i.mx7d: Change IRQ number type from hwaddr to int

The qdev_get_gpio_in() function accept an int as second parameter.

Signed-off-by: Jean-Christophe Dubois <jcd@tribudubois.net>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

i.mx7d: Change SRC unimplemented device name from sdma to src

Signed-off-by: Jean-Christophe Dubois <jcd@tribudubois.net>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

i.mx7d: Remove unused header files

Signed-off-by: Jean-Christophe Dubois <jcd@tribudubois.net>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: support reading of CNT[VCT|FRQ]_EL0 from user-space

Since kernel commit a86bd139f2 (arm64: arch_timer: Enable CNTVCT_EL0
trap..), released in kernel version v4.12, user-space has been able
to read these system registers. As we can't use QEMUTimer's in
linux-user mode we just directly call cpu_get_clock().

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180625160009.17437-2-alex.bennee@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement ARMv8.2-DotProd

We've already added the helpers with an SVE patch, all that remains
is to wire up the aa64 and aa32 translators.  Enable the feature
within -cpu max for CONFIG_USER_ONLY.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-36-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Enable SVE for aarch64-linux-user

Enable ARM_FEATURE_SVE for the generic "max" cpu.

Tested-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-35-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE dot product (indexed)

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20180627043328.11531-34-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE dot product (vectors)

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-33-richard.henderson@linaro.org
[PMM: moved 'ra=%reg_movprfx' here from following patch]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE fp complex multiply add (indexed)

Enhance the existing helpers to support SVE, which takes the
index from each 128-bit segment.  The change has no effect
for AdvSIMD, since there is only one such segment.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20180627043328.11531-32-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Pass index to AdvSIMD FCMLA (indexed)

For aa64 advsimd, we had been passing the pre-indexed vector.
However, sve applies the index to each 128-bit segment, so we
need to pass in the index separately.

For aa32 advsimd, the fp32 operation always has index 0, but
we failed to interpret the fp16 index correctly.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20180627043328.11531-31-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE fp complex multiply add

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-30-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE floating-point complex add

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-29-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE MOVPRFX

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-28-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE floating-point unary operations

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-27-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE floating-point round to integral value

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-26-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE floating-point convert to integer

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-25-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE floating-point convert precision

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-24-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE floating-point trig multiply-add coefficient

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-23-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE FP Compare with Zero Group

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-22-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE Floating Point Unary Operations - Unpredicated Group

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-21-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE FP Fast Reduction Group

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-20-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE Floating Point Multiply Indexed Group

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-19-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE floating-point arithmetic with immediate

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-18-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE floating-point compare vectors

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-17-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE scatter store vector immediate

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-16-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE first-fault gather loads

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-15-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE gather loads

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20180627043328.11531-14-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE prefetches

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-13-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE scatter stores

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-12-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE store vector/predicate register

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-11-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE load and broadcast element

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-10-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE Floating Point Accumulating Reduction Group

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-9-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE FP Multiply-Add Group

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20180627043328.11531-8-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE floating-point arithmetic (predicated)

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-7-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE integer convert to floating-point

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
[PMM: fixed typo]
Message-id: 20180627043328.11531-6-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE load and broadcast quadword

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-5-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE Memory Contiguous Store Group

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-4-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE Contiguous Load, first-fault and no-fault

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Tested-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20180627043328.11531-3-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement SVE Memory Contiguous Load Group

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180627043328.11531-2-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hw/arm/virt: Silence dtc /memory warning

When running dtc on the guest /proc/device-tree we get the
following warning: Warning (unit_address_vs_reg): Node /memory
has a reg or ranges property, but no unit name".

Let's fix that by adding the unit address to the node name. We also
don't create the /memory node anymore in create_fdt(). We directly
create it in load_dtb. /chosen still needs to be created in create_fdt
as the uart needs it. In case the user provided his own dtb, we nop
all memory nodes found in root and create new one(s).

