Change RTC time drift IRQ re-injection (Gleb Natapov)
Currently IRQ are reinjected as soon as they are acknowledged to
the RTC, but Windows sometimes do acknowledgement in a loop with
global interrupt disabled waiting for interrupt to be cleared and
it does not mask RTC vector in PIC/APIC while doing this. In such
situation interrupt injection always fails and RTC interrupt is never
cleared.
Instead of reinjecting coalesced IRQs on acknowledgement the patch below
reinjects them by accelerating RTC clock a bit. This way RTC interrupt
is not constantly raced after coalesced interrupt.
this patch adds a buffer_alignment field to BlockDriverState and
implements a qemu_blockalign function that uses that field to allocate a
memory aligned buffer to be used by the block driver.
buffer_alignment is initialized to 512 but each block driver can set
a different value (at the moment none of them do).
This patch modifies ide.c, block-qcow.c, block-qcow2.c and block.c to
use qemu_blockalign instead of qemu_memalign.
There is only one place left that still uses qemu_memalign to allocate
buffers used by block drivers that is posix-aio-compat:handle_aiocb_rw
because it is not possible to get the BlockDriverState from that
function. However I think it is not important because posix-aio-compat
already deals with driver specific code so it is supposed to know its
own needs.
This adds domain building support for paravirtual domains to qemu.
This allows booting xen guests directly with qemu, without Xend
and the management stack.
xen: blk & nic configuration via cmd line. (Gerd Hoffmann)
This patch makes qemu create backend and frontend device entries in
xenstore for devices configured on the command line. It will use
qdisk and qnic backend names, so the qemu internal backends will
be used.
Disks can be created using -drive if=xen,file=...
Nics can be created using -net nic,macaddr=...
This patch adds a network interface backend driver to qemu. It is a pure
userspace implemention using the gntdev interface. It uses "qnet" as
backend name in xenstore so it doesn't interfere with the netback
backend (aka "vnif").
The network backend is hooked into the corrosponding qemu vlan, i.e.
vif 0 is hooked into vlan 0. To make the packages actually arrive
somewhere you additionally have to link the vlan to the outside world
using the usual qemu command line options such as "-net tap,...".
This patch adds a block device backend driver to qemu. It is a pure
userspace implemention using the gntdev interface. It uses "qdisk" as
backend name in xenstore so it doesn't interfere with the other existing
backends (blkback aka "vbd" and tapdisk aka "tap").
This patch adds a frsamebuffer (and kbd+mouse) backend driver. It
it based on current xen-unstable code. It has been changed to make
use of the common backend driver code. It also has been changed to
compile with xen headers older than release 3.3
This patch adds a xenconsole backend driver. It it based on current
xen-unstable code. It has been changed to make use of the common
backend driver code.
This patch adds infrastructure for xen backend drivers living in qemu,
so drivers don't need to implement common stuff on their own. It's
mostly xenbus management stuff: some functions to access xentore,
setting up xenstore watches, callbacks on device discovery and state
changes, handle event channel, ...
This is purely cosmetical changes to make the code easier to read. Move L2
table processing from a deeply nested block to its own function, add some
comments.
Introduce a new bdrv_check function pointer for block drivers. Modify qcow2 to
return an error status in check_refcounts(), so it can implement bdrv_check.
This code is currently only compiled when DEBUG_ALLOC is defined, so you
usually don't see compiler warnings on it. This patch series wants to enable
the code, so fix the format string warnings first.
While we're at it, let's print error messages to stderr.
uses the QEMU firmware configuration interfacce to send the NUMA
topology to the BIOS, which has to setup the tables. Only one firmware
configuration channel is used.
adds an "info numa" command to the monitor to output the current
topology. Since NUMA is advertised via static ACPI tables, no changes are
possible during runtime.
adds a -numa command line parameter and sets a QEMU global array with
the memory sizes. The CPU-to-node assignemnt is written into the
CPUState. If no specific values for memory and CPUs are given,
all resources will be split equally across all nodes.
This code currently support only up to 64 virtual CPUs.
The slirp stack is full of global variables which prevents instantiating
it more than once. Catch this during net_slirp_init to prevent more harm
later on.
Allow to establish a TCP/UDP connection redirection also via a monitor
command 'host_net_redir'. Moreover, assume TCP as connection type if
that parameter is omitted.
Queue packets that are send during an ongoing packet delivery. This
ensures that packets will always arrive in their logical order at each
client of a VLAN. Currently, slirp generates such immediate relies, and
e.g. packet-sniffing clients on the same VLAN may get confused.
