Commit 0d8d769085 introduced
a regression in virtio-net performance because it looks
into the ring aggressively while we really only care
about a single packet worth of buffers.
Reported as bugzilla 1066055 in launchpad.
To fix, add parameters limiting lookahead, and
use in virtqueue_avail_bytes.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reported-by: Edivaldo de Araujo Pereira <edivaldoapereira@yahoo.com.br>
Tested-by: Edivaldo de Araujo Pereira <edivaldoapereira@yahoo.com.br>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Alexander Larsson found irq injection to Windows guests stopped after a
migration. The symptom was the mouse stopped working.
Reproduction steps are:
1. On src, start qemu with a virtio-serial port without any backend
2. On dest, start qemu with a virtio-serial port with a backend
3. Migrate.
Upon migration, the older code detected the change in backend connection
status, and sent a notification to the guest. However, it's not
guaranteed that the apic is ready to inject irqs into the guest, and the
irq line remained high, resulting in any future interrupts going
unnoticed by the guest as well.
Add a new timer based on vm_clock for 1 ns in the future from post_load
to do the event send in case host_connected differs between migration
source and target.
RHBZ: 867366
Signed-off-by: Alon Levy <alevy@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com> # verbose commit log
Using the virtqueue_avail_bytes() function had an unnecessarily
crippling effect on the number of bytes needed by the guest as reported
to the chardev layer in the can_read() callback.
Using the new virtqueue_get_avail_bytes() function will let us advertise
the exact number of bytes we can send to the guest.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
* mjt/mjt-iov2:
rewrite iov_send_recv() and move it to iov.c
cleanup qemu_co_sendv(), qemu_co_recvv() and friends
export iov_send_recv() and use it in iov_send() and iov_recv()
rename qemu_sendv to iov_send, change proto and move declarations to iov.h
change qemu_iovec_to_buf() to match other to,from_buf functions
consolidate qemu_iovec_copy() and qemu_iovec_concat() and make them consistent
allow qemu_iovec_from_buffer() to specify offset from which to start copying
consolidate qemu_iovec_memset{,_skip}() into single function and use existing iov_memset()
rewrite iov_* functions
change iov_* function prototypes to be more appropriate
virtio-serial-bus: use correct lengths in control_out() message
Conflicts:
tests/Makefile
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This is far less interesting than it sounds. We simply add an Object to each
BusState and then register the types appropriately. Most of the interesting
refactoring will follow in the next patches.
Since we're changing fundamental type names (BusInfo -> BusClass), it all needs
to convert at once. Fortunately, not a lot of code is affected.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
[AF: Made all new bus TypeInfos static const.]
[AF: Made qbus_free() call object_delete(), required {qom,glib}_allocated]
Signed-off-by: Andreas Färber <afaerber@suse.de>
In qdev, each bus in practice identified an abstract superclass, but
this was mostly hidden. In QOM, instead, these abstract classes are
explicit so we can move bus properties there.
All bus property walks are removed, and all device property walks
are changed to look along the class hierarchy instead.
We would have duplicates if class A defines some properties and its
subclass B does not define any, because class_b->props will be
left equal to class_a->props.
The solution here is to reintroduce the class_base_init TypeInfo
callback, that was present in one of the early QOM versions but
removed (on my request...) before committing.
This breaks global bus properties, an obscure feature when used
with the command-line which is actually useful and used when used by
backwards-compatible machine types. So this patch also adjusts the
global bus properties in hw/pc_piix.c to refer to the abstract class.
Globals and other properties must be modified in the same patch to
avoid complications related to initialization ordering.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Simple code movement in order to simplify future refactoring.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Reorder arguments to be more natural, readable and
consistent with other iov_* functions, and change
argument names, from:
iov_from_buf(iov, iov_cnt, buf, iov_off, size)
to
iov_from_buf(iov, iov_cnt, offset, buf, bytes)
The result becomes natural English:
copy data to this `iov' vector with `iov_cnt'
elements starting at byte offset `offset'
from memory buffer `buf', processing `bytes'
bytes max.
(Try to read the original prototype this way).
Also change iov_clear() to more general iov_memset()
(it uses memset() internally anyway).
While at it, add comments to the header file
describing what the routines actually does.
The patch only renames argumens in the header, but
keeps old names in the implementation. The next
patch will touch actual code to match.
Now, it might look wrong to pay so much attention
to so small things. But we've so many badly designed
interfaces already so the whole thing becomes rather
confusing or error prone. One example of this is
previous commit and small discussion which emerged
from it, with an outcome that the utility functions
like these aren't well-understdandable, leading to
strange usage cases. That's why I paid quite some
attention to this set of functions and a few
others in subsequent patches.
