tests/acceptance: add reverse debugging test

This is a test for GDB reverse debugging commands: reverse step and reverse continue.
Every test in this suite consists of two phases: record and replay.
Recording saves the execution of some instructions and makes an initial
VM snapshot to allow reverse execution.
Replay saves the order of the first instructions and then checks that they
are executed backwards in the correct order.
After that the execution is replayed to the end, and reverse continue
command is checked by setting several breakpoints, and asserting
that the execution is stopped at the last of them.

Signed-off-by: Pavel Dovgalyuk <Pavel.Dovgalyuk@ispras.ru>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Willian Rampazzo <willianr@redhat.com>
--

v5:
 - disabled (as some other tests) when running on gitlab
   due to the unidentified timeout problem
Message-Id: <160174524678.12451.13258942849173670277.stgit@pasha-ThinkPad-X280>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

replay: create temporary snapshot at debugger connection

When record/replay does not uses overlays for storing the snapshots,
user is not capable of issuing reverse debugging commands.
This patch adds creation of the VM snapshot on the temporary
overlay image, when the debugger connects to QEMU.
Therefore the execution can be rewind to the moment
of the debugger connection while debugging the virtual machine.

Signed-off-by: Pavel Dovgalyuk <Pavel.Dovgalyuk@ispras.ru>
--

v6:
 - dropped unused error processing (suggested by Philippe Mathieu-Daudé)
Message-Id: <160174524096.12451.11651270339216758643.stgit@pasha-ThinkPad-X280>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

replay: describe reverse debugging in docs/replay.txt

This patch updates the documentation and describes usage of the reverse
debugging in QEMU+GDB.

Signed-off-by: Pavel Dovgalyuk <Pavel.Dovgalyuk@ispras.ru>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
--

v4 changes:
 - added an example of the command line for reverse debugging of
   the diskless machine
Message-Id: <160174523509.12451.1409905901920738979.stgit@pasha-ThinkPad-X280>

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

gdbstub: add reverse continue support in replay mode

This patch adds support of the reverse continue operation for gdbstub.
Reverse continue finds the last breakpoint that would happen in normal
execution from the beginning to the current moment.
Implementation of the reverse continue replays the execution twice:
to find the breakpoints that were hit and to seek to the last breakpoint.
Reverse continue loads the previous snapshot and tries to find the breakpoint
since that moment. If there are no such breakpoints, it proceeds to
the earlier snapshot, and so on. When no breakpoints or watchpoints were
hit at all, execution stops at the beginning of the replay log.

Signed-off-by: Pavel Dovgalyuk <Pavel.Dovgalyuk@ispras.ru>
Message-Id: <160174522930.12451.6994758004725016836.stgit@pasha-ThinkPad-X280>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

gdbstub: add reverse step support in replay mode

GDB remote protocol supports two reverse debugging commands:
reverse step and reverse continue.
This patch adds support of the first one to the gdbstub.
Reverse step is intended to step one instruction in the backwards
direction. This is not possible in regular execution.
But replayed execution is deterministic, therefore we can load one of
the prior snapshots and proceed to the desired step. It is equivalent
to stepping one instruction back.
There should be at least one snapshot preceding the debugged part of
the replay log.

Signed-off-by: Pavel Dovgalyuk <Pavel.Dovgalyuk@ispras.ru>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
--

v4 changes:
 - inverted condition in cpu_handle_guest_debug (suggested by Alex Bennée)
Message-Id: <160174522341.12451.1498758422543765253.stgit@pasha-ThinkPad-X280>

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

replay: flush rr queue before loading the vmstate

Non-empty record/replay queue prevents saving and loading the VM state,
because it includes pending bottom halves and block coroutines.
But when the new VM state is loaded, we don't have to preserve the consistency
of the current state anymore. Therefore this patch just flushes the queue
allowing the coroutines to finish and removes checking for empty rr queue
for load_snapshot function.

Signed-off-by: Pavel Dovgalyuk <Pavel.Dovgalyuk@ispras.ru>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <160174521762.12451.15752448887893855757.stgit@pasha-ThinkPad-X280>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

replay: implement replay-seek command

This patch adds hmp/qmp commands replay_seek/replay-seek that proceed
the execution to the specified instruction count.
The command automatically loads nearest snapshot and replays the execution
to find the desired instruction count.

