Re-arrange how we process frames / increase frnum / report pending interrupts,
to avoid a 1 ms delay in interrupt reporting to the guest. This increases
the packet throughput for cases where the guest submits a single packet,
then waits for its completion then re-submits from 500 pkts / sec to
1000 pkts / sec. This impacts for example the use of redirected / virtual
usb to serial convertors.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
ehci_raise_irq(s, USBSTS_PCD), gets applied immediately so there is no need
to call commit_irq after it.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
I tried lowering the time between raising an interrupt and rescanning the
async schedule to see if the guest has queued a new transfer before, but
that did not have any positive effect. I now believe the cause for this is
that lowering this time made it more likely to hit the 1 ms interrupt
threshold penalty for the next packet, as described in my
"ehci: Use uframe precision for interrupt threshold checking" commit.
Now that we do interrupt threshold handling with uframe precision, futher
lowering this time from .5 to .25 ms gives an extra 15% improvement in speed
(MB/s) reading from a simple USB-2.0 thumb-drive.
While at it also properly set the int_req_by_async flag for short packet
completions.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Before this patch, the following could happen:
1) Transfer completes, raises interrupt
2) .5 ms later we check if the guest has queued up any new transfers
3) We find and execute a new transfer
4) .2 ms later the new transfer completes
5) We re-run our frame_timer to write back the completion, but less then
1 ms has passed since our last run, so frindex is not changed, so the
interrupt threshold code delays the interrupt
6) 1 ms from the re-run our frame-timer runs again and finally delivers
the interrupt
This leads to unnecessary large delays of interrupts, this code fixes this
by changing frindex to uframe precision and using that for interrupt threshold
control, making the interrupt fire at step 5 for guest which have low interrupt
threshold settings (like Linux).
Note that the guest still sees the frindex move in steps of 8 for migration
compatibility.
This boosts Linux read speed of a simple cheap USB thumb drive by 6 %.
Changes in v2:
-Make the guest see frindex move in steps of 8 by modifying ehci_opreg_read,
rather then using a shadow variable
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
ehci_fill_queue assumes that there is a one on one relationship between an ep
and a qh, this patch adds a check to ensure this.
Note I don't expect this to ever trigger, this is just something I noticed
the guest might do while working on other stuff. The only way this check can
trigger is if a guest mixes in and out qtd-s in a single qh for a non
control ep.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Remove the short-circuiting of fetchqtd in fetchqh, so that the
qtd gets properly verified before completing the transaction.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
This is not allowed, except for clearing active on cancellation, so don't
warn when the new token does not have its active bit set.
This unifies the cancellation path for modified qtd-s, and prepares
ehci_verify_qtd to be used ad an extra check inside
ehci_writeback_async_complete_packet().
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Also drop the warning printf, which was there mainly because this was an
untested code path (as the previous bug fixes to it show), but that no
longer is the case now :)
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
pc-testdev.c cannot be compiled with MinGW (and other non POSIX hosts):
CC i386-softmmu/hw/i386/../pc-testdev.o
qemu/hw/i386/../pc-testdev.c:38:22: warning: sys/mman.h: file not found
qemu/hw/i386/../pc-testdev.c: In function ‘test_flush_page’:
qemu/hw/i386/../pc-testdev.c:103: warning: implicit declaration of function ‘mprotect’
...
Signed-off-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
This typically reduces the size from 512 bytes to 128 bytes.
Signed-off-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
sys/mman.h is not needed (tested on Linux) and unavailable for MinGW,
so remove it.
Signed-off-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
pc_fw_add_pflash_drv() ignores qemu_find_file() failure, and happily
creates a drive without a medium.
When pc_system_flash_init() asks for its size, bdrv_getlength() fails
with -ENOMEDIUM, which isn't checked either. It fails relatively
cleanly only because -ENOMEDIUM isn't a multiple of 4096:
$ qemu-system-x86_64 -S -vnc :0 -bios nonexistant
qemu: PC system firmware (pflash) must be a multiple of 0x1000
[Exit 1 ]
Fix by handling the qemu_find_file() failure.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Prehistoric leftover, zap it. We poweroff via acpi these days.
And having a port (0x501,0x502) where any random guest write will make
qemu exit -- with no way to turn it off -- is a bad joke anyway.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Add a test device which supports the kvmctl ioports,
so one can run the KVM unittest suite.
Intended Usage:
qemu-system-x86_64 -nographic \
-device pc-testdev \
-device isa-debug-exit,iobase=0xf4,iosize=0x04 \
-kernel /path/to/kvm/unittests/msr.flat
Where msr.flat is one of the KVM unittests, present on a
separate repo,
git://git.kernel.org/pub/scm/virt/kvm/kvm-unit-tests.git
[ kraxel: more memory api + qom fixes ]
CC: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Lucas Meneghel Rodrigues <lmr@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
When present it makes qemu exit on any write.
