qmp: Fix tracing of non-string command IDs

Tracepoints monitor_qmp_cmd_in_band and
monitor_qmp_cmd_out_of_band (commit cf869d5317 "qmp: support
out-of-band (oob) execution") treat non-string "id" like absent "id".
Fix that.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201211171152.146877-10-armbru@redhat.com>

qobject: Move internals to qobject-internal.h

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201211171152.146877-9-armbru@redhat.com>

hw/rdma: Replace QList by GQueue

RdmaProtectedQList provides a thread-safe queue of int64_t on top of a
QList.

rdma_protected_qlist_destroy() calls qlist_destroy_obj() directly.
qlist_destroy_obj() is actually for use by qobject_destroy() only.
The next commit will make that obvious.

The minimal fix would be calling qobject_unref() instead.  But QList
is actually a bad fit here.  It's designed for representing JSON
arrays.  We're better off with a GQueue here.  Replace.

Cc: Yuval Shaia <yuval.shaia.ml@gmail.com>
Cc: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201211171152.146877-8-armbru@redhat.com>

Revert "qstring: add qstring_free()"

This reverts commit 164c374b75f87c6765a705c4418ab7005a2d356f.

A free function for a reference-counted object is in bad taste.
Fortunately, this one is now also unused.  Drop it.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201211171152.146877-7-armbru@redhat.com>

qobject: Change qobject_to_json()'s value to GString

qobject_to_json() and qobject_to_json_pretty() build a GString, then
covert it to QString.  Just one of the callers actually needs a
QString: qemu_rbd_parse_filename().  A few others need a string they
can modify: qmp_send_response(), qga's send_response(), to_json_str(),
and qmp_fd_vsend_fds().  The remainder just need a string.

Change qobject_to_json() and qobject_to_json_pretty() to return the
GString.

qemu_rbd_parse_filename() now has to convert to QString.  All others
save a QString temporary.  to_json_str() actually becomes a bit
simpler, because GString provides more convenient modification
functions.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201211171152.146877-6-armbru@redhat.com>

qobject: Use GString instead of QString to accumulate JSON

QString supports modifying its string, but it's quite limited: you can
only append.  The remaining callers use it for building an initial
string, never for modifying it later.

Use of GString for building the initial string is actually more
convenient here.  Change qobject_to_json() & friends to do that.

Once all such uses are replaced this way, QString can become immutable.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201211171152.146877-5-armbru@redhat.com>

qobject: Make qobject_to_json_pretty() take a pretty argument

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201211171152.146877-4-armbru@redhat.com>

monitor: Use GString instead of QString for output buffer

GString has a richer set of string operations than QString.  It should
be preferred to QString except where we need a QObject or reference
counting.  We don't here.  Switch to GString, and put its richer
interface to use.

Cc: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201211171152.146877-3-armbru@redhat.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>

hmp: Simplify how qmp_human_monitor_command() gets output

Commit 48c043d0d1 "hmp: human-monitor-command: stop using the Memory
chardev driver" left us "if string is non-empty, duplicate it, else
duplicate the empty string".  Meh.  Duplicate it unconditionally.

Cc: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201211171152.146877-2-armbru@redhat.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>

test-visitor-serialization: Clean up test_primitives()

test_primitives() uses union member intmax_t max to compare the
integer members.  Unspecified behavior.  Has worked fine for many
years, though.  Clean it up.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201210161452.2813491-11-armbru@redhat.com>

test-visitor-serialization: Drop insufficient precision workaround

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201210161452.2813491-10-armbru@redhat.com>

string-output-visitor: Fix to use sufficient precision

The string output visitor should serialize numbers so that the string
input visitor deserializes them back to the same number.  It fails to
do so.

print_type_number() uses format %f.  This is prone to nasty rounding
errors.  For instance, numbers between 0 and 0.0000005 get flushed to
zero.

We currently use this visitor only for HMP info migrate, info network,
info qtree, and info memdev.  No double values occur there as far as I
can tell.

Fix anyway by formatting with %.17g.  17 decimal digits always suffice
for IEEE double.

See also recent commit "qobject: Fix qnum_to_string() to use
sufficient precision".

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201210161452.2813491-9-armbru@redhat.com>

test-string-output-visitor: Cover "unround" number

This demonstrates rounding error due to insufficient precision: double
3.1415926535897932 gets converted to JSON 3.141593.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201210161452.2813491-8-armbru@redhat.com>

qobject: Fix qnum_to_string() to use sufficient precision

We should serialize numbers to JSON so that they deserialize back to
the same number.  We fail to do so.

The culprit is qnum_to_string(): it uses format %f with trailing '0'
trimmed.  Results in pretty output for "nice" numbers, but is prone to
nasty rounding errors.  For instance, numbers between 0 and 0.0000005
get flushed to zero.

Where exactly the incorrect rounding can bite is tiresome to gauge.
Here's my take.

* In QMP output, type 'number':

  - query-blockstats value avg_rd_queue_depth

  - QMP query-migrate values mbps, cache-miss-rate, encoding-rate,
    busy-rate, compression-rate.

  Relatively harmless, I guess.

* In tracing QMP input.  Harmless.

* In qemu-ga output, type 'number': guest-get-users value login-time.
  Harmless.

* In output of HMP qom-get.  Harmless.

