MAINTAINERS: self-appoint me as reviewer in build/test automation

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

MAINTAINERS: add Shippable automation platform URL

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

shippable: add mipsel target

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
[AJB: fixups after dropping powerpc]
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

shippable: add armeb-linux-user target

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

shippable: be verbose while building docker images

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

shippable: do not initialize submodules automatically

instead do it in the 'ci' target when needed.

for mips64el-softmmu target:
use dtc submodule if distrib packages are too old.

example with outdated libfdt on mips64el-softmmu target (required is >= 1.4.2):
 # dpkg-query --showformat='${Version}\n' --show libfdt-dev
 1.4.0+dfsg-1

shippable output:
----------------
  LINK    mips64el-softmmu/qemu-system-mips64el
../hw/core/loader-fit.o: In function `load_fit':
/root/src/github.com/philmd/qemu/hw/core/loader-fit.c:278: undefined reference to `fdt_first_subnode'
/root/src/github.com/philmd/qemu/hw/core/loader-fit.c:286: undefined reference to `fdt_next_subnode'
/root/src/github.com/philmd/qemu/hw/core/loader-fit.c:277: undefined reference to `fdt_first_subnode'
collect2: error: ld returned 1 exit status
Makefile:201: recipe for target 'qemu-system-mips64el' failed
make[1]: *** [qemu-system-mips64el] Error 1
Makefile:327: recipe for target 'subdir-mips64el-softmmu' failed
make: *** [subdir-mips64el-softmmu] Error 2

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

shippable: build using all available cpus

As of this commit:

$ echo "container proc:" `getconf _NPROCESSORS_ONLN` `getconf _NPROCESSORS_CONF`
container proc: 2 2

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

shippable: use C locale to simplify console output

remove this noise:

perl: warning: Setting locale failed.
perl: warning: Please check that your locale settings:
    LANGUAGE = (unset),
    LC_ALL = "en_US.UTF-8",
    LC_CTYPE = "en_US.UTF-8",
    LANG = "en_US.UTF-8"
    are supported and installed on your system.

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

docker: add mipsel build target

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
[AJB: remove apt-fake kludge]
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

docker: add extra libs to s390x target to extend codebase coverage

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

docker: add extra libs to arm64 target to extend codebase coverage

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

docker: add extra libs to armhf target to extend codebase coverage

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

docker: use eatmydata in debian arm64 image

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

docker: use eatmydata in debian armhf image

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

docker: use eatmydata, install common build packages in base image

The common build packages are: build-essential clang git bison flex

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
[AJB: fixups following stretch update]
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

docker: use better regex to generate deb-src entries

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
[AJB: fixed up following dropping emdebian]
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

docker: install ca-certificates package in base image

Resolve SSL verification issue at shippable container's git_sync stage:

shippable logs:
--------------
git_sync
- ssh-agent bash -c 'ssh-add /tmp/ssh/01_deploy; git clone https://github.com/philmd/qemu.git /root/src/github.com/philmd/qemu'
Identity added: /tmp/ssh/01_deploy (rsa w/o comment)
Cloning into '/root/src/github.com/philmd/qemu'...
fatal: unable to access 'https://github.com/philmd/qemu.git/': Problem with the SSL CA cert (path? access rights?)
retrying 1 of 3 times...

Suggested-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
[AJB: fixed re-base conflict following stretch updates]
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

docker: rebuild image if 'extra files' checksum does not match

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

docker: add --include-files argument to 'build' command

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

docker: let _copy_with_mkdir() sub_path argument be optional

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

docker: update qemu:debian base following stretch release

Debian has now released Stretch as its new stable. As we track
debian:stable-slim this has a few consequences. For one thing we can
now drop the emdebian hacks as cross compilers are part of the
official repositories now. However we do loose the ability to build
against powerpc (not ppc64) since that is no longer a release
architecture.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

tcg/tci: fix tcg-interpreter build

fix regression from commit 244f144134:

    $ make subdir-arm-softmmu
    make[1]: *** No rule to make target 'tci.o', needed by 'qemu-system-arm'.  Stop.
    Makefile:328: recipe for target 'subdir-arm-softmmu' failed
    make: *** [subdir-arm-softmmu] Error 2

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20170620163009.21764-1-f4bug@amsat.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/mst/tags/for_upstream' into staging

pc: fixes, cleanups, features

Some fixes and cleanups. Extended TSEG sizes.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
# gpg: Signature made Fri 16 Jun 2017 16:45:07 BST
# gpg:                using RSA key 0x281F0DB8D28D5469
# gpg: Good signature from "Michael S. Tsirkin <mst@kernel.org>"
# gpg:                 aka "Michael S. Tsirkin <mst@redhat.com>"
# Primary key fingerprint: 0270 606B 6F3C DF3D 0B17  0970 C350 3912 AFBE 8E67
#      Subkey fingerprint: 5D09 FD08 71C8 F85B 94CA  8A0D 281F 0DB8 D28D 5469

