libdecnumber: Eliminate redundant declarations

Eliminate redundant declarations of symbols DPD2BIN and BIN2DPD in
various .c source files.  These symbols are already declared in decDPD.h and
thus will trigger 'redundant redeclaration of ?XXX?' warnings, which, of
course, may fail QEMU compilation.

Signed-off-by: Tom Musta <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

libdecnumber: Change gstdint.h to stdint.h

Replace the inclusion of gstdint.h with the standard stdint.h
header file.

Signed-off-by: Tom Musta <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

libdecnumber: Modify dconfig.h to Integrate with QEMU

Modify the dconfig.h header file so that libdecnumber code integrates QEMU
configuration.   Specifically:

  - the WORDS_BIGENDIAN preprocessor macro is used in libdecnumber code to
    determines endianness.  It is derived from the existing QEMU macro
    HOST_WORDS_BIGENDIAN which is defined in config-host.h.

  - the DECPUN macro determines the number of decimal digits (aka declets) per
    unit (byte).  This is 3 for PowerPC DFP.

Signed-off-by: Tom Musta <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

libdecnumber: Prepare libdecnumber for QEMU include structure

Consistent with other libraries in QEMU, the libdecnumber header files were
placed in include/libdecnumber, separate from the C code.  This is different
from the original libdecnumber source, where they were co-located.

Change the libdecnumber source code so that it reflects this split.  Specifically,
modify directives of the form:

    #include "xxx.h"

to look like:

    #include "libdecnumber/xxx.h"

Signed-off-by: Tom Musta <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

libdecnumber: Eliminate #include *Symbols.h

The various *Symbols.h files were not copied from the original GCC libdecnumber
library; they are not necessary for use in QEMU.  Remove all instances of

    #include "*Symbols.h"

Signed-off-by: Tom Musta <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

libdecnumber: Introduce libdecnumber Code

Add files from the libdecnumber decimal floating point library to QEMU.  The libdecnumber
library was originally part of GCC and contains code that is useful in emulating the PowerPC
decimal floating point (DFP) instructions.  This particular copy of the source comes from
GCC 4.3 and is licensed at GPLv2+.

Signed-off-by: Tom Musta <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

mac99: Added FW_CFG_PPC_BUSFREQ to match CLOCKFREQ and TBFREQ already there

While there, also moved the hard coded value for CLOCKFREQ to a #define.

Signed-off-by: BALATON Zoltan <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

target-ppc: Remove PVR check from migration

Currently migration fails if CPU version (PVR register) is different
even a bit. This check is performed at the very end of migration when
device states are sent. This is too late for management software and
we need to provide a way for the user to make sure that migration
will succeed if QEMU is started with appropritate command line parameters.

This removes the PVR check.

This resets PVR to the default value as the existing VMSTATE record
for SPR array sends all 1024 registers unconditionally and overwrites
the destination PVR.

If the user wants some guarantees for migration to succeed, then
a CPU name or "host" CPU with a "compat" option (on its way to upsteam)
should be used and KVM or TCG is expected to fail on unsupported values
at the moment of QEMU start.

Signed-off-by: Alexey Kardashevskiy <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

target-ppc: Eliminate Magic Number MSR Masks

Use MSR mnemonics from cpu.h instead of magic numbers for the CPUPPCState.msr_mask
initialization.

There is one bit in the 401x2 (and subsequent) model that I could not find any
documentation for.  It is open coded at little endian bit position 20:

    pcc->msr_mask = (1ull << 20) |
                    (1ull << MSR_KEY) |
                    (1ull << MSR_POW) |
                    (1ull << MSR_CE) |
                    ...

Signed-off-by: Tom Musta <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

spapr_pci: Fix number of returned vectors in ibm, change-msi

Current guest kernels try allocating as many vectors as the quota is.
For example, in the case of virtio-net (which has just 3 vectors)
the guest requests 4 vectors (that is the quota in the test) and
the existing ibm,change-msi handler returns 4. But before it returns,
it calls msix_set_message() in a loop and corrupts memory behind
the end of msix_table.

This limits the number of vectors returned by ibm,change-msi to
the maximum supported by the actual device.

Signed-off-by: Alexey Kardashevskiy <[email protected]>
Cc: [email protected]
[agraf: squash in bugfix from aik]
Signed-off-by: Alexander Graf <[email protected]>

spapr-pci: remove io ports workaround

In the past, IO space could not be mapped into the memory address space
so we introduced a workaround for that. Nowadays it does not look
necessary so we can remove the workaround and make sPAPR PCI
configuration simplier.

Signed-off-by: Greg Kurz <[email protected]>
Acked-by: Alexey Kardashevskiy <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

target-ppc: Remove redundant POWER7 declarations

At the moment there are 3 versions of POWER7 CPUs defined. However
we do not emulate these CPUs diffent and it does not make much
sense to keep them all.

This removes POWER7_v2.0 and POWER7_v2.1 and leaves just one versioned
CPU per family which is POWER7_v2.3 with POWER7 alias.

Signed-off-by: Alexey Kardashevskiy <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

target-ppc: Move alias lookup after class lookup

This moves aliases lookup after CPU class lookup. This is to let new generic
CPU to be found first if it is present and only if it is not (TCG case), use
aliases.

Signed-off-by: Alexey Kardashevskiy <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

target-ppc: Create versionless CPU class per family if KVM

At the moment generic version-less CPUs are supported via hardcoded aliases.
For example, POWER7 is an alias for POWER7_v2.1. So when QEMU is started
with -cpu POWER7, the POWER7_v2.1 class instance is created.

