While doing various performance tests of reading from USB mass storage devices
I noticed the following::
1) When an async handled packet completes, we don't immediately report an
interrupt to the guest, instead we wait for the frame-timer to run and
report it from there
2) If 1) has been fixed and an async handled packet takes a while to complete,
then async_stepdown will become a high value, which means that there
will be a large latency before any new packets queued by the guest in
response to the interrupt get seen
1) was done deliberately as part of commit f0ad01f92:
http://www.kraxel.org/cgit/qemu/commit/?h=usb.57&id=f0ad01f92ca02eee7cadbfd225c5de753ebd5fce
Since setting the interrupt immediately on async packet completion was causing
issues with Linux guests, I believe this recently fixed Linux bug explains
why this is happening:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commitdiff;h=361aabf395e4a23cf554cf4ec0c0c6963b8beb01
Note that we can *not* count on this fix being present in all Linux guests!
I was hoping that the recently added support for Interrupt Threshold Control
would fix the issues with Linux guests, but adding a simple ehci_commit_irq()
call to ehci_async_bh() still caused problems with Linux guests.
The problem is, that when doing ehci_commit_irq() from ehci_async_bh(),
the "old" frindex value is used to calculate usbsts_frindex, and when
the frame-timer then runs possibly very shortly after ehci_async_bh(),
it increases the frame-timer, and thus any interrupts raised from that
frame-timer run, will also get reported to the guest immediately, rather
then being delayed to the next frame-timer run.
Luckily the solution for this is simple, this means that we need to
increase frindex before calling ehci_commit_irq() from ehci_async_bh(),
which in the end boils down to simple calling ehci_frame_timer() instead
of ehci_async_bh() from the bh.
This may seem like it causes a lot of extra work to be done, but this
is not true. Any work done from the frame-timer processing the periodic
schedule is work which then does not need to be done the next time the
frame timer runs, also the frame-timer will re-arm itself at (possibly)
a later time then it was armed for saving a vmexit at that time.
As an additional advantage moving to simply calling the frame-timer also
fixes 2) as the packet completion will set async_stepdown to 0, and the
re-arming of the timer with an async_stepdown of 0 ensures that any
newly queued up packets get seen in a reasonable amount of time.
This improves the speed (MB/s) of a Linux guest reading from a USB mass
storage device by a factor of 1.5 - 1.7 with input pipelining disabled,
and by a factor of 1.8 with input pipelining enabled.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
According to 4.15.1.2 an interrupt must be raised when a short packet
is received. If we don't do this it may take a significant time for
the guest to notice a short trasnfer has completed, since only the last td
will have its IOC flag set, and a short transfer may complete in an earlier
packet.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
This field is used in some places to track the tbytes field of the token, but
in other places the field is used directly, use it directly everywhere for
consistency.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Rather then having a special check to start queuing after the first packet,
and then another check for the other packets in uhci_fill_queue(), simply
check the previous packet beforehand in uhci_fill_queue()
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Packets with an invalid pid, or which were cancelled have
usb_packet_map() called on them on init, but not usb_packet_unmap()
before being freed.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
target_phys_addr_t is unwieldly, violates the C standard (_t suffixes are
reserved) and its purpose doesn't match the name (most target_phys_addr_t
addresses are not target specific). Replace it with a finger-friendly,
standards conformant hwaddr.
Outstanding patchsets can be fixed up with the command
git rebase -i --exec 'find -name "*.[ch]"
| xargs s/target_phys_addr_t/hwaddr/g' origin
Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
* qemu-kvm/memory/urgent:
memory: abort if a memory region is destroyed during a transaction
i440fx: avoid destroying memory regions within a transaction
memory: Make eventfd adhere to device endianness
Add multiport serial card implementation, with two variants, one
featuring two and one featuring four ports.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Split serial.c into serial.c, serial.h and serial-isa.c. While being at
creating a serial.h header file move the serial prototypes from pc.h to
the new serial.h. The latter leads to s/pc.h/serial.h/ in tons of
boards which just want the serial bits from pc.h
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
* quintela/migration-next-20121017: (41 commits)
cpus: create qemu_in_vcpu_thread()
savevm: make qemu_file_put_notify() return errors
savevm: un-export qemu_file_set_error()
block-migration: handle errors with the return codes correctly
block-migration: Switch meaning of return value
block-migration: make flush_blks() return errors
buffered_file: buffered_put_buffer() don't need to set last_error
savevm: Only qemu_fflush() can generate errors
savevm: make qemu_fill_buffer() be consistent
savevm: unexport qemu_ftell()
savevm: unfold qemu_fclose_internal()
savevm: make qemu_fflush() return an error code
savevm: Remove qemu_fseek()
virtio-net: use qemu_get_buffer() in a temp buffer
savevm: unexport qemu_fflush
migration: make migrate_fd_wait_for_unfreeze() return errors
buffered_file: make buffered_flush return the error code
buffered_file: callers of buffered_flush() already check for errors
buffered_file: We can access directly to bandwidth_limit
buffered_file: unfold migrate_fd_close
...
* qemu-kvm/memory/dma: (23 commits)
pci: honor PCI_COMMAND_MASTER
pci: give each device its own address space
memory: add address_space_destroy()
dma: make dma access its own address space
memory: per-AddressSpace dispatch
s390: avoid reaching into memory core internals
memory: use AddressSpace for MemoryListener filtering
memory: move tcg flush into a tcg memory listener
memory: move address_space_memory and address_space_io out of memory core
memory: manage coalesced mmio via a MemoryListener
xen: drop no-op MemoryListener callbacks
kvm: drop no-op MemoryListener callbacks
xen_pt: drop no-op MemoryListener callbacks
vfio: drop no-op MemoryListener callbacks
memory: drop no-op MemoryListener callbacks
memory: provide defaults for MemoryListener operations
memory: maintain a list of address spaces
memory: export AddressSpace
memory: prepare AddressSpace for exporting
xen_pt: use separate MemoryListeners for memory and I/O
...