* remotes/mst/tags/for_upstream:
  hw/i386: fix nvdimm check error path
  intel_iommu: cleanup vtd_interrupt_remap_msi()
  intel_iommu: cleanup vtd_{do_}iommu_translate()
  intel_iommu: switching the rest DPRINTF to trace
  tests/q35-test: add TSEG size checks
  tests/q35-test: push down qtest_start / qtest_end to test case(s)
  q35/mch: implement extended TSEG sizes

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/famz/tags/docker-and-block-pull-request' into staging

# gpg: Signature made Fri 16 Jun 2017 01:18:46 BST
# gpg:                using RSA key 0xCA35624C6A9171C6
# gpg: Good signature from "Fam Zheng <famz@redhat.com>"
# gpg: WARNING: This key is not certified with sufficiently trusted signatures!
# gpg:          It is not certain that the signature belongs to the owner.
# Primary key fingerprint: 5003 7CB7 9706 0F76 F021  AD56 CA35 624C 6A91 71C6

* remotes/famz/tags/docker-and-block-pull-request: (23 commits)
  block: make accounting thread-safe
  block: split BlockAcctStats creation and setup
  block: introduce block_account_one_io
  block: protect modification of dirty bitmaps with a mutex
  migration/block: reset dirty bitmap before reading
  block: introduce dirty_bitmap_mutex
  block: protect tracked_requests and flush_queue with reqs_lock
  block: access write_gen with atomics
  block: use Stat64 for wr_highest_offset
  util: add stats64 module
  throttle-groups: protect throttled requests with a CoMutex
  throttle-groups: do not use qemu_co_enter_next
  throttle-groups: only start one coroutine from drained_begin
  block: access io_plugged with atomic ops
  block: access wakeup with atomic ops
  block: access serialising_in_flight with atomic ops
  block: access io_limits_disabled with atomic ops
  block: access quiesce_counter with atomic ops
  block: access copy_on_read with atomic ops
  docker: Add flex and bison to centos6 image
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/bonzini/tags/for-upstream' into staging

* nbd and qemu-nbd fixes (Eric, Max)
* nbd refactoring (Vladimir)
* vhost-user-scsi, take N+1 (Felipe)
* replace memory_region_set_fd with memory_region_init_ram_from_fd (Marc-André)
* docs/ movement (Paolo)
* megasas TOCTOU fixes (Paolo)
* make async_safe_run_on_cpu work on kvm/hax accelerators (Paolo)
* Build system and poison.h improvements (Thomas)
* -accel thread=xxx fix (Thomas)
* move files to accel/ (Yang Zhong)

# gpg: Signature made Thu 15 Jun 2017 10:51:55 BST
# gpg:                using RSA key 0xBFFBD25F78C7AE83
# gpg: Good signature from "Paolo Bonzini <bonzini@gnu.org>"
# gpg:                 aka "Paolo Bonzini <pbonzini@redhat.com>"
# Primary key fingerprint: 46F5 9FBD 57D6 12E7 BFD4  E2F7 7E15 100C CD36 69B1
#      Subkey fingerprint: F133 3857 4B66 2389 866C  7682 BFFB D25F 78C7 AE83

* remotes/bonzini/tags/for-upstream: (41 commits)
  vhost-user-scsi: Introduce a vhost-user-scsi sample application
  vhost-user-scsi: Introduce vhost-user-scsi host device
  qemu-doc: include version number
  docs: create interop/ subdirectory
  include/exec/poison: Mark some CONFIG defines as poisoned, too
  include/exec/poison: Add missing TARGET defines
  nbd/server: refactor nbd_trip
  nbd/server: rename rc to ret
  nbd/server: get rid of fail: return rc
  nbd/server: nbd_negotiate: fix error path
  nbd/server: remove NBDClientNewData
  nbd/server: refactor nbd_co_receive_request
  nbd/server: get rid of EAGAIN dead code
  nbd/server: refactor nbd_co_send_reply
  nbd/server: get rid of ssize_t
  nbd/server: get rid of nbd_negotiate_read and friends
  nbd: make nbd_drop public
  nbd: rename read_sync and friends
  accel: move kvm related accelerator files into accel/
  tcg: move tcg backend files into accel/tcg/
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/vivier/tags/m68k-for-2.10-pull-request' into staging

# gpg: Signature made Thu 15 Jun 2017 09:16:31 BST
# gpg:                using RSA key 0xF30C38BD3F2FBE3C
# gpg: Good signature from "Laurent Vivier <lvivier@redhat.com>"
# gpg:                 aka "Laurent Vivier <laurent@vivier.eu>"
# gpg:                 aka "Laurent Vivier (Red Hat) <lvivier@redhat.com>"
# Primary key fingerprint: CD2F 75DD C8E3 A4DC 2E4F  5173 F30C 38BD 3F2F BE3C

* remotes/vivier/tags/m68k-for-2.10-pull-request:
  target-m68k: define ext_opsize
  target-m68k: move FPU helpers to fpu_helper.c
  softfloat: define 680x0 specific values
  target/m68k: fix V flag for CC_OP_SUBx

