injection/acking data structures with a separate irq_lock
mutex. This fixes the following deadlock:
CPU A CPU B
kvm_vm_ioctl_deassign_dev_irq()
mutex_lock(&kvm->lock); worker_thread()
-> kvm_deassign_irq() -> kvm_assigned_dev_interrupt_work_handler()
-> deassign_host_irq() mutex_lock(&kvm->lock);
-> cancel_work_sync() [blocked]
[gleb: fix ia64 path]
Reported-by: Alex Williamson <alex.williamson@hp.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
KVM provides a complete virtual system environment for guests, including
support for injecting interrupts modeled after the real exception/interrupt
facilities present on the native platform (such as the IDT on x86).
Virtual interrupts can come from a variety of sources (emulated devices,
pass-through devices, etc) but all must be injected to the guest via
the KVM infrastructure. This patch adds a new mechanism to inject a specific
interrupt to a guest using a decoupled eventfd mechnanism: Any legal signal
on the irqfd (using eventfd semantics from either userspace or kernel) will
translate into an injected interrupt in the guest at the next available
interrupt window.
Signed-off-by: Gregory Haskins <ghaskins@novell.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Sasha Levin