Not affected, because double values don't actually occur there (I
think):

* QMP output, type 'any':

  * qom-get value

  * qom-list, qom-list-properties value default-value

  * query-cpu-model-comparison, query-cpu-model-baseline,
    query-cpu-model-expansion value props.

* qemu-img --output json output.

* "json:" pseudo-filenames generated by bdrv_refresh_filename().

* The rbd block driver's "=keyvalue-pairs" hack.

* In -object help on property default values.  Aside: use of JSON
  feels inappropriate here.

* Output of HMP qom-get.

* Argument conversion to QemuOpts for qdev_device_add() and HMP with
  qemu_opts_from_qdict()

  QMP and HMP device_add, virtio-net failover primary creation,
  xen-usb "usb-host" creation, HMP netdev_add, object_add.

* The uses of qobject_input_visitor_new_flat_confused()

  As far as I can tell, none of the visited types contain double
  values.

* Dumping ImageInfoSpecific with dump_qobject()

Fix by formatting with %.17g.  17 decimal digits always suffice for
IEEE double.

The change to expected test output illustrates the effect: the
rounding errors are gone, but some seemingly "nice" numbers now get
converted to not so nice strings, e.g. 0.42 to "0.41999999999999998".
This is because 0.42 is not representable exactly in double.  It's
more accurate in this example than strictly necessary, though.

If ugly accuracy bothers us, we can we can try using the least number
of digits that still converts back to the same double.  In this
example, "0.42" would do.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201210161452.2813491-7-armbru@redhat.com>

tests/check-qnum: Cover qnum_to_string() for "unround" argument

qnum_to_string() has a FIXME comment about rounding errors due to
insufficient precision.  Cover it: 2.718281828459045 gets converted to
"2.718282".  The next commit will fix it.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201210161452.2813491-6-armbru@redhat.com>

tests/check-qjson: Replace redundant large_number()

Move one of large_number()'s three checks to uint_number(), and the
other two to float_number().

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201210161452.2813491-5-armbru@redhat.com>

tests/check-qjson: Cover number 2^63

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201210161452.2813491-4-armbru@redhat.com>

tests/check-qjson: Examine QNum more thoroughly

simple_number() checks only qnum_get_try_int().  Also check
qnum_get_try_uint() and qnum_get_double().

float_number() checks only qnum_get_double().  Also check
qnum_get_try_int() and qnum_get_try_uint().

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201210161452.2813491-3-armbru@redhat.com>

tests/check-qjson: Don't skip funny QNumber to JSON conversions

simple_number() and float_number() convert from JSON to QNumber and
back.

simple_number() tests "-0", but skips the conversion back to JSON,
because it yields "0", not "-0".  Works as intended, so better cover
it: don't skip, but expect the funny result.

float_number() tests "-32.20e-10", but skips the conversion back to
JSON, because it yields "-0".  This is a known bug in
qnum_to_string(), marked FIXME there.  Cover the bug: don't skip, but
expect the funny result.

While there, switch from g_assert() to g_assert_cmpstr() & friends for
friendlier test failures.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201210161452.2813491-2-armbru@redhat.com>

qapi: Use QAPI_LIST_PREPEND() where possible

Anywhere we create a list of just one item or by prepending items
(typically because order doesn't matter), we can use
QAPI_LIST_PREPEND().  But places where we must keep the list in order
by appending remain open-coded until later patches.

Note that as a side effect, this also performs a cleanup of two minor
issues in qga/commands-posix.c: the old code was performing
 new = g_malloc0(sizeof(*ret));
which 1) is confusing because you have to verify whether 'new' and
'ret' are variables with the same type, and 2) would conflict with C++
compilation (not an actual problem for this file, but makes
copy-and-paste harder).

Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <20201113011340.463563-5-eblake@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
[Straightforward conflicts due to commit a8aa94b5f8 "qga: update
schema for guest-get-disks 'dependents' field" and commit a10b453a52
"target/mips: Move mips_cpu_add_definition() from helper.c to cpu.c"
resolved.  Commit message tweaked.]
Signed-off-by: Markus Armbruster <armbru@redhat.com>

migration: Refactor migrate_cap_add

Instead of taking a list parameter and returning a new head at a
distance, just return the new item for the caller to insert into a
list via QAPI_LIST_PREPEND.

Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <20201113011340.463563-4-eblake@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

rocker: Revamp fp_port_get_info

Instead of modifying the value member of a list element passed as a
parameter, and open-coding the manipulation of that list, it's nicer
to just return a freshly allocated value to be prepended to a list
using QAPI_LIST_PREPEND.

Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <20201113011340.463563-3-eblake@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

Merge remote-tracking branch 'remotes/alistair/tags/pull-riscv-to-apply-20201217-1' into staging

A collection of RISC-V improvements:
 - Improve the sifive_u DTB generation
 - Add QSPI NOR flash to Microchip PFSoC
 - Fix a bug in the Hypervisor HLVX/HLV/HSV instructions
 - Fix some mstatus mask defines
 - Ibex PLIC improvements
 - OpenTitan memory layout update
 - Initial steps towards support for 32-bit CPUs on 64-bit builds

# gpg: Signature made Fri 18 Dec 2020 05:59:42 GMT
# gpg:                using RSA key F6C4AC46D4934868D3B8CE8F21E10D29DF977054
# gpg: Good signature from "Alistair Francis <alistair@alistair23.me>" [full]
# Primary key fingerprint: F6C4 AC46 D493 4868 D3B8  CE8F 21E1 0D29 DF97 7054

