Usually the guest can tell the host to flush data to disk. In some cases we
don't want to flush though, but try to keep everything in cache.
So let's add a new cache value to -drive that allows us to set the cache
policy to most aggressive, disabling flushes. We call this mode "unsafe",
as guest data is not guaranteed to survive host crashes anymore.
This patch also adds a noop function for aio, so we can do nothing in AIO
fashion.
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
We need to close the file even in error case. While at it, make the callers
catch all kind of errors. ENOENT is allowed for default config files, they
are optional.
Reported-by: Luiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
chardev_init functions use socket,so socket_init() shoud be placed at
the front of chardev_init on win32.
Signed-off-by: TeLeMan <geleman@gmail.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
It's emitted when the Virtual Machine resumes execution.
We currently have the STOP event but don't have the matching
RESUME one, this means that clients are notified when the VM
is stopped but don't get anything when it resumes.
Let's fix that as it's already causing some trouble to libvirt.
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Currently the commandline to create a virtual-filesystem pass-through between
the guest and the host is as follows:
#qemu -fsdev fstype,id=ID,path=path/to/share \
-device virtio-9p-pci,fsdev=ID,mount_tag=tag \
This patch provides a syntactic short-cut to achieve the same as follows:
#qemu -virtfs fstype,path=path/to/share,mount_tag=tag
This will be internally expanded as:
#qemu -fsdev fstype,id=tag,path=path/to/share, \
-device virtio-9p-pci,fsdev=tag,mount_tag=tag \
Signed-off-by: Gautham R Shenoy <ego@in.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This patch creates a new command line option named -fsdev to hold any file
system specific information.
The option will currently hold the following attributes:
-fsdev fstype id=id,path=path_to_share
where
fstype: Type of the file system.
id: Identifier used to refer to this fsdev
path: The path on the host that is identified by this fsdev.
[aneesh.kumar@linux.vnet.ibm.com: Abstraction using FsContext]
Signed-off-by: Gautham R Shenoy <ego@in.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
The boot once options seems to have gotten broken since it originally
went in. We need to wait until the second time restore_boot_devices()
gets called before restoring the standard boot order and removing itself
from the reset list.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
--
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
1) Qemu is not only a PC emulator.
2) "image image" has already been changed to "disk image" in qemu-doc.texi
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
So far a multiplexed monitor started disabled. Restore this property for
the new way of configuring by moving the monitor initialization before
all devices (the last one to attach to a char-mux will gain the focus).
Once we have a real use case for that, we may also consider assigning
the initial focus explicitly.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Canonicalize the ID assignment when creating monitor devices via the
legacy switch and use less easily colliding names.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
The 'quit' Monitor command (implemented by do_quit()) calls
exit() directly, this is problematic under QMP because QEMU
exits before having a chance to send the ok response.
Clients don't know if QEMU exited because of a problem or
because the 'quit' command has been executed.
This commit fixes that by moving the exit() call to the main
loop, so that do_quit() requests the system to quit, instead
of calling exit() directly.
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Assign directly to the bdrv_flags variable instead of using
magic numbers before translating to the BDRV_O_* options.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
What is known today as bdrv_open2 becomes the new bdrv_open. All remaining
callers of the old function are converted to the new one. In some places they
even know the right format, so they should have used bdrv_open2 from the
beginning.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Introduce a new function qemu_read_config_file which reads the VM configuration
from a config file. Unlike qemu_config_parse it doesn't take a open file but a
filename and reduces code duplication as a side effect.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This allows limited use of kvm functions (which will return ENOSYS)
even in once-compiled modules. The patch also improves a bit the error
messages for KVM initialization.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
[blauwirbel@gmail.com: fixed Win32 build]
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Both functions report errors nicely enough now, no need for additional
messages.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
In addition to removing the variable, this also renames the parse_bootdevices()
function to validate_bootdevices(), as we don't need its return value anymore.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Acked-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
There are many problems with net_set_boot_mask():
1) It is broken when using the device model instead of "-net nic". Example:
$ qemu-system-x86_64 -device rtl8139,vlan=0,id=net0,mac=52:54:00:82:41:fd,bus=pci.0,addr=0x4 -net user,vlan=0,name=hostnet0 -vnc 0.0.0.0:0 -boot n
Cannot boot from non-existent NIC
$
2) The mask was previously used to set which boot ROMs were supposed to be
loaded, but this was changed long time ago. Now all ROM images are loaded,
and SeaBIOS takes care of jumping to the right boot entry point depending on
the boot settings.
