Blue Swirl [Sat, 13 Jun 2009 08:44:31 +0000 (08:44 +0000)]
Fix mingw32 build warnings
Work around buffer and ioctlsocket argument type signedness problems
Suppress a prototype which is unused on mingw32
Expand a macro to avoid warnings from some GCC versions
Anthony Liguori [Wed, 10 Jun 2009 23:05:55 +0000 (18:05 -0500)]
Merge branch 'net-queue'
* net-queue: (28 commits)
virtio-net: Increase filter and control limits
virtio-net: Add new RX filter controls
virtio-net: MAC filter optimization
virtio-net: Fix MAC filter overflow handling
virtio-net: reorganize receive_filter()
virtio-net: Use a byte to store RX mode flags
virtio-net: Add version_id 7 placeholder for vnet header support
virtio-net: implement rx packet queueing
net: make use of async packet sending API in tap client
net: add qemu_send_packet_async()
net: split out packet queueing and flushing into separate functions
net: return status from qemu_deliver_packet()
net: add return value to packet receive handler
net: pass VLANClientState* as first arg to receive handlers
net: re-name vc->fd_read() to vc->receive()
net: add fd_readv() handler to qemu_new_vlan_client() args
net: only read from tapfd when we can send
net: vlan clients with no fd_can_read() can always receive
net: move the tap buffer into TAPState
net: factor tap_read_packet() out of tap_send()
...
Luiz Capitulino [Tue, 9 Jun 2009 21:24:57 +0000 (18:24 -0300)]
Fix "defined but not used" warning
The function qemu_calculate_timeout() is only used when CONFIG_IOTHREAD
is not defined. When CONFIG_IOTHREAD is defined, we have the following
warning:
vl.c:4389: warning: ‘qemu_calculate_timeout’ defined but not used
This change fixes that by moving the #ifdef/#endif from main_loop()
into qemu_calculate_timeout(). This encapsulates the logic and allow
us to use qemu_calculate_timeout() when CONFIG_IOTHREAD is defined
or not (suggested by Glauber Costa).
Alex Williamson [Fri, 5 Jun 2009 20:47:23 +0000 (14:47 -0600)]
virtio-net: Increase filter and control limits
Increase the size of the perfect filter table and control queue depth.
This should give us more headroom in the MAC filter and is known to be
needed by at least one guest user. Increasing the control queue depth
allows a guest to feed several commands back to back if they so desire
rather than using the send and wait approach Linux uses.
Alex Williamson [Fri, 5 Jun 2009 20:47:13 +0000 (14:47 -0600)]
virtio-net: MAC filter optimization
The MAC filter table is received from the guest as two separate
buffers, one with unicast entries, the other with multicast
entries. If we track the index dividing the two sets, we can
avoid searching the part of the table with the wrong type of
entries.
We could store this index as part of the save image, but its
trivially easy to discover it on load.
Alex Williamson [Fri, 5 Jun 2009 20:47:08 +0000 (14:47 -0600)]
virtio-net: Fix MAC filter overflow handling
Overloading the promisc and allmulti flags for indicating filter
table overflow makes it difficult to track the actual requested
operating mode. Split these out into separate flags.
Alex Williamson [Fri, 5 Jun 2009 20:47:02 +0000 (14:47 -0600)]
virtio-net: reorganize receive_filter()
Reorganize receive_filter to better handle the split between
unicast and multicast filtering. This allows us to skip the
broadcast check on unicast packets and leads to more opportunities
for optimization.
Mark McLoughlin [Wed, 29 Apr 2009 12:40:02 +0000 (13:40 +0100)]
virtio-net: implement rx packet queueing
If we don't have room to receive a packet, we return zero
from virtio_net_receive() and call qemu_flush_queued_packets()
as soon as space becomes available.
Mark McLoughlin [Wed, 29 Apr 2009 11:15:26 +0000 (12:15 +0100)]
net: add qemu_send_packet_async()
Add a qemu_send_packet() variant which will queue up the packet
if it cannot be sent when all client queues are full. It later
invokes the supplied callback when the packet has been sent.
If qemu_send_packet_async() returns zero, the caller is expected
to not send any more packets until the queued packet has been
sent.
