VIRTIO_BLK_F_SCSI is supposed to mean whether the host can *parse*
SCSI requests, not *execute* them.  You could run QEMU with scsi=on
and a file-backed disk, and QEMU would fail all SCSI requests even
though it advertises VIRTIO_BLK_F_SCSI.

Because we need to do this to fix a migration compatibility problem
related to how QEMU is invoked by management, we must do this
unconditionally even on older machine types.  This more or less assumes
that no one ever invoked QEMU with scsi=off.

Here is how testing goes:

- old QEMU, scsi=on -> new QEMU, scsi=on
- new QEMU, scsi=on -> old QEMU, scsi=on
- old QEMU, scsi=off -> new QEMU, scsi=on
- new QEMU, scsi=off -> old QEMU, scsi=on
        ok (new QEMU has VIRTIO_BLK_F_SCSI, adding host features is fine)

- old QEMU, scsi=off -> new QEMU, scsi=off
        ok (new QEMU has VIRTIO_BLK_F_SCSI, adding host features is fine)

- old QEMU, scsi=on -> new QEMU, scsi=off
        ok, bug fixed

- new QEMU, scsi=on -> old QEMU, scsi=off
        doesn't work (same as: old QEMU, scsi=on -> old QEMU, scsi=off)

- new QEMU, scsi=off -> old QEMU, scsi=off
        broken by the patch
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NAnthony Liguori <aliguori@us.ibm.com>

We will have to add another field to the virtio-blk configuration in
the next patch.  Avoid a proliferation of arguments to virtio_blk_init.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NAnthony Liguori <aliguori@us.ibm.com>

Move it from virtio_blk_exit_pci to virtio_blk_exit.

This is included here because the next patch removes proxy->block.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NAnthony Liguori <aliguori@us.ibm.com>

Linux really looks only at scsi->errors for SG_IO requests; it does
not look at the virtio request status at all.  Because of this, when
a SG_IO request is failed early with virtio_blk_req_complete(req,
VIRTIO_BLK_S_UNSUPP), without writing hdr.status, it will look like
a success to the guest.

This is their bug, but we can make it safe for older guests now by
forcing scsi->errors to have a non-zero value whenever a request
has to be failed.

But if we fix the bug in the guest driver, we will have another problem
because QEMU returns VIRTIO_BLK_S_IOERR if the status is non-zero, and
Linux translates that to -EIO.  Rather, the guest should succeed the
request and pass the non-zero status via the userspace-provided SG_IO
structure.  So, remove the case where virtio_blk_handle_scsi can
return VIRTIO_BLK_S_IOERR.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NAnthony Liguori <aliguori@us.ibm.com>

Allow load_image_targphys to load files on systems with more than 2G of
emulated memory by changing the max_sz parameter from an int to an
uint64_t.
Reviewed-by: NAndreas F=E4rber <afaerber@suse.de>
Acked-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NMark Langsdorf <mark.langsdorf@calxeda.com>
Signed-off-by: NAnthony Liguori <aliguori@us.ibm.com>

* mdroth/qga-pull-5-15-12:
  qemu-ga: align versioning with QEMU_VERSION
  qemu-ga: fix segv after failure to open log file
  qemu-ga: guest-shutdown: use only async-signal-safe functions
  qemu-ga: guest-shutdown: become synchronous
  qemu-ga: guest-suspend: make the API synchronous
  qemu-ga: become_daemon(): reopen standard fds to /dev/null
  qemu-ga: make reopen_fd_to_null() public
  qemu-ga: guest-suspend-hybrid: don't emit a success response
  qemu-ga: guest-suspend-ram: don't emit a success response
  qemu-ga: guest-suspend-disk: don't emit a success response
  qemu-ga: guest-shutdown: don't emit a success response
  qemu-ga: don't warn on no command return
  qapi: add support for command options

Commit 93e9eb68 added fdc-test,
but accidentally removed rtc-test because check-qtest-i386-y was
not enhanced but set twice.

This patch adds rtc-test again (and sorts both tests alphabetically).
Signed-off-by: NStefan Weil <sw@weilnetz.de>
Signed-off-by: NAnthony Liguori <aliguori@us.ibm.com>

When QEMU was built with the simple trace backend, linking failed:

  LINK  tests/fdc-test
oslib-posix.o: In function `trace_qemu_memalign':
qemu/bin/debug/x86/./trace.h:31: undefined reference to `trace3'
oslib-posix.o: In function `trace_qemu_vmalloc':
qemu/bin/debug/x86/./trace.h:35: undefined reference to `trace2'
oslib-posix.o: In function `trace_qemu_vfree':
qemu/bin/debug/x86/./trace.h:39: undefined reference to `trace1'
collect2: error: ld returned 1 exit status
make: *** [tests/fdc-test] Fehler 1
Signed-off-by: NStefan Weil <sw@weilnetz.de>
Reviewed-by: NKevin Wolf <kwolf@redhat.com>
Signed-off-by: NBlue Swirl <blauwirbel@gmail.com>

There's some dodgy application of De Morgan's law in the emulation
of the MIPS BC1ANY[24]F instructions: they end up branching only
if all CCs are false, rather than if one CC is.