* remotes/alistair/tags/pull-riscv-to-apply-20201217-1: (23 commits)
  riscv/opentitan: Update the OpenTitan memory layout
  hw/riscv: Use the CPU to determine if 32-bit
  target/riscv: cpu: Set XLEN independently from target
  target/riscv: csr: Remove compile time XLEN checks
  target/riscv: cpu_helper: Remove compile time XLEN checks
  target/riscv: cpu: Remove compile time XLEN checks
  target/riscv: Specify the XLEN for CPUs
  target/riscv: Add a riscv_cpu_is_32bit() helper function
  target/riscv: fpu_helper: Match function defs in HELPER macros
  hw/riscv: sifive_u: Remove compile time XLEN checks
  hw/riscv: spike: Remove compile time XLEN checks
  hw/riscv: virt: Remove compile time XLEN checks
  hw/riscv: boot: Remove compile time XLEN checks
  riscv: virt: Remove target macro conditionals
  riscv: spike: Remove target macro conditionals
  target/riscv: Add a TYPE_RISCV_CPU_BASE CPU
  hw/riscv: Expand the is 32-bit check to support more CPUs
  intc/ibex_plic: Clear interrupts that occur during claim process
  target/riscv: Fix definition of MSTATUS_TW and MSTATUS_TSR
  target/riscv: Fix the bug of HLVX/HLV/HSV
  ...

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

riscv/opentitan: Update the OpenTitan memory layout

OpenTitan is currently only avalible on an FPGA platform and the memory
addresses have changed. Update to use the new memory addresses.

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 8eb65314830a75d0fea3fccf77bc45b8ddd01c42.1607982831.git.alistair.francis@wdc.com

hw/riscv: Use the CPU to determine if 32-bit

Instead of using string compares to determine if a RISC-V machine is
using 32-bit or 64-bit CPUs we can use the initalised CPUs. This avoids
us having to maintain a list of CPU names to compare against.

This commit also fixes the name of the function to match the
riscv_cpu_is_32bit() function.

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 8ab7614e5df93ab5267788b73dcd75f9f5615e82.1608142916.git.alistair.francis@wdc.com

target/riscv: cpu: Set XLEN independently from target

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Bin Meng <bin.meng@windriver.com>
Tested-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Palmer Dabbelt <palmerdabbelt@google.com>
Acked-by: Palmer Dabbelt <palmerdabbelt@google.com>
Message-id: 7eddba45b5d223321c031431849fdd42eceb514b.1608142916.git.alistair.francis@wdc.com

target/riscv: csr: Remove compile time XLEN checks

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Palmer Dabbelt <palmerdabbelt@google.com>
Acked-by: Palmer Dabbelt <palmerdabbelt@google.com>
Tested-by: Bin Meng <bin.meng@windriver.com>
Message-id: 7371180970b7db310d3a1da21d03d33499c2beb0.1608142916.git.alistair.francis@wdc.com

target/riscv: cpu_helper: Remove compile time XLEN checks

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Palmer Dabbelt <palmerdabbelt@google.com>
Acked-by: Palmer Dabbelt <palmerdabbelt@google.com>
Reviewed-by: Bin Meng <bin.meng@windriver.com>
Tested-by: Bin Meng <bin.meng@windriver.com>
Message-id: 872d2dfcd1c7c3914655d677e911b9432eb8f340.1608142916.git.alistair.francis@wdc.com

target/riscv: cpu: Remove compile time XLEN checks

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Palmer Dabbelt <palmerdabbelt@google.com>
Acked-by: Palmer Dabbelt <palmerdabbelt@google.com>
Tested-by: Bin Meng <bin.meng@windriver.com>
Message-id: a426ead44db5065a0790066d43e91245683509d7.1608142916.git.alistair.francis@wdc.com

target/riscv: Specify the XLEN for CPUs

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Bin Meng <bin.meng@windriver.com>
Tested-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Palmer Dabbelt <palmerdabbelt@google.com>
Acked-by: Palmer Dabbelt <palmerdabbelt@google.com>
Message-id: c1da66affbb83ec4a2fbeb0194293bd24d65f5dc.1608142916.git.alistair.francis@wdc.com

target/riscv: Add a riscv_cpu_is_32bit() helper function

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Bin Meng <bin.meng@windriver.com>
Tested-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Palmer Dabbelt <palmerdabbelt@google.com>
Acked-by: Palmer Dabbelt <palmerdabbelt@google.com>
Message-id: ebd37b237a8cbe457335b948bd57f487b6b31869.1608142916.git.alistair.francis@wdc.com

target/riscv: fpu_helper: Match function defs in HELPER macros

Update the function definitions generated in helper.h to match the
actual function implementations.

Also remove all compile time XLEN checks when building.