Signed-off-by: Eric Auger <eric.auger@redhat.com>
Message-id: 1530044492-24921-4-git-send-email-eric.auger@redhat.com
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hw/arm/virt: Silence dtc /intc warnings

When running dtc on the guest /proc/device-tree we get the
following warnings: "Warning (unit_address_vs_reg): Node <name>
has a reg or ranges property, but no unit name", with name:
/intc, /intc/its, /intc/v2m.

Nodes should have a name in the form <name>[@<unit-address>] where
unit-address is the primary address used to access the device, listed
in the node's reg property. This fix seems to make dtc happy.

Signed-off-by: Eric Auger <eric.auger@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1530044492-24921-3-git-send-email-eric.auger@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

device_tree: Add qemu_fdt_node_unit_path

This helper allows to retrieve the paths of nodes whose name
match node-name or node-name@unit-address patterns.

Signed-off-by: Eric Auger <eric.auger@redhat.com>
Message-id: 1530044492-24921-2-git-send-email-eric.auger@redhat.com
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

device_tree: Replace error_setg(&error_fatal) by error_report() + exit()

Use error_report() + exit() instead of error_setg(&error_fatal),
as suggested by the "qapi/error.h" documentation:

   Please don't error_setg(&error_fatal, ...), use error_report() and
   exit(), because that's more obvious.

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Message-id: 20180625165749.3910-4-f4bug@amsat.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hw/arm/sysbus-fdt: Replace error_setg(&error_fatal) by error_report() + exit()

Use error_report() + exit() instead of error_setg(&error_fatal),
as suggested by the "qapi/error.h" documentation:

   Please don't error_setg(&error_fatal, ...), use error_report() and
   exit(), because that's more obvious.

This fixes CID 1352173:
    "Passing null pointer dt_name to qemu_fdt_node_path, which dereferences it."

And this also fixes:

    hw/arm/sysbus-fdt.c:322:9: warning: Array access (from variable 'node_path') results in a null pointer dereference
        if (node_path[1]) {
            ^~~~~~~~~~~~

Fixes: Coverity CID 1352173 (Dereference after null check)
Suggested-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Message-id: 20180625165749.3910-3-f4bug@amsat.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hw/block/fdc: Replace error_setg(&error_abort) by assert()

Use assert() instead of error_setg(&error_abort),
as suggested by the "qapi/error.h" documentation:

    Please don't error_setg(&error_fatal, ...), use error_report() and
    exit(), because that's more obvious.
    Likewise, don't error_setg(&error_abort, ...), use assert().

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Acked-by: John Snow <jsnow@redhat.com>
Message-id: 20180625165749.3910-2-f4bug@amsat.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

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

* "info mtree" improvements (Alexey)
* fake VPD block limits for SCSI passthrough (Daniel Barboza)
* chardev and main loop fixes (Daniel Berrangé, Sergio, Stefan)
* help fixes (Eduardo)
* pc-dimm refactoring (David)
* tests improvements and fixes (Emilio, Thomas)
* SVM emulation fixes (Jan)
* MemoryRegionCache fix (Eric)
* WHPX improvements (Justin)
* ESP cleanup (Mark)
* -overcommit option (Michael)
* qemu-pr-helper fixes (me)
* "info pic" improvements for x86 (Peter)
* x86 TCG emulation fixes (Richard)
* KVM slot handling fix (Shannon)
* Next round of deprecation (Thomas)
* Windows dump format support (Viktor)

# gpg: Signature made Fri 29 Jun 2018 12:03:05 BST
# gpg:                using RSA key BFFBD25F78C7AE83
# 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: (60 commits)
  tests/boot-serial: Do not delete the output file in case of errors
  hw/scsi: add VPD Block Limits emulation
  hw/scsi: centralize SG_IO calls into single function
  hw/scsi: cleanups before VPD BL emulation
  dump: add Windows live system dump
  dump: add fallback KDBG using in Windows dump
  dump: use system context in Windows dump
  dump: add Windows dump format to dump-guest-memory
  i386/cpu: make -cpu host support monitor/mwait
  kvm: support -overcommit cpu-pm=on|off
  hmp: obsolete "info ioapic"
  ioapic: support "info irq"
  ioapic: some proper indents when dump info
  ioapic: support "info pic"
  doc: another fix to "info pic"
  target-i386: Mark cpu_vmexit noreturn
  target-i386: Allow interrupt injection after STGI
  target-i386: Add NMI interception to SVM
  memory/hmp: Print owners/parents in "info mtree"
  WHPX: register for unrecognized MSR exits
  ...

