OSDN Git Service

qmiga/qemu.git
6 years agoMerge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20180309' into...
Peter Maydell [Mon, 12 Mar 2018 11:47:52 +0000 (11:47 +0000)]
Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20180309' into staging

target-arm queue:
 * i.MX: Add i.MX7 SOC implementation and i.MX7 Sabre board
 * Report the correct core count in A53 L2CTLR on the ZynqMP board
 * linux-user: preliminary SVE support work (signal handling)
 * hw/arm/boot: fix memory leak in case of error loading ELF file
 * hw/arm/boot: avoid reading off end of buffer if passed very
   small image file
 * hw/arm: Use more CONFIG switches for the object files
 * target/arm: Add "-cpu max" support
 * hw/arm/virt: Support -machine gic-version=max
 * hw/sd: improve debug tracing
 * hw/sd: sdcard: Add the Tuning Command (CMD 19)
 * MAINTAINERS: add Philippe as odd-fixes maintainer for SD

# gpg: Signature made Fri 09 Mar 2018 17:24:23 GMT
# gpg:                using RSA key 3C2525ED14360CDE
# gpg: Good signature from "Peter Maydell <peter.maydell@linaro.org>"
# gpg:                 aka "Peter Maydell <pmaydell@gmail.com>"
# gpg:                 aka "Peter Maydell <pmaydell@chiark.greenend.org.uk>"
# Primary key fingerprint: E1A5 C593 CD41 9DE2 8E83  15CF 3C25 25ED 1436 0CDE

* remotes/pmaydell/tags/pull-target-arm-20180309: (25 commits)
  MAINTAINERS: Add entries for SD (SDHCI, SDBus, SDCard)
  sdhci: Fix a typo in comment
  sdcard: Add the Tuning Command (CMD19)
  sdcard: Display which protocol is used when tracing (SD or SPI)
  sdcard: Display command name when tracing CMD/ACMD
  sdcard: Do not trace CMD55, except when we already expect an ACMD
  hw/arm/virt: Support -machine gic-version=max
  hw/arm/virt: Add "max" to the list of CPU types "virt" supports
  target/arm: Make 'any' CPU just an alias for 'max'
  target/arm: Add "-cpu max" support
  target/arm: Move definition of 'host' cpu type into cpu.c
  target/arm: Query host CPU features on-demand at instance init
  arm: avoid heap-buffer-overflow in load_aarch64_image
  arm: fix load ELF error leak
  hw/arm: Use more CONFIG switches for the object files
  aarch64-linux-user: Add support for SVE signal frame records
  aarch64-linux-user: Add support for EXTRA signal frame records
  aarch64-linux-user: Remove struct target_aux_context
  aarch64-linux-user: Split out helpers for guest signal handling
  linux-user: Implement aarch64 PR_SVE_SET/GET_VL
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agoMerge remote-tracking branch 'remotes/kevin/tags/for-upstream' into staging
Peter Maydell [Mon, 12 Mar 2018 10:08:09 +0000 (10:08 +0000)]
Merge remote-tracking branch 'remotes/kevin/tags/for-upstream' into staging

Block layer patches

# gpg: Signature made Fri 09 Mar 2018 15:09:20 GMT
# gpg:                using RSA key 7F09B272C88F2FD6
# gpg: Good signature from "Kevin Wolf <kwolf@redhat.com>"
# Primary key fingerprint: DC3D EB15 9A9A F95D 3D74  56FE 7F09 B272 C88F 2FD6

* remotes/kevin/tags/for-upstream: (56 commits)
  qemu-iotests: fix 203 migration completion race
  iotests: Tweak 030 in order to trigger a race condition with parallel jobs
  iotests: Skip test for ENOMEM error
  iotests: Mark all tests executable
  iotests: Test creating overlay when guest running
  qemu-iotests: Test ssh image creation over QMP
  qemu-iotests: Test qcow2 over file image creation with QMP
  block: Fail bdrv_truncate() with negative size
  file-posix: Fix no-op bdrv_truncate() with falloc preallocation
  ssh: Support .bdrv_co_create
  ssh: Pass BlockdevOptionsSsh to connect_to_ssh()
  ssh: QAPIfy host-key-check option
  ssh: Use QAPI BlockdevOptionsSsh object
  sheepdog: Support .bdrv_co_create
  sheepdog: QAPIfy "redundancy" create option
  nfs: Support .bdrv_co_create
  nfs: Use QAPI options in nfs_client_open()
  rbd: Use qemu_rbd_connect() in qemu_rbd_do_create()
  rbd: Assign s->snap/image_name in qemu_rbd_open()
  rbd: Support .bdrv_co_create
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agoMerge remote-tracking branch 'remotes/vivier/tags/m68k-for-2.12-pull-request' into...
Peter Maydell [Fri, 9 Mar 2018 18:49:27 +0000 (18:49 +0000)]
Merge remote-tracking branch 'remotes/vivier/tags/m68k-for-2.12-pull-request' into staging

# gpg: Signature made Fri 09 Mar 2018 14:54:33 GMT
# gpg:                using RSA key F30C38BD3F2FBE3C
# 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-2.12-pull-request:
  target/m68k: implement ftentox
  target/m68k: implement ftwotox
  target/m68k: implement fetox
  target/m68k: implement flog2
  target/m68k: implement flog10
  target/m68k: implement flogn
  target/m68k: implement flognp1
  target/m68k: define floatx80_move()

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agoMerge remote-tracking branch 'remotes/stefanha/tags/block-pull-request' into staging
Peter Maydell [Fri, 9 Mar 2018 17:28:16 +0000 (17:28 +0000)]
Merge remote-tracking branch 'remotes/stefanha/tags/block-pull-request' into staging

# gpg: Signature made Fri 09 Mar 2018 13:19:02 GMT
# gpg:                using RSA key 9CA4ABB381AB73C8
# gpg: Good signature from "Stefan Hajnoczi <stefanha@redhat.com>"
# gpg:                 aka "Stefan Hajnoczi <stefanha@gmail.com>"
# Primary key fingerprint: 8695 A8BF D3F9 7CDA AC35  775A 9CA4 ABB3 81AB 73C8

* remotes/stefanha/tags/block-pull-request:
  vl: introduce vm_shutdown()
  virtio-scsi: fix race between .ioeventfd_stop() and vq handler
  virtio-blk: fix race between .ioeventfd_stop() and vq handler
  block: add aio_wait_bh_oneshot()
  virtio-blk: dataplane: Don't batch notifications if EVENT_IDX is present
  README: Fix typo 'git-publish'
  block: Fix qemu crash when using scsi-block

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agoMAINTAINERS: Add entries for SD (SDHCI, SDBus, SDCard)
Philippe Mathieu-Daudé [Fri, 9 Mar 2018 17:09:45 +0000 (17:09 +0000)]
MAINTAINERS: Add entries for SD (SDHCI, SDBus, SDCard)

After spending months studying all the different SD Specifications
from the SD Association, voluntarily add myself as maintainer
for the SD code.

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20180309153654.13518-9-f4bug@amsat.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agosdhci: Fix a typo in comment
Philippe Mathieu-Daudé [Fri, 9 Mar 2018 17:09:45 +0000 (17:09 +0000)]
sdhci: Fix a typo in comment

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20180309153654.13518-8-f4bug@amsat.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agosdcard: Add the Tuning Command (CMD19)
Philippe Mathieu-Daudé [Fri, 9 Mar 2018 17:09:44 +0000 (17:09 +0000)]
sdcard: Add the Tuning Command (CMD19)

From the "Physical Layer Simplified Specification Version 3.01":

  A known data block ("Tuning block") can be used to tune sampling
  point for tuning required hosts. [...]
  This procedure gives the system optimal timing for each specific
  host and card combination and compensates for static delays in
  the timing budget including process, voltage and different PCB
  loads and skews. [...]
  Data block, carried by DAT[3:0], contains a pattern for tuning
  sampling position to receive data on the CMD and DAT[3:0] line.

[based on a patch from Alistair Francis <alistair.francis@xilinx.com>
 from qemu/xilinx tag xilinx-v2015.2]
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alistair Francis <alistair.francis@xilinx.com>
Message-id: 20180309153654.13518-5-f4bug@amsat.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agosdcard: Display which protocol is used when tracing (SD or SPI)
Philippe Mathieu-Daudé [Fri, 9 Mar 2018 17:09:44 +0000 (17:09 +0000)]
sdcard: Display which protocol is used when tracing (SD or SPI)

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20180309153654.13518-4-f4bug@amsat.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agosdcard: Display command name when tracing CMD/ACMD
Philippe Mathieu-Daudé [Fri, 9 Mar 2018 17:09:44 +0000 (17:09 +0000)]
sdcard: Display command name when tracing CMD/ACMD

The SDBus will reuse these functions, so we put them in a new source file.

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180309153654.13518-3-f4bug@amsat.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
[PMM: slight wordsmithing of comments, added note that string
 returned does not need to be freed]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agosdcard: Do not trace CMD55, except when we already expect an ACMD
Philippe Mathieu-Daudé [Fri, 9 Mar 2018 17:09:44 +0000 (17:09 +0000)]
sdcard: Do not trace CMD55, except when we already expect an ACMD

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Acked-by: Alistair Francis <alistair.francis@xilinx.com>
Message-id: 20180309153654.13518-2-f4bug@amsat.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agohw/arm/virt: Support -machine gic-version=max
Peter Maydell [Fri, 9 Mar 2018 17:09:44 +0000 (17:09 +0000)]
hw/arm/virt: Support -machine gic-version=max

Add support for passing 'max' to -machine gic-version. By analogy
with the -cpu max option, this picks the "best available" GIC version
whether you're using KVM or TCG, so it behaves like 'host' when
using KVM, and gives you GICv3 when using TCG.