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 614c369cb0000d070873a647b8aac7e023cba145.1608142916.git.alistair.francis@wdc.com

hw/riscv: sifive_u: Remove compile time XLEN checks

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Palmer Dabbelt <palmerdabbelt@google.com>
Acked-by: Palmer Dabbelt <palmerdabbelt@google.com>
Reviewed-by: Bin Meng <bin.meng@windriver.com>
Tested-by: Bin Meng <bin.meng@windriver.com>
Message-id: 40d6df4dd05302c566e419be3a1fef7799e57c2e.1608142916.git.alistair.francis@wdc.com

hw/riscv: spike: Remove compile time XLEN checks

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Palmer Dabbelt <palmerdabbelt@google.com>
Acked-by: Palmer Dabbelt <palmerdabbelt@google.com>
Message-id: ac75037dd58061486de421a0fcd9ac8a92014607.1608142916.git.alistair.francis@wdc.com

hw/riscv: virt: Remove compile time XLEN checks

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Bin Meng <bin.meng@windriver.com>
Tested-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Palmer Dabbelt <palmerdabbelt@google.com>
Acked-by: Palmer Dabbelt <palmerdabbelt@google.com>
Message-id: d7ca1aca672515e6a4aa0d41716238b055f3f25c.1608142916.git.alistair.francis@wdc.com

hw/riscv: boot: Remove compile time XLEN checks

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Bin Meng <bin.meng@windriver.com>
Tested-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Palmer Dabbelt <palmerdabbelt@google.com>
Acked-by: Palmer Dabbelt <palmerdabbelt@google.com>
Message-id: 51e9842dbed1acceebad7f97bd3aae69aa1ac19e.1608142916.git.alistair.francis@wdc.com

riscv: virt: Remove target macro conditionals

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Bin Meng <bin.meng@windriver.com>
Tested-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Palmer Dabbelt <palmerdabbelt@google.com>
Acked-by: Palmer Dabbelt <palmerdabbelt@google.com>
Message-id: aed1174c2efd2f050fa5bd8f524d68795b12c0e4.1608142916.git.alistair.francis@wdc.com

riscv: spike: Remove target macro conditionals

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Palmer Dabbelt <palmerdabbelt@google.com>
Acked-by: Palmer Dabbelt <palmerdabbelt@google.com>
Message-id: 04ac7fba2348c92f296a5e6a9959ac72b77ae4c6.1608142916.git.alistair.francis@wdc.com

target/riscv: Add a TYPE_RISCV_CPU_BASE CPU

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Bin Meng <bin.meng@windriver.com>
Tested-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Palmer Dabbelt <palmerdabbelt@google.com>
Acked-by: Palmer Dabbelt <palmerdabbelt@google.com>
Message-id: 86e5ccd9eae2f5d8c2257679c6ccf6078a5d51af.1608142916.git.alistair.francis@wdc.com

hw/riscv: Expand the is 32-bit check to support more CPUs

Currently the riscv_is_32_bit() function only supports the generic rv32
CPUs. Extend the function to support the SiFive and LowRISC CPUs as
well.

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Palmer Dabbelt <palmerdabbelt@google.com>
Acked-by: Palmer Dabbelt <palmerdabbelt@google.com>
Message-id: 9a13764115ba78688ba61b56526c6de65fc3ef42.1608142916.git.alistair.francis@wdc.com

intc/ibex_plic: Clear interrupts that occur during claim process

Previously if an interrupt occured during the claim process (after the
interrupt is claimed but before it's completed) it would never be
cleared.
This patch ensures that we also clear the hidden_pending bits as well.

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Jackie Ke <jackieke724@hotmail.com>
Message-id: 4e9786084a86f220689123cc8a7837af8fa071cf.1607100423.git.alistair.francis@wdc.com

target/riscv: Fix definition of MSTATUS_TW and MSTATUS_TSR

The TW and TSR fields should be bits 21 and 22 and not 30/29.
This was found while comparing QEMU behaviour against the sail formal
model (https://github.com/rems-project/sail-riscv/).

Signed-off-by: Alex Richardson <Alexander.Richardson@cl.cam.ac.uk>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20201130170117.71281-1-Alexander.Richardson@cl.cam.ac.uk
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Fix the bug of HLVX/HLV/HSV

We found that the hypervisor virtual-machine load and store instructions,
included HLVX/HLV/HSV, couldn't access guest userspace memory.

In the riscv-privileged spec, HLVX/HLV/HSV is defined as follow:
"As usual when V=1, two-stage address translation is applied, and
the HS-level sstatus.SUM is ignored."

But get_physical_address() doesn't ignore sstatus.SUM, when HLVX/HLV/HSV
accesses guest userspace memory. So this patch fixes it.

Signed-off-by: Yifei Jiang <jiangyifei@huawei.com>
Signed-off-by: Yipeng Yin <yinyipeng1@huawei.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20201130012810.899-1-jiangyifei@huawei.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/core/register.c: Don't use '#' flag of printf format

Fix code style. Don't use '#' flag of printf format ('%#') in
format strings, use '0x' prefix instead

Signed-off-by: Xinhao Zhang <zhangxinhao1@huawei.com>
Signed-off-by: Kai Deng <dengkai1@huawei.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20201116140148.2850128-1-zhangxinhao1@huawei.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv: microchip_pfsoc: add QSPI NOR flash

Add QSPI NOR flash definition for Microchip PolarFire SoC.

Signed-off-by: Vitaly Wool <vitaly.wool@konsulko.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Bin Meng <bin.meng@windriver.com>
Message-id: 20201112074950.33283-1-vitaly.wool@konsulko.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv: sifive_u: Add UART1 DT node in the generated DTB

The sifive_u machine emulates two UARTs but we have only UART0 DT
node in the generated DTB so this patch adds UART1 DT node in the
generated DTB.