Signed-off-by: Pavel Dovgalyuk <Pavel.Dovgalyuk@ispras.ru>
Acked-by: Markus Armbruster <armbru@redhat.com>
--

v4 changes:
 - fixed HMP command description indent
 - removed useless error_free call
Message-Id: <160174521180.12451.14033112911009278753.stgit@pasha-ThinkPad-X280>

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

replay: introduce breakpoint at the specified step

This patch introduces replay_break, replay_delete_break
qmp and hmp commands.
These commands allow stopping at the specified instruction.
It may be useful for debugging when there are some known
events that should be investigated.
replay_break command has one argument - number of instructions
executed since the start of the replay.
replay_delete_break removes previously set breakpoint.

Signed-off-by: Pavel Dovgalyuk <Pavel.Dovgalyuk@ispras.ru>
Acked-by: Markus Armbruster <armbru@redhat.com>
--

v4 changes:
 - removed useless error_free call
Message-Id: <160174520606.12451.7056879546045599378.stgit@pasha-ThinkPad-X280>

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

replay: introduce info hmp/qmp command

This patch introduces 'info replay' monitor command and
corresponding qmp request.
These commands request the current record/replay mode, replay log file
name, and the instruction count (number of recorded/replayed
instructions).  The instruction count can be used with the
replay_seek/replay_break commands added in the next two patches.

Signed-off-by: Pavel Dovgalyuk <Pavel.Dovgalyuk@ispras.ru>
Acked-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Acked-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <160174520026.12451.13112161947433306561.stgit@pasha-ThinkPad-X280>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

qapi: introduce replay.json for record/replay-related stuff

This patch adds replay.json file. It will be
used for adding record/replay-related data structures and commands.

Signed-off-by: Pavel Dovgalyuk <pavel.dovgalyuk@ispras.ru>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <160174519444.12451.3472949430004845434.stgit@pasha-ThinkPad-X280>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

migration: introduce icount field for snapshots

Saving icount as a parameters of the snapshot allows navigation between
them in the execution replay scenario.
This information can be used for finding a specific snapshot for proceeding
the recorded execution to the specific moment of the time.
E.g., 'reverse step' action (introduced in one of the following patches)
needs to load the nearest snapshot which is prior to the current moment
of time.
This patch also updates snapshot test which verifies qemu monitor output.

Signed-off-by: Pavel Dovgalyuk <Pavel.Dovgalyuk@ispras.ru>
Acked-by: Markus Armbruster <armbru@redhat.com>
Acked-by: Kevin Wolf <kwolf@redhat.com>
--

v4 changes:
 - squashed format update with test output update
v7 changes:
 - introduced the spaces between the fields in snapshot info output
 - updated the test to match new field widths
Message-Id: <160174518865.12451.14327573383978752463.stgit@pasha-ThinkPad-X280>

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

qcow2: introduce icount field for snapshots

This patch introduces the icount field for saving within the snapshot.
It is required for navigation between the snapshots in record/replay mode.

Signed-off-by: Pavel Dovgalyuk <Pavel.Dovgalyuk@ispras.ru>
Acked-by: Kevin Wolf <kwolf@redhat.com>
--

v7 changes:
 - also fix the test which checks qcow2 snapshot extra data
Message-Id: <160174518284.12451.2301137308458777398.stgit@pasha-ThinkPad-X280>

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

replay: provide an accessor for rr filename

This patch adds an accessor function for the name of the record/replay
log file. Adding an accessor instead of making variable global,
prevents accidental modification of this variable by other modules.

Signed-off-by: Pavel Dovgalyuk <Pavel.Dovgalyuk@ispras.ru>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <160174517710.12451.17645787545733927488.stgit@pasha-ThinkPad-X280>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

replay: don't record interrupt poll

Interrupt poll is not a real interrupt event. It is needed only for
thread safety. This interrupt is used for i386 and converted
to hardware interrupt by cpu_handle_interrupt function.
Therefore it is not needed to be recorded, because hardware
interrupt will be recorded after converting.

Signed-off-by: Pavel Dovgalyuk <Pavel.Dovgalyuk@ispras.ru>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
--

v4 changes:
 - Condition check refactoring (suggested by Alex Bennée)
Message-Id: <160174517124.12451.12983410242461131737.stgit@pasha-ThinkPad-X280>

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

configure: don't enable ASLR for --enable-debug Windows builds

Unlike other OSs it is not possible for gdb to temporarily disable ASLR when
debugging executables on Windows which causes gdb to fail with memory access
errors when trying to debug QEMU.