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

tests/boot-serial: Do not delete the output file in case of errors

Peter reported that the boot-serial tester sometimes runs into timeouts
with SPARC guests. It's currently completely unclear whether this is due
to too much load on the host machine (so that the guest really just ran
too slow), or whether there is something wrong with the guest's firmware
boot. For further debugging, we need the serial output of the guest in
case of errors, so instead of unlinking the file immediately, this is
now only done in case of success. In case of error, print the name of the
file with the serial output via g_error() (which then also calls abort()
internally to mark the test as failed).

Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <1526977831-31129-1-git-send-email-thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/scsi: add VPD Block Limits emulation

The VPD Block Limits Inquiry page is optional, allowing SCSI devices
to not implement it. This is the case for devices like the MegaRAID
SAS 9361-8i and Microsemi PM8069.

In case of SCSI passthrough, the response of this request is used by
the QEMU SCSI layer to set the max_io_sectors that the guest
device will support, based on the value of the max_sectors_kb that
the device has set in the host at that time. Without this response,
the guest kernel is free to assume any value of max_io_sectors
for the SCSI device. If this value is greater than the value from
the host, SCSI Sense errors will occur because the guest will send
read/write requests that are larger than the underlying host device
is configured to support. An example of this behavior can be seen
in [1].

A workaround is to set the max_sectors_kb host value back in the guest
kernel (a process that can be automated using rc.local startup scripts
and the like), but this has several drawbacks:

- it can be troublesome if the guest has many passthrough devices that
needs this tuning;

- if a change in max_sectors_kb is made in the host side, manual change
in the guests will also be required;

- during an OS install it is difficult, and sometimes not possible, to
go to a terminal and change the max_sectors_kb prior to the installation.
This means that the disk can't be used during the install process. The
easiest alternative here is to roll back to scsi-hd, install the guest
and then go back to SCSI passthrough when the installation is done and
max_sectors_kb can be set.

An easier way would be to QEMU handle the absence of the Block Limits
VPD device response, setting max_io_sectors accordingly and allowing
the guest to use the device without the hassle.

This patch adds emulation of the Block Limits VPD response for
SCSI passthrough devices of type TYPE_DISK that doesn't support
it. The following changes were made:

- scsi_handle_inquiry_reply will now check the available VPD
pages from the Inquiry EVPD reply. In case the device does not

- a new function called scsi_generic_set_vpd_bl_emulation,
that is called during device realize,  was created to set a
new flag 'needs_vpd_bl_emulation' of the device. This function
retrieves the Inquiry EVPD response of the device to check for
VPD BL support.

- scsi_handle_inquiry_reply will now check the available VPD
pages from the Inquiry EVPD reply in case the device needs
VPD BL emulation, adding the Block Limits page (0xb0) to
the list. This will make the guest kernel aware of the
support that we're now providing by emulation.

- a new function scsi_emulate_block_limits creates the
emulated Block Limits response. This function is called
inside scsi_read_complete in case the device requires
Block Limits VPD emulation and we detected a SCSI Sense
error in the VPD Block Limits reply that was issued
from the guest kernel to the device. This error is
expected: we're reporting support from our side, but
the device isn't aware of it.

With this patch, the guest now queries the Block Limits
page during the device configuration because it is being
advertised in the Supported Pages response. It will either
receive the Block Limits page from the hardware, if it supports
it, or will receive an emulated response from QEMU. At any rate,
the guest now has the information to set the max_sectors_kb
parameter accordingly, sparing the user of SCSI sense errors
that would happen without the emulated response and in the
absence of Block Limits support from the hardware.