Signed-off-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20201111094725.3768755-1-anup.patel@wdc.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

Merge remote-tracking branch 'remotes/ehabkost-gl/tags/x86-next-pull-request' into staging

x86 queue, 2020-12-17

Features:
* AVX512_FP16 feature (Cathy Zhang)

Cleanups:
* accel code cleanup (Claudio Fontana)
* hyperv initialization cleanup (Vitaly Kuznetsov)

# gpg: Signature made Thu 17 Dec 2020 18:44:45 GMT
# gpg:                using RSA key 5A322FD5ABC4D3DBACCFD1AA2807936F984DC5A6
# gpg:                issuer "ehabkost@redhat.com"
# gpg: Good signature from "Eduardo Habkost <ehabkost@redhat.com>" [full]
# Primary key fingerprint: 5A32 2FD5 ABC4 D3DB ACCF  D1AA 2807 936F 984D C5A6

* remotes/ehabkost-gl/tags/x86-next-pull-request:
  cpu: Remove unnecessary noop methods
  tcg: Make CPUClass.debug_excp_handler optional
  tcg: make CPUClass.cpu_exec_* optional
  tcg: cpu_exec_{enter,exit} helpers
  i386: tcg: remove inline from cpu_load_eflags
  i386: move TCG cpu class initialization to tcg/
  x86/cpu: Add AVX512_FP16 cpu feature
  i386: move hyperv_limits initialization to x86_cpu_realizefn()
  i386: move hyperv_version_id initialization to x86_cpu_realizefn()
  i386: move hyperv_interface_id initialization to x86_cpu_realizefn()
  i386: move hyperv_vendor_id initialization to x86_cpu_realizefn()
  i386: move cpu dump out of helper.c into cpu-dump.c
  i386: move TCG accel files into tcg/
  i386: hvf: remove stale MAINTAINERS entry for old hvf stubs
  i386: move hax accel files into hax/
  i386: move whpx accel files into whpx/
  i386: move kvm accel files into kvm/

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

cpu: Remove unnecessary noop methods

In the previous commits we made cpu_exec_* and debug_excp_handler
optional, so we can now remove these no-op handlers.

Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20201212155530.23098-13-cfontana@suse.de>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

tcg: Make CPUClass.debug_excp_handler optional

Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20201212155530.23098-12-cfontana@suse.de>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

tcg: make CPUClass.cpu_exec_* optional

This will let us simplify the code that initializes CPU class
methods, when we move cpu_exec_*() to a separate struct.

Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20201212155530.23098-11-cfontana@suse.de>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

tcg: cpu_exec_{enter,exit} helpers

Move invocation of CPUClass.cpu_exec_*() to separate helpers,
to make it easier to refactor that code later.

Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20201212155530.23098-10-cfontana@suse.de>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

i386: tcg: remove inline from cpu_load_eflags

make it a regular function.

Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20201212155530.23098-9-cfontana@suse.de>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

i386: move TCG cpu class initialization to tcg/

to do this, we need to take code out of cpu.c and helper.c,
and also move some prototypes from cpu.h, for code that is
needed in tcg/xxx_helper.c, and which in turn is part of the
callbacks registered by the class initialization.

Therefore, do some shuffling of the parts of cpu.h that
are only relevant for tcg/, and put them in tcg/helper-tcg.h

For FT0 and similar macros, put them in tcg/fpu-helper.c
since they are used only there.

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20201212155530.23098-8-cfontana@suse.de>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

x86/cpu: Add AVX512_FP16 cpu feature

AVX512 Half-precision floating point (FP16) has better performance
compared to FP32 if the presicion or magnitude requirements are met.
It's defined as CPUID.(EAX=7,ECX=0):EDX[bit 23].

Refer to
https://software.intel.com/content/www/us/en/develop/download/\
intel-architecture-instruction-set-extensions-programming-reference.html

Signed-off-by: Cathy Zhang <cathy.zhang@intel.com>
Message-Id: <20201216224002.32677-1-cathy.zhang@intel.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

i386: move hyperv_limits initialization to x86_cpu_realizefn()

As a preparation to expanding Hyper-V CPU features early, move
hyperv_limits initialization to x86_cpu_realizefn().

Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Message-Id: <20201119103221.1665171-5-vkuznets@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

i386: move hyperv_version_id initialization to x86_cpu_realizefn()

As a preparation to expanding Hyper-V CPU features early, move
hyperv_version_id initialization to x86_cpu_realizefn().

Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Message-Id: <20201119103221.1665171-4-vkuznets@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

i386: move hyperv_interface_id initialization to x86_cpu_realizefn()

As a preparation to expanding Hyper-V CPU features early, move
hyperv_interface_id initialization to x86_cpu_realizefn().

Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Message-Id: <20201119103221.1665171-3-vkuznets@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

i386: move hyperv_vendor_id initialization to x86_cpu_realizefn()

As a preparation to expanding Hyper-V CPU features early, move
hyperv_vendor_id initialization to x86_cpu_realizefn(). Introduce
x86_cpu_hyperv_realize() to not not pollute x86_cpu_realizefn()
itself.

Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Message-Id: <20201119103221.1665171-2-vkuznets@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

i386: move cpu dump out of helper.c into cpu-dump.c

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20201212155530.23098-7-cfontana@suse.de>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

i386: move TCG accel files into tcg/

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
[claudio: moved cc_helper_template.h to tcg/ too]

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Message-Id: <20201212155530.23098-6-cfontana@suse.de>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

i386: hvf: remove stale MAINTAINERS entry for old hvf stubs

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Roman Bolshakov <r.bolshakov@yadro.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20201212155530.23098-5-cfontana@suse.de>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

i386: move hax accel files into hax/

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20201212155530.23098-4-cfontana@suse.de>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

i386: move whpx accel files into whpx/

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20201212155530.23098-3-cfontana@suse.de>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

i386: move kvm accel files into kvm/

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20201212155530.23098-2-cfontana@suse.de>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

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

* New -action option and set-action QMP command (Alejandro)
* More vl.c cleanup (myself with help from Daniel and Igor)
* Remove deprecated options (Philippe, Thomas)
* Dirty bitmap fix (Zenghui)
* icount caching speedup (Pavel)
* SCSI race fix (Maxim)
* Remove pre-GCC 4.8 code (Marc-André)

# gpg: Signature made Tue 15 Dec 2020 17:53:24 GMT
# gpg:                using RSA key F13338574B662389866C7682BFFBD25F78C7AE83
# gpg:                issuer "pbonzini@redhat.com"
# gpg: Good signature from "Paolo Bonzini <bonzini@gnu.org>" [full]
# gpg:                 aka "Paolo Bonzini <pbonzini@redhat.com>" [full]
# 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-gitlab/tags/for-upstream: (45 commits)
  build: -no-pie is no functional linker flag
  scripts/git.orderfile: Keep files with .inc extension sorted
  compiler.h: remove QEMU_GNUC_PREREQ
  linux-user: remove GNUC check
  compiler: remove GNUC check
  xen: remove GNUC check
  poison: remove GNUC check
  compiler.h: explicit case for Clang printf attribute
  virtiofsd: replace _Static_assert with QEMU_BUILD_BUG_ON
  tests: remove GCC < 4 fallbacks
  qemu-plugin.h: remove GCC < 4
  compiler.h: remove GCC < 3 __builtin_expect fallback
  accel/tcg: Remove special case for GCC < 4.6
  qemu/atomic: Drop special case for unsupported compiler
  hw/core: Restrict 'fw-path-provider.c' to system mode emulation
  docs: set CONFDIR when running sphinx
  vl: rename local variable in configure_accelerators
  qemu-option: pass QemuOptsList to opts_accepts_any
  qemu-option: simplify search for end of key
  kvm: Take into account the unaligned section size when preparing bitmap
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
# Conflicts:
# softmmu/vl.c

build: -no-pie is no functional linker flag

Recent binutils changes dropping unsupported options [1] caused a build
issue in regard to the optionroms.

  ld -m elf_i386 -T /<<PKGBUILDDIR>>/pc-bios/optionrom//flat.lds -no-pie \
    -s -o multiboot.img multiboot.o
  ld.bfd: Error: unable to disambiguate: -no-pie (did you mean --no-pie ?)

This isn't really a regression in ld.bfd, filing the bug upstream
revealed that this never worked as a ld flag [2] - in fact it seems we
were by accident setting --nmagic).

Since it never had the wanted effect this usage of LDFLAGS_NOPIE, should be
droppable without any effect. This also is the only use-case of LDFLAGS_NOPIE
in .mak, therefore we can also remove it from being added there.

[1]: https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=983d925d
[2]: https://sourceware.org/bugzilla/show_bug.cgi?id=27050#c5

Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
Message-Id: <20201214150938.1297512-1-christian.ehrhardt@canonical.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

scripts/git.orderfile: Keep files with .inc extension sorted

Sort .inc files along with the extension including them.

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20201213205132.243628-1-f4bug@amsat.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

compiler.h: remove QEMU_GNUC_PREREQ

When needed, the G_GNUC_CHECK_VERSION() glib macro can be used instead.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20201210134752.780923-14-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

linux-user: remove GNUC check

QEMU requires Clang or GCC, that define and support __GNUC__ extensions.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20201210134752.780923-13-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

compiler: remove GNUC check

QEMU requires Clang or GCC, that define and support __GNUC__ extensions.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-Id: <20201210134752.780923-12-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

xen: remove GNUC check

QEMU requires Clang or GCC, that define and support __GNUC__ extensions

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Acked-by: Stefano Stabellini <sstabellini@kernel.org>
Message-Id: <20201210134752.780923-11-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

poison: remove GNUC check

QEMU requires Clang or GCC, that define and support __GNUC__ extensions

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20201210134752.780923-10-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

compiler.h: explicit case for Clang printf attribute

Since commit efc6c07 ("configure: Add a test for the minimum compiler
version"), QEMU explicitely depends on GCC >= 4.8, we could thus drop
earlier version checks. Except clang advertizes itself as GCC 4.2.1.

Since clang doesn't support gnu_printf, make that case explicitely and
drop GCC version check.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-Id: <20201210134752.780923-8-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

virtiofsd: replace _Static_assert with QEMU_BUILD_BUG_ON

This allows to get rid of a check for older GCC version (which was a bit
bogus too since it was falling back on c++ version..)

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Message-Id: <20201210134752.780923-7-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

tests: remove GCC < 4 fallbacks

Since commit efc6c07 ("configure: Add a test for the minimum compiler
version"), QEMU explicitely depends on GCC >= 4.8.