Also like '-cpu host', using -machine gic-version=max' means there
is no guarantee of migration compatibility between QEMU versions;
in future 'max' might mean '4'.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20180308130626.12393-7-peter.maydell@linaro.org
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
6 years agohw/arm/virt: Add "max" to the list of CPU types "virt" supports
Peter Maydell [Fri, 9 Mar 2018 17:09:44 +0000 (17:09 +0000)]
hw/arm/virt: Add "max" to the list of CPU types "virt" supports

Allow the virt board to support '-cpu max' in the same way
it already handles '-cpu host'.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20180308130626.12393-6-peter.maydell@linaro.org
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
6 years agotarget/arm: Make 'any' CPU just an alias for 'max'
Peter Maydell [Fri, 9 Mar 2018 17:09:44 +0000 (17:09 +0000)]
target/arm: Make 'any' CPU just an alias for 'max'

Now we have a working '-cpu max', the linux-user-only
'any' CPU is pretty much the same thing, so implement it
that way.

For the moment we don't add any of the extra feature bits
to the system-emulation "max", because we don't set the
ID register bits we would need to to advertise those
features as present.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20180308130626.12393-5-peter.maydell@linaro.org
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
6 years agotarget/arm: Add "-cpu max" support
Peter Maydell [Fri, 9 Mar 2018 17:09:44 +0000 (17:09 +0000)]
target/arm: Add "-cpu max" support

Add support for "-cpu max" for ARM guests. This CPU type behaves
like "-cpu host" when KVM is enabled, and like a system CPU with
the maximum possible feature set otherwise. (Note that this means
it won't be migratable across versions, as we will likely add
features to it in future.)

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20180308130626.12393-4-peter.maydell@linaro.org

6 years agotarget/arm: Move definition of 'host' cpu type into cpu.c
Peter Maydell [Fri, 9 Mar 2018 17:09:44 +0000 (17:09 +0000)]
target/arm: Move definition of 'host' cpu type into cpu.c

Move the definition of the 'host' cpu type into cpu.c, where all the
other CPU types are defined.  We can do this now we've decoupled it
from the KVM-specific host feature probing.  This means we now create
the type unconditionally (assuming we were built with KVM support at
all), but if you try to use it without -enable-kvm this will end
up in the "host cpu probe failed and KVM not enabled" path in
arm_cpu_realizefn(), for an appropriate error message.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20180308130626.12393-3-peter.maydell@linaro.org

6 years agotarget/arm: Query host CPU features on-demand at instance init
Peter Maydell [Fri, 9 Mar 2018 17:09:44 +0000 (17:09 +0000)]
target/arm: Query host CPU features on-demand at instance init

Currently we query the host CPU features in the class init function
for the TYPE_ARM_HOST_CPU class, so that we can later copy them
from the class object into the instance object in the object
instance init function. This is awkward for implementing "-cpu max",
which should work like "-cpu host" for KVM but like "cpu with all
implemented features" for TCG.

Move the place where we store the information about the host CPU from
a class object to static variables in kvm.c, and then in the instance
init function call a new kvm_arm_set_cpu_features_from_host()
function which will query the host kernel if necessary and then
fill in the CPU instance fields.

This allows us to drop the special class struct and class init
function for TYPE_ARM_HOST_CPU entirely.

We can't delay the probe until realize, because the ARM
instance_post_init hook needs to look at the feature bits we
set, so we need to do it in the initfn. This is safe because
the probing doesn't affect the actual VM state (it creates a
separate scratch VM to do its testing), but the probe might fail.
Because we can't report errors in retrieving the host features
in the initfn, we check this belatedly in the realize function
(the intervening code will be able to cope with the relevant
fields in the CPU structure being zero).

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20180308130626.12393-2-peter.maydell@linaro.org

6 years agoarm: avoid heap-buffer-overflow in load_aarch64_image
Marc-André Lureau [Fri, 9 Mar 2018 17:09:44 +0000 (17:09 +0000)]
arm: avoid heap-buffer-overflow in load_aarch64_image

Spotted by ASAN:

elmarco@boraha:~/src/qemu/build (master *%)$ QTEST_QEMU_BINARY=aarch64-softmmu/qemu-system-aarch64 tests/boot-serial-test
/aarch64/boot-serial/virt: ** (process:19740): DEBUG: 18:39:30.275: foo /tmp/qtest-boot-serial-cXaS94D
=================================================================
==19740==ERROR: AddressSanitizer: heap-buffer-overflow on address 0x603000069648 at pc 0x7f1d2201cc54 bp 0x7fff331f6a40 sp 0x7fff331f61e8
READ of size 4 at 0x603000069648 thread T0
    #0 0x7f1d2201cc53  (/lib64/libasan.so.4+0xafc53)
    #1 0x55bc86685ee3 in load_aarch64_image /home/elmarco/src/qemu/hw/arm/boot.c:894
    #2 0x55bc86687217 in arm_load_kernel_notify /home/elmarco/src/qemu/hw/arm/boot.c:1047
    #3 0x55bc877363b5 in notifier_list_notify /home/elmarco/src/qemu/util/notify.c:40
    #4 0x55bc869331ea in qemu_run_machine_init_done_notifiers /home/elmarco/src/qemu/vl.c:2716
    #5 0x55bc8693bc39 in main /home/elmarco/src/qemu/vl.c:4679
    #6 0x7f1d1652c009 in __libc_start_main (/lib64/libc.so.6+0x21009)
    #7 0x55bc86255cc9 in _start (/home/elmarco/src/qemu/build/aarch64-softmmu/qemu-system-aarch64+0x1ae5cc9)

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agoarm: fix load ELF error leak
Marc-André Lureau [Fri, 9 Mar 2018 17:09:44 +0000 (17:09 +0000)]
arm: fix load ELF error leak

Spotted by ASAN:
QTEST_QEMU_BINARY=aarch64-softmmu/qemu-system-aarch64 tests/boot-serial-test

Direct leak of 48 byte(s) in 1 object(s) allocated from:
    #0 0x7ff8a9b0ca38 in __interceptor_calloc (/lib64/libasan.so.4+0xdea38)
    #1 0x7ff8a8ea7f75 in g_malloc0 ../glib/gmem.c:124
    #2 0x55fef3d99129 in error_setv /home/elmarco/src/qemu/util/error.c:59
    #3 0x55fef3d99738 in error_setg_internal /home/elmarco/src/qemu/util/error.c:95
    #4 0x55fef323acb2 in load_elf_hdr /home/elmarco/src/qemu/hw/core/loader.c:393
    #5 0x55fef2d15776 in arm_load_elf /home/elmarco/src/qemu/hw/arm/boot.c:830
    #6 0x55fef2d16d39 in arm_load_kernel_notify /home/elmarco/src/qemu/hw/arm/boot.c:1022
    #7 0x55fef3dc634d in notifier_list_notify /home/elmarco/src/qemu/util/notify.c:40
    #8 0x55fef2fc3182 in qemu_run_machine_init_done_notifiers /home/elmarco/src/qemu/vl.c:2716
    #9 0x55fef2fcbbd1 in main /home/elmarco/src/qemu/vl.c:4679
    #10 0x7ff89dfed009 in __libc_start_main (/lib64/libc.so.6+0x21009)

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agohw/arm: Use more CONFIG switches for the object files
Thomas Huth [Fri, 9 Mar 2018 17:09:44 +0000 (17:09 +0000)]
hw/arm: Use more CONFIG switches for the object files

A lot of ARM object files are linked into the executable unconditionally,
even though we have corresponding CONFIG switches like CONFIG_PXA2XX or
CONFIG_OMAP. We should make sure to use these switches in the Makefile so
that the users can disable certain unwanted boards and devices more easily.
While we're at it, also add some new switches for the boards that do not
have a CONFIG option yet.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-id: 1520266949-29817-1-git-send-email-thuth@redhat.com
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agoaarch64-linux-user: Add support for SVE signal frame records
Richard Henderson [Fri, 9 Mar 2018 17:09:44 +0000 (17:09 +0000)]
aarch64-linux-user: Add support for SVE signal frame records

Depending on the currently selected size of the SVE vector registers,
we can either store the data within the "standard" allocation, or we
may beedn to allocate additional space with an EXTRA record.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180303143823.27055-6-richard.henderson@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agoaarch64-linux-user: Add support for EXTRA signal frame records
Richard Henderson [Fri, 9 Mar 2018 17:09:43 +0000 (17:09 +0000)]
aarch64-linux-user: Add support for EXTRA signal frame records

The EXTRA record allows for additional space to be allocated
beyon what is currently reserved.  Add code to emit and read
this record type.

Nothing uses extra space yet.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180303143823.27055-5-richard.henderson@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agoaarch64-linux-user: Remove struct target_aux_context
Richard Henderson [Fri, 9 Mar 2018 17:09:43 +0000 (17:09 +0000)]
aarch64-linux-user: Remove struct target_aux_context

This changes the qemu signal frame layout to be more like the kernel's,
in that the various records are dynamically allocated rather than fixed
in place by a structure.