Tested on mips64-linux-gnu, where it fixes the GCC MIPS3D tests.
Signed-off-by: NRichard Sandiford <rdsandiford@googlemail.com>
Reviewed-by: NRichard Henderson <rth@twiddle.net>
Signed-off-by: NBlue Swirl <blauwirbel@gmail.com>

If we execute linux-user code that does the following:

  * A = mmap()
  * execute code in A
  * munmap(A)
  * B = mmap(), but mmap returns the same address as A
  * execute code in B

we end up executing a stale cached tb that contains translated code
from A, while we want new code from B.

This patch adds a TB flush for mmap'ed regions, before we return them,
avoiding the whole issue. It also adds a flush for munmap, so that we
don't execute stale TBs instead of getting a segfault.
Reported-by: NPeter Maydell <peter.maydell@linaro.org>
Signed-off-by: NAlexander Graf <agraf@suse.de>
Reviewed-by: NPeter Maydell <peter.maydell@linaro.org>
Acked-by: NRiku Voipio <riku.voipio@linaro.org>
Signed-off-by: NBlue Swirl <blauwirbel@gmail.com>

There are no outside references to virtio_portio.
Add missing 'static' specifier.
Reviewed-by: NStefan Weil <sw@weilnetz.de>
Signed-off-by: NBlue Swirl <blauwirbel@gmail.com>

Initrd load address is too low, it conflicts with kernel load
address:
rom: requested regions overlap (rom phdr #0: /tmp/vmlinux-debian-6.0.4-sparc64. free=0x0000000000742519, addr=0x0000000000400000)
rom loading failed

Fix by making the initrd address variable, load initrd after kernel
image. Use 64 bit variables instead of longs or 32 bit types.
Tested-by: NArtyom Tarasenko <atar4qemu@gmail.com>
Signed-off-by: NBlue Swirl <blauwirbel@gmail.com>

Split IN_T into BSIZE and ITYPE, to avoid expansion if the OS has
defined macros for the intX_t and uintX_t types. The IN_T constant is
then defined in mixeng_template.h so it can be used by the
functions/macros on this header file.

This change has been tested successfully under Debian Linux and NetBSD
6.0BETA.

Cc: Vassili Karpov (malc) <av1474@comtv.ru>
Signed-off-by: NRoger Pau Monne <roger.pau@citrix.com>
Signed-off-by: Nmalc <av1474@comtv.ru>

Implemented the swapb and swaph byte/halfword reversal instructions added
to microblaze v8.30
Signed-off-by: NPeter A. G. Crosthwaite <peter.crosthwaite@petalogix.com>
Signed-off-by: NEdgar E. Iglesias <edgar.iglesias@gmail.com>

Previously qemu-ga version was defined seperately. Since it is aligned
with QEMU releases, use QEMU_VERSION instead. This also implies the
version bump for 1.1[-rcN] release of qemu-ga.
Reviewed-by: NMichal Privoznik <mprivozn@redhat.com>
Acked-by: NLuiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: NMichael Roth <mdroth@linux.vnet.ibm.com>

Currently, if we fail to open the specified log file (generally due to a
permissions issue), we'll assign NULL to the logfile handle (stderr,
initially) used by the logging routines, which can cause a segfault to
occur when we attempt to report the error before exiting.

Instead, only re-assign if the open() was successful.
Reviewed-by: NMichal Privoznik <mprivozn@redhat.com>
Signed-off-by: NMichael Roth <mdroth@linux.vnet.ibm.com>

POSIX mandates[1] that a child process of a multi-thread program uses
only async-signal-safe functions before exec(). We consider qemu-ga
to be multi-thread, because it uses glib.

