monitor: Use QERR_INVALID_PARAMETER instead of QERR_INVALID_CPU_INDEX

This changes the error message from "Invalid CPU index" to "Invalid
parameter index" in the human monitor.

Signed-off-by: Markus Armbruster <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

monitor: convert do_migrate_set_downtime() to QObject

Signed-off-by: Markus Armbruster <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

monitor: Use argument type 'T' for migrate_set_downtime

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 <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

monitor: New argument type 'T'

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 <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

monitor: convert do_migrate_set_speed() to QObject

Signed-off-by: Markus Armbruster <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

monitor: Use argument type 'b' for migrate_set_speed

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 <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

monitor: New argument type 'b'

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 <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

QDict: New qdict_get_double()

Helper function just like qdict_get_int(), just for QFloat/double.

Signed-off-by: Markus Armbruster <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

monitor: Document argument type 'M'

Was forgotten in commit b6e098d7.

Signed-off-by: Markus Armbruster <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

seabios: use correct submodule hash

Commit 392ecf543dd2b219828374850d833296a55567ea 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 <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

bsd/darwin-user: mmap_frag() users only check for -1 error

See also ee636500d6eab44b83f09cb730b67226b70423b1.

Signed-off-by: Blue Swirl <[email protected]>

cris: Prettify register dumps.

Use friendly names for registers in the dumps.

Signed-off-by: Edgar E. Iglesias <[email protected]>

sparc32 don't mark page dirty when failing

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 <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

sparc64: implement basic IOMMU/interrupt handling

Also add reset control.

Signed-off-by: Blue Swirl <[email protected]>

Sparc64: byte swap IO port access

This requires an updated OpenBIOS image.

Signed-off-by: Blue Swirl <[email protected]>

virtio-blk: Fix error cases which ignored rerror/werror

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 <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

virtio-blk: Fix restart after read error

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 <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

virtio_blk: Factor virtio_blk_handle_request out

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 <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

cirrus: Properly re-register cirrus_linear_io_addr on vram unmap

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 <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

virtio-serial-bus: Fix bus initialisation and allow for bus identification

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 <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

softmmu: Dont clobber retaddr in slow_ldx().

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 <[email protected]>

Solaris: test for presence of commands with has()

Signed-off-by: Loïc Minier <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

Add and use has() and path_of() funcs

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 <[email protected]
Signed-off-by: Blue Swirl <[email protected]>

Check for sdl-config before calling it

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 <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

Monitor: Fix command execution regression

Function is_async_return() added by commit 940cc30d0d4 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 <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

sparc64: reimplement tick timers v4

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 <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

sparc64: correct write extra bits to cwp

- correctly fit to cwp if provided window number is out of range

Signed-off-by: Igor V. Kovalenko <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

Fix BSD build

<sys/wait.h> must be included in order to use WIF* macros.

Signed-off-by: Blue Swirl <[email protected]>

Fix regression in option parsing

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 <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

pflash: Buffer block writes

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 <[email protected]>

virtio: Add memory statistics reporting to the balloon driver

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 <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

New API for asynchronous monitor commands

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 <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

block: avoid creating too large iovecs in multiwrite_merge

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 <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

vnc_refresh: calling vnc_update_client might free vs

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 <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

block/raw-posix: Abort on pread beyond end of non-growable file

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 <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

win32: pair qemu_memalign() with qemu_vfree()

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 <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

win32/sdl: Fix toggle full screen

Toggle full screen on Win32/SDL 1.2.13 was broken by commit
c18a2c360e3100bbd71162cf922dcd8c429a8b71. Re-add the call to
do_sdl_resize() which was removed in this revision

Signed-off-by: Herve Poussineau <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

Tell users about out-of-memory errors

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 <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

PCI: fix bridge configuration

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 <[email protected]>

fix savevm command without id or tag

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 <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

qcow2: rename two QCowAIOCB members

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 <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

block: kill BDRV_O_CREAT

The BDRV_O_CREAT option is unused inside qemu and partially duplicates
the bdrv_create method.  Remove it, and the -C option to qemu-io which
isn't used in qemu-iotests anyway.

Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

block: clean up bdrv_open2 structure a bit

Check the whitelist as early as possible instead of continuing the
setup, and move all the error handling code to the end of the
function.

Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

virtio-console: Automatically use virtio-serial-bus for the older -virtioconsole invocation

These hunks got dropped off mysteriously during the rebasing of my
virtio-serial series. Thanks go to Markus for noticing it.

Without these fixes, -virtioconsole doesn't actually have any effect.