For now, all of the allocation is out of uc.tuc_mcontext.__reserved,
so the allocation is actually trivial.  That will change with SVE support.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180303143823.27055-4-richard.henderson@linaro.org
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agoaarch64-linux-user: Split out helpers for guest signal handling
Richard Henderson [Fri, 9 Mar 2018 17:09:43 +0000 (17:09 +0000)]
aarch64-linux-user: Split out helpers for guest signal handling

Split out helpers from target_setup_frame and target_restore_sigframe
for dealing with general registers, fpsimd registers, and the end record.

When we add support for sve registers, the relative positions of
these will change.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180303143823.27055-3-richard.henderson@linaro.org
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agolinux-user: Implement aarch64 PR_SVE_SET/GET_VL
Richard Henderson [Fri, 9 Mar 2018 17:09:43 +0000 (17:09 +0000)]
linux-user: Implement aarch64 PR_SVE_SET/GET_VL

As an implementation choice, widening VL has zeroed the
previously inaccessible portion of the sve registers.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Acked-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20180303143823.27055-2-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agoImplement support for i.MX7 Sabre board
Andrey Smirnov [Fri, 9 Mar 2018 17:09:43 +0000 (17:09 +0000)]
Implement support for i.MX7 Sabre board

Implement code needed to set up emulation of MCIMX7SABRE board from
NXP. For more info about the HW see:

https://www.nxp.com/support/developer-resources/hardware-development-tools/sabre-development-system/sabre-board-for-smart-devices-based-on-the-i.mx-7dual-applications-processors:MCIMX7SABRE

Cc: Peter Maydell <peter.maydell@linaro.org>
Cc: Jason Wang <jasowang@redhat.com>
Cc: Philippe Mathieu-Daudé <f4bug@amsat.org>
Cc: Marcel Apfelbaum <marcel.apfelbaum@zoho.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: qemu-devel@nongnu.org
Cc: qemu-arm@nongnu.org
Cc: yurovsky@gmail.com
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agoi.MX: Add i.MX7 SOC implementation.
Andrey Smirnov [Fri, 9 Mar 2018 17:09:43 +0000 (17:09 +0000)]
i.MX: Add i.MX7 SOC implementation.

The following interfaces are partially or fully emulated:

    * up to 2 Cortex A9 cores (SMP works with PSCI)
    * A7 MPCORE (identical to A15 MPCORE)
    * 4 GPTs modules
    * 7 GPIO controllers
    * 2 IOMUXC controllers
    * 1 CCM module
    * 1 SVNS module
    * 1 SRC module
    * 1 GPCv2 controller
    * 4 eCSPI controllers
    * 4 I2C controllers
    * 7 i.MX UART controllers
    * 2 FlexCAN controllers
    * 2 Ethernet controllers (FEC)
    * 3 SD controllers (USDHC)
    * 4 WDT modules
    * 1 SDMA module
    * 1 GPR module
    * 2 USBMISC modules
    * 2 ADC modules
    * 1 PCIe controller

Tested to boot and work with upstream Linux (4.13+) guest.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
[PMM: folded a couple of long lines]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agopci: Add support for Designware IP block
Andrey Smirnov [Fri, 9 Mar 2018 17:09:43 +0000 (17:09 +0000)]
pci: Add support for Designware IP block

Add code needed to get a functional PCI subsytem when using in
conjunction with upstream Linux guest (4.13+). Tested to work against
"e1000e" (network adapter, using MSI interrupts) as well as
"usb-ehci" (USB controller, using legacy PCI interrupts).

Based on "i.MX6 Applications Processor Reference Manual" (Document
Number: IMX6DQRM Rev. 4) as well as corresponding dirver in Linux
kernel (circa 4.13 - 4.16 found in drivers/pci/dwc/*)

Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agohw/arm: Set the core count for Xilinx's ZynqMP
Alistair Francis [Fri, 9 Mar 2018 17:09:43 +0000 (17:09 +0000)]
hw/arm: Set the core count for Xilinx's ZynqMP

Set the ARM CPU core count property for the A53's attached to the Xilnx
ZynqMP machine.

Signed-off-by: Alistair Francis <alistair.francis@xilinx.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: fe0dd90b85ac73f9fc9548c253bededa70a07006.1520018138.git.alistair.francis@xilinx.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agotarget/arm: Add a core count property
Alistair Francis [Fri, 9 Mar 2018 17:09:43 +0000 (17:09 +0000)]
target/arm: Add a core count property

The cortex A53 TRM specifies that bits 24 and 25 of the L2CTLR register
specify the number of cores in the processor, not the total number of
cores in the system. To report this correctly on machines with multiple
CPU clusters (ARM's big.LITTLE or Xilinx's ZynqMP) we need to allow
the machine to overwrite this value. To do this let's add an optional
property.

Signed-off-by: Alistair Francis <alistair.francis@xilinx.com>
Message-id: ef01d95c0759e88f47f22d11b14c91512a658b4f.1520018138.git.alistair.francis@xilinx.com
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agomemory: fix flatview_access_valid RCU read lock/unlock imbalance
Paolo Bonzini [Wed, 7 Mar 2018 13:02:38 +0000 (14:02 +0100)]
memory: fix flatview_access_valid RCU read lock/unlock imbalance

Fixes: 11e732a5ed46903f997985bed4c3767ca28a7eb6
Reported-by: Cornelia Huck <cohuck@redhat.com>
Reported-by: luigi burdo <intermediadc@hotmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Tested-by: Cornelia Huck <cohuck@redhat.com>
Tested-by: Thomas Huth <thuth@redhat.com>
Message-id: 20180307130238.19358-1-pbonzini@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agoMerge remote-tracking branch 'mreitz/tags/pull-block-2018-03-09' into queue-block
Kevin Wolf [Fri, 9 Mar 2018 15:09:06 +0000 (16:09 +0100)]
Merge remote-tracking branch 'mreitz/tags/pull-block-2018-03-09' into queue-block

Block patches

# gpg: Signature made Fri Mar  9 15:40:32 2018 CET
# gpg:                using RSA key F407DB0061D5CF40
# gpg: Good signature from "Max Reitz <mreitz@redhat.com>"
# Primary key fingerprint: 91BE B60A 30DB 3E88 57D1  1829 F407 DB00 61D5 CF40

* mreitz/tags/pull-block-2018-03-09:
  qemu-iotests: fix 203 migration completion race
  iotests: Tweak 030 in order to trigger a race condition with parallel jobs
  iotests: Skip test for ENOMEM error
  iotests: Mark all tests executable
  iotests: Test creating overlay when guest running

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
6 years agotarget/m68k: implement ftentox
Laurent Vivier [Mon, 5 Mar 2018 20:39:10 +0000 (21:39 +0100)]
target/m68k: implement ftentox

Using a local m68k floatx80_tentox()
[copied from previous:
Written by Andreas Grabher for Previous, NeXT Computer Emulator.]

Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20180305203910.10391-9-laurent@vivier.eu>

6 years agotarget/m68k: implement ftwotox
Laurent Vivier [Mon, 5 Mar 2018 20:39:09 +0000 (21:39 +0100)]
target/m68k: implement ftwotox

Using a local m68k floatx80_twotox()
[copied from previous:
Written by Andreas Grabher for Previous, NeXT Computer Emulator.]

Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20180305203910.10391-8-laurent@vivier.eu>

6 years agotarget/m68k: implement fetox
Laurent Vivier [Mon, 5 Mar 2018 20:39:08 +0000 (21:39 +0100)]
target/m68k: implement fetox

Using a local m68k floatx80_etox()
[copied from previous:
Written by Andreas Grabher for Previous, NeXT Computer Emulator.]

Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20180305203910.10391-7-laurent@vivier.eu>

6 years agotarget/m68k: implement flog2
Laurent Vivier [Mon, 5 Mar 2018 20:39:07 +0000 (21:39 +0100)]
target/m68k: implement flog2

Using a local m68k floatx80_log2()
[copied from previous:
Written by Andreas Grabher for Previous, NeXT Computer Emulator.]

Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20180305203910.10391-6-laurent@vivier.eu>

6 years agotarget/m68k: implement flog10
Laurent Vivier [Mon, 5 Mar 2018 20:39:06 +0000 (21:39 +0100)]
target/m68k: implement flog10

Using a local m68k floatx80_log10()
[copied from previous:
Written by Andreas Grabher for Previous, NeXT Computer Emulator.]

Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20180305203910.10391-5-laurent@vivier.eu>

6 years agotarget/m68k: implement flogn
Laurent Vivier [Mon, 5 Mar 2018 20:39:05 +0000 (21:39 +0100)]
target/m68k: implement flogn

Using a local m68k floatx80_logn()
[copied from previous:
Written by Andreas Grabher for Previous, NeXT Computer Emulator.]

Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20180305203910.10391-4-laurent@vivier.eu>

6 years agotarget/m68k: implement flognp1
Laurent Vivier [Mon, 5 Mar 2018 20:39:04 +0000 (21:39 +0100)]
target/m68k: implement flognp1

Using a local m68k floatx80_lognp1()
[copied from previous:
Written by Andreas Grabher for Previous, NeXT Computer Emulator.]

Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20180305203910.10391-3-laurent@vivier.eu>

6 years agoqemu-iotests: fix 203 migration completion race
Stefan Hajnoczi [Mon, 5 Mar 2018 15:59:26 +0000 (15:59 +0000)]
qemu-iotests: fix 203 migration completion race

There is a race between the test's 'query-migrate' QMP command after the
QMP 'STOP' event and completing the migration:

The test case invokes 'query-migrate' upon receiving 'STOP'.  At this
point the migration thread may still be in the process of completing.
Therefore 'query-migrate' can return 'status': 'active' for a brief
window of time instead of 'status': 'completed'.  This results in
qemu-iotests 203 hanging.

Solve the race by enabling the 'events' migration capability, which
causes QEMU to emit migration-specific QMP events that do not suffer
from this race condition.  Wait for the QMP 'MIGRATION' event with
'status': 'completed'.

Reported-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 20180305155926.25858-1-stefanha@redhat.com
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
6 years agoiotests: Tweak 030 in order to trigger a race condition with parallel jobs
Alberto Garcia [Tue, 6 Mar 2018 13:01:21 +0000 (15:01 +0200)]
iotests: Tweak 030 in order to trigger a race condition with parallel jobs

This patch tweaks TestParallelOps in iotest 030 so it allocates data
in smaller regions (256KB/512KB instead of 512KB/1MB) and the
block-stream job in test_stream_commit() only needs to copy data that
is at the very end of the image.

This way when the block-stream job is awakened it will finish right
away without any chance of being stopped by block_job_sleep_ns(). This
triggers the bug that was fixed by 3d5d319e1221082974711af1d09d82f and
1a63a907507fbbcfaee3f622907ec24 and is therefore a more useful test
case for parallel block jobs.

After this patch the aforementiond bug can also be reproduced with the
test_stream_parallel() test case.

Since with this change the stream job in test_stream_commit() finishes
early, this patch introduces a similar test case where both jobs are
slowed down so they can actually run in parallel.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Cc: John Snow <jsnow@redhat.com>
Message-id: 20180306130121.30243-1-berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
6 years agoiotests: Skip test for ENOMEM error
Fam Zheng [Thu, 1 Mar 2018 01:14:13 +0000 (09:14 +0800)]
iotests: Skip test for ENOMEM error

The AFL image is to exercise the code validating image size, which
doesn't work on 32 bit or when out of memory (there is a large
allocation before the interesting point). So check that and skip the
test, instead of faking the result.

Signed-off-by: Fam Zheng <famz@redhat.com>
Message-id: 20180301011413.11531-1-famz@redhat.com
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
6 years agoiotests: Mark all tests executable
Eric Blake [Mon, 5 Mar 2018 16:18:24 +0000 (10:18 -0600)]
iotests: Mark all tests executable

The majority of our iotests have the executable bit set; fix the
few outliers for consistency.

Signed-off-by: Eric Blake <eblake@redhat.com>
Message-id: 20180305161824.7188-1-eblake@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
6 years agoiotests: Test creating overlay when guest running
Fam Zheng [Mon, 25 Dec 2017 02:51:07 +0000 (10:51 +0800)]
iotests: Test creating overlay when guest running

Signed-off-by: Fam Zheng <famz@redhat.com>
Message-id: 20171225025107.23985-1-famz@redhat.com
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
6 years agosoftfloat: fix crash on int conversion of SNaN
Stef O'Rear [Sat, 3 Mar 2018 11:46:00 +0000 (03:46 -0800)]
softfloat: fix crash on int conversion of SNaN

Signed-off-by: Stef O'Rear <sorear2@gmail.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agoqemu-iotests: Test ssh image creation over QMP
Kevin Wolf [Thu, 8 Feb 2018 19:06:41 +0000 (20:06 +0100)]
qemu-iotests: Test ssh image creation over QMP

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agoqemu-iotests: Test qcow2 over file image creation with QMP
Kevin Wolf [Wed, 7 Feb 2018 15:42:44 +0000 (16:42 +0100)]
qemu-iotests: Test qcow2 over file image creation with QMP

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agoblock: Fail bdrv_truncate() with negative size
Kevin Wolf [Wed, 7 Feb 2018 16:58:47 +0000 (17:58 +0100)]
block: Fail bdrv_truncate() with negative size

Most callers have their own checks, but something like this should also
be checked centrally. As it happens, x-blockdev-create can pass negative
image sizes to format drivers (because there is no QAPI type that would
reject negative numbers) and triggers the check added by this patch.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
6 years agofile-posix: Fix no-op bdrv_truncate() with falloc preallocation
Kevin Wolf [Wed, 7 Feb 2018 15:38:05 +0000 (16:38 +0100)]
file-posix: Fix no-op bdrv_truncate() with falloc preallocation

If bdrv_truncate() is called, but the requested size is the same as
before, don't call posix_fallocate(), which returns -EINVAL for length
zero and would therefore make bdrv_truncate() fail.

The problem can be triggered by creating a zero-sized raw image with
'falloc' preallocation mode.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
6 years agossh: Support .bdrv_co_create
Kevin Wolf [Mon, 5 Feb 2018 15:24:32 +0000 (16:24 +0100)]
ssh: Support .bdrv_co_create

This adds the .bdrv_co_create driver callback to ssh, which enables
image creation over QMP.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agossh: Pass BlockdevOptionsSsh to connect_to_ssh()
Kevin Wolf [Mon, 5 Feb 2018 14:28:14 +0000 (15:28 +0100)]
ssh: Pass BlockdevOptionsSsh to connect_to_ssh()

Move the parsing of the QDict options up to the callers, in preparation
for the .bdrv_co_create implementation that directly gets a QAPI type.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agossh: QAPIfy host-key-check option
Kevin Wolf [Mon, 5 Feb 2018 13:59:05 +0000 (14:59 +0100)]
ssh: QAPIfy host-key-check option

This makes the host-key-check option available in blockdev-add.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agossh: Use QAPI BlockdevOptionsSsh object
Kevin Wolf [Fri, 2 Feb 2018 15:12:18 +0000 (16:12 +0100)]
ssh: Use QAPI BlockdevOptionsSsh object

Create a BlockdevOptionsSsh object in connect_to_ssh() and take the
options from there. 'host_key_check' is still processed separately
because it's not in the schema yet.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agosheepdog: Support .bdrv_co_create
Kevin Wolf [Thu, 1 Feb 2018 12:20:44 +0000 (13:20 +0100)]
sheepdog: Support .bdrv_co_create

This adds the .bdrv_co_create driver callback to sheepdog, which enables
image creation over QMP.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agosheepdog: QAPIfy "redundancy" create option
Kevin Wolf [Thu, 1 Feb 2018 16:26:27 +0000 (17:26 +0100)]
sheepdog: QAPIfy "redundancy" create option

The "redundancy" option for Sheepdog image creation is currently a
string that can encode one or two integers depending on its format,
which at the same time implicitly selects a mode.

This patch turns it into a QAPI union and converts the string into such
a QAPI object before interpreting the values.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agonfs: Support .bdrv_co_create
Kevin Wolf [Wed, 31 Jan 2018 15:27:38 +0000 (16:27 +0100)]
nfs: Support .bdrv_co_create

This adds the .bdrv_co_create driver callback to nfs, which enables
image creation over QMP.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agonfs: Use QAPI options in nfs_client_open()
Kevin Wolf [Wed, 31 Jan 2018 18:47:48 +0000 (19:47 +0100)]
nfs: Use QAPI options in nfs_client_open()

Using the QAPI visitor to turn all options into QAPI BlockdevOptionsNfs
simplifies the code a lot. It will also be useful for implementing the
QAPI based .bdrv_co_create callback.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agorbd: Use qemu_rbd_connect() in qemu_rbd_do_create()
Kevin Wolf [Fri, 16 Feb 2018 17:48:25 +0000 (18:48 +0100)]
rbd: Use qemu_rbd_connect() in qemu_rbd_do_create()

This is almost exactly the same code. The differences are that
qemu_rbd_connect() supports BlockdevOptionsRbd.server and that the cache
mode is set explicitly.

Supporting 'server' is a welcome new feature for image creation.
Caching is disabled by default, so leave it that way.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agorbd: Assign s->snap/image_name in qemu_rbd_open()
Kevin Wolf [Fri, 16 Feb 2018 17:54:52 +0000 (18:54 +0100)]
rbd: Assign s->snap/image_name in qemu_rbd_open()

Now that the options are already available in qemu_rbd_open() and not
only parsed in qemu_rbd_connect(), we can assign s->snap and
s->image_name there instead of passing the fields by reference to
qemu_rbd_connect().