Keep ASLR enabled by default on Windows via the --dynamicbase compiler flag
except for --enable-debug builds when there is a clear expectation that a
functioning gdb is expected at the cost of slightly less security.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Message-Id: <20201005133434.12614-1-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

configure: consistently pass CFLAGS/CXXFLAGS/LDFLAGS to meson

Environment variables like CFLAGS are easy to accidentally change.  Meson
warns if that happens, but in a project with a lot of configuration that
is easy to lose.  It is also surprising behavior since meson caches -D
options and remembers those on reconfiguration (which we rely on,
since configure options become -D options).

By placing the user-provided CFLAGS, CXXFLAGS and LDFLAGS in the
cross file, we at least get consistent behavior.  These environment
variables are still ugly and not really recommended, but there are
distros that rely on them.  For the gory details, refer to
https://github.com/mesonbuild/meson/issues/4664.

Tested-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20200923092617.1593722-5-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

configure: do not clobber environment CFLAGS/CXXFLAGS/LDFLAGS

If the CFLAGS, CXXFLAGS or LDFLAGS variables are present in the environment,
any modification made within the configure script is passed down to Meson.
This is particularly undesirable for the "-pie" option, since it overrides
"-shared" and thus messes up the linker flags for shared modules.

Using a separate variable therefore fixes the bug, while clarifying that
the scope of these CFLAGS is just the configure script.

We also do not need to pass those variables in config-host.mak; they
were only used for printing the summary now that all submodules are
built with handwritten Meson rules).  For now synthesize CFLAGS in the
configuration summary, the next patch will also pass them in a cleaner
way using the cross file.

Reported-by: Frederic Bezies
Analyzed-by: Toolybird
Tested-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20200923092617.1593722-4-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

dtc: Convert Makefile bits to meson bits

Build the library via the main meson.build just like for capstone.
This improves the current state of affairs in that we will re-link
the qemu executables against a changed libfdt.a, which we wouldn't
do before-hand, and lets us remove the whole recursive make machinery.

Tested-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

slirp: Convert Makefile bits to meson bits

SLIRP uses Meson so it could become a subproject in the future,
but our choice of configure options is not yet supported in Meson
(https://github.com/mesonbuild/meson/pull/7740).

For now, build the library via the main meson.build just like for
capstone.

This improves the current state of affairs in that we will re-link
the qemu executables against a changed libslirp.a, which we wouldn't
do before-hand.

Tested-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

accel/tcg: use current_machine as it is always set for softmmu

current_machine is always set before accelerators are initialized,
so use that instead of MACHINE(qdev_get_machine()).

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

kvm: kvm_init_vcpu take Error pointer

Clean up the error handling in kvm_init_vcpu so we can see what went
wrong more easily.

Make it take an Error ** and fill it out with what failed, including
the cpu id, so you can tell if it only fails at a given ID.

Replace the remaining DPRINTF by a trace.

This turns a:
kvm_init_vcpu failed: Invalid argument

into:
kvm_init_vcpu: kvm_get_vcpu failed (256): Invalid argument

and with the trace you then get to see:

19049@1595520414.310107:kvm_init_vcpu index: 169 id: 212
19050@1595520414.310635:kvm_init_vcpu index: 170 id: 256
qemu-system-x86_64: kvm_init_vcpu: kvm_get_vcpu failed (256): Invalid argument

which makes stuff a lot more obvious.

Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20200723160915.129069-1-dgilbert@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

kvm: remove kvm specific functions from global includes

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hax: remove hax specific functions from global includes

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

whpx: remove whpx specific functions from global includes

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hvf: remove hvf specific functions from global includes

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Roman Bolshakov <r.bolshakov@yadro.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

cpus: add handle_interrupt to the CpusAccel interface

kvm: uses the generic handler
qtest: uses the generic handler
whpx: changed to use the generic handler (identical implementation)
hax: changed to use the generic handler (identical implementation)
hvf: changed to use the generic handler (identical implementation)
tcg: adapt tcg-cpus to point to the tcg-specific handler

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

cpus: remove checks for non-NULL cpus_accel

now that all accelerators support the CpusAccel interface,
we can remove most checks for non-NULL cpus_accel,
we just add a sanity check/assert at vcpu creation.

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

cpus: cleanup now unneeded includes

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

cpus: extract out hvf-specific code to target/i386/hvf/

register a "CpusAccel" interface for HVF as well.