However, qmp_guest_shutdown() uses functions that are not
async-signal-safe. Fix it the following way:

- fclose() -> reopen_fd_to_null()
- execl() -> execle()
- exit() -> _exit()
- drop slog() usage (which is not safe)

  [1] http://pubs.opengroup.org/onlinepubs/009695399/functions/fork.htmlSigned-off-by: NLuiz Capitulino <lcapitulino@redhat.com>
Reviewed-by: NEric Blake <eblake@redhat.com>
Signed-off-by: NMichael Roth <mdroth@linux.vnet.ibm.com>

Last commit dropped qemu-ga's SIGCHLD handler, used to automatically
reap terminated children processes. This introduced a bug to
qmp_guest_shutdown(): it will generate zombies.

This problem probably doesn't matter in the success case, as the VM
will shutdown anyway, but let's do the right thing and reap the
created process. This ultimately means that guest-shutdown is now a
synchronous command.

An interesting side effect is that guest-shutdown is now able to
report an error to the client if shutting down fails.
Signed-off-by: NLuiz Capitulino <lcapitulino@redhat.com>
Reviewed-by: NEric Blake <eblake@redhat.com>
Signed-off-by: NMichael Roth <mdroth@linux.vnet.ibm.com>

Currently, qemu-ga has a SIGCHLD handler that automatically reaps terminated
children processes. The idea is to avoid having qemu-ga commands blocked
waiting for children to terminate.

That approach has two problems:

 1. qemu-ga is unable to detect errors in the child, meaning that qemu-ga
    returns success even if the child fails to perform its task

 2. if a command does depend on the child exit status, the command has to
    play tricks to bypass the automatic reaper

Case 2 impacts the guest-suspend-* API, because it has to execute an external
program to check for suspend support. Today, to bypass the automatic reaper,
suspend code has to double fork and pass exit status information through a
pipe. Besides being complex, this is prone to race condition bugs. Indeed,
the current code does have such bugs.

Making the guest-suspend-* API synchronous (ie. by dropping the SIGCHLD
handler and calling waitpid() from commands) is a much simpler approach,
which fixes current race conditions bugs and enables commands to detect
errors in the child.

This commit does just that. There's a side effect though, guest-shutdown
will generate zombies if shutting down fails. This will be fixed by the
next commit.
Signed-off-by: NLuiz Capitulino <lcapitulino@redhat.com>
Reviewed-by: NEric Blake <eblake@redhat.com>
Signed-off-by: NMichael Roth <mdroth@linux.vnet.ibm.com>

This fixes a bug where qemu-ga doesn't suspend the guest because it
fails to detect suspend support even when the guest does support
suspend. This happens because of the way qemu-ga fds are managed in
daemon mode.

When starting qemu-ga with --daemon, become_daemon() will close all
standard fds. This will cause qemu-ga to end up with the following
fds (if started with 'qemu-ga --daemon'):

    0 -> /dev/vport0p1
    3 -> /run/qemu-ga.pid

Then a guest-suspend-* function is issued. They call bios_supports_mode(),
which will call pipe(), and qemu-ga's fd will be:

    0 -> /dev/vport0p1
    1 -> pipe:[16247]
    2 -> pipe:[16247]
    3 -> /run/qemu-ga.pid

bios_supports_mode() forks off a child and blocks waiting for the child
to write something to the pipe. The child, however, closes its reading
end of the pipe _and_ reopen all standard fds to /dev/null. This will
cause the child's fds to be:

    0 -> /dev/null
    1 -> /dev/null
    2 -> /dev/null
    3 -> /run/qemu-ga.pid

In other words, the child's writing end of the pipe is now /dev/null.
It writes there and exits. The parent process (blocked on read()) will
get an EOF and interpret this as "something unexpected happened in
the child, let's assume the guest doesn't support suspend". And suspend
will fail.

To solve this problem we have to reopen standard fds to /dev/null
in become_daemon(), instead of closing them.
Signed-off-by: NLuiz Capitulino <lcapitulino@redhat.com>
Reviewed-by: NEric Blake <eblake@redhat.com>
Signed-off-by: NMichael Roth <mdroth@linux.vnet.ibm.com>

The next commit wants to use it.
Signed-off-by: NLuiz Capitulino <lcapitulino@redhat.com>
Reviewed-by: NEric Blake <eblake@redhat.com>
Signed-off-by: NMichael Roth <mdroth@linux.vnet.ibm.com>

Today, qemu-ga may not be able to emit a success response when
guest-suspend-hybrid completes. This happens because the VM may
suspend before qemu-ga is able to emit a response.

This semantic is a bit confusing, as it's not clear for clients if
they should wait for a response or how they should check for success.

This commit solves that problem by changing guest-suspend-hybrid to
never emit a success response and suggests in the documentation
what clients should do to check for success.
Signed-off-by: NLuiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: NMichael Roth <mdroth@linux.vnet.ibm.com>

Today, qemu-ga may not be able to emit a success response when
guest-suspend-ram completes. This happens because the VM may
suspend before qemu-ga is able to emit a response.