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agorbd: Support .bdrv_co_create
Kevin Wolf [Wed, 31 Jan 2018 15:27:38 +0000 (16:27 +0100)]
rbd: Support .bdrv_co_create

This adds the .bdrv_co_create driver callback to rbd, which enables
image creation over QMP.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agorbd: Pass BlockdevOptionsRbd to qemu_rbd_connect()
Kevin Wolf [Thu, 15 Feb 2018 19:58:24 +0000 (20:58 +0100)]
rbd: Pass BlockdevOptionsRbd to qemu_rbd_connect()

With the conversion to a QAPI options object, the function is now
prepared to be used in a .bdrv_co_create implementation.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agorbd: Remove non-schema options from runtime_opts
Kevin Wolf [Thu, 15 Feb 2018 19:31:04 +0000 (20:31 +0100)]
rbd: Remove non-schema options from runtime_opts

Instead of the QemuOpts in qemu_rbd_connect(), we want to use QAPI
objects. As a preparation, fetch those options directly from the QDict
that .bdrv_open() supports in the rbd driver and that are not in the
schema.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agorbd: Factor out qemu_rbd_connect()
Kevin Wolf [Thu, 15 Feb 2018 18:13:47 +0000 (19:13 +0100)]
rbd: Factor out qemu_rbd_connect()

The code to establish an RBD connection is duplicated between open and
create. In order to be able to share the code, factor out the code from
qemu_rbd_open() as a first step.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agorbd: Fix use after free in qemu_rbd_set_keypairs() error path
Kevin Wolf [Fri, 16 Feb 2018 18:14:55 +0000 (19:14 +0100)]
rbd: Fix use after free in qemu_rbd_set_keypairs() error path

If we want to include the invalid option name in the error message, we
can't free the string earlier than that.

Cc: qemu-stable@nongnu.org
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
6 years agogluster: Support .bdrv_co_create
Kevin Wolf [Wed, 31 Jan 2018 15:27:38 +0000 (16:27 +0100)]
gluster: Support .bdrv_co_create

This adds the .bdrv_co_create driver callback to gluster, which enables
image creation over QMP.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
6 years agofile-win32: Support .bdrv_co_create
Kevin Wolf [Mon, 5 Feb 2018 15:24:32 +0000 (16:24 +0100)]
file-win32: Support .bdrv_co_create

This adds the .bdrv_co_create driver callback to file-win32, which
enables image creation over QMP.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
6 years agofile-posix: Support .bdrv_co_create
Kevin Wolf [Tue, 16 Jan 2018 15:04:21 +0000 (16:04 +0100)]
file-posix: Support .bdrv_co_create

This adds the .bdrv_co_create driver callback to file, which enables
image creation over QMP.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
6 years agoblock: x-blockdev-create QMP command
Kevin Wolf [Tue, 9 Jan 2018 15:50:57 +0000 (16:50 +0100)]
block: x-blockdev-create QMP command

This adds a synchronous x-blockdev-create QMP command that can create
qcow2 images on a given node name.

We don't want to block while creating an image, so this is not the final
interface in all aspects, but BlockdevCreateOptionsQcow2 and
.bdrv_co_create() are what they actually might look like in the end. In
any case, this should be good enough to test whether we interpret
BlockdevCreateOptions as we should.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
6 years agoblock: Make bdrv_is_whitelisted() public
Kevin Wolf [Wed, 21 Feb 2018 10:47:43 +0000 (11:47 +0100)]
block: Make bdrv_is_whitelisted() public

We'll use a separate source file for image creation, and we need to
check there whether the requested driver is whitelisted.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
6 years agoqcow2: Use visitor for options in qcow2_create()
Kevin Wolf [Thu, 11 Jan 2018 15:18:08 +0000 (16:18 +0100)]
qcow2: Use visitor for options in qcow2_create()

Instead of manually creating the BlockdevCreateOptions object, use a
visitor to parse the given options into the QAPI object.

This involves translation from the old command line syntax to the syntax
mandated by the QAPI schema. Option names are still checked against
qcow2_create_opts, so only the old option names are allowed on the
command line, even if they are translated in qcow2_create().

In contrast, new option values are optionally recognised besides the old
values: 'compat' accepts 'v2'/'v3' as an alias for '0.10'/'1.1', and
'encrypt.format' accepts 'qcow' as an alias for 'aes' now.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
6 years agoqdict: Introduce qdict_rename_keys()
Kevin Wolf [Thu, 1 Feb 2018 19:00:44 +0000 (20:00 +0100)]
qdict: Introduce qdict_rename_keys()

A few block drivers will need to rename .bdrv_create options for their
QAPIfication, so let's have a helper function for that.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
6 years agotest-qemu-opts: Test qemu_opts_to_qdict_filtered()
Kevin Wolf [Mon, 19 Feb 2018 18:05:30 +0000 (19:05 +0100)]
test-qemu-opts: Test qemu_opts_to_qdict_filtered()

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agotest-qemu-opts: Test qemu_opts_append()
Kevin Wolf [Mon, 19 Feb 2018 16:28:08 +0000 (17:28 +0100)]
test-qemu-opts: Test qemu_opts_append()

Basic test for merging two QemuOptsLists.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agoutil: Add qemu_opts_to_qdict_filtered()
Kevin Wolf [Thu, 11 Jan 2018 16:37:23 +0000 (17:37 +0100)]
util: Add qemu_opts_to_qdict_filtered()

This allows, given a QemuOpts for a QemuOptsList that was merged from
multiple QemuOptsList, to only consider those options that exist in one
specific list. Block drivers need this to separate format-layer create
options from protocol-level options.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agoqcow2: Handle full/falloc preallocation in qcow2_co_create()
Kevin Wolf [Thu, 11 Jan 2018 13:38:34 +0000 (14:38 +0100)]
qcow2: Handle full/falloc preallocation in qcow2_co_create()

Once qcow2_co_create() can be called directly on an already existing
node, we must provide the 'full' and 'falloc' preallocation modes
outside of creating the image on the protocol layer. Fortunately, we
have preallocated truncate now which can provide this functionality.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agoqcow2: Use QCryptoBlockCreateOptions in qcow2_co_create()
Kevin Wolf [Wed, 10 Jan 2018 16:55:16 +0000 (17:55 +0100)]
qcow2: Use QCryptoBlockCreateOptions in qcow2_co_create()

Instead of passing the encryption format name and the QemuOpts down, use
the QCryptoBlockCreateOptions contained in BlockdevCreateOptions.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agoqcow2: Use BlockdevRef in qcow2_co_create()
Kevin Wolf [Wed, 10 Jan 2018 14:52:33 +0000 (15:52 +0100)]
qcow2: Use BlockdevRef in qcow2_co_create()

Instead of passing a separate BlockDriverState* into qcow2_co_create(),
make use of the BlockdevRef that is included in BlockdevCreateOptions.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agoqcow2: Pass BlockdevCreateOptions to qcow2_co_create()
Kevin Wolf [Tue, 9 Jan 2018 18:44:33 +0000 (19:44 +0100)]
qcow2: Pass BlockdevCreateOptions to qcow2_co_create()

All of the simple options are now passed to qcow2_co_create() in a
BlockdevCreateOptions object. Still missing: node-name and the
encryption options.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
6 years agoqcow2: Let qcow2_create() handle protocol layer
Kevin Wolf [Tue, 9 Jan 2018 16:35:26 +0000 (17:35 +0100)]
qcow2: Let qcow2_create() handle protocol layer

Currently, qcow2_create() only parses the QemuOpts and then calls
qcow2_co_create() for the actual image creation, which includes both the
creation of the actual file on the file system and writing a valid empty
qcow2 image into that file.

The plan is that qcow2_co_create() becomes the function that implements
the functionality for a future 'blockdev-create' QMP command, which only
creates the qcow2 layer on an already opened file node.

This is a first step towards that goal: Let's move out anything that
deals with the protocol layer from qcow2_co_create() into
qcow2_create().  This means that qcow2_co_create() doesn't need a file
name any more.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agoqcow2: Rename qcow2_co_create2() to qcow2_co_create()
Kevin Wolf [Thu, 1 Mar 2018 16:45:23 +0000 (17:45 +0100)]
qcow2: Rename qcow2_co_create2() to qcow2_co_create()

The functions originally known as qcow2_create() and qcow2_create2()
are now called qcow2_co_create_opts() and qcow2_co_create(), which
matches the names of the BlockDriver callbacks that they will implement
at the end of this patch series.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agoblock/qapi: Add qcow2 create options to schema
Kevin Wolf [Fri, 24 Nov 2017 15:01:07 +0000 (16:01 +0100)]
block/qapi: Add qcow2 create options to schema

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agoblock/qapi: Introduce BlockdevCreateOptions
Kevin Wolf [Fri, 24 Nov 2017 15:01:07 +0000 (16:01 +0100)]
block/qapi: Introduce BlockdevCreateOptions

This creates a BlockdevCreateOptions union type that will contain all of
the options for image creation. We'll start out with an empty struct
type BlockdevCreateNotSupported for all drivers.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
6 years agoqcow2: Make qemu-img check detect corrupted L1 tables in snapshots
Alberto Garcia [Tue, 6 Mar 2018 16:14:12 +0000 (18:14 +0200)]
qcow2: Make qemu-img check detect corrupted L1 tables in snapshots

'qemu-img check' cannot detect if a snapshot's L1 table is corrupted.
This patch checks the table's offset and size and reports corruption
if the values are not valid.

This patch doesn't add code to fix that corruption yet, only to detect
and report it.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
6 years agoqcow2: Check snapshot L1 table in qcow2_snapshot_delete()
Alberto Garcia [Tue, 6 Mar 2018 16:14:11 +0000 (18:14 +0200)]
qcow2: Check snapshot L1 table in qcow2_snapshot_delete()

This function deletes a snapshot from disk, removing its entry from
the snapshot table, freeing its L1 table and decreasing the refcounts
of all clusters.

The L1 table offset and size are however not validated. If we use
invalid values in this function we'll probably corrupt the image even
more, so we should return an error instead.