This approach works for TCG and KVMs other than HV KVM. HV KVM cannot emulate
PVR value so the guest always sees the real PVR. HV KVM will not allow setting
PVR other that the host PVR because of that (the kernel patch for it is on
its way). So in most cases it is impossible to run QEMU with -cpu POWER7
unless the host PVR is exactly the same as the one from the alias (which
is now POWER7_v2.3). It was decided that under HV KVM QEMU should use
-cpu host.

Using "host" CPU type creates a problem for management tools such as libvirt
because they want to know in advance if the destination guest can possibly
run on the destination. Since the "host" type is really not a type and will
always work with any KVM, there is no way for libvirt to know if the migration
will success.

This registers additional CPU class derived from the host CPU family.
The name for it is taken from @desc field of the CPU family class.

Signed-off-by: Alexey Kardashevskiy <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

target-ppc: gdbstub allow byte swapping for reading/writing registers

This patch allows registers to be properly read from and written to
when using the gdbstub to debug a ppc guest running in little
endian mode.

Reviewed-by: Andreas Färber <[email protected]>
Signed-off-by: Thomas Falcon <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

target-ppc: extract register length calculation in gdbstub

This patch extracts the method to determine a register's size
into a separate function.

Reviewed-by: Andreas Färber <[email protected]>
Signed-off-by: Thomas Falcon <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

spapr_nvram: Correct max nvram size

Currently it is UINT16_MAX*16 = 65536*16 = 1048560 which is not
a round number and therefore a bit confusing.

This defines MAX_NVRAM_SIZE precisely as 1MB.

Suggested-by: Thomas Huth <[email protected]>
Signed-off-by: Alexey Kardashevskiy <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

Fix typo in eTSEC Ethernet controller

IRQ are lowered when ievent bit is cleared, so irq_pulse makes no sense
here...

Signed-off-by: Fabien Chouteau <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

monitor: QEMU Monitor Instruction Disassembly Incorrect for PowerPC LE Mode

The monitor support for disassembling instructions does not honor the MSR[LE]
bit for PowerPC processors.

This change enhances the monitor_disas() routine by supporting a flag bit
for Little Endian mode.  Bit 16 is used since that bit was used in the
analagous guest disassembly routine target_disas().

Also, to be consistent with target_disas(), the disassembler bfd_mach field
can be passed in the flags argument.

Reported-by: Anton Blanchard <[email protected]>
Signed-off-by: Tom Musta <[email protected]>
Reviewed-by: Peter Maydell <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

target-ppc: Fix target_disas

Inspect only bit 16 for the Little Endian test.  Correct comment preceding
the target_disas() function.  Correct grammar in comment for flags processing.

Signed-off-by: Tom Musta <[email protected]>
Reviewed-by: Peter Maydell <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

Merge remote-tracking branch 'remotes/juanquintela/tags/migration/20140616' into staging

migration/next for 20140616

# gpg: Signature made Mon 16 Jun 2014 04:10:18 BST using RSA key ID 5872D723
# gpg: Can't check signature: public key not found

* remotes/juanquintela/tags/migration/20140616:
  migration: catch unknown flags in ram_load
  rdma: Fix block during rdma migration
  migration: Increase default max_downtime from 30ms to 300ms
  vmstate: Refactor opening of files
  savevm: Remove all the unneeded version_minimum_id_old (x86)
  savevm: Remove all the unneeded version_minimum_id_old (ppc)

Signed-off-by: Peter Maydell <[email protected]>

migration: catch unknown flags in ram_load

if a saved vm has unknown flags in the memory data qemu
currently simply ignores this flag and continues which
yields in an unpredictable result.

This patch catches all unknown flags and aborts the
loading of the vm. Additionally error reports are thrown
if the migration aborts abnormally.

Signed-off-by: Peter Lieven <[email protected]>
Signed-off-by: Juan Quintela <[email protected]>

rdma: Fix block during rdma migration

If the networking break or there's something wrong with rdma
device(ib0 with no IP) during rdma migration, the main_loop of
qemu will be blocked in rdma_destroy_id. I add rdma_ack_cm_event
to fix this bug.

Signed-off-by: Mo Yuxiang <[email protected]>
Signed-off-by: Gonglei <[email protected]>
Reviewed-by: Michael R. Hines <[email protected]>
Signed-off-by: Juan Quintela <[email protected]>

migration: Increase default max_downtime from 30ms to 300ms

The existing timeout is 30ms which on 100MB/s (1Gbit) gives us
3MB/s rate maximum. If we put some load on the guest, it is easy to
get page dirtying rate too big so live migration will never complete.
In the case of libvirt that means that the guest will be stopped
anyway after a timeout specified in the "virsh migrate" command and
this normally generates even bigger delay.

This changes max_downtime to 300ms which seems to be more
reasonable value.

Signed-off-by: Alexey Kardashevskiy <[email protected]>
Signed-off-by: Juan Quintela <[email protected]>

vmstate: Refactor opening of files

Signed-off-by: Juan Quintela <[email protected]>
Reviewed-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Amit Shah <[email protected]>

savevm: Remove all the unneeded version_minimum_id_old (x86)

After previous Peter patch, they are redundant.  This way we don't
assign them except when needed.  Once there, there were lots of case
where the ".fields" indentation was wrong:

     .fields = (VMStateField []) {
and
     .fields =      (VMStateField []) {

Change all the combinations to:

     .fields = (VMStateField[]){

The biggest problem (appart from aesthetics) was that checkpatch complained
when we copy&pasted the code from one place to another.