Packets are queued iff a receive() handler returns zero (indicating
queue full) and the caller has provided a sent notification callback
(indicating it will stop and start its own queue).
We need the packet sending API to support queueing because:
- a sending client should process all available packets in one go
(e.g. virtio-net emptying its tx ring)
- a receiving client may not be able to handle the packet
(e.g. -EAGAIN from write() to tapfd)
- the sending client could detect this condition in advance
(e.g. by select() for writable on tapfd)
- that's too much overhead (e.g. a select() call per packet)
- therefore the sending client must handle the condition by
dropping the packet or queueing it
- dropping packets is poor form; we should queue.
However, we don't want queueing to be completely transparent. We
want the sending client to stop sending packets as soon as a
packet is queued. This allows the sending client to be throttled
by the receiver.
Jan Kiszka [Fri, 8 May 2009 10:34:18 +0000 (12:34 +0200)]
slirp: Reorder initialization
This patch reorders the initialization of slirp itself as well as its
associated features smb and redirection. So far the first reference to
slirp triggered the initialization, independent of the actual -net user
option which may carry additional parameters. Now we save any request to
add a smb export or some redirections until the actual initialization of
the stack. This also allows to move a few parameters that were passed
via global variable into the argument list of net_slirp_init.
Jan Kiszka [Fri, 8 May 2009 10:34:18 +0000 (12:34 +0200)]
net: Improve parameter error reporting
As host network devices can also be instantiated via the monitor, errors
should then be reported to the related monitor instead of stderr. This
requires larger refactoring, so this patch starts small with introducing
a helper to catch both cases and convert net_client_init as well as
net_slirp_redir.
Jan Kiszka [Fri, 8 May 2009 10:34:17 +0000 (12:34 +0200)]
net: Real fix for check_params users
OK, last try: 8e4416af45 broke -net socket, ffad4116b9 tried to fix it
but broke error reporting of invalid parameters. So this patch widely
reverts ffad4116b9 again and intead fixes those callers of check_params
that originally suffered from overwritten buffers by using separate
ones.
fixed that final problem, but causing us to lose some
error reporting information in the process.
Meanwhile Jan posted a patch to mostly re-do ffad4116b9
in a way that fixes the original issue, but without
losing the error reporting information. So, let's revert 8cf07dcbe7 and apply Jan's patch.
Jan Kiszka [Fri, 8 May 2009 10:34:17 +0000 (12:34 +0200)]
slirp: Avoid zombie processes after fork_exec
Slirp uses fork_exec for spawning service processes, and QEMU uses this
for running smbd. As SIGCHLD is not handled, these processes become
zombies on termination. Fix this by installing a proper signal handler,
but also make sure we disable the signal while waiting on forked network
setup/shutdown scripts.
Kevin Wolf [Thu, 4 Jun 2009 13:39:39 +0000 (15:39 +0200)]
Document changes in qemu-img interface
Update the documentation to reflect the introduction of format specific options
with -o. Don't advertise -e or -6 any more, they exist only for compatibility
reasons and can be replaced by the corresponding -o options.
Nathan Froyd [Wed, 3 Jun 2009 18:33:08 +0000 (11:33 -0700)]
fix gdbstub support for multiple threads in usermode, v3
When debugging multi-threaded programs, QEMU's gdb stub would report the
correct number of threads (the qfThreadInfo and qsThreadInfo packets).
However, the stub was unable to actually switch between threads (the T
packet), since it would report every thread except the first as being
dead. Furthermore, the stub relied upon cpu_index as a reliable means
of assigning IDs to the threads. This was a bad idea; if you have this
sequence of events:
initial thread created
new thread #1
new thread #2
thread #1 exits
new thread #3
thread #3 will have the same cpu_index as thread #1, which would confuse
GDB. (This problem is partly due to the remote protocol not having a
good way to send thread creation/destruction events.)
We fix this by using the host thread ID for the identifier passed to GDB
when debugging a multi-threaded userspace program. The thread ID might
wrap, but the same sort of problems with wrapping thread IDs would come
up with debugging programs natively, so this doesn't represent a
problem.