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Roman Bolshakov <r.bolshakov@yadro.com>
[added const]
Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

cpus: extract out whpx-specific code to target/i386/

register a "CpusAccel" interface for WHPX as well.

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

cpus: extract out hax-specific code to target/i386/

register a "CpusAccel" interface for HAX as well.

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

cpus: extract out kvm-specific code to accel/kvm

register a "CpusAccel" interface for KVM as well.

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
[added const]
Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

cpus: extract out qtest-specific code to accel/qtest

register a "CpusAccel" interface for qtest as well.

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

cpus: extract out TCG-specific code to accel/tcg

TCG is the first accelerator to register a "CpusAccel" interface
on initialization, providing functions for starting a vcpu,
kicking a vcpu, sychronizing state and getting virtual clock
and ticks.

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
[added const]
Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

cpus: prepare new CpusAccel cpu accelerator interface

The new interface starts unused, will start being used by the
next patches.

It provides methods for each accelerator to start a vcpu, kick a vcpu,
synchronize state, get cpu virtual clock and elapsed ticks.

In qemu_wait_io_event, make it clear that APC is used only for HAX
on Windows.

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

icount: rename functions to be consistent with the module name

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

cpu-timers, icount: new modules

refactoring of cpus.c continues with cpu timer state extraction.

cpu-timers: responsible for the softmmu cpu timers state,
            including cpu clocks and ticks.

icount: counts the TCG instructions executed. As such it is specific to
the TCG accelerator. Therefore, it is built only under CONFIG_TCG.

One complication is due to qtest, which uses an icount field to warp time
as part of qtest (qtest_clock_warp).

In order to solve this problem, provide a separate counter for qtest.

This requires fixing assumptions scattered in the code that
qtest_enabled() implies icount_enabled(), checking each specific case.

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
[remove redundant initialization with qemu_spice_init]
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
[fix lingering calls to icount_get]
Signed-off-by: Claudio Fontana <cfontana@suse.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

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

* move target configuration to default-configs/targets (myself)
* Memory failure event (Zhenwei)

# gpg: Signature made Mon 05 Oct 2020 08:14:29 BST
# 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:
  dockerfiles: add diffutils to Fedora
  tests: tcg: do not use implicit rules
  target-i386: post memory failure event to QMP
  qapi/run-state.json: introduce memory failure event
  target-i386: seperate MCIP & MCE_MASK error reason
  meson: move sparse detection to Meson and rewrite check_sparse.py
  default-configs: remove redundant keys
  default-configs: use TARGET_ARCH key
  configure: move OpenBSD W^X test to meson
  default-configs: remove default-configs/devices for user-mode targets
  configure: remove target configuration
  configure: remove useless config-target.mak symbols
  configure: compute derivatives of target name in meson
  configure: remove dead variable
  configure: move accelerator logic to meson
  configure: rewrite accelerator defaults as tests
  configure: convert accelerator variables to meson options
  default-configs: move files to default-configs/devices/
  travis: remove TCI test

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

Merge remote-tracking branch 'remotes/rth-gitlab/tags/pull-cap-20201003' into staging

Update capstone submodule from v3.0.5 to v5 ("next").
Convert submodule build to meson.
Enable capstone disassembly for s390x.
Code cleanups in disas.c

# gpg: Signature made Sat 03 Oct 2020 10:33:44 BST
# gpg:                using RSA key 7A481E78868B4DB6A85A05C064DF38E8AF7E215F
# gpg:                issuer "richard.henderson@linaro.org"
# gpg: Good signature from "Richard Henderson <richard.henderson@linaro.org>" [full]
# Primary key fingerprint: 7A48 1E78 868B 4DB6 A85A  05C0 64DF 38E8 AF7E 215F

* remotes/rth-gitlab/tags/pull-cap-20201003:
  disas/capstone: Add skipdata hook for s390x
  disas: Enable capstone disassembly for s390x
  disas: Split out capstone code to disas/capstone.c
  disas: Configure capstone for aarch64 host without libvixl
  disas: Cleanup plugin_disas
  disas: Use qemu/bswap.h for bfd endian loads
  disas: Clean up CPUDebug initialization
  disas: Move host asm annotations to tb_gen_code
  capstone: Require version 4.0 from a system library
  capstone: Update to upstream "next" branch
  capstone: Convert Makefile bits to meson bits

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

dockerfiles: add diffutils to Fedora

For some reason diffutils is not included in the Fedora containers anymore,
causing the build to fail.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Neal Gompa <ngompa13@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20201003085054.332992-2-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

tests: tcg: do not use implicit rules

Use pattern rules to clarify which targets are going to match the
rule and to provide clearer error messages.