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/cody/tags/block-pull-request' into staging

# gpg: Signature made Wed 14 Jun 2017 22:54:41 BST
# gpg:                using RSA key 0xBDBE7B27C0DE3057
# gpg: Good signature from "Jeffrey Cody <jcody@redhat.com>"
# gpg:                 aka "Jeffrey Cody <jeff@codyprime.org>"
# gpg:                 aka "Jeffrey Cody <codyprime@gmail.com>"
# Primary key fingerprint: 9957 4B4D 3474 90E7 9D98  D624 BDBE 7B27 C0DE 3057

* remotes/cody/tags/block-pull-request:
  block/iscsi: enable filename option and parsing
  block/rbd: enable filename option and parsing

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/kraxel/tags/pull-ui-20170614-1' into staging

ui: prefer gtk3 and sdl2, various fixes.

# gpg: Signature made Wed 14 Jun 2017 08:54:22 BST
# gpg:                using RSA key 0x4CB6D8EED3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <kraxel@redhat.com>"
# gpg:                 aka "Gerd Hoffmann <gerd@kraxel.org>"
# gpg:                 aka "Gerd Hoffmann (private) <kraxel@gmail.com>"
# Primary key fingerprint: A032 8CFF B93A 17A7 9901  FE7D 4CB6 D8EE D3E8 7138

* remotes/kraxel/tags/pull-ui-20170614-1:
  spice: don't enter opengl mode in case another UI provides opengl support
  sdl: prefer sdl2 over sdl1
  gtk: prefer gtk3 over gtk2
  spice: Use proper enum type for kbd led state
  Improve Cocoa modifier key handling

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/kraxel/tags/pull-usb-20170614-1' into staging

usb: bugfixes for ehci and xhci

# gpg: Signature made Wed 14 Jun 2017 08:44:01 BST
# gpg:                using RSA key 0x4CB6D8EED3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <kraxel@redhat.com>"
# gpg:                 aka "Gerd Hoffmann <gerd@kraxel.org>"
# gpg:                 aka "Gerd Hoffmann (private) <kraxel@gmail.com>"
# Primary key fingerprint: A032 8CFF B93A 17A7 9901  FE7D 4CB6 D8EE D3E8 7138

* remotes/kraxel/tags/pull-usb-20170614-1:
  ehci: stop recursive calls to ehci_work_bh
  xhci: only update dequeue ptr on completed transfers

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

migration: Fix race of image locking between src and dst

Previously, dst side will immediately try to lock the write byte upon
receiving QEMU_VM_EOF, but at src side, bdrv_inactivate_all() is only
done after sending it. If the src host is under load, dst may fail to
acquire the lock due to racing with the src unlocking it.

Fix this by hoisting the bdrv_inactivate_all() operation before
QEMU_VM_EOF.

N.B. A further improvement could possibly be done to cleanly handover
locks between src and dst, so that there is no window where a third QEMU
could steal the locks and prevent src and dst from running.

N.B. This commit includes a minor improvement to the error handling
by using qemu_file_set_error().

Reported-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Fam Zheng <famz@redhat.com>
Message-id: 20170616160658.32290-1-famz@redhat.com
Reviewed-by: Daniel P. Berrange <berrange@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
[PMM: noted qemu_file_set_error() use in commit as suggested by Daniel]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hw/i386: fix nvdimm check error path

Commit e987c37aee1752177906847630d32477da57e705 ("hw/i386: check if
nvdimm is enabled before plugging") introduced a check to reject nvdimm
hotplug if -machine pc,nvdimm=on was not given.

This check executes after pc_dimm_memory_plug() has already completed
and does not reverse the effect of this function in the case of failure.

Perform the check before calling pc_dimm_memory_plug().  This fixes the
following abort:

  $ qemu -M accel=kvm -m 1G,slots=4,maxmem=8G \
         -object memory-backend-file,id=mem1,share=on,mem-path=nvdimm.dat,size=1G
  (qemu) device_add nvdimm,memdev=mem1
  nvdimm is not enabled: missing 'nvdimm' in '-M'
  (qemu) device_add nvdimm,memdev=mem1
  Core dumped

The backtrace is:

