The default action of coalesced MMIO is, cache the writing in buffer, until:
1. The buffer is full.
2. Or the exit to QEmu due to other reasons.
But this would result in a very late writing in some condition.
1. The each time write to MMIO content is small.
2. The writing interval is big.
3. No need for input or accessing other devices frequently.
This issue was observed in a experimental embbed system. The test image
simply print "test" every 1 seconds. The output in QEmu meets expectation,
but the output in KVM is delayed for seconds.
Per Avi's suggestion, I hooked flushing coalesced MMIO buffer in VGA update
handler. By this way, We don't need vcpu explicit exit to QEmu to
handle this issue.
Signed-off-by: Sheng Yang <sheng@linux.intel.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
This provides the same information as reverted commit 2ba6edf0. Not
much, just better than nothing.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
While there, use "property" rather than "option", for consistency with
-global.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Option "-device DRIVER,?" and monitor command "device_add DRIVER,?"
print the supported properties instead of creating a device. The
former also terminates the program.
This is commit 2ba6edf0 (just reverted) done right.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This reverts commit 2ba6edf0dd.
The commit has two issues:
* When it runs from the monitor, e.g. "device_add e1000,?", it prints
to stderr instead of the monitor.
* Help looks to callers just like failed device creation. This makes
main() exit unsuccessfully on "-device e1000,?".
We need to do this differently.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This reverts commit 922910ce42.
The commit has four issues:
* When it runs from the monitor, e.g. "device_add e1000,mac=?", it
prints to stderr instead of the monitor.
* Help looks to callers just like failed device creation. This makes
main() exit unsuccessfully on "-device e1000,mac=?".
* It has an undocumented side effect on -global: "-global e1000.mac=?"
prints help, but only when we actually add an e1000 device.
* It does not work for properties that accept the value "?".
We need to do this differently.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Help was shoehorned into device creation, qdev_device_add(). Since
help doesn't create a device, it returns NULL, which looks to callers
just like failed device creation. Monitor handler do_device_add()
doesn't care, but main() exits unsuccessfully.
Move help out of device creation, into new qdev_device_help().
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Return the bitwise inclusive or of all return values instead of the
last call's value. This lets you find out whether any of the calls
returned a non-zero value.
No functional change, as existing users either don't care for the
value, or pass non-zero abort_on_failure, which breaks the loop on the
first non-zero return value.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This changes the error message from "Invalid CPU index" to "Invalid
parameter index" in the human monitor.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Before, it used type 's', which strips quotes and interprets escapes,
and is quite inappropriate for QMP.
Negative arguments are no flushed to zero. Before, they were cast to
uint32_t, which wrecked the sign.
Ridiculously large arguments including infinities are now rejected.
Before, they were interpreted as zero. Same for NaN.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This is a double value with optional suffixes ms, us, ns. We'll need
this to get migrate_set_downtime() QMP-ready.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Before, it used type 's', which strips quotes and interprets escapes,
and is quite inappropriate for QMP.
Negative arguments are no flushed to zero. Before, they were cast to
uint32_t, which wrecked the sign.
Ridiculously large arguments including infinities are now rejected.
Before, they were interpreted as zero. Same for NaN.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This is a double value with optional suffixes G, g, M, m, K, k. We'll
need this to get migrate_set_speed() QMP-ready.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Helper function just like qdict_get_int(), just for QFloat/double.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Commit 392ecf543d introduced an accidental change
to the roms/seabios submodule. This commit partially reverts that commit to
ensure the seabios module points to the right commit.
Reported-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
if the access check fails, the page can not be modified
and shouldn't be marked dirty.
The patch fixes the "hsfs_putpage: dirty HSFS page"
error in Solaris guests.
Signed-off-by: Artyom Tarasenko <atar4qemu@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
If an I/O request fails right away instead of getting an error only in the
callback, we still need to consider rerror/werror.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Current code assumes that only write requests are ever going to be restarted.
This is wrong since rerror=stop exists. Instead of directly starting writes,
use the same request processing as used for new requests.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
We need a function that handles a single request. Create one by splitting out
code from virtio_blk_handle_output.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This fixes CONFIG_FB_CIRRUS for Linux guests and probably much more:
When switching away from linearly mapped vram, we also have to restore
the I/O handlers for the LFB.
This regression was once introduced by commit 2bec46dc97.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This commit enables one to use multiple virtio-serial devices and to
assign ports to arbitrary devices like this:
-device virtio-serial,id=foo -device virtio-serial,id=bar \
-device virtserialport,bus=foo.0,name=foo \
-device virtserialport,bus=bar.0,name=bar
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
When splitting up unaligned IO accesses, ld calls slow_ld which was
clobbering retaddr.
AFAIK the problem only shows up when running emulations with -icount
that may abort TB execution on IO accesses.
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>
Add has() and path_of() funcs and use them across configure; has()
will test whether a command or builtin is available; path_of() will
search the PATH for executables and return the full pathname if found.
Signed-off-by: Loïc Minier <lool@dooz.org
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Check whether sdl-config is available before calling it, otherwise
./configure triggers a warning:
./configure: 957: sdl-config: not found
If neither the .pc file not sdl-config are present, disable SDL support.
Signed-off-by: Loïc Minier <lool@dooz.org>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Function is_async_return() added by commit 940cc30d0d assumes
that 'data', which is returned by handlers, is always a QDict.
This is not true, as QLists can also be returned, in this case
we'll get a segfault.
Fix that by checking if 'data' is a QDict.
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
sparc64 timer has tick counter which can be set and read,
and tick compare value used as deadline to fire timer interrupt.