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

target-i386: post memory failure event to QMP

Post memory failure event through QMP to handle hardware memory corrupted
event. Rather than simply printing to the log, QEMU could report more
effective message to the client. For example, if a guest receives an MCE,
evacuating the host could be a good idea.

Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
Message-Id: <20200930100440.1060708-4-pizhenwei@bytedance.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

qapi/run-state.json: introduce memory failure event

Introduce memory failure events for hypervisor and guest. This lets
mft: Need exactly one file argument.  Try `mft --help' for more
information.

Suggested by Peter Maydell, rename events name&description to make
them architecture-neutral; and suggested by Paolo, add more info to
distinguish a mce is AR/AO, and if a previous MCE was still being
processed in the guest.

Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
Message-Id: <20200930100440.1060708-3-pizhenwei@bytedance.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target-i386: seperate MCIP & MCE_MASK error reason

Previously we would only get a simple string "Triple fault" in qemu
log. Add detailed message for the two reasons to describe why qemu
has to reset the guest.

Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
Message-Id: <20200930100440.1060708-2-pizhenwei@bytedance.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

meson: move sparse detection to Meson and rewrite check_sparse.py

Pass the path to the program to scripts/check_sparse.py, which
previously was not included in config-host.mak.  Change
scripts/check_sparse.py to work with cgcc, which seems to
work better with sparse 0.6.x.

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

default-configs: remove redundant keys

TARGET_BASE_ARCH and TARGET_ABI_DIR are 99% of the time the same
as TARGET_ARCH, remove them if so.

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

default-configs: use TARGET_ARCH key

Replace the individual TARGET_*=y lines with TARGET_ARCH,
similar to how TARGET_BASE_ARCH is handled already.

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

configure: move OpenBSD W^X test to meson

The TCG option is now passed through to Meson.

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

default-configs: remove default-configs/devices for user-mode targets

We no longer need dummy files to detect targets, since
default-configs/targets/ exists.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

configure: remove target configuration

The config-target.mak files are small constant, we can therefore just
write them down explicitly.

This removes a pretty large part of the configure script, including the
whole logic to detect which accelerators are supported by each target.

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

configure: remove useless config-target.mak symbols

Omit symbols that are not needed by softmmu or bsd-user targets,
in preparation for moving the generated config-target.mak files
into the source tree.

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

configure: compute derivatives of target name in meson

Several CONFIG_* symbols in config-target.mak are easily computed from just
the target name.  We do not need them in config-target.mak, and can instead
place them in the config_target dictionary only.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

configure: remove dead variable

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

configure: move accelerator logic to meson

Move to meson the code to detect the presence of accelerators, and
to define accelerator-specific config-target.h symbols.

The logic for now is duplicated in configure because it is still
in use to build the list of targets (which is in turn used to
create the config-target.mak files).  The next patches remove it.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

configure: rewrite accelerator defaults as tests

Prepare to process "auto" in meson rather than configure: standardize the
shape of the code that changes "auto" to enabled/disabled, to ease the review
when it will be moved to meson.

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

configure: convert accelerator variables to meson options

Prepare for moving the tests to meson.  For now they only have
enabled/disabled as the possible values when meson is invoked,
but "auto" will be a possibility later, when configure will only
parse the command line options.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

default-configs: move files to default-configs/devices/

Make room for target files in default-configs/targets/

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

travis: remove TCI test

TCI is already covered on gitlab CI, so we can remove it.

Cc: Thomas Huth <thuth@redhat.com>
Cc: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

disas/capstone: Add skipdata hook for s390x

It is always possible to tell the length of an insn, even if the
actual insn is unknown.  Skip the correct number of bytes, so that
we stay in sync with the instruction stream.

Acked-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

disas: Enable capstone disassembly for s390x

Enable s390x, aka SYSZ, in the git submodule build.
Set the capstone parameters for both s390x host and guest.

Tested-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Acked-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

disas: Split out capstone code to disas/capstone.c

There is nothing target-specific about this code, so it
can be added to common_ss.  This also requires that the
base capstone dependency be added to common_ss, so that
we get the correct include paths added to CFLAGS.