Signed-off-by: Juan Quintela <[email protected]>
Acked-by: Alexey Kardashevskiy <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>

savevm: Remove all the unneeded version_minimum_id_old (ppc)

After previous Peter patch, they are redundant.  This way we don't
assign them except when needed.  Once there, there were lots of case
where the ".fields" indentation was wrong:

     .fields = (VMStateField []) {
and
     .fields =      (VMStateField []) {

Change all the combinations to:

     .fields = (VMStateField[]){

The biggest problem (appart from aesthetics) was that checkpatch complained
when we copy&pasted the code from one place to another.

Signed-off-by: Juan Quintela <[email protected]>
Acked-by: Alexey Kardashevskiy <[email protected]>

Merge remote-tracking branch 'remotes/kraxel/tags/pull-usb-20140613-1' into staging

usb-host: add range checks for usb-host parameters

# gpg: Signature made Fri 13 Jun 2014 12:33:05 BST using RSA key ID D3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <[email protected]>"
# gpg:                 aka "Gerd Hoffmann <[email protected]>"
# gpg:                 aka "Gerd Hoffmann (private) <[email protected]>"

* remotes/kraxel/tags/pull-usb-20140613-1:
  usb-host: add range checks for usb-host parameters

Signed-off-by: Peter Maydell <[email protected]>

Merge remote-tracking branch 'remotes/kraxel/tags/pull-trivial-20140613-1' into staging

inet_listen_opts: add error checking

# gpg: Signature made Fri 13 Jun 2014 12:29:43 BST using RSA key ID D3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <[email protected]>"
# gpg:                 aka "Gerd Hoffmann <[email protected]>"
# gpg:                 aka "Gerd Hoffmann (private) <[email protected]>"

* remotes/kraxel/tags/pull-trivial-20140613-1:
  inet_listen_opts: add error checking

Signed-off-by: Peter Maydell <[email protected]>

Merge remote-tracking branch 'remotes/spice/tags/pull-spice-20140613-1' into staging

spice: add mouse cursor support
qxl-render: add sanity check

# gpg: Signature made Fri 13 Jun 2014 12:22:45 BST using RSA key ID D3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <[email protected]>"
# gpg:                 aka "Gerd Hoffmann <[email protected]>"
# gpg:                 aka "Gerd Hoffmann (private) <[email protected]>"

* remotes/spice/tags/pull-spice-20140613-1:
  qxl-render: add sanity check
  spice: add mouse cursor support

Signed-off-by: Peter Maydell <[email protected]>

Merge remote-tracking branch 'remotes/kraxel/tags/pull-chardev-20140613-1' into staging

char: fix avail_connections init in qemu_chr_open_eventfd()

# gpg: Signature made Fri 13 Jun 2014 12:16:50 BST using RSA key ID D3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <[email protected]>"
# gpg:                 aka "Gerd Hoffmann <[email protected]>"
# gpg:                 aka "Gerd Hoffmann (private) <[email protected]>"

* remotes/kraxel/tags/pull-chardev-20140613-1:
  char: fix avail_connections init in qemu_chr_open_eventfd()

Signed-off-by: Peter Maydell <[email protected]>

Merge remote-tracking branch 'remotes/kraxel/tags/pull-audio-20140613-1' into staging

audio: Drop superfluous conditionals around g_free()

# gpg: Signature made Fri 13 Jun 2014 12:14:24 BST using RSA key ID D3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <[email protected]>"
# gpg:                 aka "Gerd Hoffmann <[email protected]>"
# gpg:                 aka "Gerd Hoffmann (private) <[email protected]>"

* remotes/kraxel/tags/pull-audio-20140613-1:
  audio: Drop superfluous conditionals around g_free()

Signed-off-by: Peter Maydell <[email protected]>

usb-host: add range checks for usb-host parameters

Signed-off-by: Gerd Hoffmann <[email protected]>

inet_listen_opts: add error checking

Don't use atoi() function which doesn't detect errors, switch to
strtol and error out on failures.  Also add a range check while
being at it.

Signed-off-by: Gerd Hoffmann <[email protected]>
Reviewed-by: Markus Armbruster <[email protected]>

qxl-render: add sanity check

Verify dirty rectangle is completely within the primary surface,
just ignore it in case it isn't.

Signed-off-by: Gerd Hoffmann <[email protected]>

spice: add mouse cursor support

So you'll have a mouse pointer when running non-qxl gfx cards with
mouse pointer support (virtio-gpu, IIRC vmware too).

Signed-off-by: Gerd Hoffmann <[email protected]>

char: fix avail_connections init in qemu_chr_open_eventfd()

When trying to use a ivshmem server with qemu, ivshmem init code tries to
create a CharDriverState object for each eventfd retrieved from the server.
To create this object, a call to qemu_chr_open_eventfd() is done.
Right after this, before adding a frontend, qemu_chr_fe_claim_no_fail() is
called.
qemu_chr_open_eventfd() does not set avail_connections to 1, so no frontend can
be associated because qemu_chr_fe_claim_no_fail() makes qemu stop right away.

This problem comes from 456d60692310e7ac25cf822cc1e98192ad636ece
"qemu-char: Call fe_claim / fe_release when not using qdev chr properties".

Fix this, by setting avail_connections to 1 in qemu_chr_open_eventfd().

Signed-off-by: David Marchand <[email protected]>
Signed-off-by: Gerd Hoffmann <[email protected]>

audio: Drop superfluous conditionals around g_free()

Signed-off-by: Markus Armbruster <[email protected]>
Signed-off-by: Gerd Hoffmann <[email protected]>

Merge remote-tracking branch 'remotes/kraxel/tags/pull-gtk-20140611-1' into staging

gtk: misc fixes & cleanups.