Currently we ignore PCI_COMMAND_MASTER completely: DMA succeeds even when
the bit is clear.
Honor PCI_COMMAND_MASTER by inserting a memory region into the device's
bus master address space, and tying its enable status to PCI_COMMAND_MASTER.
Tested using
setpci -s 03 COMMAND=3
while a ping was running on a NIC in slot 3. The kernel (Linux) detected
the stall and recovered after the command
setpci -s 03 COMMAND=7
was issued.
Signed-off-by: Avi Kivity <avi@redhat.com>
Accesses from different devices can resolve differently
(depending on bridge settings, iommus, and PCI_COMMAND_MASTER), so
set up an address space for each device.
Currently iommus are expressed outside the memory API, so this doesn't
work if an iommu is present.
Signed-off-by: Avi Kivity <avi@redhat.com>
Instead of accessing the cpu address space, use an address space
configured by the caller.
Eventually all dma functionality will be folded into AddressSpace,
but we have to start from something.
Reviewed-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Using the AddressSpace type reduces confusion, as you can't accidentally
supply the MemoryRegion you're interested in.
Reviewed-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Use LOG_GUEST_ERROR rather than hw_error or direct fprintf.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Use LOG_UNIMP and LOG_GUEST_ERROR where appropriate rather
than hw_error().
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Use the new LOG_UNIMP and LOG_GUEST_ERROR logging types rather
than hw_error().
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
If the guest attempts an offset to a nonexistent register, just
log this via LOG_GUEST_ERROR rather than killing QEMU with a hw_error.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Use the new LOG_UNIMP tracing to report unimplemented
features.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Rather than a mix of direct printing to stderr and aborting
via hw_error(), use LOG_UNIMP and LOG_GUEST_ERROR.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Add an include of qemu-log.h to hw.h, so that device model
code has access to these logging functions without the need
to directly include qemu-log.h.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
This should help us to:
- More easily add or remove machine initialization arguments without
having to change every single machine init function;
- More easily make mechanical changes involving the machine init
functions in the future;
- Let machine initialization forward the init arguments to other
functions more easily.
This change was half-mechanical process: first the struct was added with
the local ram_size, boot_device, kernel_*, initrd_*, and cpu_model local
variable initialization to all functions. Then the compiler helped me
locate the local variables that are unused, so they could be removed.
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
This patch adds a mmio bar to the qemu standard vga which allows to
access the standard vga registers and bochs dispi interface registers
via mmio.
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
qemu_fseek() is known to be wrong. Would be removed on the next
commit. This code should never been used (value has been
MAC_TABLE_ENTRIES since 2009).
Signed-off-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Calling memory_region_destroy() within a transaction is illegal, since
the memory API is allowed to continue to dispatch to a region until the
transaction commits. 440fx does that however when managing PAM registers.
This bug is benign, since the regions are all aliases (which the memory
core tends to throw anyway), and since we don't do concurrent dispatch yet,
but instead of relying on that, tighten ship ahead of the coming concurrency
storm.
Fix by having a predefined set of regions, of which one will be enabled at
any time.
Signed-off-by: Avi Kivity <avi@redhat.com>
Using an unfiltered memory listener will cause regions to be reported
fails multiple times if we have more than two address spaces. Use a separate
listener for memory and I/O, and utilize MemoryListener's address space
filtering to fix this.
Signed-off-by: Avi Kivity <avi@redhat.com>
Needed for changing qemu_cpu_kick() argument type to CPUState.
Signed-off-by: Andreas Färber <afaerber@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Needed for changing cpu_kick_irq() argument type to SPARCCPU.
Signed-off-by: Andreas Färber <afaerber@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Needed for changing qemu_cpu_kick() argument type to CPUState.
Signed-off-by: Andreas Färber <afaerber@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Needed for changing cpu_kick_irq() argument type to SPARCCPU.
Signed-off-by: Andreas Färber <afaerber@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
* pmaydell/arm-devs.for-upstream:
arm_gic: Rename gic_state to GICState
zynq_slcr: Fixed ResetValues enum
versatilepb: add gpio pl061 support
hw/ds1338: Implement state save/restore
hw/ds1338: Remove 'now' field from state struct
hw/ds1338: Recapture current time when register pointer wraps around
hw/ds1338: Fix mishandling of register pointer
hw/arm_gic.c: Fix improper DPRINTF output.
cadence_ttc: Fix 'clear on read' behavior
* kraxel/usb.67:
uhci: Raise interrupt when requested even for non active tds
usb-redir: Don't make migration fail in none seamless case
usb-redir: Change usbredir_open_chardev into usbredir_create_parser
* stefanha/net:
net: consolidate NetClientState header files into one
virtio-net: update nc.link_down in virtio_net_load()
e1000: update nc.link_down in e1000_post_load()
rtl8139: implement 8139cp link status
* spice/spice.v61:
qxl: set default revision to 4
spice: raise requirement to 0.12
hw/qxl: qxl_dirty_surfaces: use uintptr_t
hw/qxl: fix condition for exiting guest_bug
hw/qxl: exit on failure to register qxl interface
qxl: fix range check for rev3 io commands.
qxl/update_area_io: cleanup invalid parameters handling
qxl: always update displaysurface on resize