This semantic is a bit confusing, as it's not clear for clients if
they should wait for a response or how they should check for success.

This commit solves that problem by changing guest-suspend-ram to
never emit a success response and suggests in the documentation
what clients should do to check for success.
Signed-off-by: NLuiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: NMichael Roth <mdroth@linux.vnet.ibm.com>

Today, qemu-ga may not be able to emit a success response when
guest-suspend-disk completes. This happens because the VM may
vanish before qemu-ga is able to emit a response.

This semantic is a bit confusing, as it's not clear for clients if
they should wait for a response or how they should check for success.

This commit solves that problem by changing guest-suspend-disk to
never emit a success response and suggests in the documentation
what clients could do to check for success.
Signed-off-by: NLuiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: NMichael Roth <mdroth@linux.vnet.ibm.com>

Today, qemu-ga may not be able to emit a success response when
guest-shutdown completes. This happens because the VM may vanish
before qemu-ga is able to emit a response.

This semantic is a bit confusing, as it's not clear for clients if
they should wait for a response or how they should check for success.

This commit solves that problem by changing guest-shutdown to never
emit a success response and suggests in the documentation what
clients could do to check for success.
Signed-off-by: NLuiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: NMichael Roth <mdroth@linux.vnet.ibm.com>

This is a valid condition when a command chooses to not emit a
success response.
Signed-off-by: NLuiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: NMichael Roth <mdroth@linux.vnet.ibm.com>

Options allow for changes in commands behavior. This commit introduces
the QCO_NO_SUCCESS_RESP option, which causes a command to not emit a
success response.

This is needed by commands such as qemu-ga's guest-shutdown, which
may not be able to complete before the VM vanishes. In this case, it's
useful and simpler not to bother sending a success response.
Signed-off-by: NLuiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: NMichael Roth <mdroth@linux.vnet.ibm.com>

Signed-off-by: NAnthony Liguori <aliguori@us.ibm.com>

In qemu_ld/st load the registers for the helper calls directly rather
than rotating them around afterwards for AREG0.

Also clobber the additional register.
Signed-off-by: NAndreas F?rber <afaerber@suse.de>
Signed-off-by: Nmalc <av1474@comtv.ru>

Facilitates using r3 for prepended AREG0.
Signed-off-by: NAndreas F?rber <afaerber@suse.de>
Signed-off-by: Nmalc <av1474@comtv.ru>

* kwolf/for-anthony:
  qemu-img: Fix segmentation fault
  qcow2: Don't ignore failure to clear autoclear flags
  coroutine: Fix setup of sigaltstack coroutines

* qmp/queue/qmp:
  qapi: QMP input visitor, handle floats parsed as ints

* afaerber-or/qom-1.1:
  mips_fulong2e: Don't register "cpu" VMState twice
  pc: Add back PCI.rombar compat property
  qdev: Fix adding of ptr properties
  qdev: Use object_property_print() in info qtree
  target-i386: Defer MCE init
  qom: Documentation addition for object_class_by_name()
  target-mips: Remove commented-out function declaration

* kraxel/usb.50:
  usb-host: handle guest-issued clear halt

* spice/spice.v55:
  qxl: set size of PCI IO BAR correctly (16) for revision 2

* sweil/for-1.1:
  qemu-doc: Use QEMU instead of qemu for product name
  qemu-doc: Fix executable name in examples
  qemu-doc: Add missing parameter in description of -D option
  configure: Use QEMU instead of Qemu
  fix some common typos
  qemu-timer: Fix wrong error message

Since most property types do not have a parse property now, this was
broken.  Fix it by looking at the setter instead.
Reviewed-by: NAnthony Liguori <aliguori@us.ibm.com>
Acked-by: NAndreas F=E4rber <afaerber@suse.de>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NAnthony Liguori <aliguori@us.ibm.com>

The following command generates a segmentation fault.
qemu-img convert -O vpc -o ? test test2
This is because the 'goto out;' statement calls qemu_progress_end
before qemu_progress_init is called resulting in a NULL pointer
invocation.
Signed-off-by: NCharles Arnold <carnold@suse.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

Signed-off-by: NKevin Wolf <kwolf@redhat.com>

Use pthread_kill instead of process-wide kill to invoke the signal
handler used for stack switching. This may fix spurious lock-ups with
this backend, easily triggerable by extending the time window between
kill and sigsuspend.
Signed-off-by: NJan Kiszka <jan.kiszka@siemens.com>
Reviewed-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>