  #0  0x00007fffdb5b191f in raise () at /lib64/libc.so.6
  #1  0x00007fffdb5b351a in abort () at /lib64/libc.so.6
  #2  0x00007fffdb5a9da7 in __assert_fail_base () at /lib64/libc.so.6
  #3  0x00007fffdb5a9e52 in  () at /lib64/libc.so.6
  #4  0x000055555577a5fa in qemu_ram_set_idstr (new_block=0x555556747a00, name=<optimized out>, dev=dev@entry=0x555556705590) at qemu/exec.c:1709
  #5  0x0000555555a0fe86 in vmstate_register_ram (mr=mr@entry=0x55555673a0e0, dev=dev@entry=0x555556705590) at migration/savevm.c:2293
  #6  0x0000555555965088 in pc_dimm_memory_plug (dev=dev@entry=0x555556705590, hpms=hpms@entry=0x5555566bb0e0, mr=mr@entry=0x555556705630, align=<optimized out>, errp=errp@entry=0x7fffffffc660)
      at hw/mem/pc-dimm.c:110
  #7  0x000055555581d89b in pc_dimm_plug (errp=0x7fffffffc6c0, dev=0x555556705590, hotplug_dev=<optimized out>) at qemu/hw/i386/pc.c:1713
  #8  0x000055555581d89b in pc_machine_device_plug_cb (hotplug_dev=<optimized out>, dev=0x555556705590, errp=0x7fffffffc6c0) at qemu/hw/i386/pc.c:2004
  #9  0x0000555555914da6 in device_set_realized (obj=<optimized out>, value=<optimized out>, errp=0x7fffffffc7e8) at hw/core/qdev.c:926

Cc: Haozhong Zhang <haozhong.zhang@intel.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Haozhong Zhang <haozhong.zhang@intel.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

intel_iommu: cleanup vtd_interrupt_remap_msi()

Move the memcpy upper into where needed, then share the trace so that we
trace every correct remapping.

Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

intel_iommu: cleanup vtd_{do_}iommu_translate()

First, let vtd_do_iommu_translate() return a status, so that we
explicitly knows whether error occured. Meanwhile, we make sure that
IOMMUTLBEntry is filled in in that.

Then, cleanup vtd_iommu_translate a bit. So even with PT we'll get a log
now. Also, remove useless assignments.

Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

intel_iommu: switching the rest DPRINTF to trace

We have converted many of the DPRINTF() into traces. This patch does the
last 100+ ones.

To debug VT-d when error happens, let's try enable:

  -trace enable="vtd_err*"

This should works just like the old GENERAL but of course better, since
we don't need to recompile.

Similar rules apply to the other modules. I was trying to make the
prefix good enough for sub-module debugging.

Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

tests/q35-test: add TSEG size checks

These checks verify that the guest RAM turns from read-write to
"blackhole" when crossing the low boundary of the TSEG. Both the standard
1MB/2MB/8MB TSEG sizes and an extended (16MB) TSEG size are tested.

Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

tests/q35-test: push down qtest_start / qtest_end to test case(s)

A test program can start up QEMU several times, with different command
lines. For such cases, qtest_start() and qtest_end() are called from
within the individual test functions. Examples: "virtio-console-test.c",
"numa-test.c", and many others.

Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

q35/mch: implement extended TSEG sizes

The q35 machine type currently lets the guest firmware select a 1MB, 2MB
or 8MB TSEG (basically, SMRAM) size. In edk2/OVMF, we use 8MB, but even
that is not enough when a lot of VCPUs (more than approx. 224) are
configured -- SMRAM footprint scales largely proportionally with VCPU
count.

Introduce a new property for "mch" called "extended-tseg-mbytes", which
expresses (in megabytes) the user's choice of TSEG (SMRAM) size.

Invent a new, QEMU-specific register in the config space of the DRAM
Controller, at offset 0x50, in order to allow guest firmware to query the
TSEG (SMRAM) size.

According to Intel Document Number 316966-002, Table 5-1 "DRAM Controller
Register Address Map (D0:F0)":

    Warning: Address locations that are not listed are considered Intel
             Reserved registers locations. Reads to Reserved registers may
             return non-zero values. Writes to reserved locations may
             cause system failures.

             All registers that are defined in the PCI 2.3 specification,
             but are not necessary or implemented in this component are
             simply not included in this document. The
             reserved/unimplemented space in the PCI configuration header
             space is not documented as such in this summary.

Offsets 0x50 and 0x51 are not listed in Table 5-1. They are also not part
of the standard PCI config space header. And they precede the capability
list as well, which starts at 0xe0 for this device.

When the guest writes value 0xffff to this register, the value that can be
read back is that of "mch.extended-tseg-mbytes" -- unless it remains
0xffff. The guest is required to write 0xffff first (as opposed to a
read-only register) because PCI config space is generally not cleared on
QEMU reset, and after S3 resume or reboot, new guest firmware running on
old QEMU could read a guest OS-injected value from this register.

After reading the available "extended" TSEG size, the guest firmware may
actually request that TSEG size by writing pattern 11b to the ESMRAMC
register's TSEG_SZ bit-field. (The Intel spec referenced above defines
only patterns 00b (1MB), 01b (2MB) and 10b (8MB); 11b is reserved.)

On the QEMU command line, the value can be set with

  -global mch.extended-tseg-mbytes=N

The default value for 2.10+ q35 machine types is 16. The value is limited
to 0xfff (4095) at the moment, purely so that the product (4095 MB) can be
stored to the uint32_t variable "tseg_size" in mch_update_smram(). Users
are responsible for choosing sensible TSEG sizes.

On 2.9 and earlier q35 machine types, the default value is 0. This lets
the 11b bit pattern in ESMRAMC.TSEG_SZ, and the register at offset 0x50,
keep their original behavior.