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

disas: Configure capstone for aarch64 host without libvixl

The ifdef tangle failed to set cap_arch if libvixl itself
was not configured (e.g. due to lack of c++ compiler).

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

disas: Cleanup plugin_disas

Do not retain a GString in thread-local storage.  Allocate a
new one and free it on every invocation.  Do not g_strdup the
result; return the buffer from the GString.  Do not use
warn_report.

Using cs_disasm allocated memory via the &insn parameter, but
that was never freed.  Use cs_disasm_iter so that we use the
memory that we've already allocated, and so that we only try
to disassemble one insn, as desired.  Do not allocate 1k to
hold the bytes for a single instruction.

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

disas: Use qemu/bswap.h for bfd endian loads

Use the routines we have already instead of open-coding.

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

disas: Clean up CPUDebug initialization

Rename several functions, dropping "generic" and making "host"
vs "target" clearer.  Make a bunch of functions static that are
not used outside this file. Replace INIT_DISASSEMBLE_INFO with
a trio of functions.

Acked-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

disas: Move host asm annotations to tb_gen_code

Instead of creating GStrings and passing them into log_disas,
just print the annotations directly in tb_gen_code.

Fix the annotations for the slow paths of the TB, after the
part implementing the final guest instruction.

Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

capstone: Require version 4.0 from a system library

We're about to use a portion of the 4.0 API.
Reject a system library version prior to that.

Tested-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

capstone: Update to upstream "next" branch

This branch contains a number of improvements over master,
including making all of the disassembler data constant.

We are skipping past the 4.0 branchpoint, which changed
the location of the includes within the source directory.

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

capstone: Convert Makefile bits to meson bits

There are better ways to do this, e.g. meson cmake subproject,
but that requires cmake 3.7 and some of our CI environments
only provide cmake 3.5.

Nor can we add a meson.build file to capstone/, because the git
submodule would then always report "untracked files".  Fixing that
would require creating our own branch on the qemu git mirror, at
which point we could just as easily create a native meson subproject.

Instead, build the library via the main meson.build.

This improves the current state of affairs in that we will re-link
the qemu executables against a changed libcapstone.a, which we wouldn't
do before-hand.  In addition, the use of the configuration header file
instead of command-line -DEFINES means that we will rebuild the
capstone objects with changes to meson.build.

Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

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

Block layer patches:

- Add block export infrastructure
- iotests improvements
- Document the throttle block filter
- Misc code cleanups

# gpg: Signature made Fri 02 Oct 2020 15:36:53 BST
# gpg:                using RSA key DC3DEB159A9AF95D3D7456FE7F09B272C88F2FD6
# gpg:                issuer "kwolf@redhat.com"
# gpg: Good signature from "Kevin Wolf <kwolf@redhat.com>" [full]
# Primary key fingerprint: DC3D EB15 9A9A F95D 3D74  56FE 7F09 B272 C88F 2FD6

* remotes/kevin/tags/for-upstream: (37 commits)
  qcow2: Use L1E_SIZE in qcow2_write_l1_entry()
  qemu-storage-daemon: Fix help line for --export
  iotests: Test block-export-* QMP interface
  iotests: Allow supported and unsupported formats at the same time
  iotests: Introduce qemu_nbd_list_log()
  iotests: Factor out qemu_tool_pipe_and_status()
  nbd: Deprecate nbd-server-add/remove
  nbd: Merge nbd_export_new() and nbd_export_create()
  block/export: Move writable to BlockExportOptions
  block/export: Add query-block-exports
  block/export: Create BlockBackend in blk_exp_add()
  block/export: Move blk to BlockExport
  block/export: Add BLOCK_EXPORT_DELETED event
  block/export: Add block-export-del
  block/export: Move strong user reference to block_exports
  block/export: Add 'id' option to block-export-add
  block/export: Add blk_exp_close_all(_type)
  block/export: Allocate BlockExport in blk_exp_add()
  block/export: Add node-name to BlockExportOptions
  block/export: Move AioContext from NBDExport to BlockExport
  ...

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

qcow2: Use L1E_SIZE in qcow2_write_l1_entry()

We overlooked these in 02b1ecfa100e7ecc2306560cd27a4a2622bfeb04

Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-Id: <20200928162333.14998-1-berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

qemu-storage-daemon: Fix help line for --export

Commit 5f479a8d renamed the 'device' option of --export into
'node-name', but forgot to update the help in qemu-storage-daemon.