3) Interpretation and validation of the boot parameter letters is done on
the machine type code. Examples: PC accepts only a,b,c,d,n as valid boot
device letters. mac99 accepts only a,b,c,d,e,f.
As a side-effect of this change, qemu-kvm won't abort anymore if using "-boot n"
on a machine with no network devices. Checking if the requested boot device is
valid is now a task for the BIOS or the machine-type code.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Acked-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Move target specific functions and RAM handling to arch_init.c.
Add a flag to QEMUOptions structure to indicate for which
architectures the option is allowed, check the flag
in run time and remove conditional code in option handling.
Now that no target dependencies remain, compile vl.c only once
for all targets.
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Make win2k install hack unconditional as it is still restricted to
x86 only in vl.c.
Replace TARGET_PAGE_SIZE and 4096 with PAGE_SIZE.
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
This reverts commit d7234f4d7e.
Conflicts:
hw/xen_machine_pv.c
This should have never been committed.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
mkdir() only takes path argument on mingw32:
CC i386-softmmu/vl.o
/src/qemu/vl.c: In function 'qmp_add_default':
/src/qemu/vl.c:3763: error: too many arguments to function 'mkdir'
/src/qemu/vl.c:3769: error: too many arguments to function 'mkdir'
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Basically, -qmp unix:%{home}/.qemu/qmp/%{uuid}.sock,server,nowait
%{uuid} will be -uuid if it's specified, otherwise, if libuuid is available,
we generate a uuid. If it's not available, we don't create one.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Instead, we introduce a default_qmp flag. We don't use it yet, but will in the
next patch.
This has a user-visible impact as specifying just -qmp will now also show a
monitor on the 'vc'.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
All of these users have global state so we really don't see a benefit from
exit_notifier. However, using exit_notifier means that there's one less
justification for having global state in the first place.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Just tell main_loop_wait whether to be blocking or nonblocking, so that
there is no need to call qemu_cpus_have_work from the timer subsystem.
Instead, tcg_cpu_exec can say "we want the main loop not to block because
we have stuff to do".
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Tweaking the rounding in qemu_next_deadline ensures that there's
no change whatsoever.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
A simple patch to place together all handling of -icount.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
By adding the possibility to turn on/off a clock, yet another
incestuous relationship between timers and CPUs can be disentangled.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Make the timer subsystem register its own callback instead.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Instead of testing specially next_cpu in host_alarm_handler, just do
that in qemu_notify_event. The idea is, if we are not running (or
not yet running) target CPU code, prepare things so that the execution
loop is exited asap; just make that clear.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
qemu_notify_event in the non-iothread case is only stopping the current
CPU. However, if the CPU is idle and the main loop is in the select
call then a call to qemu_event_increment is needed too (as done in
host_alarm_handler). Since in general one doesn't know whether the CPU
is executing or not, it is a safe bet to always do qemu_event_increment.
Another way to see it: after this patch qemu_event_increment is the
"common part" of qemu_notify_event for both the CONFIG_IOTHREAD and
!CONFIG_IOTHREAD cases, which makes sense.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
The timer_alarm_pending variable is related to the alarm timer but not
placed in the struct. Also, in qemu_mod_timer the wrong flag was being
tested: the timer is rearmed in the alarm timer "bottom half", so the
right flag to test there is the "pending" flag.
Finally, I hoisted the NULL checks from alarm_has_dynticks to
host_alarm_handler.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
The ALARM_FLAG_DYNTICKS can be testing simply by checking if there is
a rearm function.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
The TIME_ONESHOT and TIME_PERIODIC flags are mutually exclusive.
The code after the patch matches the flags used in win32_start_timer.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
The code is initializing an unsigned int to UINT_MAX using "-1", so that
the following always-true comparison seems to be always-false at a
first look. Since alarm timer initializations are never nested, it is
simpler to unconditionally store the result of timeGetDevCaps into
data->period.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
We sometimes permit omitting the first option name, for example
-device foo is short for -device driver=foo. The name to use
("driver" in the example) is passed as argument to qemu_opts_parse().