Mapped to port 0x501 by default.
Without this patch Anthony doesn't allow me to
remove the bochs bios debug ports because his
test suite uses this.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The hw/dataplane/vring.c code includes linux/virtio_ring.h. Ensure that
we use linux-headers/ instead of the system-wide headers, which may be
out-of-date on older distros.
This resolves the following build error on Debian 6:
CC hw/dataplane/vring.o
cc1: warnings being treated as errors
hw/dataplane/vring.c: In function 'vring_enable_notification':
hw/dataplane/vring.c:71: error: implicit declaration of function 'vring_avail_event'
hw/dataplane/vring.c:71: error: nested extern declaration of 'vring_avail_event'
hw/dataplane/vring.c:71: error: lvalue required as left operand of assignment
Note that we now build dataplane/ for each target instead of only once.
There is no way around this since linux-headers/ is only available for
per-target objects - and it's how virtio, vfio, kvm, and friends are
built.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Currently, all unknown requests are treated as VIRTIO_BLK_T_IN
Signed-off-by: Alexey Zaytsev <alexey.zaytsev@gmail.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The virtio-blk-data-plane feature is easy to integrate into
hw/virtio-blk.c. The data plane can be started and stopped similar to
vhost-net.
Users can take advantage of the virtio-blk-data-plane feature using the
new -device virtio-blk-pci,x-data-plane=on property.
The x-data-plane name was chosen because at this stage the feature is
experimental and likely to see changes in the future.
If the VM configuration does not support virtio-blk-data-plane an error
message is printed. Although we could fall back to regular virtio-blk,
I prefer the explicit approach since it prompts the user to fix their
configuration if they want the performance benefit of
virtio-blk-data-plane.
Limitations:
* Only format=raw is supported
* Live migration is not supported
* Block jobs, hot unplug, and other operations fail with -EBUSY
* I/O throttling limits are ignored
* Only Linux hosts are supported due to Linux AIO usage
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
virtio-blk-data-plane is a subset implementation of virtio-blk. It only
handles read, write, and flush requests. It does this using a dedicated
thread that executes an epoll(2)-based event loop and processes I/O
using Linux AIO.
This approach performs very well but can be used for raw image files
only. The number of IOPS achieved has been reported to be several times
higher than the existing virtio-blk implementation.
Eventually it should be possible to unify virtio-blk-data-plane with the
main body of QEMU code once the block layer and hardware emulation is
able to run outside the global mutex.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Two slightly different versions of a patch to conditionally set
VIRTIO_BLK_F_CONFIG_WCE through the "config-wce" qdev property have been
applied (ea776abca and eec7f96c2). David Gibson
<david@gibson.dropbear.id.au> noticed that the "config-wce"
property is broken as a result and fixed it recently.
The fix sets the host_features VIRTIO_BLK_F_CONFIG_WCE bit from a qdev
property. Unfortunately, the virtio device then has no chance to test
for the presence of the feature bit during virtio_blk_init().
Therefore, reinstate the VirtIOBlkConf->config_wce flag. Drop the
duplicate qdev property to set the host_features bit. The
VirtIOBlkConf->config_wce flag will be used by virtio-blk-data-plane in
a later patch.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The IOQueue has a pool of iocb structs and a function to add new
read/write requests. Multiple requests can be added before calling the
submit function to actually tell the host kernel to begin I/O. This
allows callers to batch requests and submit them in one go.
The actual I/O is performed using Linux AIO.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Outside the safety of the global mutex we need to poll on file
descriptors. I found epoll(2) is a convenient way to do that, although
other options could replace this module in the future (such as an
AioContext-based loop or glib's GMainLoop).
One important feature of this small event loop implementation is that
the loop can be terminated in a thread-safe way. This allows QEMU to
stop the data plane thread cleanly.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The virtio-blk-data-plane cannot access memory using the usual QEMU
functions since it executes outside the global mutex and the memory APIs
are this time are not thread-safe.
This patch introduces a virtqueue module based on the kernel's vhost
vring code. The trick is that we map guest memory ahead of time and
access it cheaply outside the global mutex.
Once the hardware emulation code can execute outside the global mutex it
will be possible to drop this code.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The data plane thread needs to map guest physical addresses to host
pointers. Normally this is done with cpu_physical_memory_map() but the
function assumes the global mutex is held. The data plane thread does
not touch the global mutex and therefore needs a thread-safe memory
mapping mechanism.