Fixes: 5f479a8dc086bfa42c9f94e9ab69962f256e207f
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20200930133909.58820-1-kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

iotests: Test block-export-* QMP interface

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20200924152717.287415-32-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

iotests: Allow supported and unsupported formats at the same time

This is useful for specifying 'generic' as supported (which includes
only writable image formats), but still excluding some incompatible
writable formats.

It also removes more lines than it adds.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20200924152717.287415-31-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

iotests: Introduce qemu_nbd_list_log()

Add a function to list the NBD exports offered by an NBD server.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20200924152717.287415-30-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

iotests: Factor out qemu_tool_pipe_and_status()

We have three almost identical functions that call an external process
and return its output and return code. Refactor them into small wrappers
around a common function.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20200924152717.287415-29-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

nbd: Deprecate nbd-server-add/remove

These QMP commands are replaced by block-export-add/del.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20200924152717.287415-28-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

nbd: Merge nbd_export_new() and nbd_export_create()

There is no real reason any more why nbd_export_new() and
nbd_export_create() should be separate functions. The latter only
performs a few checks before it calls the former.

What makes the current state stand out is that it's the only function in
BlockExportDriver that is not a static function inside nbd/server.c, but
a small wrapper in blockdev-nbd.c that then calls back into nbd/server.c
for the real functionality.

Move all the checks to nbd/server.c and make the resulting function
static to improve readability.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200924152717.287415-27-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block/export: Move writable to BlockExportOptions

The 'writable' option is a basic option that will probably be applicable
to most if not all export types that we will implement. Move it from NBD
to the generic BlockExport layer.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200924152717.287415-26-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block/export: Add query-block-exports

This adds a simple QMP command to query the list of block exports.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200924152717.287415-25-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block/export: Create BlockBackend in blk_exp_add()

Every export type will need a BlockBackend, so creating it centrally in
blk_exp_add() instead of the .create driver callback avoids duplication.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200924152717.287415-24-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block/export: Move blk to BlockExport

Every block export has a BlockBackend representing the disk that is
exported. It should live in BlockExport therefore.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200924152717.287415-23-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block/export: Add BLOCK_EXPORT_DELETED event

Clients may want to know when an export has finally disappeard
(block-export-del returns earlier than that in the general case), so add
a QAPI event for it.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20200924152717.287415-22-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block/export: Add block-export-del

Implement a new QMP command block-export-del and make nbd-server-remove
a wrapper around it.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200924152717.287415-21-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block/export: Move strong user reference to block_exports

The reference owned by the user/monitor that is created when adding the
export and dropped when removing it was tied to the 'exports' list in
nbd/server.c. Every block export will have a user reference, so move it
to the block export level and tie it to the 'block_exports' list in
block/export/export.c instead. This is necessary for introducing a QMP
command for removing exports.

Note that exports are present in block_exports even after the user has
requested shutdown. This is different from NBD's exports where exports
are immediately removed on a shutdown request, even if they are still in
the process of shutting down. In order to avoid that the user still
interacts with an export that is shutting down (and possibly removes it
a second time), we need to remember if the user actually still owns it.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200924152717.287415-20-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block/export: Add 'id' option to block-export-add

We'll need an id to identify block exports in monitor commands. This
adds one.

Note that this is different from the 'name' option in the NBD server,
which is the externally visible export name. While block export ids need
to be unique in the whole process, export names must be unique only for
the same server. Different export types or (potentially in the future)
multiple NBD servers can have the same export name externally, but still
need different block export ids internally.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200924152717.287415-19-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block/export: Add blk_exp_close_all(_type)

This adds a function to shut down all block exports, and another one to
shut down the block exports of a single type. The latter is used for now
when stopping the NBD server. As soon as we implement support for
multiple NBD servers, we'll need a per-server list of exports and it
will be replaced by a function using that.

As a side effect, the BlockExport layer has a list tracking all existing
exports now. closed_exports loses its only user and can go away.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200924152717.287415-18-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block/export: Allocate BlockExport in blk_exp_add()

Instead of letting the driver allocate and return the BlockExport
object, allocate it already in blk_exp_add() and pass it. This allows us
to initialise the generic part before calling into the driver so that
the driver can just use these values instead of having to parse the
options a second time.

For symmetry, move freeing the BlockExport to blk_exp_unref().

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200924152717.287415-17-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block/export: Add node-name to BlockExportOptions

Every block export needs a block node to export, so add a 'node-name'
option to BlockExportOptions and remove the replaced option 'device'
from BlockExportOptionsNbd.