When "extended-tseg-mbytes" is nonzero, the new register at offset 0x50 is
set to that value on reset, for completeness.

PCI config space is migrated automatically, so no VMSD changes are
necessary.

Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Ref: https://bugzilla.redhat.com/show_bug.cgi?id=1447027
Ref: https://lists.01.org/pipermail/edk2-devel/2017-May/010456.html
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

block: make accounting thread-safe

I'm not trying too hard yet.  Later, with multiqueue support,
this may cause mutex contention or cacheline bouncing.

Cc: Alberto Garcia <berto@igalia.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20170605123908.18777-20-pbonzini@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

block: split BlockAcctStats creation and setup

block_acct_destroy is called unconditionally in blk_delete, but there is
no BlockAcctStats function that is called unconditionally in blk_new.
Split block_acct_init in two, so that it will be possible to create a
QemuMutex in block_acct_init and destroy it in block_acct_cleanup.

Cc: Alberto Garcia <berto@igalia.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20170605123908.18777-19-pbonzini@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

block: introduce block_account_one_io

This is the common code to account operations that produced actual I/O.

Reviewed-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20170605123908.18777-18-pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

block: protect modification of dirty bitmaps with a mutex

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20170605123908.18777-17-pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

migration/block: reset dirty bitmap before reading

Any data that is returned by read may be stale already, the bitmap
has to be cleared before issuing the read.

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20170605123908.18777-16-pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

block: introduce dirty_bitmap_mutex

It protects only the list of dirty bitmaps; in the next patch we will
also protect their content.

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20170605123908.18777-15-pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

block: protect tracked_requests and flush_queue with reqs_lock

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20170605123908.18777-14-pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

block: access write_gen with atomics

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20170605123908.18777-13-pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

block: use Stat64 for wr_highest_offset

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20170605123908.18777-12-pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

util: add stats64 module

This module provides fast paths for 64-bit atomic operations on machines
that only have 32-bit atomic access.

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>
Message-Id: <20170605123908.18777-11-pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

throttle-groups: protect throttled requests with a CoMutex

Another possibility is to use tg->lock, which we're holding anyway in
both schedule_next_request and throttle_group_co_io_limits_intercept.
This would require open-coding the CoQueue however, so I've chosen this
alternative.

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20170605123908.18777-10-pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

throttle-groups: do not use qemu_co_enter_next

Prepare for removing this function; always restart throttled requests
from coroutine context.  This will matter when restarting throttled
requests will have to acquire a CoMutex.

Reviewed-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20170605123908.18777-9-pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

throttle-groups: only start one coroutine from drained_begin

Starting all waiting coroutines from bdrv_drain_all is unnecessary;
throttle_group_co_io_limits_intercept calls schedule_next_request as
soon as the coroutine restarts, which in turn will restart the next
request if possible.

If we only start the first request and let the coroutines dance from
there the code is simpler and there is more reuse between
throttle_group_config, throttle_group_restart_blk and timer_cb.  The
next patch will benefit from this.

We also stop accessing from throttle_group_restart_blk the
blkp->throttled_reqs CoQueues even when there was no
attached throttling group.  This worked but is not pretty.

The only thing that can interrupt the dance is the QEMU_CLOCK_VIRTUAL
timer when switching from one block device to the next, because the
timer is set to "now + 1" but QEMU_CLOCK_VIRTUAL might not be running.
Set that timer to point in the present ("now") rather than the future
and things work.

Reviewed-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20170605123908.18777-8-pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

block: access io_plugged with atomic ops

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20170605123908.18777-7-pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

block: access wakeup with atomic ops

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20170605123908.18777-6-pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

block: access serialising_in_flight with atomic ops

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20170605123908.18777-5-pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

block: access io_limits_disabled with atomic ops

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20170605123908.18777-4-pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

block: access quiesce_counter with atomic ops

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20170605123908.18777-3-pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

block: access copy_on_read with atomic ops

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20170605123908.18777-2-pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

docker: Add flex and bison to centos6 image

Currently there are warnings about flex and bison being missing when
building in the centos6 image:

    make[1]: flex: Command not found
             BISON dtc-parser.tab.c
    make[1]: bison: Command not found

Add them.

Reported-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>
Message-Id: <20170524005206.31916-1-famz@redhat.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Fam Zheng <famz@redhat.com>

docker: Add libaio to fedora image

Signed-off-by: Fam Zheng <famz@redhat.com>
Message-Id: <20170505032340.26467-5-famz@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

docker: Add bzip2 and hostname to fedora image

It is used by qemu-iotests.

Signed-off-by: Fam Zheng <famz@redhat.com>
Message-Id: <20170505032340.26467-3-famz@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

docker: Run tests with current user

We've used --add-current-user to create a user in the image, use it to
run tests, because root has too much priviledge, and can surprise test
cases.