Signed-off-by: Amit Shah <[email protected]>
Reported-by: Markus Armbruster <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

virtio-console: Rename virtio-serial.c back to virtio-console.c

This file was renamed to ease the reviews of the recent changes
that went in.

Now that the changes are done, rename the file back to its original
name.

Signed-off-by: Amit Shah <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

Fix generation of config-host.h

This patch improves Anthony patch a6a853c86275efd89996ce59612a000c5873db5d

Once there, it improves handling of object files for qemu tools

cc: Andreas Färber <[email protected]>
Signed-off-by: Juan Quintela <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

ide save/restore current transfer fields

If migration takes place between write of the bmdma address register and
write of the command register (to initiate DMA), the destination will
not properly start the DMA op, hanging the guest:

ata1.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x6 frozen
ata1.00: cmd c8/00:16:41:00:00/00:00:00:00:00/e0 tag 0 dma 11264 in
         res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 (timeout)
ata1.00: status: { DRDY }

Fix by sending current transfer information in the migration data.

We need to update ide version to 4 for this to work.  As we don't
have subsectios, we need to chain the update increase until
vmstate_ide_pci (quintela)

Signed-off-by: Marcelo Tosatti <[email protected]>
Signed-off-by: Juan Quintela <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

PCI: do_pci_info(): PCI bridge devices support

This commit completes the do_pci_info() conversion to
QObject by adding support to PCI bridge devices.

This is done by recursively adding devices in the
"pci_bridge" key.

IMPORTANT: This code is being added separately because I could
NOT test it properly. According to Michael Tsirkin, it depends
on ultrasparc and it would take time to do the proper setup.

Signed-off-by: Luiz Capitulino <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

PCI: Convert pci_info() to QObject

The returned QObject is a QList of all buses. Each bus is
represented by a QDict, which has a key with a QList of all
PCI devices attached to it. Each device is represented by
a QDict.

As has happended to other complex conversions, it's hard to
split this commit as part of it are new functions which are
called by each other.

IMPORTANT: support for printing PCI bridge attached devices
is NOT part of this commit, it's going to be added by the
next commit, as it's untested.

Signed-off-by: Luiz Capitulino <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

QDict: Introduce qdict_get_qdict()

A helper to retrieve a QDict from a QDict.

Signed-off-by: Luiz Capitulino <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

QList: Introduce QLIST_FOREACH_ENTRY()

Iterate over QList entries, it's needed to call qlist_entry_obj()
to retrieve the stored QObject.

I'm not sure if it's ok to have this, because it's not as easy as
qlist_iter() and the QListEntry data type is now exposed to the
users, which means we have one more struct to be maintained when
we have libqmp.

Adding anyway, as it's more compact and people are asking for it.

Signed-off-by: Luiz Capitulino <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

Read-only device changed to opens it's file for read-only.

Signed-off-by: Naphtali Sprei <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

Ask for read-write permissions when opening files

Found some places that seems needs this explicitly, now that
read-write is not the default.

Signed-off-by: Naphtali Sprei <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

No need anymoe for bdrv_set_read_only

Signed-off-by: Naphtali Sprei <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

Documentation: Add missing texi description for command line options

Some more command line options had entries for command line help,
but documentation for texi and derived formats (man, html, info)
was missing.

For conditional options, the texi documentation was added
unconditionally.

This seems reasonable because typically man pages are
shared, and html users expect to see one documentation
(not several nearly identical documents for the different
systems).

Signed-off-by: Stefan Weil <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

Documentation: Improve command line help for -device option

* Fix column for help text.

* Give some more help, especially for the new '?' parameters.

Signed-off-by: Stefan Weil <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

qdev: Add help for property value

When called with property value "?",
a help text will be printed (instead of an error message).

This is useful for command lines like
     qemu -device e1000,mac=?
and is already standard for other command line options.

A better help text could be provided by extending
the Property structure with a desc field.

Signed-off-by: Stefan Weil <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

qdev: Add help for device properties

When called with property "?", a list of supported
properties will be printed (instead of an error message).

This is useful for command lines like
qemu -device e1000,?
and was already standard for other options like model=?

Signed-off-by: Stefan Weil <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

mmap_frag() users only check for -1 error

Signed-off-by: Juan Quintela <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

Check availavility of -fstack-protector-all

Signed-off-by: Juan Quintela <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

Enable _FORTIFY_SOURCE=2

_FORTIFY_SOURCE is a Glibc feature which adds memory and string function
protection.