For each QemuOptsList, we use at most one such name.
Move the name into QemuOptsList, and pass whether to permit the
abbreviation. This ensures continued consistency, and simplifies the
commit after next in this series.
New LOC_CMDLINE. Use it for tracking option with argument in
lookup_opt(). We now report errors like this
qemu: -device smbus-eeprom: Did not find I2C bus for smbus-eeprom
New LOC_FILE. Use it for tracking file name and line number in
qemu_config_parse(). We now report errors like
qemu:foo.conf:42: Did not find I2C bus for smbus-eeprom
In particular, gems like this message:
-device: no driver specified
become almost nice now:
qemu:foo.conf:44: -device: no driver specified
(A later commit will get rid of the bogus -device:)
error_report() terminates the message with a newline. Strip it it
from its arguments.
This fixes a few error messages lacking a newline:
net_handle_fd_param()'s "No file descriptor named %s found", and
tap_open()'s "vnet_hdr=1 requested, but no kernel support for
IFF_VNET_HDR available" (all three versions).
There's one place that passes arguments without newlines
intentionally: load_vmstate(). Fix it up.
qemu_error_sink can either point to a monitor or a file. In practice,
it always points to the current monitor if we have one, else to
stderr. Simply route errors to the current monitor or else to stderr,
and remove qemu_error_sink along with the functions to control it.
Actually, the old code switches the sink slightly later, in
handle_user_command() and handle_qmp_command(), than it gets switched
now, implicitly, by setting the current monitor in monitor_read() and
monitor_control_read(). Likewise, it switches back slightly earlier
(same places). Doesn't make a difference, because there are no calls
of qemu_error() in between.
Something bad has happened in the merge of commit 0ee44250, as
the log message says it's supposed to be in qemu_system_reset()
but it is do_vm_stop().
Possibly, it was a problem with the conflict resolution with
ea375f9a (which has been merged first).
This commit moves (again) the RESET event into qemu_system_reset().
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This event has been introduced in the first round of QMP commits,
turns out that it's based on the usage of the EXCP_DEBUG macro,
which has discussable semantics when exposed through QMP.
As libvirt doesn't use this, let's just drop it.
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Nothing will change as that function is currently only called by
the main loop code, but it's the right place for the RESET event,
as it's where the reset is actually performed.
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
I've introduced the STOP event in the main loop, this is wrong
as it will be only emitted if the io thread is enabled.
This fixes that by moving the STOP event to do_vm_stop().
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This grand cleanup drops all reset and vmsave/load related
synchronization points in favor of four(!) generic hooks:
- cpu_synchronize_all_states in qemu_savevm_state_complete
(initial sync from kernel before vmsave)
- cpu_synchronize_all_post_init in qemu_loadvm_state
(writeback after vmload)
- cpu_synchronize_all_post_init in main after machine init
- cpu_synchronize_all_post_reset in qemu_system_reset
(writeback after system reset)
These writeback points + the existing one of VCPU exec after
cpu_synchronize_state map on three levels of writeback:
- KVM_PUT_RUNTIME_STATE (during runtime, other VCPUs continue to run)
- KVM_PUT_RESET_STATE (on synchronous system reset, all VCPUs stopped)
- KVM_PUT_FULL_STATE (on init or vmload, all VCPUs stopped as well)
This level is passed to the arch-specific VCPU state writing function
that will decide which concrete substates need to be written. That way,
no writer of load, save or reset functions that interact with in-kernel
KVM states will ever have to worry about synchronization again. That
also means that a lot of reasons for races, segfaults and deadlocks are
eliminated.
cpu_synchronize_state remains untouched, just as Anthony suggested. We
continue to need it before reading or writing of VCPU states that are
also tracked by in-kernel KVM subsystems.
Consequently, this patch removes many cpu_synchronize_state calls that
are now redundant, just like remaining explicit register syncs.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Port qemu-kvm's -mem-path and -mem-prealloc options. These are useful
for backing guest memory with huge pages via hugetlbfs.
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
CC: john cooper <john.cooper@redhat.com>
This is a reimplementation of prior versions which adds
the ability to define cpu models for contemporary processors.
The added models are likewise selected via -cpu <name>,
and are intended to displace the existing convention
of "-cpu qemu64" augmented with a series of feature flags.