We now have a function to take care of this, so let's use it.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
6 years agoqcow2: Check snapshot L1 table in qcow2_snapshot_goto()
Alberto Garcia [Tue, 6 Mar 2018 16:14:10 +0000 (18:14 +0200)]
qcow2: Check snapshot L1 table in qcow2_snapshot_goto()

This function copies a snapshot's L1 table into the active one without
validating it first.

We now have a function to take care of this, so let's use it.

Cc: Eric Blake <eblake@redhat.com>
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
6 years agoqcow2: Check snapshot L1 tables in qcow2_check_metadata_overlap()
Alberto Garcia [Tue, 6 Mar 2018 16:14:09 +0000 (18:14 +0200)]
qcow2: Check snapshot L1 tables in qcow2_check_metadata_overlap()

The inactive-l2 overlap check iterates uses the L1 tables from all
snapshots, but it does not validate them first.

We now have a function to take care of this, so let's use it.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
6 years agoqcow2: Check L1 table parameters in qcow2_expand_zero_clusters()
Alberto Garcia [Tue, 6 Mar 2018 16:14:08 +0000 (18:14 +0200)]
qcow2: Check L1 table parameters in qcow2_expand_zero_clusters()

This function iterates over all snapshots of a qcow2 file in order to
expand all zero clusters, but it does not validate the snapshots' L1
tables first.

We now have a function to take care of this, so let's use it.

We can also take the opportunity to replace the sector-based
bdrv_read() with bdrv_pread().

Cc: Eric Blake <eblake@redhat.com>
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
6 years agoqcow2: Check L1 table offset in qcow2_snapshot_load_tmp()
Alberto Garcia [Tue, 6 Mar 2018 16:14:07 +0000 (18:14 +0200)]
qcow2: Check L1 table offset in qcow2_snapshot_load_tmp()

This function checks that the size of a snapshot's L1 table is not too
large, but it doesn't validate the offset.

We now have a function to take care of this, so let's use it.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
6 years agoqcow2: Generalize validate_table_offset() into qcow2_validate_table()
Alberto Garcia [Tue, 6 Mar 2018 16:14:06 +0000 (18:14 +0200)]
qcow2: Generalize validate_table_offset() into qcow2_validate_table()

This function checks that the offset and size of a table are valid.

While the offset checks are fine, the size check is too generic, since
it only verifies that the total size in bytes fits in a 64-bit
integer. In practice all tables used in qcow2 have much smaller size
limits, so the size needs to be checked again for each table using its
actual limit.

This patch generalizes this function by allowing the caller to specify
the maximum size for that table. In addition to that it allows passing
an Error variable.

The function is also renamed and made public since we're going to use
it in other parts of the code.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
6 years agoblock: convert bdrv_check callback to coroutine_fn
Paolo Bonzini [Thu, 1 Mar 2018 16:36:19 +0000 (17:36 +0100)]
block: convert bdrv_check callback to coroutine_fn

Suggested-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <1516279431-30424-8-git-send-email-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
6 years agoblock: convert bdrv_invalidate_cache callback to coroutine_fn
Paolo Bonzini [Thu, 1 Mar 2018 16:36:18 +0000 (17:36 +0100)]
block: convert bdrv_invalidate_cache callback to coroutine_fn

QED's bdrv_invalidate_cache implementation would like to reuse functions
that acquire/release the metadata locks.  Call it from coroutine context
to simplify the logic.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <1516279431-30424-6-git-send-email-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
6 years agoqed: make bdrv_qed_do_open a coroutine_fn
Paolo Bonzini [Thu, 1 Mar 2018 16:36:17 +0000 (17:36 +0100)]
qed: make bdrv_qed_do_open a coroutine_fn

It is called from qed_invalidate_cache in coroutine context (incoming
migration runs in a coroutine), so it's cleaner if metadata is always
loaded from a coroutine.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <1516279431-30424-5-git-send-email-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
6 years agoqcow2: make qcow2_do_open a coroutine_fn
Paolo Bonzini [Thu, 1 Mar 2018 16:36:16 +0000 (17:36 +0100)]
qcow2: make qcow2_do_open a coroutine_fn

It is called from qcow2_invalidate_cache in coroutine context (incoming
migration runs in a coroutine), so it's cleaner if metadata is always
loaded from a coroutine.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <1516279431-30424-4-git-send-email-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
6 years agoqcow2: fix flushing after dirty bitmap metadata writes
Paolo Bonzini [Thu, 1 Mar 2018 16:36:15 +0000 (17:36 +0100)]
qcow2: fix flushing after dirty bitmap metadata writes

update_header_sync itself does not need to flush the caches to disk.
The only paths that allocate clusters are:

- bitmap_list_store with in_place=false, called by update_ext_header_and_dir

- store_bitmap_data, called by store_bitmap

- store_bitmap, called by qcow2_store_persistent_dirty_bitmaps and
  followed by update_ext_header_and_dir

So in the end the central place where we need to flush the caches
is update_ext_header_and_dir.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
6 years agoqcow2: introduce qcow2_write_caches and qcow2_flush_caches
Paolo Bonzini [Thu, 1 Mar 2018 16:36:14 +0000 (17:36 +0100)]
qcow2: introduce qcow2_write_caches and qcow2_flush_caches

They will be used to avoid recursively taking s->lock during
bdrv_open or bdrv_check.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <1516279431-30424-7-git-send-email-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
6 years agoblock: implement the bdrv_reopen_prepare helper for LUKS driver
Daniel P. Berrange [Thu, 18 Jan 2018 10:31:43 +0000 (10:31 +0000)]
block: implement the bdrv_reopen_prepare helper for LUKS driver

If the bdrv_reopen_prepare helper isn't provided, the qemu-img commit
command fails to re-open the base layer after committing changes into
it. Provide a no-op implementation for the LUKS driver, since there
is not any custom work that needs doing to re-open it.

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
6 years agotarget/m68k: define floatx80_move()
Laurent Vivier [Mon, 5 Mar 2018 20:39:03 +0000 (21:39 +0100)]
target/m68k: define floatx80_move()

This functions is needed by upcoming m68k softfloat functions.

Source code copied for WinUAE (tag 3500)
(The WinUAE file has been copied from QEMU and has
the QEMU licensing notice)

Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20180305203910.10391-2-laurent@vivier.eu>

6 years agoMerge remote-tracking branch 'remotes/riscv/tags/riscv-qemu-upstream-v8.2' into staging
Peter Maydell [Fri, 9 Mar 2018 10:58:57 +0000 (10:58 +0000)]
Merge remote-tracking branch 'remotes/riscv/tags/riscv-qemu-upstream-v8.2' into staging

QEMU RISC-V Emulation Support (RV64GC, RV32GC)

This release renames the SiFive machines to sifive_e and sifive_u
to represent the SiFive Everywhere and SiFive Unleashed platforms.
SiFive has configurable soft-core IP, so it is intended that these
machines will be extended to enable a variety of SiFive IP blocks.
The CPU definition infrastructure has been improved and there are
now vendor CPU modules including the SiFiVe E31, E51, U34 and U54
cores. The emulation accuracy for the E series has been improved
by disabling the MMU for the E series. S mode has been disabled on
cores that only support M mode and U mode. The two Spike machines
that support two privileged ISA versions have been coalesced into
one file. This series has Signed-off-by from the core contributors.

*** Known Issues ***

* Disassembler has some checkpatch warnings for the sake of code brevity
* scripts/qemu-binfmt-conf.sh has checkpatch warnings due to line length
* PMP (Physical Memory Protection) is as-of-yet unused and needs testing

*** Changelog ***

v8.2

* Rebase

v8.1

* Fix missed case of renaming spike_v1.9 to spike_v1.9.1

v8

* Added linux-user/riscv/target_elf.h during rebase
* Make resetvec configurable and clear mpp and mie on reset
* Use SiFive E31, E51, U34 and U54 cores in SiFive machines
* Define SiFive E31, E51, U34 and U54 cores
* Refactor CPU core definition in preparation for vendor cores
* Prevent S or U mode unless S or U extensions are present
* SiFive E Series cores have no MMU
* SiFive E Series cores have U mode
* Make privileged ISA v1.10 implicit in CPU types
* Remove DRAM_BASE and EXT_IO_BASE as they vary by machine
* Correctly handle mtvec and stvec alignment with respect to RVC
* Print more machine mode state in riscv_cpu_dump_state
* Make riscv_isa_string use compact extension order method
* Fix bug introduced in v6 RISCV_CPU_TYPE_NAME macro change
* Parameterize spike v1.9.1 config string
* Coalesce spike_v1.9.1 and spike_v1.10 machines
* Rename sifive_e300 to sifive_e, and sifive_u500 to sifive_u

v7

* Make spike_v1.10 the default machine
* Rename spike_v1.9 to spike_v1.9.1 to match privileged spec version
* Remove empty target/riscv/trace-events file
* Monitor ROM 32-bit reset code needs to be target endian
* Add TARGET_TIOCGPTPEER to linux-user/riscv/termbits.h
* Add -initrd support to the virt board
* Fix naming in spike machine interface header
* Update copyright notice on RISC-V Spike machines
* Update copyright notice on RISC-V HTIF Console device
* Change CPU Core and translator to GPLv2+
* Change RISC-V Disassembler to GPLv2+
* Change SiFive Test Finisher to GPLv2+
* Change SiFive CLINT to GPLv2+
* Change SiFive PRCI to GPLv2+
* Change SiFive PLIC to GPLv2+
* Change RISC-V spike machines to GPLv2+
* Change RISC-V virt machine to GPLv2+
* Change SiFive E300 machine to GPLv2+
* Change SiFive U500 machine to GPLv2+
* Change RISC-V Hart Array to GPLv2+
* Change RISC-V HTIF device to GPLv2+
* Change SiFiveUART device to GPLv2+