Hostmem registers a MemoryListener similar to how vhost collects and
pushes memory region information into the kernel. There is a
fine-grained lock on the regions list which is held during lookup and
when installing a new regions list.
When the physical memory map changes the MemoryListener callbacks are
invoked. They build up a new list of memory regions which is finally
installed when the list has been completed.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
This optimizes MSIX handling in virtio-pci.
Also included is pci express capability bugfix.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)
iQEcBAABAgAGBQJQ2tD2AAoJECgfDbjSjVRpUNcIAKN2c+3iiutUWFBBII2TWppc
QAQ4Q5HK7gCtAnwNrlQMAIXcUzHBd5s6BW74BaFBZYymf/tqe4CsvmIH15qQyvm0
McdJAba3FLk0+TELG/Fmf4+faM/kr3gl5Cve3YJC69NHpcq3gi8V4696sP8cGfUt
atA+NR8AITBJDmQlcq6Vwfp+t+B1MY9D9SROT/BmfO+/kY3krkhlPL2pdcoinBa2
zKJLz+jE0tjz7kZ99bmbb2uzKImvtFwxCVZjhD0UINjDOWd9k6ao2pWQIEftv56z
zwz/L8TKCFdM2350XXPg99f4WbrvBqmg3Slb4vrsIYEuAWvArI8sUSYG3rC4fS4=
=8Jun
-----END PGP SIGNATURE-----
Merge remote-tracking branch 'mst/tags/for_anthony' into staging
pci,virtio
This optimizes MSIX handling in virtio-pci.
Also included is pci express capability bugfix.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
* mst/tags/for_anthony:
virtio-pci: don't poll masked vectors
msix: expose access to masked/pending state
msi: add API to get notified about pending bit poll
pcie: Fix bug in pcie_ext_cap_set_next
virtio: make bindings typesafe
There are several ARM and MIPS boards which are manufactured with
either Intel (pflash_cfi01.c) or AMD (pflash_cfi02.c) flash memory.
The Linux kernel supports both and first probes for AMD flash which
resulted in one or two warnings from the Intel flash emulation:
pflash_write: Unimplemented flash cmd sequence (offset 0000000000000000, wcycle 0x0 cmd 0x0 value 0xf000f0)
pflash_write: Unimplemented flash cmd sequence (offset 0000000000000000, wcycle 0x0 cmd 0x0 value 0xf0)
These warnings confuse users, so suppress them.
Signed-off-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
* 'qom-cpu' of git://repo.or.cz/qemu/afaerber:
MAINTAINERS: Include X86CPU in CPU maintenance area
cpu: Move kvm_run into CPUState
cpu: Move kvm_state field into CPUState
ppc_booke: Pass PowerPCCPU to ppc_booke_timers_init()
ppc4xx_devs: Return PowerPCCPU from ppc4xx_init()
ppc_booke: Pass PowerPCCPU to {decr,fit,wdt} timer callbacks
ppc: Pass PowerPCCPU to [h]decr timer callbacks
ppc: Pass PowerPCCPU to [h]decr callbacks
ppc: Pass PowerPCCPU to ppc_set_irq()
kvm: Pass CPUState to kvm_vcpu_ioctl()
kvm: Pass CPUState to kvm_arch_*
cpu: Move kvm_fd into CPUState
qdev-properties.c: Separate core from the code used only by qemu-system-*
qdev: Coding style fixes
cpu: Introduce CPUListState struct
target-alpha: Add support for -cpu ?
target-alpha: Turn CPU definitions into subclasses
target-alpha: Avoid leaking the alarm timer over reset
alpha: Pass AlphaCPU array to Typhoon
target-alpha: Let cpu_alpha_init() return AlphaCPU
At the moment, when irqfd is in use but a vector is masked,
qemu will poll it and handle vector masks in userspace.
Since almost no one ever looks at the pending bits,
it is better to defer this until pending bits
are actually read.
Implement this optimization using the new poll notifier.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Upper 16 bits of the PCIe Extended Capability Header was truncated during update,
also breaking pcie_add_capability.
Signed-off-by: Knut Omang <knut.omang@oracle.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Enable 64 bits bar emulation.
Test pass with the current seabios which already support 64bit pci bars.
Signed-off-by: Xudong Hao <xudong.hao@intel.com>
Reviewed-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
* Define enum for TMP105 registers
* Move tmp105_set() from I2C to TMP105 header
* Document units and range of temperature as preconditions
Reviewed-by: Andreas Färber <afaerber@suse.de>
Signed-off-by: Alex Horn <alex.horn@cs.ox.ac.uk>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>