Signed-off-by: Kirill A. Shutemov <[email protected]>
Signed-off-by: Juan Quintela <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

check pipe() return value

Signed-off-by: Juan Quintela <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

linux-user/mmap.c: fix warnings with _FORTIFY_SOURCE

CC    i386-linux-user/mmap.o
cc1: warnings being treated as errors
/usr/src/RPM/BUILD/qemu-0.11.92/linux-user/mmap.c: In function 'mmap_frag':
/usr/src/RPM/BUILD/qemu-0.11.92/linux-user/mmap.c:253: error: ignoring return value of 'pread', declared with attribute warn_unused_result
/usr/src/RPM/BUILD/qemu-0.11.92/linux-user/mmap.c: In function 'target_mmap':
/usr/src/RPM/BUILD/qemu-0.11.92/linux-user/mmap.c:477: error: ignoring return value of 'pread', declared with attribute warn_unused_result
make[1]: *** [mmap.o] Error 1

Signed-off-by: Kirill A. Shutemov <[email protected]>
Signed-off-by: Juan Quintela <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

monitor.c: fix warnings with _FORTIFY_SOURCE

CC    i386-softmmu/monitor.o
cc1: warnings being treated as errors
/usr/src/RPM/BUILD/qemu-0.11.92/monitor.c: In function 'do_memory_save':
/usr/src/RPM/BUILD/qemu-0.11.92/monitor.c:1318: error: ignoring return value of 'fwrite', declared with attribute warn_unused_result
/usr/src/RPM/BUILD/qemu-0.11.92/monitor.c: In function 'do_physical_memory_save':
/usr/src/RPM/BUILD/qemu-0.11.92/monitor.c:1345: error: ignoring return value of 'fwrite', declared with attribute warn_unused_result
make[1]: *** [monitor.o] Error 1

Signed-off-by: Kirill A. Shutemov <[email protected]>
Signed-off-by: Juan Quintela <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

vl.c: fix warning with _FORTIFY_SOURCE

CC    i386-softmmu/vl.o
cc1: warnings being treated as errors
/usr/src/RPM/BUILD/qemu-0.11.92/vl.c: In function 'qemu_event_increment':
/usr/src/RPM/BUILD/qemu-0.11.92/vl.c:3404: error: ignoring return value of 'write', declared with attribute warn_unused_result
/usr/src/RPM/BUILD/qemu-0.11.92/vl.c: In function 'main':
/usr/src/RPM/BUILD/qemu-0.11.92/vl.c:5774: error: ignoring return value of 'write', declared with attribute warn_unused_result
/usr/src/RPM/BUILD/qemu-0.11.92/vl.c:6064: error: ignoring return value of 'chdir', declared with attribute warn_unused_result
/usr/src/RPM/BUILD/qemu-0.11.92/vl.c:6083: error: ignoring return value of 'chdir', declared with attribute warn_unused_result
make[1]: *** [vl.o] Error 1

Signed-off-by: Kirill A. Shutemov <[email protected]>
Signed-off-by: Juan Quintela <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

usb-linux.c: fix warning with _FORTIFY_SOURCE

CC    usb-linux.o
cc1: warnings being treated as errors
usb-linux.c: In function 'usb_host_read_file':
usb-linux.c:1204: error: ignoring return value of 'fgets', declared with attribute warn_unused_result
make: *** [usb-linux.o] Error 1

Signed-off-by: Kirill A. Shutemov <[email protected]>
Signed-off-by: Juan Quintela <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

net/slirp.c: fix warning with _FORTIFY_SOURCE

CC    net/slirp.o
cc1: warnings being treated as errors
net/slirp.c: In function 'slirp_smb_cleanup':
net/slirp.c:470: error: ignoring return value of 'system', declared with attribute warn_unused_result
make: *** [net/slirp.o] Error 1

Signed-off-by: Kirill A. Shutemov <[email protected]>
Signed-off-by: Juan Quintela <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

block/qcow2.c: fix warnings with _FORTIFY_SOURCE

CC    block/qcow2.o
cc1: warnings being treated as errors
block/qcow2.c: In function 'qcow_create2':
block/qcow2.c:829: error: ignoring return value of 'write', declared with attribute warn_unused_result
block/qcow2.c:838: error: ignoring return value of 'write', declared with attribute warn_unused_result
block/qcow2.c:839: error: ignoring return value of 'write', declared with attribute warn_unused_result
block/qcow2.c:841: error: ignoring return value of 'write', declared with attribute warn_unused_result
block/qcow2.c:844: error: ignoring return value of 'write', declared with attribute warn_unused_result
block/qcow2.c:849: error: ignoring return value of 'write', declared with attribute warn_unused_result
block/qcow2.c:852: error: ignoring return value of 'write', declared with attribute warn_unused_result
block/qcow2.c:855: error: ignoring return value of 'write', declared with attribute warn_unused_result
make: *** [block/qcow2.o] Error 1