[1] https://bugzilla.redhat.com/show_bug.cgi?id=1566195

Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1566195
Reported-by: Dac Nguyen <dacng@us.ibm.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20180627172432.11120-4-danielhb413@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/scsi: centralize SG_IO calls into single function

For the VPD Block Limits emulation with SCSI passthrough,
we'll issue an Inquiry request with EVPD set to retrieve
the available VPD pages of the device. This would be done in
a way similar of what scsi_generic_read_device_identification
does: create a SCSI command and a reply buffer, fill in the
sg_io_hdr_t structure, call blk_ioctl, check if an error
occurred, process the response.

This same process is done in other 2 functions, get_device_type
and get_stream_blocksize. They differ in the command/reply
buffer and post-processing, everything else is almost a
copy/paste.

Instead of adding a forth copy/pasted-ish code when adding
the passthrough VPD BL emulation, this patch extirpates
this repetition of those 3 functions and put it into
a new one called scsi_SG_IO_FROM_DEV. Any future code that
wants to execute an SG_DXFER_FROM_DEV to the device can
use it, avoiding filling sg_io_hdr_t again and et cetera.

Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20180627172432.11120-3-danielhb413@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/scsi: cleanups before VPD BL emulation

To add support for the emulation of Block Limits VPD page
for passthrough devices, a few adjustments in the current code
base is required to avoid repetition and improve clarity.

In scsi-generic.c, detach the Inquiry handling from
scsi_read_complete and put it into a new function called
scsi_handle_inquiry_reply. This change aims to avoid
cluttering of scsi_read_complete when we more logic in the
Inquiry response handling is added in the next patches,
centralizing the changes in the new function.

In scsi-disk.c, take the build of all emulated VPD pages
from scsi_disk_emulate_inquiry and make it available to
other files into a non-static function called
scsi_disk_emulate_vpd_page. Making it public will allow
the future VPD BL emulation code for passthrough devices
to use it from scsi-generic.c, avoiding copy/pasting this
code solely for that purpose. It also has the advantage of
providing emulation of all VPD pages in case we need to
emulate other pages in other scenarios. As a bonus,
scsi_disk_emulate_inquiry got tidier.

Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20180627172432.11120-2-danielhb413@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

dump: add Windows live system dump

Unlike dying Windows, live system memory doesn't contain
correct register contexts. But they can be populated with QEMU register
values.
After this patch, QEMU will be able to produce guest Windows live system
dump.

Signed-off-by: Viktor Prutyanov <viktor.prutyanov@virtuozzo.com>
Message-Id: <20180517162342.4330-5-viktor.prutyanov@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

dump: add fallback KDBG using in Windows dump

KdDebuggerDataBlock may be encrypted in guest memory and dump will be
useless in this case. But guest driver can obtain decrypted KDBG and
expose its address through BugcheckParameter1 field in raw header.
After this patch, QEMU will be able to use fallback KdDebuggerDataBlock.

Signed-off-by: Viktor Prutyanov <viktor.prutyanov@virtuozzo.com>
Message-Id: <20180517162342.4330-4-viktor.prutyanov@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

dump: use system context in Windows dump

We use CPU #0 to access guest virtual memory, but it can execute user
thread at that moment. So, switch CR3 to PageDirectoryBase from header
and restore original value at the end.

Signed-off-by: Viktor Prutyanov <viktor.prutyanov@virtuozzo.com>
Message-Id: <20180517162342.4330-3-viktor.prutyanov@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

dump: add Windows dump format to dump-guest-memory

This patch adds Windows crashdumping feature. Now QEMU can produce ELF-dump
containing Windows crashdump header, which can help to convert to a valid
WinDbg-understandable crashdump file, or immediately create such file.
The crashdump will be obtained by joining physical memory dump and 8K header
exposed through vmcoreinfo/fw_cfg device by guest driver at BSOD time. Option
'-w' was added to dump-guest-memory command. At the moment, only x64
configuration is supported.
Suitable driver can be found at
https://github.com/virtio-win/kvm-guest-drivers-windows/tree/master/fwcfg64

Signed-off-by: Viktor Prutyanov <viktor.prutyanov@virtuozzo.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20180517162342.4330-2-viktor.prutyanov@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/cpu: make -cpu host support monitor/mwait

When guest CPU PM is enabled, and with -cpu host, expose the host CPU
MWAIT leaf in the CPUID so guest can make good PM decisions.