(clang >= 3.4 advertizes itself as GCC >= 4.2 compatible)

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Acked-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20201210134752.780923-6-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

qemu-plugin.h: remove GCC < 4

Since commit efc6c07 ("configure: Add a test for the minimum compiler
version"), QEMU explicitely depends on GCC >= 4.8.

(clang >= 3.4 advertizes itself as GCC >= 4.2 compatible)

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Acked-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20201210134752.780923-5-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

compiler.h: remove GCC < 3 __builtin_expect fallback

Since commit efc6c07 ("configure: Add a test for the minimum compiler
version"), QEMU explicitely depends on GCC >= 4.8.

(clang >= 3.4 advertizes itself as GCC >= 4.2 compatible and supports
__builtin_expect too)

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20201210134752.780923-4-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

accel/tcg: Remove special case for GCC < 4.6

Since commit efc6c070aca ("configure: Add a test for the
minimum compiler version") the minimum compiler version
required for GCC is 4.8.

We can safely remove the special case for GCC 4.6 introduced
in commit 0448f5f8b81 ("cpu-exec: Fix compiler warning
(-Werror=clobbered)").
No change for Clang as we don't know.

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20201210134752.780923-3-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

qemu/atomic: Drop special case for unsupported compiler

Since commit efc6c070aca ("configure: Add a test for the
minimum compiler version") the minimum compiler version
required for GCC is 4.8, which has the GCC BZ#36793 bug fixed.

We can safely remove the special case introduced in commit
a281ebc11a6 ("virtio: add missing mb() on notification").

With clang 3.4, __ATOMIC_RELAXED is defined, so the chunk to
remove (which is x86-specific), isn't reached either.

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20201210134752.780923-2-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/core: Restrict 'fw-path-provider.c' to system mode emulation

fw-path-provider.c is only consumed by qdev-fw.c, which itself
is in softmmu_ss[], so we can restrict fw-path-provider.c to
softmmu too.

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20201207220709.4017938-1-philmd@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

docs: set CONFDIR when running sphinx

The default configuration path /etc/qemu can be overriden with configure
options, and the generated documentation used to reflect it.

Fixes regression introduced in commit
f8aa24ea9a82da38370470c6bc0eaa393999edfe ("meson: sphinx-build").

Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1902537
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20201201183704.299697-1-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

vl: rename local variable in configure_accelerators

Silly patch extracted from the next one, which is already big enough.

Because there are already local variables named "accel", we will name
the global vl.c variable for "-M accel" accelerators instead.  Rename
it already in configure_accelerators to be ready.

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

qemu-option: pass QemuOptsList to opts_accepts_any

A QemuOptsList can be of one of two kinds: either it is pre-validated, or
it accepts any key and validation happens somewhere else (typically in
a Visitor or against a list of QOM properties).  opts_accepts_any
returns true if a QemuOpts instance was created from a QemuOptsList of
the latter kind, but there is no function to do the check on a QemuOptsList.

Since this property comes from the QemuOptsList and almost all callers of
opts_accepts_any use opts->list anyway, modify the function to accept
QemuOptsList.

Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

qemu-option: simplify search for end of key

Use strcspn to find an equal or comma value, and pass the result directly
to get_opt_name to avoid another strchr.

Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

kvm: Take into account the unaligned section size when preparing bitmap

The kernel KVM_CLEAR_DIRTY_LOG interface has align requirement on both the
start and the size of the given range of pages. We have been careful to
handle the unaligned cases when performing CLEAR on one slot. But it seems
that we forget to take the unaligned *size* case into account when
preparing bitmap for the interface, and we may end up clearing dirty status
for pages outside of [start, start + size).

If the size is unaligned, let's go through the slow path to manipulate a
temp bitmap for the interface so that we won't bother with those unaligned
bits at the end of bitmap.

I don't think this can happen in practice since the upper layer would
provide us with the alignment guarantee. I'm not sure if kvm-all could rely
on it. And this patch is mainly intended to address correctness of the
specific algorithm used inside kvm_log_clear_one_slot().

Signed-off-by: Zenghui Yu <yuzenghui@huawei.com>
Message-Id: <20201208114013.875-1-yuzenghui@huawei.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

scsi: fix device removal race vs IO restart callback on resume

There is (mostly theoretical) race between removal of a scsi device and
scsi_dma_restart_bh.

It used to be easier to hit this race prior to my / Paulo's patch series
that added rcu to scsi bus device handling code, but IMHO this race
should still be possible to hit, at least in theory.

Buglink: https://bugzilla.redhat.com/show_bug.cgi?id=1854811
Fix it anyway with a patch that was proposed by Paulo in the above bugzilla.

Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
Message-Id: <20201210125929.1136390-2-mlevitsk@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

icount: improve exec nocache usage

cpu-exec tries to execute TB without caching when current
icount budget is over. But sometimes refilled budget is big
enough to try executing cached blocks.
This patch checks that instruction budget is big enough
for next block execution instead of just running cpu_exec_nocache.
It halves the number of calls of cpu_exec_nocache function
during tested OS boot scenario.