Fix the documentation of the host_net_add monitor command and allow the
user to pass no options at all. Moreover, inform the user on the
monitor terminal if a request failed.
This patch is derived from Tristan Gingold's patch. It adds a new VLAN
client type that writes all traffic on the VLAN it is attached to into a
pcap file. Such a file can then be analyzed offline with Wireshark or
tcpdump.
Besides rebasing and some minor cleanups, the major differences to the
original version are:
- support for enabling/disabling via the monitor (host_net_add/remove)
- no special ordering of VLAN client list, qemu_send_packet now takes
care of properly ordered packets
- 64k default capturing limit (I hate tcpdump's default)
slirp: Handle DHCP requests for specific IP (Jan Kiszka)
This adds proper handling of the ciaddr field as well as the "Requested
IP Address" option to slirp's DHCP server. If the client requests an
invalid or used IP, a NAK reply is sent, if it requests a specific but
valid IP, this is now respected.
NAK'ing invalid IPs is specifically useful when changing the slirp IP
range via '-net user,ip=...' while the client saved its previously used
address and tries to reacquire it. Now this will be NAK'ed and the
client will start a new discovery round.
net: Add parameter checks for VLAN clients (Jan Kiszka)
This aims at helping the user to find typos or other mistakes in
parameter lists passed for VLAN client initialization. The existing
parsing infrastructure does not allow a leaner approach, but this is
better than nothing IMHO.
In case no symbolic name is provided when requesting VLAN connection via
listening TCP socket ('-net socket,listen=...'), qemu crashes. This
fixes the cause.
Also fixes a register corruption bug in do_sigreturn. When "returning"
from sigreturn we are actually restoring the virtual cpu state from the
signal frame. This is actually surprisingly hard to observe in practice.
Typically an thread be blocked in a FUTEX_WAIT call when the signal arrives,
so the effect is a spurious syscall success and the introduction of a
subtle race condition.
On x86/arm a syscall modifies a single word sized register, so
do_sigreturn can just return that value. On MIPS a syscall clobbers
multiple registers, so we need additional smarts. My solution is to
invent a magic errno value that means "don't touch CPU state".
Adds SM501 usb host emulation feature.
It makes usb keyboard available for sh4/r2d system emulation.
The changes for "hw/usb-ohci.c" are as follows.
- 'localmem_base' is introduced as OHCIState struct member.
SM501 has a local memory, and it is used to pass and receive data with
OHCI driver. OHCI driver accesses it with SH4 physical memory address,
and SM501 accesses it with SM501 local address. 'localmem_base' holds
where the SM501 local memory is mapped into SH4 physical address space.
- Memory access functions modified to adjust address with 'localmem_base'.
The functions are, ohci_read_*(), ohci_put_*(), and ohci_copy_*().
- ohci_read_hcca() and ohci_put_hcca() are introduced for more consistent
implementation.
For other source files, it does,
- introduces usb_ohci_init_sm501().
- adds irq argument for SM501 initialization, to emulate USB interrupts.
kqemu.o is compiled even if kqemu support is disabled. This is useless
(kqemu.o should provide nothing that is actually used in that case) and
slightly confusing. So introduce CONFIG_KQEMU for optionally compiling
kqemu.o.
qemu-io: Verify read data by patterns (Kevin Wolf)
This patch adds a -P option to read and readv which allows to compare the read
data to a given pattern. This can be used to verify data written by write -P.
Align some monitor help texts to the related command parameter
definitions. host_net_add is skipped intentionally, will be slightly
reworked in a separate patch later.
x86: Enhanced dump of segment registers (Jan Kiszka)
Parse the descriptor flags that segment registers refer to and show the
result in a more human-friendly format. The output of info registers eg.
then looks like this:
Changes in this version:
- refactoring so that only a single helper is used for dumping the
segment descriptor cache
- tiny typo fixed that broke 64-bit segment type names
hpet: Fix emulation of HPET_TN_SETVAL (Jan Kiszka)
While Intel's spec is not that clear here, latest changes to Linux' HPET
code (commit c23e253e67c9d8a91a0ffa33c1f571a17f0a2403, "x86: hpet: stop
HPET_COUNTER when programming periodic mode") strongly suggest that
HPET_TN_SETVAL rather means: Set _both_ the comparator value and
register.