The timer is not used as periodic timer, instead deadline
is set each time new timer interrupt is needed.
v3 -> v4:
- coding style
v2 -> v3:
- added missing timer debug output macro
- CPUTimer struct and typedef moved to cpu.h
- change CPU_SAVE_VERSION to 6, older save formats not supported
v1 -> v2:
- new conversion helpers cpu_to_timer_ticks and timer_to_cpu_ticks
- save offset from clock source to implement cpu_tick_set_count
- renamed struct sun4u_timer to CPUTimer
- load and save cpu timers
v0 -> v1:
- coding style
Signed-off-by: Igor V. Kovalenko <igor.v.kovalenko@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
- correctly fit to cwp if provided window number is out of range
Signed-off-by: Igor V. Kovalenko <igor.v.kovalenko@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Commit ec229bbe7 broke invocation without a specific -hda. IOW, qemu foo.img.
The lack of an optind update caused an infinite loop.
Reported-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Buffer block writes to avoid flushing every word access onto backing
storage device. This significantly speeds up flash emulation for flashes
connected through an 8 or 16-bit bus combined with backing storage (-pflash).
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@petalogix.com>
When using ballooning to manage overcommitted memory on a host, a system for
guests to communicate their memory usage to the host can provide information
that will minimize the impact of ballooning on the guests. The current method
employs a daemon running in each guest that communicates memory statistics to a
host daemon at a specified time interval. The host daemon aggregates this
information and inflates and/or deflates balloons according to the level of
host memory pressure. This approach is effective but overly complex since a
daemon must be installed inside each guest and coordinated to communicate with
the host. A simpler approach is to collect memory statistics in the virtio
balloon driver and communicate them directly to the hypervisor.
Signed-off-by: Adam Litke <agl@us.ibm.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Qemu has a number of commands that can operate asynchronously (savevm, migrate,
etc) and it will be getting more. For these commands, the user monitor needs
to be suspended, but QMP monitors could continue to to accept other commands.
This patch introduces a new command API that isolates the details of handling
different monitor types from the actual command execution.
A monitor command can use this API by implementing the mhandler.cmd_async
handler (or info_async if appropriate). This function is responsible for
submitting the command and does not return any data although it may raise
errors. When the command completes, the QMPCompletion callback should be
invoked with its opaque data and the command result.
The process for submitting and completing an asynchronous command is different
for QMP and user monitors. A user monitor must be suspended at submit time and
resumed at completion time. The user_print() function must be passed to the
QMPCompletion callback so the result can be displayed properly. QMP monitors
are simpler. No submit time setup is required. When the command completes,
monitor_protocol_emitter() writes the result in JSON format.
This API can also be used to implement synchronous commands. In this case, the
cmd_async handler should immediately call the QMPCompletion callback. It is my
hope that this new interface will work for all commands, leading to a
drastically simplified monitor.c once all commands are ported.
Signed-off-by: Adam Litke <agl@us.ibm.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
If we go over the maximum number of iovecs support by syscall we get
back EINVAL from the kernel which translate to I/O errors for the guest.
Add a MAX_IOV defintion for platforms that don't have it. For now we use
the same 1024 define that's used on Linux and various other platforms,
but until the windows block backend implements some kind of vectored I/O
it doesn't matter.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Hi all,
this patch fixes another bug in vnc_refresh: calling vnc_update_client
might cause vs to be free()ed, in this case we cannot access vs->next
right after to examine the next item on the list.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This shouldn't happen under any normal circumstances. However, it looks like
it's possible to achieve this with corrupted images. Without this patch
raw_pread is hanging in an endless loop in such cases.
The patch is not affecting growable files, for which such reads happen in
normal use cases. raw_pread_aligned already handles these cases and won't
return zero in the first place.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Win32 suffers from a very big memory leak when dealing with SCSI devices.
Each read/write request allocates memory with qemu_memalign (ie
VirtualAlloc) but frees it with qemu_free (ie free).
Pair all qemu_memalign() calls with qemu_vfree() to prevent such leaks.
Signed-off-by: Herve Poussineau <hpoussin@reactos.org>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Toggle full screen on Win32/SDL 1.2.13 was broken by commit
c18a2c360e. Re-add the call to
do_sdl_resize() which was removed in this revision
Signed-off-by: Herve Poussineau <hpoussin@reactos.org>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Aborting without an error message when memory is short
is not helpful, so print the reason for the abort.
Try
qemu -m 1000000
or
qemu -m 2000 (win32)
to force an out-of-memory error.
v2:
* Fix error message for win32.
* Fix error message for posix_memalign.
Thanks to malc for the hints.
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
PCI bridges' qdev info structures must indicate bridge header type,
otherwise critical bridge registers (esp. PCI_PRIMARY_BUS,
PCI_SECONDARY_BUS, PCI_SUBORDINATE_BUS) will not be writable.
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
savevm without id or tag segfaults in:
(gdb) bt
#0 0x00007f600a83bf8a in __strcmp_sse42 () from /lib64/libc.so.6
#1 0x00000000004745b6 in bdrv_snapshot_find (bs=<value optimized out>,
sn_info=0x7fff996be280, name=0x0) at savevm.c:1631
#2 0x0000000000475c80 in del_existing_snapshots (name=<value optimized out>,
mon=<value optimized out>) at savevm.c:1654
#3 do_savevm (name=<value optimized out>, mon=<value optimized out>)
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
The n member is not very descriptive and very hard to grep, rename it to
cur_nr_sectors to better indicate what it is used for. Also rename
nb_sectors to remaining_sectors as that is what it is used for.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