Signed-off-by: Pavel Dovgalyuk <pavel.dovgalyuk@ispras.ru>
Message-Id: <160741865825.348476.7169239332367828943.stgit@pasha-ThinkPad-X280>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

Remove the deprecated -show-cursor option

It has been marked as deprecated since QEMU v5.0, replaced by the
corresponding parameter of the -display option.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20201210155808.233895-5-thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

Remove the deprecated -realtime option

It has been marked as deprecated since QEMU v4.2, replaced by
the -overcommit option. Time to remove it now.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20201210155808.233895-4-thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

docs/system: Move the list of removed features to a separate file

Otherwise there is a chance that new deprecated features get added
to the list of removed features at the end of the file by accident.
It's way less confusing if the removed features reside in a separate
file.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20201210155808.233895-3-thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

accel/tcg: Remove deprecated '-tb-size' option

The '-tb-size' option (replaced by '-accel tcg,tb-size') is
deprecated since 5.0 (commit fe174132478). Remove it.

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20201202112714.1223783-1-philmd@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20201210155808.233895-2-thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

memory: clamp cached translation in case it points to an MMIO region

In using the address_space_translate_internal API, address_space_cache_init
forgot one piece of advice that can be found in the code for
address_space_translate_internal:

    /* MMIO registers can be expected to perform full-width accesses based only
     * on their address, without considering adjacent registers that could
     * decode to completely different MemoryRegions.  When such registers
     * exist (e.g. I/O ports 0xcf8 and 0xcf9 on most PC chipsets), MMIO
     * regions overlap wildly.  For this reason we cannot clamp the accesses
     * here.
     *
     * If the length is small (as is the case for address_space_ldl/stl),
     * everything works fine.  If the incoming length is large, however,
     * the caller really has to do the clamping through memory_access_size.
     */

address_space_cache_init is exactly one such case where "the incoming length
is large", therefore we need to clamp the resulting length---not to
memory_access_size though, since we are not doing an access yet, but to
the size of the resulting section.  This ensures that subsequent accesses
to the cached MemoryRegionSection will be in range.

With this patch, the enclosed testcase notices that the used ring does
not fit into the MSI-X table and prints a "qemu-system-x86_64: Cannot map used"
error.

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

msix: assert that accesses are within bounds

This makes the testcase from the next patch fail.

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

qtest/pvpanic: Test panic option that allows VM to continue

Test the scenario where the -action panic=none parameter is used to
signal that the VM must continue executing after a guest panic
occurs.

Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Alejandro Jimenez <alejandro.j.jimenez@oracle.com>
Message-Id: <1607705564-26264-5-git-send-email-alejandro.j.jimenez@oracle.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

vl: Add option to avoid stopping VM upon guest panic

The current default action of pausing a guest after a panic event
is received leaves the responsibility to resume guest execution to the
management layer. The reasons for this behavior are discussed here:
https://lore.kernel.org/qemu-devel/52148F88.5000509@redhat.com/

However, in instances like the case of older guests (Linux and
Windows) using a pvpanic device but missing support for the
PVPANIC_CRASHLOADED event, and Windows guests using the hv-crash
enlightenment, it is desirable to allow the guests to continue
running after sending a PVPANIC_PANICKED event. This allows such
guests to proceed to capture a crash dump and automatically reboot
without intervention of a management layer.

Add an option to avoid stopping a VM after a panic event is received,
by passing:

-action panic=none

in the command line arguments, or during runtime by using an upcoming
QMP command.

Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Alejandro Jimenez <alejandro.j.jimenez@oracle.com>
Message-Id: <1607705564-26264-3-git-send-email-alejandro.j.jimenez@oracle.com>
[Do not fix panic action in the variable, instead modify -no-shutdown. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

vl: Add an -action option specifying response to guest events

Several command line options currently in use are meant to modify
the behavior of QEMU in response to certain guest events like:
-no-reboot, -no-shutdown, -watchdog-action.

These can be grouped into a single option of the form:

-action event=action

Which can be used to specify the existing options above in the
following format:

-action reboot=none|shutdown
-action shutdown=poweroff|pause
-action watchdog=reset|shutdown|poweroff|pause|debug|none|inject-nmi

This is done in preparation for adding yet another option of this
type, which modifies the QEMU behavior when a guest panic occurs.

Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Alejandro Jimenez <alejandro.j.jimenez@oracle.com>
Message-Id: <1607705564-26264-2-git-send-email-alejandro.j.jimenez@oracle.com>
[Use QemuOpts help support, invoke QMP command. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

qmp: generalize watchdog-set-action to -no-reboot/-no-shutdown

Add a QMP command to allow for the behaviors specified by the
-no-reboot and -no-shutdown command line option to be set at runtime.
The new command is named set-action and takes optional arguments, named
after an event, that provide a corresponding action to take.

Example:

-> { "execute": "set-action",
     "arguments": {
"reboot": "none",
"shutdown": "poweroff",
"watchdog": "debug" } }
<- { "return": {} }

Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Alejandro Jimenez <alejandro.j.jimenez@oracle.com>
Message-Id: <1607705564-26264-4-git-send-email-alejandro.j.jimenez@oracle.com>
[Split the series differently, with -action based on the QMP command. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

monitor: allow quitting while in preconfig state

Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

memory: allow creating MemoryRegions before accelerators

Compute the DIRTY_MEMORY_CODE bit in memory_region_get_dirty_log_mask
instead of memory_region_init_*.  This makes it possible to allocate
memory backend objects at any time.

Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

plugin: propagate errors

qemu_finish_machine_init currently can only exit QEMU if it fails.
Prepare for giving it proper error propagation, and possibly for
adding a plugin_add monitor command that calls an accelerator
method.

While at it, make all errors from plugin_load look the same.

Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>