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Original code has one thing to process (cur_len), requests to
convert from iovec to buf another thing (len which is actually max_len),
and processes something else (copied). Whole thing is very difficult
to understand, even if it does a right thing. The iov_to_buf()
conversion in this case will always return cur_len, because it is
the length of the iovec it was asked to process, and the size we
asked to convert is the same or larger, and iov_to_buf() will stop
at reaching either iov or buf.
Make the code saner by doing the only sane thing: dropping `copied'
which is always the same as `cur_len' but just introduces questions.
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
When a guest driver resets the virtio status to not ready, or when qemu
is reset, reset all ports' guest_connected bit and let port users know
of this event if they have the guest_close() callback registered.
Reviewed-by: Alon Levy <alevy@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
guest_connected should be false before guest driver initialization, and
true after, both for multiport aware and non multiport aware drivers.
Don't set it before the guest_features are available; instead use
set_status which is called by io to VIRTIO_PCI_STATUS with
VIRTIO_CONFIG_S_DRIVER_OK by even older non multiport drivers.
[Amit: Add comment, tweak summary, only set guest_connected and not
reset it as a side-effect.]
Signed-off-by: Alon Levy <alevy@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Replace device_init() with generalized type_init().
While at it, unify naming convention: type_init([$prefix_]register_types)
Also, type_init() is a function, so add preceding blank line where
necessary and don't put a semicolon after the closing brace.
Signed-off-by: Andreas Färber <afaerber@suse.de>
Cc: Anthony Liguori <anthony@codemonkey.ws>
Cc: malc <av1474@comtv.ru>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This was done in a mostly automated fashion. I did it in three steps and then
rebased it into a single step which avoids repeatedly touching every file in
the tree.
The first step was a sed-based addition of the parent type to the subclass
registration functions.
The second step was another sed-based removal of subclass registration functions
while also adding virtual functions from the base class into a class_init
function as appropriate.
Finally, a python script was used to convert the DeviceInfo structures and
qdev_register_subclass functions to TypeInfo structures, class_init functions,
and type_register_static calls.
We are almost fully converted to QOM after this commit.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Right now, DeviceInfo acts as the class for qdev. In order to switch to a
proper ObjectClass derivative, we need to ween all of the callers off of
interacting directly with the info pointer.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
All files under GPLv2 will get GPLv2+ changes starting tomorrow.
event_notifier.c and exec-obsolete.h were only ever touched by Red Hat
employees and can be relicensed now.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
when I tried qemu with -virtio-console pty the guest hangs and attaching
on /dev/pts/<x> does not return anything if the attachment is too late.
This results in pty_chr_write() returning 0, which causes the port to
get throttled. This results in the guest getting frozen as the
guest->host virtio_console writes don't return until the host releases
the vq element back to the guest.
For the virtio-serial use case we don't want to lose data but for the
console case we better drop data instead of "killing" the guest
console. If we get chardev->frontend notification and a better behaving
virtio-console we can revert this fix.
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
There's no need to check if ports can accept any incoming data from the
guest each time the guest sends data. Check if the port implements such
functionality during port initialisation.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Old version looks like this in info qtree (last four lines):
dev: virtconsole, id ""
dev-prop: is_console = 1
dev-prop: nr = 0
dev-prop: chardev = <null>
dev-prop: name = <null>
dev-prop-int: id: 0
dev-prop-int: guest_connected: 1
dev-prop-int: host_connected: 0
dev-prop-int: throttled: 0
Indentation is off, and "dev-prop-int" suggests these are properties
you can configure with -device, which isn't the case. The other
buses' print_dev() callbacks don't do that. For instance, PCI's
output looks like this:
class Ethernet controller, addr 00:03.0, pci id 1af4:1000 (sub 1af4:0001)
bar 0: i/o at 0xffffffffffffffff [0x1e]
bar 1: mem at 0xffffffffffffffff [0xffe]
bar 6: mem at 0xffffffffffffffff [0xfffe]
Change virtser_bus_dev_print() to that style. Result:
dev: virtconsole, id ""
dev-prop: is_console = 1
dev-prop: nr = 0
dev-prop: chardev = <null>
dev-prop: name = <null>
port 0, guest on, host off, throttle off
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
iov_to_buf() has an 'offset' parameter, iov_from_buf() hasn't.
This patch adds the missing parameter to iov_from_buf().