Signed-off-by: Fam Zheng <famz@redhat.com>
Message-Id: <20170505032340.26467-2-famz@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>

Merge remote-tracking branch 'remotes/rth/tags/pull-s390-20170613' into staging

Queued s390 patches

# gpg: Signature made Tue 13 Jun 2017 21:22:41 BST
# gpg:                using RSA key 0xAD1270CC4DD0279B
# gpg: Good signature from "Richard Henderson <rth7680@gmail.com>"
# gpg:                 aka "Richard Henderson <rth@redhat.com>"
# gpg:                 aka "Richard Henderson <rth@twiddle.net>"
# Primary key fingerprint: 9CB1 8DDA F8E8 49AD 2AFC  16A4 AD12 70CC 4DD0 279B

* remotes/rth/tags/pull-s390-20170613:
  s390x/cpumodel: wire up cpu type + id for TCG
  target/s390x: rework PGM interrupt psw.addr handling
  target/s390x: correctly indicate PER nullification

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

vhost-user-scsi: Introduce a vhost-user-scsi sample application

This commit introduces a vhost-user-scsi backend sample application. It
must be linked with libiscsi and libvhost-user.

To use it, compile with:
  $ make vhost-user-scsi

And run as follows:
  $ ./vhost-user-scsi -u vus.sock -i iscsi://uri_to_target/
  $ qemu-system-x86_64 --enable-kvm -m 512 \
      -object memory-backend-file,id=mem,size=512m,share=on,mem-path=guestmem \
      -numa node,memdev=mem \
      -chardev socket,id=vhost-user-scsi,path=vus.sock \
      -device vhost-user-scsi-pci,chardev=vhost-user-scsi \

The application is currently limited at one LUN only and it processes
requests synchronously (therefore only achieving QD1). The purpose of
the code is to show how a backend can be implemented and to test the
vhost-user-scsi Qemu implementation.

If a different instance of this vhost-user-scsi application is executed
at a remote host, a VM can be live migrated to such a host.

Signed-off-by: Felipe Franciosi <felipe@nutanix.com>
Message-Id: <1488479153-21203-5-git-send-email-felipe@nutanix.com>

vhost-user-scsi: Introduce vhost-user-scsi host device

This commit introduces a vhost-user device for SCSI. This is based
on the existing vhost-scsi implementation, but done over vhost-user
instead. It also uses a chardev to connect to the backend. Unlike
vhost-scsi (today), VMs using vhost-user-scsi can be live migrated.

To use it, start Qemu with a command line equivalent to:

qemu-system-x86_64 \
       -chardev socket,id=vus0,path=/tmp/vus.sock \
       -device vhost-user-scsi-pci,chardev=vus0,bus=pci.0,addr=...

A separate commit presents a sample application linked with libiscsi to
provide a backend for vhost-user-scsi.

Signed-off-by: Felipe Franciosi <felipe@nutanix.com>
Message-Id: <1488479153-21203-4-git-send-email-felipe@nutanix.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

qemu-doc: include version number

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

docs: create interop/ subdirectory

This is for the future interoperability & management guide.  It includes
the QAPI docs, including the automatically generated ones, other socket
protocols (vhost-user, VNC), and the qcow2 file format.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

include/exec/poison: Mark some CONFIG defines as poisoned, too

These are defined in config-target.h and thus should never be
used in common code.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <1497468113-2874-3-git-send-email-thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

include/exec/poison: Add missing TARGET defines

Since we've got some new CPU targets in QEMU during the last months
and years, we've got some new TARGET_xxx defines now which should
be marked as poisoned for common code.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <1497468113-2874-2-git-send-email-thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

nbd/server: refactor nbd_trip

- do not use 'goto error_reply' outside a switch to jump into the
  middle of the switch's default case label
- reduce code duplication

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20170602150150.258222-13-vsementsov@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

nbd/server: rename rc to ret

For consistency use 'ret' name for saving return code everywhere
in the file.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20170602150150.258222-12-vsementsov@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

nbd/server: get rid of fail: return rc

"goto fail" error handling scheme is not needed for just returning
error code. Better is return it immediately.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20170602150150.258222-11-vsementsov@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

nbd/server: nbd_negotiate: fix error path

Current code will return 0 on this nbd_write fail, as rc is 0
after successful nbd_negotiate_options. Fix this.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20170602150150.258222-10-vsementsov@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

nbd/server: remove NBDClientNewData

"co" field of NBDClientNewData has never been used, all the way back to
its declaration in commit 1a6245a5. So let's just use client pointer
instead of extra structure.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20170602150150.258222-9-vsementsov@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

nbd/server: refactor nbd_co_receive_request

Move function tail, about receiving next request out of the function.
Error path is simplified and nbd_co_receive_request becomes more
corresponding to its name.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20170602150150.258222-8-vsementsov@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

nbd/server: get rid of EAGAIN dead code

For now nbd_read never returns EAGAIN. So, don't handle it.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20170602150150.258222-7-vsementsov@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

nbd/server: refactor nbd_co_send_reply

As nbd_write never returns value > 0, we can get rid of extra ret.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20170602150150.258222-6-vsementsov@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