Signed-off-by: Kirill A. Shutemov <[email protected]>
Signed-off-by: Juan Quintela <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

block/vvfat.c: fix warnings with _FORTIFY_SOURCE

CC    block/vvfat.o
cc1: warnings being treated as errors
block/vvfat.c: In function 'commit_one_file':
block/vvfat.c:2259: error: ignoring return value of 'ftruncate', declared with attribute warn_unused_result
make: *** [block/vvfat.o] Error 1
  CC    block/vvfat.o
In file included from /usr/include/stdio.h:912,
                 from ./qemu-common.h:19,
                 from block/vvfat.c:27:
In function 'snprintf',
    inlined from 'init_directories' at block/vvfat.c:871,
    inlined from 'vvfat_open' at block/vvfat.c:1068:
/usr/include/bits/stdio2.h:65: error: call to __builtin___snprintf_chk will always overflow destination buffer
make: *** [block/vvfat.o] Error 1

Signed-off-by: Kirill A. Shutemov <[email protected]>
Signed-off-by: Juan Quintela <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

block/vmdk.o: fix warnings with _FORTIFY_SOURCE

CC    block/vmdk.o
cc1: warnings being treated as errors
block/vmdk.c: In function 'vmdk_snapshot_create':
block/vmdk.c:236: error: ignoring return value of 'ftruncate', declared with attribute warn_unused_result
block/vmdk.c: In function 'vmdk_create':
block/vmdk.c:775: error: ignoring return value of 'write', declared with attribute warn_unused_result
block/vmdk.c:776: error: ignoring return value of 'write', declared with attribute warn_unused_result
block/vmdk.c:778: error: ignoring return value of 'ftruncate', declared with attribute warn_unused_result
block/vmdk.c:784: error: ignoring return value of 'write', declared with attribute warn_unused_result
block/vmdk.c:790: error: ignoring return value of 'write', declared with attribute warn_unused_result
block/vmdk.c:807: error: ignoring return value of 'write', declared with attribute warn_unused_result
make: *** [block/vmdk.o] Error 1

Signed-off-by: Kirill A. Shutemov <[email protected]>
Signed-off-by: Juan Quintela <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

block/qcow.c: fix warnings with _FORTIFY_SOURCE

CC    block/qcow.o
cc1: warnings being treated as errors
block/qcow.c: In function 'qcow_create':
block/qcow.c:804: error: ignoring return value of 'write', declared with attribute warn_unused_result
block/qcow.c:806: error: ignoring return value of 'write', declared with attribute warn_unused_result
block/qcow.c:811: error: ignoring return value of 'write', declared with attribute warn_unused_result
make: *** [block/qcow.o] Error 1

Signed-off-by: Kirill A. Shutemov <[email protected]>
Signed-off-by: Juan Quintela <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

block/cow.c: fix warnings with _FORTIFY_SOURCE

CC    block/cow.o
cc1: warnings being treated as errors
block/cow.c: In function 'cow_create':
block/cow.c:251: error: ignoring return value of 'write', declared with attribute warn_unused_result
block/cow.c:253: error: ignoring return value of 'ftruncate', declared with attribute warn_unused_result
make: *** [block/cow.o] Error 1

Signed-off-by: Kirill A. Shutemov <[email protected]>
Signed-off-by: Juan Quintela <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

posix-aio-compat.c: fix warning with _FORTIFY_SOURCE

CC    posix-aio-compat.o
cc1: warnings being treated as errors
posix-aio-compat.c: In function 'aio_signal_handler':
posix-aio-compat.c:505: error: ignoring return value of 'write', declared with attribute warn_unused_result
make: *** [posix-aio-compat.o] Error 1

Signed-off-by: Kirill A. Shutemov <[email protected]>
Signed-off-by: Juan Quintela <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

force to test result for qemu_write_full()

Signed-off-by: Juan Quintela <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

Introduce qemu_write_full()

A variant of write(2) which handles partial write.

Signed-off-by: Kirill A. Shutemov <[email protected]>
Signed-off-by: Juan Quintela <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

qcow2: Don't ignore qcow2_alloc_clusters return value

Now that qcow2_alloc_clusters can return error codes, we must handle them in
the callers of qcow2_alloc_clusters.

Signed-off-by: Kevin Wolf <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

qcow2: Don't ignore update_refcount return value

update_refcount can return errors that need to be handled by the callers.