Jan Kiszka [Fri, 22 May 2009 21:51:45 +0000 (23:51 +0200)]
kvm: Mark full address range dirty on live migration start
As Avi correctly noted, last_ram_offset does not mark the last physical
RAM address the guest may see (due to non-continuous memory regions).
Ensure that we catch them all by marking the full possible address range
dirty.
Chris Lalancette [Mon, 25 May 2009 14:38:23 +0000 (16:38 +0200)]
Allow monitor interaction when using migrate -exec
All,
I've recently been playing around with migration via exec. Unfortunately,
when starting the incoming qemu process with "-incoming exec:cmd", it suffers
the same problem that -incoming tcp used to suffer; namely, that you can't
interact with the monitor until after the migration has happened. This causes
problems for libvirt usage of -incoming exec, since libvirt expects to be able
to access the monitor ahead of time. This fairly simple patch allows you to
access the monitor both before and after the migration has completed using exec.
(note: developed/tested with qemu-kvm, but applies perfectly fine to qemu)
Now that we have a separate aio pool structure we can remove those
aio pool details from BlockDriver.
Every driver supporting AIO now needs to declare a static AIOPool
with the aiocb size and the cancellation method. This cleans up the
current code considerably and will make it cleaner and more obvious
to support two different aio implementations behind a single
BlockDriver.
[this one is required for [PATCH] fully split aio_pool from BlockDriver,
sorry for not sending it out earlier]
Add a qcow_aio_setup helper to qcow to shared common code between
the aio_readv and aio_writev methods. Based on the function with
the same name in qcow2.
We do need hdev_create unconditionally on all platforms so that qemu-img
create support for host device works on all platforms.
Also relax the check to allow character devices in addition to block
devices. On many Unix platforms block devices have buffered block
nodes and unbuffered character device nodes, and on FreeBSD the block
nodes don't even exist anymore. Also on Linux we do support the
/dev/sgN scsi passthrough devices through the host device driver,
and probably the old-style /dev/raw/rawN raw devices although I haven't
tested that.
raw_pread_aligned currently returns the raw return value from
lseek/read, which is always -1 in case of an error. But the
callers higher up the stack expect it to return the negated
errno just like raw_pwrite_aligned.
Pointer vs addresses a VncDisplay structure,
so it is sufficient to allocate sizeof(VncDisplay)
or sizeof(*vs) bytes instead of the much larger
sizeof(VncState).
Maybe the misleading name should be fixed, too:
the code contains many places where vs is used,
sometimes it is a VncState *, sometimes it is a
VncDisplay *. vd would be a better name.
Kevin Wolf [Tue, 26 May 2009 12:36:03 +0000 (14:36 +0200)]
qcow2: Update multiple refcounts at once
Don't write each single changed refcount block entry to the disk after it is
written, but update all entries of the block and write all of them at once.
Kevin Wolf [Tue, 26 May 2009 12:36:02 +0000 (14:36 +0200)]
qcow2: Refactor update_refcount
This is a preparation patch with no functional changes. It moves the allocation
of new refcounts block to a new function and makes update_cluster_refcount (for
one cluster) call update_refcount (for multiple clusters) instead the other way
round.
Kevin Wolf [Sat, 23 May 2009 09:21:33 +0000 (11:21 +0200)]
e1000: Ignore reset command
When a reset is requested, the current e1000 emulation never clears the
reset bit which may cause a driver to hang. This patch masks the reset
bit out when setting the control registert, so the reset is immediately
completed.
Kevin Wolf [Tue, 19 May 2009 16:51:34 +0000 (18:51 +0200)]
Fix output of uninitialized strings
Commit ffad4116b96e29e0fbe892806f97c0a6c903d30d removed the "scratch buffer"
from check_params, but didn't care for the error messages which actually
included this string to tell the user which option was wrong. Now this string
is uninitialized, so this patch removes it from the message.
This means that the user is only told the whole parameter string and has to
pick the wrong option by himself as the callers of check_params can't know this
value any more. An alternative approach would be to revert that commit and do
whatever is needed to fix the original problem without changing check_params.
projects / qemu.git / log