With this patch applied, I'm again able to boot 2.6.30-rc kernels as
they no longer panic like this (which was due to the comparator
register remaining 0):
ENABLING IO-APIC IRQs
..TIMER: vector=0x30 apic1=0 pin1=2 apic2=-1 pin2=-1
..MP-BIOS bug: 8254 timer not connected to IO-APIC
...trying to set up timer (IRQ0) through the 8259A ...
..... (found apic 0 pin 2) ...
....... failed.
...trying to set up timer as Virtual Wire IRQ...
..... failed.
...trying to set up timer as ExtINT IRQ...
..... failed :(.
Kernel panic - not syncing: IO-APIC + timer doesn't work! [...]
Fix (more or less) spurious guest boot failures due to corrupted cpuid
states. The reason was insufficient initialization of cpuid entries
before passing them to the kernel.
At this chance also fix improper entry pointer progression and simplify
the code a bit.
This patch fixes a qcow2 corruption bug introduced in SVN Rev 5861. L2 tables
are big endian, so entries must be converted before being passed to functions.
This bug is easy to trigger. The following script will create and destroy a
qcow2 image (the header is gone after three loop iterations):
#!/bin/bash
qemu-img create -f qcow2 test.qcow 1M
for i in $(seq 1 10); do
qemu-system-x86_64 -hda test.qcow -monitor stdio > /dev/null 2>&1 <<EOF
savevm test-$i
quit
EOF
done
qemu: Add support for SMBIOS command line otions (Alex Williamson)
Create a new -smbios option (x86-only) to allow binary SMBIOS entries
to be passed through to the BIOS or modify the default values of
individual fields of type 0 and 1 entries on the command line.
Binary SMBIOS entries can be generated as follows:
Don't fail PCI hotplug if no NIC model is supplied (Mark McLoughlin)
It's perfectly fine to not supply a NIC model when adding
a new NIC - we supply the default model to pci_nic_init()
and it uses that if one wasn't explicitly supplied.
In theory, there are no more quirks in the KVM slot management that
requires dirty log start/stop all over the place. We just have to start
the logging each time the mapping may have changed. This patch drops
vga_dirty_log_stop for both standard and cirrus VGA. It also reverts
#6851 as it was obviously a tribute to the old slot system.
vga: Fix inconsistent tracking of map_addr (Jan Kiszka)
Only track video RAM mapping in map_addr and use the correct RAM size.
Furthermore, make sure the reset the address in case unmapping took
place via PCI reconfiguration.
kvm: improve handling of overlapping slots (Jan Kiszka)
This reworks the slot management to handle more patterns of
cpu_register_physical_memory*, finally allowing to reset KVM guests (so
far address remapping on reset broke the slot management).
We could actually handle all possible ones without failing, but a KVM
kernel bug in older versions would force us to track all previous
fragmentations and maintain them (as that bug prevents registering
larger slots that overlap also deleted ones). To remain backward
compatible but avoid overly complicated workarounds, we apply a simpler
workaround that covers all currently used patterns.
kvm: Add sanity checks to slot management (Jan Kiszka)
Fail loudly if we run out of memory slot.
Make sure that dirty log start/stop works with consistent memory regions
by reporting invalid parameters. This reveals several inconsistencies in
the vga code, patch to fix them follows later in this series.
And, for simplicity reasons, also catch and report unaligned memory
regions passed to kvm_set_phys_mem (KVM works on page basis).
kvm: Cleanup unmap condition in kvm_set_phys_mem (Jan Kiszka)
Testing for TLB_MMIO on unmap makes no sense as A) that flag belongs to
CPUTLBEntry and not to io_memory slots or physical addresses and B) we
already use a different condition before mapping. So make this test
consistent.
The patch called "prefer glibc over direct syscalls" (commit 7118) has
replaced the getcwd syscall with a call to the glibc. With this change,
the syscall is returning -1 in error case and 0 otherwise.
This is problematic as the sys_getcwd syscall should return the number
of bytes written to the buffer including the '\0'.
check_*() functions may in fine call generate_exception(), which ends
by a call to tcg_gen_exit_tb(). As a consequence, we have to make sure
that no TCG temp variables are crossing a check_*() function.
The proper exit code for dieing from an uncaught signal is -<signal>.
The kernel doesn't allow exit() or _exit() to pass a negative value.
To get the proper exit code we need to actually die from an uncaught signal.
A default signal handler is installed, we send ourself a signal
and we wait for it to arrive.
projects / qemu.git / log