v6

* Drop IEEE 754-201x minimumNumber/maximumNumber for fmin/fmax
* Remove some unnecessary commented debug statements
* Change RISCV_CPU_TYPE_NAME to use riscv-cpu suffix
* Define all CPU variants for linux-user
* qemu_log calls require trailing \n
* Replace PLIC printfs with qemu_log
* Tear out unused HTIF code and eliminate shouting debug messages
* Fix illegal instruction when sfence.vma is passed (rs2) arguments
* Make updates to PTE accessed and dirty bits atomic
* Only require atomic PTE updates on MTTCG enabled guests
* Page fault if accessed or dirty bits can't be updated
* Fix get_physical_address PTE reads and writes on riscv32
* Remove erroneous comments from the PLIC
* Default enable MTTCG
* Make WFI less conservative
* Unify local interrupt handling
* Expunge HTIF interrupts
* Always access mstatus.mip under a lock
* Don't implement rdtime/rdtimeh in system mode (bbl emulates them)
* Implement insreth/cycleh for rv32 and always enable user-mode counters
* Add GDB stub support for reading and writing CSRs
* Rename ENABLE_CHARDEV #ifdef from HTIF code
* Replace bad HTIF ELF code with load_elf symbol callback
* Convert chained if else fault handlers to switch statements
* Use RISCV exception codes for linux-user page faults

v5

* Implement NaN-boxing for flw, set high order bits to 1
* Use float_muladd_negate_* flags to floatXX_muladd
* Use IEEE 754-201x minimumNumber/maximumNumber for fmin/fmax
* Fix TARGET_NR_syscalls
* Update linux-user/riscv/syscall_nr.h
* Fix FENCE.I, needs to terminate translation block
* Adjust unusual convention for interruptno >= 0

v4

* Add @riscv: since 2.12 to CpuInfoArch
* Remove misleading little-endian comment from load_kernel
* Rename cpu-model property to cpu-type
* Drop some unnecessary inline function attributes
* Don't allow GDB to set value of x0 register
* Remove unnecessary empty property lists
* Add Test Finisher device to implement poweroff in virt machine
* Implement priv ISA v1.10 trap and sret/mret xPIE/xIE behavior
* Store fflags data in fp_status
* Purge runtime users of helper_raise_exception
* Fix validate_csr
* Tidy gen_jalr
* Tidy immediate shifts
* Add gen_exception_inst_addr_mis
* Add gen_exception_debug
* Add gen_exception_illegal
* Tidy helper_fclass_*
* Split rounding mode setting to a new function
* Enforce MSTATUS_FS via TB flags
* Implement acquire/release barrier semantics
* Use atomic operations as required
* Fix FENCE and FENCE_I
* Remove commented code from spike machines
* PAGE_WRITE permissions can be set on loads if page is already dirty
* The result of format conversion on an NaN must be a quiet NaN
* Add missing process_queued_cpu_work to riscv linux-user
* Remove float(32|64)_classify from cpu.h
* Removed nonsensical unions aliasing the same type
* Use uintN_t instead of uintN_fast_t in fpu_helper.c
* Use macros for FPU exception values in softfloat_flags_to_riscv
* Move code to set round mode into set_fp_round_mode function
* Convert set_fp_exceptions from a macro to an inline function
* Convert round mode helper into an inline function
* Make fpu_helper ieee_rm array static const
* Include cpu_mmu_index in cpu_get_tb_cpu_state flags
* Eliminate MPRV influence on mmu_index
* Remove unrecoverable do_unassigned_access function
* Only update PTE accessed and dirty bits if necessary
* Remove unnecessary tlb_flush in set_mode as mode is in mmu_idx
* Remove buggy support for misa writes. misa writes are optional
  and are not implemented in any known hardware
* Always set PTE read or execute permissions during page walk
* Reorder helper function declarations to match order in helper.c
* Remove redundant variable declaration in get_physical_address
* Remove duplicated code from get_physical_address
* Use mmu_idx instead of mem_idx in riscv_cpu_get_phys_page_debug

v3

* Fix indentation in PMP and HTIF debug macros
* Fix disassembler checkpatch open brace '{' on next line errors
* Fix trailing statements on next line in decode_inst_decompress
* NOTE: the other checkpatch issues have been reviewed previously

v2

* Remove redundant NULL terminators from disassembler register arrays
* Change disassembler register name arrays to const
* Refine disassembler internal function names
* Update dates in disassembler copyright message
* Remove #ifdef CONFIG_USER_ONLY version of cpu_has_work
* Use ULL suffix on 64-bit constants
* Move riscv_cpu_mmu_index from cpu.h to helper.c
* Move riscv_cpu_hw_interrupts_pending from cpu.h to helper.c
* Remove redundant TARGET_HAS_ICE from cpu.h
* Use qemu_irq instead of void* for irq definition in cpu.h
* Remove duplicate typedef from struct CPURISCVState
* Remove redundant g_strdup from cpu_register
* Remove redundant tlb_flush from riscv_cpu_reset
* Remove redundant mode calculation from get_physical_address
* Remove redundant debug mode printf and dcsr comment
* Remove redundant clearing of MSB for bare physical addresses
* Use g_assert_not_reached for invalid mode in get_physical_address
* Use g_assert_not_reached for unreachable checks in get_physical_address
* Use g_assert_not_reached for unreachable type in raise_mmu_exception
* Return exception instead of aborting for misaligned fetches
* Move exception defines from cpu.h to cpu_bits.h
* Remove redundant breakpoint control definitions from cpu_bits.h
* Implement riscv_cpu_unassigned_access exception handling
* Log and raise exceptions for unimplemented CSRs
* Match Spike HTIF exit behavior - don’t print TEST-PASSED
* Make frm,fflags,fcsr writes trap when mstatus.FS is clear
* Use g_assert_not_reached for unreachable invalid mode
* Make hret,uret,dret generate illegal instructions
* Move riscv_cpu_dump_state and int/fpr regnames to cpu.c
* Lift interrupt flag and mask into constants in cpu_bits.h
* Change trap debugging to use qemu_log_mask LOG_TRACE
* Change CSR debugging to use qemu_log_mask LOG_TRACE
* Change PMP debugging to use qemu_log_mask LOG_TRACE
* Remove commented code from pmp.c
* Change CpuInfoRISCV qapi schema docs to Since 2.12
* Change RV feature macro to use target_ulong cast
* Remove riscv_feature and instead use misa extension flags
* Make riscv_flush_icache_syscall a no-op
* Undo checkpatch whitespace fixes in unrelated linux-user code
* Remove redudant constants and tidy up cpu_bits.h
* Make helper_fence_i a no-op
* Move include "exec/cpu-all" to end of cpu.h
* Rename set_privilege to riscv_set_mode
* Move redundant forward declaration for cpu_riscv_translate_address
* Remove TCGV_UNUSED from riscv_translate_init
* Add comment to pmp.c stating the code is untested and currently unused
* Use ctz to simplify decoding of PMP NAPOT address ranges
* Change pmp_is_in_range to use than equal for end addresses
* Fix off by one error in pmp_update_rule
* Rearrange PMP_DEBUG so that formatting is compile-time checked
* Rearrange trap debugging so that formatting is compile-time checked
* Rearrange PLIC debugging so that formatting is compile-time checked
* Use qemu_log/qemu_log_mask for HTIF logging and debugging
* Move exception and interrupt names into cpu.c
* Add Palmer Dabbelt as a RISC-V Maintainer
* Rebase against current qemu master branch

v1

* initial version based on forward port from riscv-qemu repository

*** Background ***

"RISC-V is an open, free ISA enabling a new era of processor innovation
through open standard collaboration. Born in academia and research,
RISC-V ISA delivers a new level of free, extensible software and
hardware freedom on architecture, paving the way for the next 50 years
of computing design and innovation."

The QEMU RISC-V port has been developed and maintained out-of-tree for
several years by Sagar Karandikar and Bastian Koppelmann. The RISC-V
Privileged specification has evolved substantially over this period but
has recently been solidifying. The RISC-V Base ISA has been frozon for
some time and the Privileged ISA, GCC toolchain and Linux ABI are now
quite stable. I have recently joined Sagar and Bastian as a RISC-V QEMU
Maintainer and hope to support upstreaming the port.

There are multiple vendors taping out, preparing to ship, or shipping
silicon that implements the RISC-V Privileged ISA Version 1.10. There
are also several RISC-V Soft-IP cores implementing Privileged ISA
Version 1.10 that run on FPGA such as SiFive's Freedom U500 Platform
and the U54‑MC RISC-V Core IP, among many more implementations from a
variety of vendors. See https://riscv.org/ for more details.

RISC-V support was upstreamed in binutils 2.28 and GCC 7.1 in the first
half of 2016. RISC-V support is now available in LLVM top-of-tree and
the RISC-V Linux port was accepted into Linux 4.15-rc1 late last year
and is available in the Linux 4.15 release. GLIBC 2.27 added support
for the RISC-V ISA running on Linux (requires at least binutils-2.30,
gcc-7.3.0, and linux-4.15). We believe it is timely to submit the
RISC-V QEMU port for upstream review with the goal of incorporating
RISC-V support into the upcoming QEMU 2.12 release.