It also renames the 'offset' parameter to 'iov_off' to
emphasize it's the offset into the iovec and not the buffer.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Acked-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
If I start qemu with:
# qemu -hda disks/test.img -enable-kvm -m 1G -snapshot \
-device virtio-serial \
-chardev socket,host=localhost,port=1234,server,nowait,id=foo \
-device virtserialport,chardev=foo,name=org.qemu.guest_agent
I get a segfault when booting a Fedora 14 guest. The backtrace says:
Program terminated with signal 11, Segmentation fault.
#0 0x0000000000420850 in handle_control_message (vser=0x3732bd0, buf=0x2c173e0, len=8) at /home/lcapitulino/src/qmp-unstable/hw/virtio-serial-bus.c:335
335 info = DO_UPCAST(VirtIOSerialPortInfo, qdev, port->dev.info);
What's happening is VIRTIO_CONSOLE_DEVICE_READY is a message for the
whole device, not for an individual port. So port is NULL. This bug was
introduced by commit a15bb0d6a9.
This commit fixes that by making the port returned by find_port_by_id()
be used only by the VIRTIO_CONSOLE_PORT_READY and
VIRTIO_CONSOLE_PORT_OPEN messages.
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
error_report() prepends location, and appends a newline. The message
constructed from the arguments should not contain a newline. Fix the
obvious offenders.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
virtio-serial-bus needs to treat "virtconsole" devices specially. It
uses VirtIOSerialPort member is_console to recognize them. It gets
its value via property initialization. Cute hack, except it lets
users mess with it: "-device virtconsole,is_console=0" isn't plugged
into port 0 as it should.
Move the flag to VirtIOSerialPortInfo. Keep the property for backward
compatibility; its value has no effect.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
virtio_serial_init() allocates the VirtIOSerialBus dynamically, but
virtio_serial_exit() doesn't free it.
Fix by getting rid of the allocation.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Instead of calling flush_queued_data when unthrottling, schedule
a bh. That way we can return immediately to the caller, and the
flush uses the same call path as a have_data for callbackee.
No migration change is required because bh are called from vm_stop.
Signed-off-by: Alon Levy <alevy@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
The virtio serial specification requres that the values in the config
space are encoded in native endian of the guest.
The qemu virtio-serial code did not do conversion to the guest endian
format what caused problems when host and guest use different format.
This patch corrects the qemu side, correctly doing host-native <->
guest-native conversions when accessing the config space. This won't
break any setups that aren't already broken, and fixes the case
of different host and guest endianness.
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
There's no code change, just re-arrangement to simplify the function
after recent modifications.
Reported-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Instead of using a single variable to pass to the virtio_serial_init
function, use a struct so that expanding the number of variables to be
passed on later is easier.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
VirtIOSerialDevice is like VirtIOSerialPort with just the first two
fields, which makes it pretty pointless. Using VirtIOSerialPort
directly works equally well and is less confusing.
[Amit: - rebase
- rename 'dev' to 'port' in function params in virtio-serial.h ]
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This can happen if a port gets unplugged before guest has chance to
initialise vqs.
Reported-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
The new fields that got added as part of not copying over the guest
buffer to the host need to be saved/restored across migration. Do that
and bump up the version number.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
This commit lets apps signal an incomplete write. When that happens,
stop sending out any more data to the app and wait for it to unthrottle
the port.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
When the guest writes something to a host, we copied over the entire
buffer first into the host and then processed it. Do away with that, it
could result in a malicious guest causing a DoS on the host.
Reported-by: Paul Brook <paul@codesourcery.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Instead of combining flush logic into the discard case and not discard
case, have one function doing discard case. This will help later when
adding flow control logic to the do_flush_queued_data() function.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
The current default of 16 buffers for the control vq is too small. We
can get more entries in there, for example when asking the guest to add
max. allowed ports.
Note: a more robust solution would involve some kind of event queueing
in host to guarantee no event loss. Added a TODO to look into
this later.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Free malloc'ed memory, unregister from savevm and clean up virtio-common
bits on device hot-unplug.
This was found performing a migration after device hot-unplug.
Reported-by: <lihuang@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Currently virtio-serial supports a maximum of 31 ports. Specifying the
'max_ports' parameter to be > 31 on the cmd line causes badness.
Ensure we initialise virtio-serial only if max_ports is within the
supported range.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
In addition to the previous fix for calling do_flush_queued_data() only
when the virtqueue is ready, ensure do_flush_queued_data() gets a vq
that's suitably initialised.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
If a virtio-serial port is removed before the guest comes up and
initialises the virtqueues, qemu exits with the message
Guest moved used index from 0 to 61440
This happens because we try to clear any pending buffers from the
virtqueue.
Ensure the virtqueue is initialised before calling any virtqueue
operations.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