Note: the result is 100% CPU utilization reported by host as host
no longer knows that the CPU is halted.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <20180622192148.178309-3-mst@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

kvm: support -overcommit cpu-pm=on|off

With this flag, kvm allows guest to control host CPU power state.  This
increases latency for other processes using same host CPU in an
unpredictable way, but if decreases idle entry/exit times for the
running VCPU, so to use it QEMU needs a hint about whether host CPU is
overcommitted, hence the flag name.

Follow-up patches will expose this capability to guest
(using mwait leaf).

Based on a patch by Wanpeng Li <kernellwp@gmail.com> .

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Message-Id: <20180622192148.178309-2-mst@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hmp: obsolete "info ioapic"

Let's start to use "info pic" just like other platforms.  For now we
keep the command for a while so that old users can know what is the new
command to use.

Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20171229073104.3810-6-peterx@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

ioapic: support "info irq"

This include both userspace and in-kernel ioapic.  Note that the numbers
can be inaccurate for kvm-ioapic.  One reason is the same with
kvm-i8259, that when irqfd is used, irqs can be delivered all inside
kernel without our notice.  Meanwhile, kvm-ioapic is specially treated
when irq numbers <ISA_NUM_IRQS, those irqs will be delivered in kernel
too via kvm-i8259 (please refer to kvm_pc_gsi_handler).

Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20171229073104.3810-5-peterx@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

ioapic: some proper indents when dump info

So that now it looks better when with other irqchips.

Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20171229073104.3810-4-peterx@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

ioapic: support "info pic"

People start to use "info pic" for all kinds of irqchip dumps.  Let x86
ioapic join the family.  It dumps the same thing as "info ioapic".

Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20171229073104.3810-3-peterx@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

doc: another fix to "info pic"

Something that commit 254316fa1f ("intc: make HMP 'info irq' and 'info
pic' commands available on all targets", 2016-10-04) forgot to touch up.

Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20171229073104.3810-2-peterx@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target-i386: Mark cpu_vmexit noreturn

It calls cpu_loop_exit in system emulation mode (and should never be
called in user emulation mode).

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Message-Id: <6f4d44ffde55d074cbceb48309c1678600abad2f.1522769774.git.jan.kiszka@web.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target-i386: Allow interrupt injection after STGI

We need to terminate the translation block after STGI so that pending
interrupts can be injected.

This fixes pending NMI injection for Jailhouse which uses "stgi; clgi"
to open a brief injection window.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Message-Id: <37939b244dda0e9cccf96ce50f2b15df1e48315d.1522769774.git.jan.kiszka@web.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target-i386: Add NMI interception to SVM

Check for SVM interception prior to injecting an NMI. Tested via the
Jailhouse hypervisor.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Message-Id: <c65877e9a011ee4962931287e59f502c482b8d0b.1522769774.git.jan.kiszka@web.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

memory/hmp: Print owners/parents in "info mtree"

This adds owners/parents (which are the same, just occasionally
owner==NULL) printing for memory regions; a new '-o' flag
enabled new output.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Message-Id: <20180604032511.6980-1-aik@ozlabs.ru>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

WHPX: register for unrecognized MSR exits

Some variations of Linux kernels end up accessing MSR's that the Windows
Hypervisor doesn't implement which causes a GP to be returned to the guest.
This fix registers QEMU for unimplemented MSR access and globally returns 0 on
reads and ignores writes. This behavior is allows the Linux kernel to probe the
MSR with a write/read/check sequence it does often without failing the access.

Signed-off-by: Justin Terry (VM) <juterry@microsoft.com>
Message-Id: <20180605221500.21674-2-juterry@microsoft.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

WHPX workaround bug in OSVW handling

Adds a workaround to an incorrect value setting
CPUID Fn8000_0001_ECX[bit 9 OSVW] = 1. This can cause a guest linux kernel
to panic when an issue to rdmsr C001_0140h returns 0. Disabling this feature
correctly allows the guest to boot without accessing the osv workarounds.