To maintain compatibility in nbd-server-add, BlockExportOptionsNbd needs
to be wrapped by a new type NbdServerAddOptions that adds 'device' back
because nbd-server-add doesn't use the BlockExportOptions base type at
all (so even without changing it to a 'node-name' option in
block-export-add, this compatibility code would be necessary).

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200924152717.287415-16-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block/export: Move AioContext from NBDExport to BlockExport

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200924152717.287415-15-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block/export: Move refcount from NBDExport to BlockExport

Having a refcount makes sense for all types of block exports. It is also
a prerequisite for keeping a list of all exports at the BlockExport
level.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200924152717.287415-14-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

nbd/server: Simplify export shutdown

Closing export is somewhat convoluted because nbd_export_close() and
nbd_export_put() call each other and the ways they actually end up being
nested is not necessarily obvious.

However, it is not really necessary to call nbd_export_close() from
nbd_export_put() when putting the last reference because it only does
three things:

1. Close all clients. We're going to refcount 0 and all clients hold a
   reference, so we know there is no active client any more.

2. Close the user reference (represented by exp->name being non-NULL).
   The same argument applies: If the export were still named, we would
   still have a reference.

3. Freeing exp->description. This is really cleanup work to be done when
   the export is finally freed. There is no reason to already clear it
   while clients are still in the process of shutting down.

So after moving the cleanup of exp->description, the code can be
simplified so that only nbd_export_close() calls nbd_export_put(), but
never the other way around.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20200924152717.287415-13-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

qemu-nbd: Use blk_exp_add() to create the export

With this change, NBD exports are now only created through the
BlockExport interface. This allows us finally to move things from the
NBD layer to the BlockExport layer if they make sense for other export
types, too.

blk_exp_add() returns only a weak reference, so the explicit
nbd_export_put() goes away.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200924152717.287415-12-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

nbd: Remove NBDExport.close callback

The export close callback is unused by the built-in NBD server. qemu-nbd
uses it only during shutdown to wait for the unrefed export to actually
go away. It can just use nbd_export_close_all() instead and do without
the callback.

This removes the close callback from nbd_export_new() and makes both
callers of it more similar.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20200924152717.287415-11-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

nbd: Add writethrough to block-export-add

qemu-nbd allows use of writethrough cache modes, which mean that write
requests made through NBD will cause a flush before they complete.
Expose the same functionality in block-export-add.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200924152717.287415-10-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

nbd: Add max-connections to nbd-server-start

This is a QMP equivalent of qemu-nbd's --shared option, limiting the
maximum number of clients that can attach at the same time.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20200924152717.287415-9-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block/export: Remove magic from block-export-add

nbd-server-add tries to be convenient and adds two questionable
features that we don't want to share in block-export-add, even for NBD
exports:

1. When requesting a writable export of a read-only device, the export
   is silently downgraded to read-only. This should be an error in the
   context of block-export-add.

2. When using a BlockBackend name, unplugging the device from the guest
   will automatically stop the NBD server, too. This may sometimes be
   what you want, but it could also be very surprising. Let's keep
   things explicit with block-export-add. If the user wants to stop the
   export, they should tell us so.

Move these things into the nbd-server-add QMP command handler so that
they apply only there.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200924152717.287415-8-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

qemu-nbd: Use raw block driver for --offset

Instead of implementing qemu-nbd --offset in the NBD code, just put a
raw block node with the requested offset on top of the user image and
rely on that doing the job.

This does not only simplify the nbd_export_new() interface and bring it
closer to the set of options that the nbd-server-add QMP command offers,
but in fact it also eliminates a potential source for bugs in the NBD
code which previously had to add the offset manually in all relevant
places.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20200924152717.287415-7-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

qemu-storage-daemon: Use qmp_block_export_add()

No reason to duplicate the functionality locally, we can now just reuse
the QMP command block-export-add for --export.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20200924152717.287415-6-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block/export: Add BlockExport infrastructure and block-export-add

We want to have a common set of commands for all types of block exports.
Currently, this is only NBD, but we're going to add more types.

This patch adds the basic BlockExport and BlockExportDriver structs and
a QMP command block-export-add that creates a new export based on the
given BlockExportOptions.

qmp_nbd_server_add() becomes a wrapper around qmp_block_export_add().

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200924152717.287415-5-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>