A primary motivation was determination of a least common
denominator within a given processor class to simplify guest
migration. It is still possible to modify an arbitrary model
via additional feature flags however the goal here was to
make doing so unnecessary in typical usage. The other
consideration was providing models names reflective of
current processors. Both AMD and Intel have reviewed the
models in terms of balancing generality of migration vs.
excessive feature downgrade relative to released silicon.
This version of the patch replaces the prior hard wired
definitions with a configuration file approach for new
models. Existing models are thus far left as-is but may
easily be transitioned to (or may be overridden by) the
configuration file representation.
Proposed new model definitions are provided here for current
AMD and Intel processors. Each model consists of a name
used to select it on the command line (-cpu <name>), and a
model_id which corresponds to a least common denominator
commercial instance of the processor class.
A table of names/model_ids may be queried via "-cpu ?model":
:
x86 Opteron_G3 AMD Opteron 23xx (Gen 3 Class Opteron)
x86 Opteron_G2 AMD Opteron 22xx (Gen 2 Class Opteron)
x86 Opteron_G1 AMD Opteron 240 (Gen 1 Class Opteron)
x86 Nehalem Intel Core i7 9xx (Nehalem Class Core i7)
x86 Penryn Intel Core 2 Duo P9xxx (Penryn Class Core 2)
x86 Conroe Intel Celeron_4x0 (Conroe/Merom Class Core 2)
:
Also added is "-cpu ?dump" which exhaustively outputs all config
data for all defined models, and "-cpu ?cpuid" which enumerates
all qemu recognized CPUID feature flags.
The pseudo cpuid flag 'check' when added to the feature flag list
will warn when feature flags (either implicit in a cpu model or
explicit on the command line) would have otherwise been quietly
unavailable to a guest:
# qemu-system-x86_64 ... -cpu Nehalem,check
warning: host cpuid 0000_0001 lacks requested flag 'sse4.2|sse4_2' [0x00100000]
warning: host cpuid 0000_0001 lacks requested flag 'popcnt' [0x00800000]
A similar 'enforce' pseudo flag exists which in addition
to the above causes qemu to error exit if requested flags are
unavailable.
Configuration data for a cpu model resides in the target config
file which by default will be installed as:
/usr/local/etc/qemu/target-<arch>.conf
The format of this file should be self explanatory given the
definitions for the above six models and essentially mimics
the structure of the static x86_def_t x86_defs.
Encoding of cpuid flags names now allows aliases for both the
configuration file and the command line which reconciles some
Intel/AMD/Linux/Qemu naming differences.
This patch was tested relative to qemu.git.
Signed-off-by: john cooper <john.cooper@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Change the way the internal qemu signal, used for communication between
iothread and vcpus, is handled.
Block and consume it with sigtimedwait on the outer vcpu loop, which
allows more precise timing control.
Change from standard signal (SIGUSR1) to real-time one, so multiple
signals are not collapsed.
Set the signal number on KVM's in-kernel allowed sigmask.
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
In KVM mode the global mutex is released when vcpus are executing,
which means acquiring the fairness mutex is not required.
Also for KVM there is one thread per vcpu, so tcg_has_work is meaningless.
Add a new qemu_wait_io_event_common function to hold common code
between TCG/KVM.
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Otherwise a vcpu thread can run the sigchild handler causing
waitpid() from iothread to fail.
Signed-off-by: Marcelo Tosatti <mtosa...@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Moving stuff in console.c to avoid the need for prototypes makes
this patch a bit bigger, but there's no change in the code.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Let register_displayallocator hand over the old width/height to the new
allocator.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Ensure initialization of a dumb display, if needed, by making
all accesses go through get_displaystate.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Call it right after -device devices get created.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Round robin vcpus in tcg_cpu_next even if the vm stopped. This
allows all cpus to enter stopped state.
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
No need to loop if less than a full buffer is read, the next
read would return EAGAIN.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Some places use get_clock directly because they want to access the
rt_clock with nanosecond precision. Add a function to do exactly that
instead of using internal interfaces.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Similar to the qemu-img.c patch, but I also have to unescape remaining
% signs in qemu-options.hx.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: malc <av1474@comtv.ru>
Since qemu-options.h is only used in vl.c, we can avoid using
brittle interpolation from a generated file.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: malc <av1474@comtv.ru>