Signed-off-by: Justin Terry (VM) <juterry@microsoft.com>
Message-Id: <20180605221500.21674-1-juterry@microsoft.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

esp: remove legacy esp_init() function

Remove the legacy esp_init() function now that there are no more remaining
users.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Message-Id: <20180613094727.11326-3-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Tested-by: Hervé Poussineau <hpoussin@reactos.org>

hw/mips/jazz: create ESP device directly via qdev

MIPS jazz is the last user of the legacy esp_init() function so move creation
of the ESP device over to use qdev.

Note that the esp_reset and dma_enable qemu_irqs are currently unused and so
we do not wire these up and instead remove the variables to prevent the
compiler emitting unused variable warnings.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Message-Id: <20180613094727.11326-2-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Tested-by: Hervé Poussineau <hpoussin@reactos.org>

pr-manager-helper: report event on connection/disconnection

Let management know if there were any problems communicating with
qemu-pr-helper.  The event is edge-triggered, and is sent every time
the connection status of the pr-manager-helper object changes.

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

pr-manager: add query-pr-managers QMP command

This command lets you query the connection status of each pr-manager-helper
object.

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

pr-manager: put stubs in .c file

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

pr-manager-helper: avoid SIGSEGV when writing to the socket fail

When writing to the qemu-pr-helper socket failed, the persistent
reservation manager was correctly disconnecting the socket, but it
did not clear pr_mgr->ioc.  So the rest of the code did not know
that the socket had been disconnected, accessed pr_mgr->ioc and
happily caused a crash.

To reproduce, it is enough to stop qemu-pr-helper between QEMU
startup and executing e.g. sg_persist -k /dev/sdb.

Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

pr-helper: fix assertion failure on failed multipath PERSISTENT RESERVE IN

The response size is expected to be zero if the SCSI status is not
"GOOD", but nothing was resetting it.

This can be reproduced simply by "sg_persist -s /dev/sdb" where /dev/sdb
in the guest is a scsi-block device corresponding to a multipath device
on the host.

Before:

  PR in (Read full status): Aborted command

and on the host:

  prh_write_response: Assertion `resp->sz == 0' failed.

After:

  PR in (Read full status): bad field in cdb or parameter list
  (perhaps unsupported service action)

Reported-by: Jiri Belka <jbelka@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>

pr-helper: fix --socket-path default in help

Currently --help shows "(default '(null)')" for the -k/--socket-path
option.  Fix it by getting the default path in /var/run.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>

Deprecate the -enable-hax option

We currently have got three ways of turning on the HAX accelerator:
"-machine accel=hax", "-accel hax" and "-enable-hax". That's really
confusing and overloaded. Since "-accel" is our preferred way to enable
an accelerator nowadays, and "-accel hax" is even less to type than
"-enable-hax", let's deprecate the "-enable-hax" option now.

Note: While "-enable-kvm" is available since a long time and can hardly be
removed since it is used in a lot of upper layer tools and scripts, the
"-enable-hax" option is still rather new and not very widespread yet, so
I think that it should be OK if we remove this in a couple of releases again
(we'll see whether someone complains after seeing the deprecation message -
then we could still reconsider to keep it if there a well-founded reasons).

Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <1529950933-28347-1-git-send-email-thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

osdep: work around Coverity parsing errors

Coverity does not like the new _Float* types that are used by
recent glibc, and croaks on every single file that includes
stdlib.h.  Add dummy typedefs to please it.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

numa: report all DIMM/NVDIMMs as plugged memory

Right now, there is some inconsistency between hotplugged and
coldplugged memory. DIMMs added via "-device" result in different stats
than DIMMs added using "device_add".