# gpg: Signature made Wed 11 Jun 2014 13:28:12 BST using RSA key ID D3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <[email protected]>"
# gpg:                 aka "Gerd Hoffmann <[email protected]>"
# gpg:                 aka "Gerd Hoffmann (private) <[email protected]>"

* remotes/kraxel/tags/pull-gtk-20140611-1:
  gtk: update window size after showing/hiding tabs
  gtk: factor out gtk3 grab into the new gd_grab_devices function
  gtk: cleanup backend dependencies
  gtk: factor out keycode mapping

Signed-off-by: Peter Maydell <[email protected]>

Merge remote-tracking branch 'remotes/qmp-unstable/queue/qmp' into staging

* remotes/qmp-unstable/queue/qmp:
  json-parser: drop superfluous assignment for token variable
  readline: Clear screen on form feed.
  monitor: Add delvm and loadvm argument completion
  monitor: Add host_net_remove arguments completion
  readline: Make completion strings always unique
  monitor: Add host_net_add device argument completion
  net: Export valid host network devices list
  monitor: Add migrate_set_capability completion
  monitor: Add watchdog_action argument completion
  monitor: Add ringbuf_write and ringbuf_read argument completion
  dump: simplify get_len_buf_out()
  dump: hoist lzo_init() from get_len_buf_out() to dump_init()
  dump: select header bitness based on ELF class, not ELF architecture
  dump: eliminate DumpState.page_size ("guest's page size")
  dump: eliminate DumpState.page_shift ("guest's page shift")
  dump: simplify write_start_flat_header()
  dump: fill in the flat header signature more pleasingly to the eye

Signed-off-by: Peter Maydell <[email protected]>

Merge remote-tracking branch 'remotes/pmaydell/tags/pull-bsd-user-20140611' into staging

bsd-user queue:
 * build fixes
 * improvements to strace

# gpg: Signature made Wed 11 Jun 2014 15:23:40 BST using RSA key ID 14360CDE
# gpg: Good signature from "Peter Maydell <[email protected]>"

* remotes/pmaydell/tags/pull-bsd-user-20140611:
  bsd-user: Fix syscall format, add strace support for more syscalls
  bsd-user: Implement strace support for thr_* syscalls
  bsd-user: Implement strace support for extattr_* syscalls
  bsd-user: Implement strace support for __acl_* syscalls
  bsd-user: Implement strace support for print_ioctl syscall
  bsd-user: Implement strace support for print_sysctl syscall
  bsd-user: GPL v2 attribution update and style
  bsd-user: add HOST_VARIANT_DIR for various *BSD dependent code
  exec: replace ffsl with ctzl
  vhost: replace ffsl with ctzl
  xen: replace ffsl with ctzl
  util/qemu-openpty: fix build with musl libc by include termios.h as fallback
  bsd-user/mmap.c: Don't try to override g_malloc/g_free
  util/hbitmap.c: Use ctpopl rather than reimplementing a local equivalent
  bsd-user: refresh freebsd system call numbers

Signed-off-by: Peter Maydell <[email protected]>

Merge remote-tracking branch 'remotes/bonzini/configure' into staging

* remotes/bonzini/configure:
  rules.mak: Rewrite unnest-vars
  configure: unset interfering variables
  configure: duplicate/incorrect order of -lrt
  libcacard: improve documentation
  libcacard: actually use symbols file
  libcacard: replace qemu thread primitives with glib ones
  vscclient: use glib thread primitives not qemu
  glib-compat.h: add new thread API emulation on top of pre-2.31 API

Signed-off-by: Peter Maydell <[email protected]>

json-parser: drop superfluous assignment for token variable

Signed-off-by: ChenLiang <[email protected]>
Signed-off-by: Gonglei <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Reviewed-by: Paolo Bonzini <[email protected]>
Signed-off-by: Luiz Capitulino <[email protected]>

readline: Clear screen on form feed.

Signed-off-by: Hani Benhabiles <[email protected]>
Signed-off-by: Luiz Capitulino <[email protected]>

monitor: Add delvm and loadvm argument completion

Signed-off-by: Hani Benhabiles <[email protected]>
Signed-off-by: Luiz Capitulino <[email protected]>

monitor: Add host_net_remove arguments completion

Relies on readline unique completion strings patch to make the added vlan/hub
completion values unique, instead of using something like a hash table.

Signed-off-by: Hani Benhabiles <[email protected]>
Reviewed-by: Stefan Hajnoczi <[email protected]>
Signed-off-by: Luiz Capitulino <[email protected]>

readline: Make completion strings always unique

There is no need to clutter the user's choices with repeating the same value
multiple times.

Signed-off-by: Hani Benhabiles <[email protected]>
Signed-off-by: Luiz Capitulino <[email protected]>

monitor: Add host_net_add device argument completion

Also fix the parameters documentation.

Signed-off-by: Hani Benhabiles <[email protected]>
Reviewed-by: Stefan Hajnoczi <[email protected]>
Signed-off-by: Luiz Capitulino <[email protected]>

net: Export valid host network devices list

Signed-off-by: Hani Benhabiles <[email protected]>
Reviewed-by: Stefan Hajnoczi <[email protected]>
Signed-off-by: Luiz Capitulino <[email protected]>

monitor: Add migrate_set_capability completion

Signed-off-by: Hani Benhabiles <[email protected]>
Signed-off-by: Luiz Capitulino <[email protected]>

monitor: Add watchdog_action argument completion

Signed-off-by: Hani Benhabiles <[email protected]>
Signed-off-by: Luiz Capitulino <[email protected]>

monitor: Add ringbuf_write and ringbuf_read argument completion

Export chr_is_ringbuf() function. Also remove left-over function prototypes
while at it.