nbd/server: get rid of ssize_t

Now nbd_read and friends return int, so get rid of ssize_t.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20170602150150.258222-5-vsementsov@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

nbd/server: get rid of nbd_negotiate_read and friends

Functions nbd_negotiate_{read,write,drop_sync} were introduced in
1a6245a5b, when nbd_rwv (was nbd_wr_sync) was working through
qemu_co_sendv_recvv (the path is nbd_wr_sync -> qemu_co_{recv/send} ->
qemu_co_send_recv -> qemu_co_sendv_recvv), which just yields, without
setting any handlers. But starting from ff82911cd nbd_rwv (was
nbd_wr_syncv) works through qio_channel_yield() which sets handlers, so
watchers are redundant in nbd_negotiate_{read,write,drop_sync}, then,
let's just use nbd_{read,write,drop} functions.

Functions nbd_{read,write,drop} has errp parameter, which is unused in
this patch. This will be fixed later.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20170602150150.258222-4-vsementsov@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

nbd: make nbd_drop public

Following commit will reuse it for nbd server too.

Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20170602150150.258222-3-vsementsov@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

nbd: rename read_sync and friends

Rename
  nbd_wr_syncv -> nbd_rwv
  read_sync -> nbd_read
  read_sync_eof -> nbd_read_eof
  write_sync -> nbd_write
  drop_sync -> nbd_drop

1. nbd_ prefix
   read_sync and write_sync are already shared, so it is good to have a
   namespace prefix. drop_sync will be shared, and read_sync_eof is
   related to read_sync, so let's rename them all.

2. _sync suffix
   _sync is related to the fact that nbd_wr_syncv doesn't return if a
   write to socket returns EAGAIN. The first implementation of
   nbd_wr_syncv (was wr_sync in 7a5ca8648b) just loops while getting
   EAGAIN, the current implementation yields in this case.
   Why we want to get rid of it:
   - it is normal for r/w functions to be synchronous, so having an
     additional suffix for it looks redundant (contrariwise, we have
     _aio suffix for async functions)
   - _sync suffix in block layer is used when function does flush (so
     using it for other thing is confusing a bit)
   - keep function names short after adding nbd_ prefix

3. for nbd_wr_syncv let's use more common notation 'rw'

Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20170602150150.258222-2-vsementsov@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

accel: move kvm related accelerator files into accel/

move kvm related accelerator files into accel/ subdirectory, also
create one stub subdirectory, which will include accelerator's stub
files.

Signed-off-by: Yang Zhong <yang.zhong@intel.com>
Message-Id: <1496383606-18060-5-git-send-email-yang.zhong@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

tcg: move tcg backend files into accel/tcg/

move tcg-runtime.c, translate-all.(ch) and translate-common.c into
accel/tcg/ subdirectory and updated related trace-events file.

Signed-off-by: Yang Zhong <yang.zhong@intel.com>
Message-Id: <1496383606-18060-4-git-send-email-yang.zhong@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

tcg: move tcg related files into accel/tcg/ subdirectory

move cputlb.c, cpu-exec-common.c and cpu-exec.c related tcg exec
file into accel/tcg/ subdirectory.

Signed-off-by: Yang Zhong <yang.zhong@intel.com>
Message-Id: <1496383606-18060-3-git-send-email-yang.zhong@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

accel: split the tcg accelerator from accel.c file

there are some types of accelerators in qemu, and all accelerators
have their own file except tcg. tcg accelerator is also defined in
accel.c file. tcg accelerator file will be splited from accel.c and
re-name to tcg-all.c. accel/ directory will be created to include
kvm and tcg related files.

Signed-off-by: Yang Zhong <yang.zhong@intel.com>
Message-Id: <1496383606-18060-2-git-send-email-yang.zhong@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

qemu-nbd: Ignore SIGPIPE

qemu proper has done so for 13 years
(8a7ddc38a60648257dc0645ab4a05b33d6040063), qemu-img and qemu-io have
done so for four years (526eda14a68d5b3596be715505289b541288ef2a).
Ignoring this signal is especially important in qemu-nbd because
otherwise a client can easily take down the qemu-nbd server by dropping
the connection when the server wants to send something, for example:

$ qemu-nbd -x foo -f raw -t null-co:// &
[1] 12726
$ qemu-io -c quit nbd://localhost/bar
can't open device nbd://localhost/bar: No export with name 'bar' available
[1]  + 12726 broken pipe  qemu-nbd -x foo -f raw -t null-co://

In this case, the client sends an NBD_OPT_ABORT and closes the
connection (because it is not required to wait for a reply), but the
server replies with an NBD_REP_ACK (because it is required to reply).

Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20170611123714.31292-1-mreitz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

nbd: Fix regression on resiliency to port scan

Back in qemu 2.5, qemu-nbd was immune to port probes (a transient
server would not quit, regardless of how many probe connections
came and went, until a connection actually negotiated).  But we
broke that in commit ee7d7aa when removing the return value to
nbd_client_new(), although that patch also introduced a bug causing
an assertion failure on a client that fails negotiation.  We then
made it worse during refactoring in commit 1a6245a (a segfault
before we could even assert); the (masked) assertion was cleaned
up in d3780c2 (still in 2.6), and just recently we finally fixed
the segfault ("nbd: Fully intialize client in case of failed
negotiation").  But that still means that ever since we added
TLS support to qemu-nbd, we have been vulnerable to an ill-timed
port-scan being able to cause a denial of service by taking down
qemu-nbd before a real client has a chance to connect.

Since negotiation is now handled asynchronously via coroutines,
we no longer have a synchronous point of return by re-adding a
return value to nbd_client_new().  So this patch instead wires
things up to pass the negotiation status through the close_fn
callback function.

Simple test across two terminals:
$ qemu-nbd -f raw -p 30001 file
$ nmap 127.0.0.1 -p 30001 && \
  qemu-io -c 'r 0 512' -f raw nbd://localhost:30001

Note that this patch does not change what constitutes successful
negotiation (thus, a client must enter transmission phase before
that client can be considered as a reason to terminate the server
when the connection ends).  Perhaps we may want to tweak things
in a later patch to also treat a client that uses NBD_OPT_ABORT
as being a 'successful' negotiation (the client correctly talked
the NBD protocol, and informed us it was not going to use our
export after all), but that's a discussion for another day.

Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1451614
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <20170608222617.20376-1-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hax-all: make async_safe_run_on_cpu safe on HAX too

While at it, drop the current_cpu assignment since this is a
per-thread variable on modern QEMU.

Cc: Vincent Palatin <vpalatin@chromium.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

kvm-all: make async_safe_run_on_cpu safe on kvm too

Wrap the bulk of kvm_cpu_exec with cpu_exec_start/end, so that kvm
version can also enjoy performing certain operations while all vCPUs are
quiescent.

Signed-off-by: Roman Kagan <rkagan@virtuozzo.com>
Message-Id: <20170606181948.16238-15-rkagan@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

vl: Fix broken thread=xxx option of the --accel parameter

Commit bde4d9205 ("Fix the -accel parameter and the documentation for
'hax'") introduced a regression by adding a new local accel_opts
variable which shadows the variable with the same name that is
declared at the beginning of the main() scope. This causes the
qemu_tcg_configure() call later to be always called with NULL, so
that the thread=xxx option gets ignored. Fix it by removing the
local accel_opts variable and use "opts" instead, which is meant
for storing temporary QemuOpts values.
And while we're at it, also change the exit(1) here to exit(0)
since asking for help is not an error.

Fixes: bde4d9205ee9def98852ff6054cdef4efd74e1f8
Reported-by: Markus Armbruster <armbru@redhat.com>
Reported-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <1496899257-25800-1-git-send-email-thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

Makefile: Do not generate files if "configure" has not been run yet

When doing a "make -j10" in the vanilla QEMU source tree (without
running "configure" first), the Makefile currently generates two
files already, qemu-version.h and qemu-options.def. This should not
happen, so let's only build the generated files if config-host.mak
is available (i.e. "configure" has been run already).

Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <1496926799-13040-1-git-send-email-thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

megasas: always store SCSIRequest* into MegasasCmd

This ensures that the request is unref'ed properly, and avoids a
segmentation fault in the new qtest testcase that is added.
This is CVE-2017-9503.

Reported-by: Zhangyanyu <zyy4013@stu.ouc.edu.cn>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

megasas: do not read SCSI req parameters more than once from frame

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

megasas: do not read command more than once from frame

Avoid TOC-TOU bugs by passing the frame_cmd down, and checking
cmd->dcmd_opcode instead of cmd->frame->header.frame_cmd.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

megasas: do not read DCMD opcode more than once from frame

Avoid TOC-TOU bugs by storing the DCMD opcode in the MegasasCmd

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

megasas: do not read iovec count more than once from frame

Avoid TOC-TOU bugs depending on how the compiler behaves.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

megasas: do not read sense length more than once from frame

Avoid TOC-TOU bugs depending on how the compiler behaves.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

megasas: add qtest

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

memory: remove memory_region_set_fd

Now unnecessary since ivshmem uses memory_region_init_ram_from_fd.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20170602141229.15326-7-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

ivshmem: use ram_from_fd()

Instead of having its own mmap handling code, reuse the code from
exec.c.

Note: memory_region_init_ram_from_fd() adds some restrictions
(check for xen, kvm sync-mmu, etc) and changes (such as size
alignment). This may actually be more correct.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20170602141229.15326-6-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

Add memory_region_init_ram_from_fd()

Add a new function to initialize a RAM memory region with a file
descriptor to be mmap-ed.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20170602141229.15326-5-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

exec: split qemu_ram_alloc_from_file()

Add qemu_ram_alloc_from_fd(), which can be use to allocate ramblock from
fd only.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20170602141229.15326-4-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>