E.g.
    [...]
    -numa node,nodeid=0,cpus=0-1 -numa node,nodeid=1,cpus=2-3 \
    -m 4G,maxmem=20G,slots=2 \
    -object memory-backend-ram,id=mem0,size=8G \
    -device pc-dimm,id=dimm0,memdev=mem0 \
    -object memory-backend-ram,id=mem1,size=8G \
    -device nvdimm,id=dimm1,memdev=mem1,node=1

Results in NUMA info
    (qemu) info numa
    info numa
    2 nodes
    node 0 cpus: 0 1
    node 0 size: 10240 MB
    node 0 plugged: 0 MB
    node 1 cpus: 2 3
    node 1 size: 10240 MB
    node 1 plugged: 0 MB

But in memory size summary:
    (qemu) info memory_size_summary
    info memory_size_summary
    base memory: 4294967296
    plugged memory: 17179869184

Make this consistent by reporting all hot and coldplugged
memory a.k.a. DIMM and NVDIMM as "plugged".

Fixes: 31959e82fb0 ("hmp: extend "info numa" with hotplugged memory information")
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180622144045.737-1-david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

pc-dimm: get_memory_region() will not fail after realize

Let's try to reduce error handling a bit. In the plug/unplug case, the
device was realized and therefore we can assume that getting access to
the memory region will not fail.

For get_vmstate_memory_region() this is already handled that way.
Document both cases.

Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180619134141.29478-13-david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

nvdimm: make get_memory_region() perform checks and initialization

We might get a call to get_memory_region() before the device has been
realized. We should return a consistent value, as the return value
will e.g. later on be used in the pre_plug handler.

To avoid duplicating too much code, factor the initialization and checks
out into a helper function.

Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180619134141.29478-12-david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

nvdimm: convert nvdimm_mr into a pointer

This way we can easily check if the region has already been inititalized
without having to rely on the size of an uninitialized region being 0.

Free the region in nvdimm_finalize() and not in unrealize() as we will
allow to create the region before realization in following patches.

Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180619134141.29478-11-david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

nvdimm: convert "unarmed" into a static property

We don't allow to modify it after realization. So we can simply turn
it into a static property.

Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180619134141.29478-10-david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

pc-dimm: merge get_(vmstate_)memory_region()

Importantly, get_vmstate_memory_region() should also fail with a proper
error if called before the device is realized. For a PCDIMM, both functions
are to return the same thing, so share the implementation.

All current users are called after the device has been realized, so we
can expect the calls to succeed.

Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180619134141.29478-9-david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hostmem: drop error variable from host_memory_backend_get_memory()

Unused, so let's remove it.

Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180619134141.29478-8-david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

nvdimm: no need to overwrite get_vmstate_memory_region()

Our parent class (PC_DIMM) provides exactly the same function.

Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180619134141.29478-7-david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

pc: factor out pc specific dimm checks into pc_memory_pre_plug()

We can perform these checks before the device is actually realized.

Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180619134141.29478-6-david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

pc-dimm: remove pc_dimm_get_free_slot() from header

Not used outside of pc-dimm.c and there shouldn't be other users. If
other devices (e.g. memory devices) ever have to also use slots, then we
will have to factor this out.

Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180619134141.29478-5-david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

pc-dimm: rename pc_dimm_memory_* to pc_dimm_*

Let's rename it to make it look more consistent.

Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180619134141.29478-4-david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

pc: rename pc_dimm_(plug|unplug|...)* into pc_memory_(plug|unplug|...)*

Use a similar naming scheme as spapr. This way, we can go ahead and
rename e.g. pc_dimm_memory_plug to pc_dimm_plug, which avoids
confusion.

Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180619134141.29478-3-david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

pc-dimm: remove leftover "struct pc_dimms_capacity"

Not needed anymore, let's drop it.

Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180619134141.29478-2-david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

qemu-options: Add missing newline to -accel help text

The newline was removed by commit c97d6d2c, and broke -help output:

Before this patch:

  $ qemu-system-x86_64 -help | grep smp
                  thread=single|multi (enable multi-threaded TCG)-smp [...]