Signed-off-by: Kevin Wolf <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

qcow2: Allow updating no refcounts

There's absolutely no problem with updating the refcounts of 0 clusters.
At least snapshot code is doing this and would fail once the result of
update_refcount isn't ignored any more.

Signed-off-by: Kevin Wolf <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

qcow2: Improve error handling in update_refcount

If update_refcount fails, try to undo any changes made so far to avoid
inconsistencies in the image file.

Signed-off-by: Kevin Wolf <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

qcow2: Fix error handling in grow_refcount_table

Return the appropriate error code instead of -EIO.

Signed-off-by: Kevin Wolf <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

block: Return original error codes in bdrv_pread/write

Don't assume -EIO but return the real error.

Signed-off-by: Kevin Wolf <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

qcow2: Return 0/-errno in qcow2_alloc_cluster_offset

Returning 0/-errno allows it to distingush different errors classes. The
cluster offset of newly allocated clusters is now returned in the QCowL2Meta
struct.

Signed-off-by: Kevin Wolf <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

qcow2: Return 0/-errno in get_cluster_table

Switching to 0/-errno allows it to distinguish different error cases.

Signed-off-by: Kevin Wolf <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

qcow2: Fix error handling in qcow_save_vmstate

Don't assume success but pass the bdrv_pwrite return value on.

Signed-off-by: Kevin Wolf <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

qcow2: Fix error handling in qcow2_grow_l1_table

Return the appropriate error value instead of always using EIO. Don't free the
L1 table on errors, we still need it.

Signed-off-by: Kevin Wolf <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

monitor: convert do_cpu_set() to QObject, QError

Signed-off-by: Markus Armbruster <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

QError: New QERR_INVALID_CPU_INDEX

Signed-off-by: Markus Armbruster <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

monitor: convert do_physical_memory_save() to QError

Signed-off-by: Markus Armbruster <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

monitor: convert do_memory_save() to QError

Signed-off-by: Markus Armbruster <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

QError: New QERR_OPEN_FILE_FAILED

Signed-off-by: Markus Armbruster <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

monitor: Don't check for mon_get_cpu() failure

mon_get_cpu() can't return null pointer, because it passes its return
value to cpu_synchronize_state() first, which crashes if its argument
is null.

Remove the (pretty cheesy) handling of this non-existing error.

Signed-off-by: Markus Armbruster <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

json: escape u0000 .. u001F when outputting json

Markus Armbruster pointed out:

JSON requires control characters in strings to be escaped.  RFC 4627
section 2.5:

   A string begins and ends with quotation marks.  All Unicode
   characters may be placed within the quotation marks except for the
   characters that must be escaped: quotation mark, reverse solidus, and
   the control characters (U+0000 through U+001F).

We've been quoting the special escape sequences that JSON defines but we
haven't been encoding the full control character range.  This patch fixes that.

Signed-off-by: Anthony Liguori <[email protected]>

Sparc64: fix initrd

Fix HdrS offsets for Sparc64. The initrd address must be offset by
KERNBASE.

Use rom_ptr mechanism to actually write to the kernel image.

Signed-off-by: Blue Swirl <[email protected]>

pflash: Dont open memarea for full IO if already done.

When wcycle is non zero the area is already opened for readable IO.
Avoiding the re-registration of the memarea significantly speeds up
the flash emulation. In particular for flashes connected through 8 or
16-bit buses.

Signed-off-by: Edgar E. Iglesias <[email protected]>

pflash: Reduce writebuf len for 8-bit flashes.

Flashes connected through an 8 bit bus cannot handle write buffers
larger than 256 bytes.

Signed-off-by: Edgar E. Iglesias <[email protected]>

pflash: Remove dead code, no functional changes.

Signed-off-by: Edgar E. Iglesias <[email protected]>

microblaze: The petalogix s3adsp board uses intel flashes

Signed-off-by: Edgar E. Iglesias <[email protected]>

Load global config files by default

A new option, -nodefconfig is introduced to prevent loading from the default
config location.  Otherwise, two configuration files will be searched for,
qemu.conf and target-<TARGET_NAME>.conf.

To ensure that the default configuration is overridden by a user specified
config, we introduce a two stage option parsing mechanism.

Signed-off-by: Anthony Liguori <[email protected]>

Move out option lookup into a separate function

Signed-off-by: Anthony Liguori <[email protected]>

Support --sysconfdir in configure to specify path to configuration files

The default value is ${prefix}/etc/qemu.  --sysconfdir can be used to override
the default to an absolute path.  The expectation is that when installed to
/usr, --sysconfdir=/etc/qemu will be used.

Signed-off-by: Anthony Liguori <[email protected]>