Signed-off-by: Hani Benhabiles <[email protected]>
Signed-off-by: Luiz Capitulino <[email protected]>

dump: simplify get_len_buf_out()

We can (and should) rely on the fact that s->flag_compress is exactly one
of DUMP_DH_COMPRESSED_ZLIB, DUMP_DH_COMPRESSED_LZO, and
DUMP_DH_COMPRESSED_SNAPPY.

This is ensured by the QMP schema and dump_init() in combination.

Suggested-by: Paolo Bonzini <[email protected]>
Signed-off-by: Laszlo Ersek <[email protected]>
Reviewed-by: Paolo Bonzini <[email protected]>
Signed-off-by: Luiz Capitulino <[email protected]>

dump: hoist lzo_init() from get_len_buf_out() to dump_init()

qmp_dump_guest_memory()
  dump_init()
    lzo_init() <---------+
  create_kdump_vmcore()  |
    write_dump_pages()   |
      get_len_buf_out()  |
        lzo_init() ------+

This patch doesn't change the fact that lzo_init() is called for every
LZO-compressed dump, but it makes get_len_buf_out() more focused (single
responsibility).

Suggested-by: Paolo Bonzini <[email protected]>
Signed-off-by: Laszlo Ersek <[email protected]>
Reviewed-by: Paolo Bonzini <[email protected]>
Signed-off-by: Luiz Capitulino <[email protected]>

dump: select header bitness based on ELF class, not ELF architecture

The specific ELF architecture (d_machine) carries Too Much Information
(TM) for deciding between create_header32() and create_header64(), use
"d_class" instead (ELFCLASS32 vs. ELFCLASS64).

This change adapts write_dump_header() to write_elf_loads(), dump_begin()
etc. that also rely on the ELF class of the target for bitness selection.

Considering the current targets that support dumping, cpu_get_dump_info()
works as follows:
- target-s390x/arch_dump.c: (EM_S390, ELFCLASS64) only
- target-ppc/arch_dump.c (EM_PPC64, ELFCLASS64) only
- target-i386/arch_dump.c: sets (EM_X86_64, ELFCLASS64) vs. (EM_386,
  ELFCLASS32) keying off the same Long Mode Active flag.

Hence no observable change.

Approximately-suggested-by: Paolo Bonzini <[email protected]>
Signed-off-by: Laszlo Ersek <[email protected]>
Reviewed-by: Paolo Bonzini <[email protected]>
Signed-off-by: Luiz Capitulino <[email protected]>

dump: eliminate DumpState.page_size ("guest's page size")

Use TARGET_PAGE_SIZE and ~TARGET_PAGE_MASK instead.

"DumpState.page_size" has type "size_t", whereas TARGET_PAGE_SIZE has type
"int". TARGET_PAGE_MASK is of type "int" and has negative value. The patch
affects the implicit type conversions as follows:

- create_header32() and create_header64(): assigned to "block_size", which
  has type "uint32_t". No change.

- get_next_page(): "block->target_start", "block->target_end" and "addr"
  have type "hwaddr" (uint64_t).

  Before the patch,
  - if "size_t" was "uint64_t", then no additional conversion was done as
    part of the usual arithmetic conversions,
  - If "size_t" was "uint32_t", then it was widened to uint64_t as part of
    the usual arithmetic conversions,
  for the remainder and addition operators.

  After the patch,
  - "~TARGET_PAGE_MASK" expands to  ~~((1 << TARGET_PAGE_BITS) - 1). It
    has type "int" and positive value (only least significant bits set).
    That's converted (widened) to "uint64_t" for the bit-ands. No visible
    change.
  - The same holds for the (addr + TARGET_PAGE_SIZE) addition.

- write_dump_pages():
  - TARGET_PAGE_SIZE passed as argument to a bunch of functions that all
    have prototypes. No change.

  - When incrementing "offset_data" (of type "off_t"): given that we never
    build for ILP32_OFF32 (see "-D_FILE_OFFSET_BITS=64" in configure),
    "off_t" is always "int64_t", and we only need to consider:
    - ILP32_OFFBIG: "size_t" is "uint32_t".
      - before: int64_t += uint32_t. Page size converted to int64_t for
        the addition.
      - after:  int64_t += int32_t. No change.
    - LP64_OFF64: "size_t" is "uint64_t".
      - before: int64_t += uint64_t. Offset converted to uint64_t for the
        addition, then the uint64_t result is converted to int64_t for
        storage.
      - after:  int64_t += int32_t. Same as the ILP32_OFFBIG/after case.
        No visible change.

  - (size_out < s->page_size) comparisons, and (size_out = s->page_size)
    assignment:
    - before: "size_out" is of type "size_t", no implicit conversion for
              either operator.
    - after: TARGET_PAGE_SIZE (of type "int" and positive value) is
             converted to "size_t" (for the relop because the latter is
             one of "uint32_t" and "uint64_t"). No visible change.

- dump_init():
  - DIV_ROUND_UP(DIV_ROUND_UP(s->max_mapnr, CHAR_BIT), s->page_size): The
    innermost "DumpState.max_mapnr" field has type uint64_t, which
    propagates through all implicit conversions at hand:

    #define DIV_ROUND_UP(n,d) (((n) + (d) - 1) / (d))

    regardless of the page size macro argument's type. In the outer macro
    replacement, the page size is converted from uint32_t and int32_t
    alike to uint64_t.

  - (tmp * s->page_size) multiplication: "tmp" has size "uint64_t"; the
    RHS is converted to that type from uint32_t and int32_t just the same
    if it's not uint64_t to begin with.

Signed-off-by: Laszlo Ersek <[email protected]>
Reviewed-by: Paolo Bonzini <[email protected]>
Signed-off-by: Luiz Capitulino <[email protected]>

dump: eliminate DumpState.page_shift ("guest's page shift")

Just use TARGET_PAGE_BITS.