After this patch:

  $ qemu-system-x86_64 -help  | grep smp
  -smp [cpus=]n[,maxcpus=cpus][,cores=cores][,threads=threads][,sockets=sockets]

Fixes: c97d6d2cdf97edb4aebe832fdba65d701ad7bcb6
Cc: Sergio Andres Gomez Del Real <sergio.g.delreal@gmail.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <20180611195607.3015-1-ehabkost@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

Replace '-enable-kvm' with '-accel kvm' in docs and help texts

The preferred way to select the KVM accelerator is to use "-accel kvm"
these days, so let's be consistent in our documentation and help texts.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <1528866321-23886-3-git-send-email-thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

configure: enable debug-mutex if debug enabled

Reviewed-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180425025459.5258-5-peterx@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

QemuMutex: support --enable-debug-mutex

We have had some tracing tools for mutex but it's not easy to use them
for e.g. dead locks.  Let's provide "--enable-debug-mutex" parameter
when configure to allow QemuMutex to store the last owner that took
specific lock.  It will be easy to use this tool to debug deadlocks
since we can directly know who took the lock then as long as we can have
a debugger attached to the process.

Reviewed-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180425025459.5258-4-peterx@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

qemu-thread: introduce qemu-thread-common.h

Introduce some hooks for the shared part of qemu thread between POSIX
and Windows implementations.  Note that in qemu_mutex_unlock_impl() we
moved the call before unlock operation which should make more sense.
And we don't need qemu_mutex_post_unlock() hook.

Put all these shared hooks into the header files.  It should be internal
to qemu-thread but not for qemu-thread users, hence put into util/
directory.

Reviewed-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180425025459.5258-3-peterx@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

tests/atomic_add-bench: add -m option to use mutexes

This allows us to use atomic-add-bench as a microbenchmark
for evaluating qemu_mutex_lock's performance.

Signed-off-by: Emilio G. Cota <cota@braap.org>
[cherry picked from https://github.com/cota/qemu/commit/f04f34df]
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180425025459.5258-2-peterx@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

kvm: Delete the slot if and only if the KVM_MEM_READONLY flag is changed

According to KVM commit 75d61fbc, it needs to delete the slot before
changing the KVM_MEM_READONLY flag. But QEMU commit 235e8982 only check
whether KVM_MEM_READONLY flag is set instead of changing. It doesn't
need to delete the slot if the KVM_MEM_READONLY flag is not changed.

This fixes a issue that migrating a VM at the OVMF startup stage and
VM is executing the codes in rom. Between the deleting and adding the
slot in kvm_set_user_memory_region, there is a chance that guest access
rom and trap to KVM, then KVM can't find the corresponding memslot.
While KVM (on ARM) injects an abort to guest due to the broken hva, then
guest will get stuck.

Signed-off-by: Shannon Zhao <zhaoshenglong@huawei.com>
Message-Id: <1526462314-19720-1-git-send-email-zhaoshenglong@huawei.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

exec: check that alignment is a power of two

Right now we can crash QEMU using e.g.

qemu-system-x86_64 -m 256M,maxmem=20G,slots=2 \
 -object memory-backend-file,id=mem0,size=12288,mem-path=/dev/zero,align=12288 \
 -device pc-dimm,id=dimm1,memdev=mem0

qemu-system-x86_64: util/mmap-alloc.c:115:
 qemu_ram_mmap: Assertion `is_power_of_2(align)' failed

Fix this by adding a proper check.

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180607154705.6316-3-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

memory-device: turn alignment assert into check

The start of the address space indicates which maximum alignment is
supported by our machine (e.g. ppc, x86 1GB). This is helpful to
catch fragmenting guest physical memory in strange fashions.

Right now we can crash QEMU by e.g. (there might be easier examples)

qemu-system-x86_64 -m 256M,maxmem=20G,slots=2 \
 -object memory-backend-file,id=mem0,size=8192M,mem-path=/dev/zero,align=8192M \
 -device pc-dimm,id=dimm1,memdev=mem0

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180607154705.6316-2-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

whpx: commit missing file

Not included by mistake in commit 327fccb288976f95808efa968082fc9d4a9ced84.

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