The RISC-V QEMU port is still under active development, mostly with
respect to device emulation, the addition of Hypervisor support as
specified in the RISC-V Draft Privileged ISA Version 1.11, and Vector
support once the first draft is finalized later this year. We believe
now is the appropriate time for RISC-V QEMU development to be carried
out in the main QEMU repository as the code will benefit from more
rigorous review. The RISC-V QEMU port currently supports all the ISA
extensions that have been finalized and frozen in the Base ISA.

Blog post about recent additions to RISC-V QEMU: https://goo.gl/fJ4zgk

The RISC-V QEMU wiki: https://github.com/riscv/riscv-qemu/wiki

Instructions for building a busybox+dropbear root image, BBL (Berkeley
Boot Loader) and linux kernel image for use with the RISC-V QEMU
'virt' machine: https://github.com/michaeljclark/busybear-linux

*** Overview ***

The RISC-V QEMU port implements the following specifications:

* RISC-V Instruction Set Manual Volume I: User-Level ISA Version 2.2
* RISC-V Instruction Set Manual Volume II: Privileged ISA Version 1.9.1
* RISC-V Instruction Set Manual Volume II: Privileged ISA Version 1.10

The RISC-V QEMU port supports the following instruction set extensions:

* RV32GC with Supervisor-mode and User-mode (RV32IMAFDCSU)
* RV64GC with Supervisor-mode and User-mode (RV64IMAFDCSU)

The RISC-V QEMU port adds the following targets to QEMU:

* riscv32-softmmu
* riscv64-softmmu
* riscv32-linux-user
* riscv64-linux-user

The RISC-V QEMU port supports the following hardware:

* HTIF Console (Host Target Interface)
* SiFive CLINT (Core Local Interruptor) for Timer interrupts and IPIs
* SiFive PLIC (Platform Level Interrupt Controller)
* SiFive Test (Test Finisher) for exiting simulation
* SiFive UART, PRCI, AON, PWM, QSPI support is partially implemented
* VirtIO MMIO (GPEX PCI support will be added in a future patch)
* Generic 16550A UART emulation using 'hw/char/serial.c'
* MTTCG and SMP support (PLIC and CLINT) on the 'virt' machine

The RISC-V QEMU full system emulator supports 5 machines:

* 'spike_v1.9.1', CLINT, PLIC, HTIF console, config-string, Priv v1.9.1
* 'spike_v1.10', CLINT, PLIC, HTIF console, device-tree, Priv v1.10
* 'sifive_e', CLINT, PLIC, SiFive UART, HiFive1 compat, Priv v1.10
* 'sifive_u', CLINT, PLIC, SiFive UART, device-tree, Priv v1.10
* 'virt', CLINT, PLIC, 16550A UART, VirtIO, device-tree, Priv v1.10

This is a list of RISC-V QEMU Port Contributors:

* Alex Suykov
* Andreas Schwab
* Antony Pavlov
* Bastian Koppelmann
* Bruce Hoult
* Chih-Min Chao
* Daire McNamara
* Darius Rad
* David Abdurachmanov
* Hesham Almatary
* Ivan Griffin
* Jim Wilson
* Kito Cheng
* Michael Clark
* Palmer Dabbelt
* Richard Henderson
* Sagar Karandikar
* Shea Levy
* Stefan O'Rear

Notes:

* contributor email addresses available off-list on request.
* checkpatch has been run on all 23 patches.
* checkpatch exceptions are noted in patches that have errors.
* passes "make check" on full build for all targets
* tested riscv-linux-4.6.2 on 'spike_v1.9.1' machine
* tested riscv-linux-4.15 on 'spike_v1.10' and 'virt' machines
* tested SiFive HiFive1 binaries in 'sifive_e' machine
* tested RV64 on 32-bit i386

This patch series includes the following patches:

# gpg: Signature made Thu 08 Mar 2018 19:40:20 GMT
# gpg:                using DSA key 6BF1D7B357EF3E4F
# gpg: Good signature from "Michael Clark <michaeljclark@mac.com>"
# gpg:                 aka "Michael Clark <mjc@sifive.com>"
# gpg:                 aka "Michael Clark <michael@metaparadigm.com>"
# 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: 7C99 930E B17C D8BA 073D  5EFA 6BF1 D7B3 57EF 3E4F

* remotes/riscv/tags/riscv-qemu-upstream-v8.2: (23 commits)
  RISC-V Build Infrastructure
  SiFive Freedom U Series RISC-V Machine
  SiFive Freedom E Series RISC-V Machine
  SiFive RISC-V PRCI Block
  SiFive RISC-V UART Device
  RISC-V VirtIO Machine
  SiFive RISC-V Test Finisher
  RISC-V Spike Machines
  SiFive RISC-V PLIC Block
  SiFive RISC-V CLINT Block
  RISC-V HART Array
  RISC-V HTIF Console
  Add symbol table callback interface to load_elf
  RISC-V Linux User Emulation
  RISC-V Physical Memory Protection
  RISC-V TCG Code Generation
  RISC-V GDB Stub
  RISC-V FPU Support
  RISC-V CPU Helpers
  RISC-V Disassembler
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agoMerge remote-tracking branch 'remotes/cohuck/tags/s390x-20180308' into staging
Peter Maydell [Thu, 8 Mar 2018 18:30:16 +0000 (18:30 +0000)]
Merge remote-tracking branch 'remotes/cohuck/tags/s390x-20180308' into staging

Fixes and cleanups for the 2.12 softfreeze.

# gpg: Signature made Thu 08 Mar 2018 17:53:14 GMT
# gpg:                using RSA key DECF6B93C6F02FAF
# gpg: Good signature from "Cornelia Huck <conny@cornelia-huck.de>"
# gpg:                 aka "Cornelia Huck <huckc@linux.vnet.ibm.com>"
# gpg:                 aka "Cornelia Huck <cornelia.huck@de.ibm.com>"
# gpg:                 aka "Cornelia Huck <cohuck@kernel.org>"
# gpg:                 aka "Cornelia Huck <cohuck@redhat.com>"
# Primary key fingerprint: C3D0 D66D C362 4FF6 A8C0  18CE DECF 6B93 C6F0 2FAF

* remotes/cohuck/tags/s390x-20180308:
  s390x/virtio: Convert virtio-ccw from *_exit to *_unrealize
  pc-bios/s390-ccw: Move string arrays from bootmap header to .c file
  s390x/sclp: clean up sclp masks
  s390x/sclp: proper support of larger send and receive masks
  vfio-ccw: license text should indicate GPL v2 or later
  s390x/sclpconsole: Remove dead code - remove exit handlers
  numa: we don't implement NUMA for s390x
  hw/s390x: Add the possibility to specify the netboot image on the command line
  target/s390x: Remove leading underscores from #defines
  s390/ipl: only print boot menu error if -boot menu=on was specified
  hw/s390x/ipl: Bail out if the network bootloader can not be found

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6 years agovl: introduce vm_shutdown()
Stefan Hajnoczi [Wed, 7 Mar 2018 14:42:05 +0000 (14:42 +0000)]
vl: introduce vm_shutdown()

Commit 00d09fdbbae5f7864ce754913efc84c12fdf9f1a ("vl: pause vcpus before
stopping iothreads") and commit dce8921b2baaf95974af8176406881872067adfa
("iothread: Stop threads before main() quits") tried to work around the
fact that emulation was still active during termination by stopping
iothreads.  They suffer from race conditions:
1. virtio_scsi_handle_cmd_vq() racing with iothread_stop_all() hits the
   virtio_scsi_ctx_check() assertion failure because the BDS AioContext
   has been modified by iothread_stop_all().
2. Guest vq kick racing with main loop termination leaves a readable
   ioeventfd that is handled by the next aio_poll() when external
   clients are enabled again, resulting in unwanted emulation activity.

This patch obsoletes those commits by fully disabling emulation activity
when vcpus are stopped.

Use the new vm_shutdown() function instead of pause_all_vcpus() so that
vm change state handlers are invoked too.  Virtio devices will now stop
their ioeventfds, preventing further emulation activity after vm_stop().

Note that vm_stop(RUN_STATE_SHUTDOWN) cannot be used because it emits a
QMP STOP event that may affect existing clients.

It is no longer necessary to call replay_disable_events() directly since
vm_shutdown() does so already.

Drop iothread_stop_all() since it is no longer used.

Cc: Fam Zheng <famz@redhat.com>
Cc: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 20180307144205.20619-5-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
6 years agovirtio-scsi: fix race between .ioeventfd_stop() and vq handler
Stefan Hajnoczi [Wed, 7 Mar 2018 14:42:04 +0000 (14:42 +0000)]
virtio-scsi: fix race between .ioeventfd_stop() and vq handler

If the main loop thread invokes .ioeventfd_stop() just as the vq handler
function begins in the IOThread then the handler may lose the race for
the AioContext lock.  By the time the vq handler is able to acquire the
AioContext lock the ioeventfd has already been removed and the handler
isn't supposed to run anymore!

Use the new aio_wait_bh_oneshot() function to perform ioeventfd removal
from within the IOThread.  This way no races with the vq handler are
possible.

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 20180307144205.20619-4-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>