"DumpState.page_shift" used to have type "uint32_t", while the replacement
TARGET_PAGE_BITS has type "int". Since "DumpState.page_shift" was only
used as bit shift counts in the paddr_to_pfn() and pfn_to_paddr() macros,
this is safe.

Suggested-by: Paolo Bonzini <[email protected]>
Signed-off-by: Laszlo Ersek <[email protected]>
Reviewed-by: Paolo Bonzini <[email protected]>
Signed-off-by: Luiz Capitulino <[email protected]>

dump: simplify write_start_flat_header()

Currently, the function
- defines and populates an auto variable of type MakedumpfileHeader
- allocates and zeroes a buffer of size MAX_SIZE_MDF_HEADER (4096)
- copies the former into the latter (covering an initial portion of the
  latter)

Fill in the MakedumpfileHeader structure in its final place (the alignment
is OK because the structure lives at the address returned by g_malloc0()).

Approximately-suggested-by: Luiz Capitulino <[email protected]>
Signed-off-by: Laszlo Ersek <[email protected]>
Reviewed-by: Paolo Bonzini <[email protected]>
Signed-off-by: Luiz Capitulino <[email protected]>

dump: fill in the flat header signature more pleasingly to the eye

The "mh.signature" array field has size 16, and is zeroed by the preceding
memset(). MAKEDUMPFILE_SIGNATURE expands to a string literal with string
length 12 (size 13). There's no need to measure the length of
MAKEDUMPFILE_SIGNATURE at runtime, nor for the extra zero-filling of
"mh.signature" with strncpy().

Use memcpy() with MIN(sizeof, sizeof) for robustness (which is an integer
constant expression, evaluable at compile time.)

Approximately-suggested-by: Paolo Bonzini <[email protected]>
Signed-off-by: Laszlo Ersek <[email protected]>
Reviewed-by: Paolo Bonzini <[email protected]>
Signed-off-by: Luiz Capitulino <[email protected]>

gtk: update window size after showing/hiding tabs

Signed-off-by: Gerd Hoffmann <[email protected]>

gtk: factor out gtk3 grab into the new gd_grab_devices function

Signed-off-by: Gerd Hoffmann <[email protected]>

gtk: cleanup backend dependencies

Make configure detect gtk x11 backend and link libX11 then.  Make
gtk backend specific code properly #ifdef'ed on the GTK_WINDOWING_*
backends at runtime).  Our gtk ui code should build and run fine on
any platform now.

This also fixes the linker failute due to the new XkbGetKeyboard call
added by commit 3158a3482b0093e41f2b2596fba50774ea31ae08.

Cc: Richard Henderson <[email protected]>
Signed-off-by: Gerd Hoffmann <[email protected]>
Reviewed-by: Daniel P. Berrange <[email protected]>

gtk: factor out keycode mapping

Signed-off-by: Gerd Hoffmann <[email protected]>

bsd-user: Fix syscall format, add strace support for more syscalls

Signed-off-by: Sean Bruno <[email protected]>
Message-id: 1402246651[email protected]
Signed-off-by: Peter Maydell <[email protected]>

bsd-user: Implement strace support for thr_* syscalls

Signed-off-by: Sean Bruno <[email protected]>
Message-id: 1402246651[email protected]
Signed-off-by: Peter Maydell <[email protected]>

bsd-user: Implement strace support for extattr_* syscalls

Signed-off-by: Sean Bruno <[email protected]>
Message-id: 1402246651[email protected]
Signed-off-by: Peter Maydell <[email protected]>

bsd-user: Implement strace support for __acl_* syscalls

Signed-off-by: Sean Bruno <[email protected]>
Message-id: 1402246651[email protected]
Signed-off-by: Peter Maydell <[email protected]>

bsd-user: Implement strace support for print_ioctl syscall

Signed-off-by: Sean Bruno <[email protected]>
Message-id: 1402246651[email protected]
Signed-off-by: Peter Maydell <[email protected]>

bsd-user: Implement strace support for print_sysctl syscall

Signed-off-by: Sean Bruno <[email protected]>
Message-id: 1402246651[email protected]
Signed-off-by: Peter Maydell <[email protected]>

bsd-user: GPL v2 attribution update and style

Signed-off-by: Sean Bruno <[email protected]>
Message-id: 1402246651[email protected]
Signed-off-by: Peter Maydell <[email protected]>

bsd-user: add HOST_VARIANT_DIR for various *BSD dependent code

This change adds HOST_VARIANT_DIR so the various BSD OS dependent
code can be separated into its own directories rather than
using #ifdef's.

This may also allow an BSD variant OS to host another BSD variant's
executable as a target.

Signed-off-by: Sean Bruno <[email protected]>
Message-id: 1402246651[email protected]
Signed-off-by: Peter Maydell <[email protected]>

exec: replace ffsl with ctzl

See commit fbeadf50 (bitops: unify bitops_ffsl with the one in
host-utils.h, call it bitops_ctzl) on why ctzl should be used instead
of ffsl.

This is also needed for musl libc which does not implement ffsl.

Signed-off-by: Natanael Copa <[email protected]>
Reviewed-by: Paolo Bonzini <[email protected]>
Signed-off-by: Peter Maydell <[email protected]>

vhost: replace ffsl with ctzl

Avoid using the GNU extesion ffsl which is not implemented in musl libc.

The atomic_xchg() means we know that vhost_log_chunk_t will never be
larger than the 'long' type, so ctzl() is always sufficient.

See also commit fbeadf50 (bitops: unify bitops_ffsl with the one in
host-utils.h, call it bitops_ctzl) on why ctzl should be used instead
of ffsl.

Signed-off-by: Natanael Copa <[email protected]>
Reviewed-by: Paolo Bonzini <[email protected]>
Signed-off-by: Peter Maydell <[email protected]>

xen: replace ffsl with ctzl

ffsl is a GNU extension and not available in musl libc.

See also commit fbeadf50 (bitops: unify bitops_ffsl with the one in
host-utils.h, call it bitops_ctzl) on why ctzl should be used instead
of ffsl.

Signed-off-by: Natanael Copa <[email protected]>
Reviewed-by: Paolo Bonzini <[email protected]>
[PMM: rebased to accommodate file rename to xen-hvm.c]
Signed-off-by: Peter Maydell <[email protected]>

util/qemu-openpty: fix build with musl libc by include termios.h as fallback

Include termios.h as POSIX fallback when not glibc, bsd or solaris.
POSIX says that termios.h should define struct termios and TCAFLUSH.
http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/termios.h.html

This fixes the following compile errors with musl libc:

util/qemu-openpty.c: In function 'qemu_openpty_raw':
util/qemu-openpty.c:112:20: error: storage size of 'tty' isn't known
     struct termios tty;
                    ^
...
util/qemu-openpty.c:128:24: error: 'TCSAFLUSH' undeclared (first use in this function)
     tcsetattr(*aslave, TCSAFLUSH, &tty);
                        ^

Signed-off-by: Natanael Copa <[email protected]>
Reviewed-by: Paolo Bonzini <[email protected]>
Signed-off-by: Peter Maydell <[email protected]>

bsd-user/mmap.c: Don't try to override g_malloc/g_free

Trying to override the implementations of g_malloc and g_free is
a really bad idea -- it means statically linked builds fail to
link (because of the multiple definitions provided by this file
and by glib), and non-statically linked builds segfault as soon
as they try to do anything more complicated than printing the
usage message. Remove these overridden versions and just use
the glib ones.

This is sufficient that bsd-user can run basic x86-64
binaries on OpenBSD again; FreeBSD and NetBSD seem to have
further issues.

Signed-off-by: Peter Maydell <[email protected]>
Reviewed-by: Sean Bruno <[email protected]>
Reviewed-by: Ed Maste <[email protected]>

util/hbitmap.c: Use ctpopl rather than reimplementing a local equivalent

The function popcountl() in hbitmap.c is effectively a reimplementation
of what host-utils.h provides as ctpopl(). Use ctpopl() directly; this fixes
a failure to compile on NetBSD (whose strings.h erroneously exposes a
system popcountl() which clashes with this one).

Reported-by: Martin Husemann <[email protected]>
Reviewed-by: Paolo Bonzini <[email protected]>
Signed-off-by: Peter Maydell <[email protected]>

bsd-user: refresh freebsd system call numbers

Update FreeBSD system call numbers in freebsd/syscall_nr.h.

Signed-off-by: Stacey Son <[email protected]>
Reviewed-by: Ed Maste <[email protected]>
Signed-off-by: Sean Bruno <[email protected]>
Message-id: 1401220104[email protected]
Reviewed-by: Peter Maydell <[email protected]>
Signed-off-by: Peter Maydell <[email protected]>

Merge remote-tracking branch 'remotes/mjt/tags/trivial-patches-2014-06-10' into staging

trivial patches for 2014-06-10

# gpg: Signature made Tue 10 Jun 2014 17:07:19 BST using RSA key ID A4C3D7DB
# gpg: Good signature from "Michael Tokarev <[email protected]>"
# gpg:                 aka "Michael Tokarev <[email protected]>"
# gpg:                 aka "Michael Tokarev <[email protected]>"
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg:          There is no indication that the signature belongs to the owner.
# Primary key fingerprint: 6EE1 95D1 886E 8FFB 810D  4324 457C E0A0 8044 65C5
#      Subkey fingerprint: 6F67 E18E 7C91 C5B1 5514  66A7 BEE5 9D74 A4C3 D7DB

* remotes/mjt/tags/trivial-patches-2014-06-10: (25 commits)
  virtio.c: fix error message
  hw: vmware_vga: don't return cursorx when the driver asks for cursory register
  migration: Plug memory leak in migrate-set-cache-size command
  libcacard: Clean up dead stores before g_free()
  libcacard: Drop superfluous conditionals around g_free()
  cpu/x86: correctly set errors in x86_cpu_parse_featurestr
  smbios: use g_free directly on NULL pointers
  vdi: remove double conversion
  apb: Fix compiler warnings (large constants)
  hw/net/ne2000-isa: Register vmstate struct
  target-microblaze: Delete unused sign_extend() function
  hw/misc/milkymist-softusb: Remove unused softusb_{read, write}_pmem()
  target-i386/translate.c: Remove unused tcg_gen_lshift()
  hw/isa/pc87312: Remove unused function is_parallel_epp()
  hw/intc/openpic: Remove unused function IRQ_testbit()
  hw/dma/xilinx_axidma: Remove unused stream_halted() function
  util/qemu-sockets.c: Avoid unused variable warnings
  hw/sd/sd.c: Drop unused sd_acmd_type[] array
  hw/i386/pc.c: Remove unused parallel_io and parallel_irq variables
  slirp: Remove unused zero_ethaddr[] variable
  ...

Signed-off-by: Peter Maydell <[email protected]>

virtio.c: fix error message

Suggested-by: Peter Maydell <[email protected]>
Signed-off-by: Michael Tokarev <[email protected]>

hw: vmware_vga: don't return cursorx when the driver asks for cursory register

hello qemu-*@nongnu.org, this is my first contribution. apologies if
something is incorrect.

this patch fixes vmware_vga.c so that it actually returns the cursory
register when asked for, instead of cursorx.

Signed-off-by: Nicolas Owens <[email protected]>
Signed-off-by: Michael Tokarev <[email protected]>

migration: Plug memory leak in migrate-set-cache-size command

We call g_free() after cache_fini() in migration_end(), but we don't
call it after cache_fini() in xbzrle_cache_resize(), leaking the
memory.

cache_init() and cache_fini() are a pair.  Since cache_init()
allocates the cache, let cache_fini() free it.  This plugs the leak.

Signed-off-by: Chen Gang <[email protected]>
Reviewed-by: Dr. David Alan Gilbert <[email protected]>
Signed-off-by: Michael Tokarev <[email protected]>

libcacard: Clean up dead stores before g_free()

Signed-off-by: Markus Armbruster <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Signed-off-by: Michael Tokarev <[email protected]>

libcacard: Drop superfluous conditionals around g_free()

Signed-off-by: Markus Armbruster <[email protected]>
Signed-off-by: Michael Tokarev <[email protected]>

cpu/x86: correctly set errors in x86_cpu_parse_featurestr

Because of the "goto out", the contents of local_err are leaked
and lost.

Signed-off-by: Paolo Bonzini <[email protected]>
Signed-off-by: Michael Tokarev <[email protected]>

smbios: use g_free directly on NULL pointers

No need to wrap it with an if.

Signed-off-by: Paolo Bonzini <[email protected]>
Reviewed-by: Benoit Canet <[email protected]>
Signed-off-by: Michael Tokarev <[email protected]>

vdi: remove double conversion

This should be a problem when running on big-endian machines.

Signed-off-by: Paolo Bonzini <[email protected]>
Reviewed-by: Benoit Canet <[email protected]>
Reviewed-by: Stefan Weil <[email protected]>
Signed-off-by: Michael Tokarev <[email protected]>

apb: Fix compiler warnings (large constants)

Both constants need more than 32 bit.

Signed-off-by: Stefan Weil <[email protected]>
Signed-off-by: Michael Tokarev <[email protected]>

hw/net/ne2000-isa: Register vmstate struct

The ne2000-isa device defines a VMState struct for migration, but
we forgot to actually register it. Correct this deficiency by
setting dc->vmsd.

Signed-off-by: Peter Maydell <[email protected]>
Signed-off-by: Michael Tokarev <[email protected]>

target-microblaze: Delete unused sign_extend() function

The sign_extend() function is unused; delete it.

Signed-off-by: Peter Maydell <[email protected]>
Reviewed-by: Edgar E. Iglesias <[email protected]>
Signed-off-by: Michael Tokarev <[email protected]>

hw/misc/milkymist-softusb: Remove unused softusb_{read, write}_pmem()

The functions softusb_read_pmem() and softusb_write_pmem() are unused;
remove them.

Signed-off-by: Peter Maydell <[email protected]>
Signed-off-by: Michael Tokarev <[email protected]>

target-i386/translate.c: Remove unused tcg_gen_lshift()

The function tcg_gen_lshift() is unused; remove it.

Signed-off-by: Peter Maydell <[email protected]>
Signed-off-by: Michael Tokarev <[email protected]>

hw/isa/pc87312: Remove unused function is_parallel_epp()

The function is_parallel_epp() is unused; remove it.

Signed-off-by: Peter Maydell <[email protected]>
Signed-off-by: Michael Tokarev <[email protected]>

hw/intc/openpic: Remove unused function IRQ_testbit()

The IRQ_testbit() function is never used; remove it.

Signed-off-by: Peter Maydell <[email protected]>
Signed-off-by: Michael Tokarev <[email protected]>

hw/dma/xilinx_axidma: Remove unused stream_halted() function

The stream_halted() function is never used; remove it.

Signed-off-by: Peter Maydell <[email protected]>
Signed-off-by: Michael Tokarev <[email protected]>

util/qemu-sockets.c: Avoid unused variable warnings

The 'on' variable is never used, and 'off' is only used
if IPV6_V6ONLY is defined; delete 'on' and move 'off' to
the point where it is used. This avoids warnings from
clang 3.4.

Signed-off-by: Peter Maydell <[email protected]>
Signed-off-by: Michael Tokarev <[email protected]>

hw/sd/sd.c: Drop unused sd_acmd_type[] array

Drop the sd_acmd_type[] array: it is never used. (The equivalent
sd_cmd_type[] array for normal commands is used to identify
those commands whose argument includes the card address in the
top 16 bits; but for app commands the card address is passed
with the APP_CMD prefix, not with the argument to the app command
itself.)

Signed-off-by: Peter Maydell <[email protected]>
Reviewed-by: Peter Crosthwaite <[email protected]>
Signed-off-by: Michael Tokarev <[email protected]>

hw/i386/pc.c: Remove unused parallel_io and parallel_irq variables

The variables parallel_io and parallel_irq are unused; delete them.

Signed-off-by: Peter Maydell <[email protected]>
Signed-off-by: Michael Tokarev <[email protected]>

slirp: Remove unused zero_ethaddr[] variable

The zero_ethaddr[] array is never used; delete it.

Signed-off-by: Peter Maydell <[email protected]>
Signed-off-by: Michael Tokarev <[email protected]>

qtest: fix hex2nib for capital characters

Signed-off-by: Sergey Fedorov <[email protected]>
Reviewed-by: Markus Armbruster <[email protected]>
Reviewed-by: Stefan Hajnoczi <[email protected]>
Signed-off-by: Michael Tokarev <[email protected]>