chardev: fix race with client connections in tcp_chr_wait_connected

When the 'reconnect' option is given for a client connection, the
qmp_chardev_open_socket_client method will run an asynchronous
connection attempt. The QIOChannel socket executes this is a single use
background thread, so the connection will succeed immediately (assuming
the server is listening). The chardev, however, won't get the result
from this background thread until the main loop starts running and
processes idle callbacks.

Thus when tcp_chr_wait_connected is run s->ioc will be NULL, but the
state will be TCP_CHARDEV_STATE_CONNECTING, and there may already
be an established connection that will be associated with the chardev
by the pending idle callback. tcp_chr_wait_connected doesn't check the
state, only s->ioc, so attempts to establish another connection
synchronously.

If the server allows multiple connections this is unhelpful but not a
fatal problem as the duplicate connection will get ignored by the
tcp_chr_new_client method when it sees the state is already connected.

If the server only supports a single connection, however, the
tcp_chr_wait_connected method will hang forever because the server will
not accept its synchronous connection attempt until the first connection
is closed.

To deal with this tcp_chr_wait_connected needs to synchronize with the
completion of the background connection task. To do this it needs to
create the QIOTask directly and use the qio_task_wait_thread method.
This will cancel the pending idle callback and directly dispatch the
task completion callback, allowing the connection to be associated
with the chardev. If the background connection failed, it can still
attempt a new synchronous connection.

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20190211182442.8542-15-berrange@redhat.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

chardev: disallow TLS/telnet/websocket with tcp_chr_wait_connected

In the previous commit

    commit 1dc8a6695c731abb7461c637b2512c3670d82be4
    Author: Marc-André Lureau <marcandre.lureau@redhat.com>
    Date:   Tue Aug 16 12:33:32 2016 +0400

      char: fix waiting for TLS and telnet connection

the tcp_chr_wait_connected() method was changed to check for a non-NULL
's->ioc' as a sign that there is already a connection present, as
opposed to checking the "connected" flag to supposedly fix handling of
TLS/telnet connections.

The original code would repeatedly call tcp_chr_wait_connected creating
many connections as 'connected' would never become true. The changed
code would still repeatedly call tcp_chr_wait_connected busy waiting
because s->ioc is set but the chardev will never see CHR_EVENT_OPENED.
IOW, the code is still broken with TLS/telnet, but in a different way.

Checking for a non-NULL 's->ioc' does not mean that a CHR_EVENT_OPENED
will be ready for a TLS/telnet connection. These protocols (and the
websocket protocol) all require the main loop to be running in order
to complete the protocol handshake before emitting CHR_EVENT_OPENED.
The tcp_chr_wait_connected() method is only used during early startup
before a main loop is running, so TLS/telnet/websock connections can
never complete initialization.

Making this work would require changing tcp_chr_wait_connected to run
a main loop. This is quite complex since we must not allow GSource's
that other parts of QEMU have registered to run yet. The current callers
of tcp_chr_wait_connected do not require use of the TLS/telnet/websocket
protocols, so the simplest option is to just forbid this combination
completely for now.

Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20190211182442.8542-14-berrange@redhat.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

chardev: honour the reconnect setting in tcp_chr_wait_connected

If establishing a client connection fails, the tcp_chr_wait_connected
method should sleep for the reconnect timeout and then retry the
attempt. This ensures the callers don't immediately abort with an
error when the initial connection fails.

Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20190211182442.8542-13-berrange@redhat.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

chardev: use a state machine for socket connection state

The socket connection state is indicated via the 'bool connected' field
in the SocketChardev struct. This variable is somewhat misleading
though, as it is only set to true once the connection has completed all
required handshakes (eg for TLS, telnet or websockets). IOW there is a
period of time in which the socket is connected, but the "connected"
flag is still false.

The socket chardev really has three states that it can be in,
disconnected, connecting and connected and those should be tracked
explicitly.

Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20190211182442.8542-12-berrange@redhat.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

chardev: split up qmp_chardev_open_socket connection code

In qmp_chardev_open_socket the code for connecting client chardevs is
split across two conditionals far apart with some server chardev code in
the middle. Split up the method so that code for client connection setup
is separate from code for server connection setup.

Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20190211182442.8542-11-berrange@redhat.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

chardev: split tcp_chr_wait_connected into two methods

The tcp_chr_wait_connected method can deal with either server or client
chardevs, but some callers only care about one of these possibilities.
The tcp_chr_wait_connected method will also need some refactoring to
reliably deal with its primary goal of allowing a device frontend to
wait for an established connection, which will interfere with other
callers.

Split it into two methods, one responsible for server initiated
connections, the other responsible for client initiated connections.
In doing this split the tcp_char_connect_async() method is renamed
to become consistent with naming of the new methods.

Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20190211182442.8542-10-berrange@redhat.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

chardev: remove unused 'sioc' variable & cleanup paths

The 'sioc' variable in qmp_chardev_open_socket was unused since

  commit 3e7d4d20d3a528b1ed10b1dc3d83119bfb0c5f24
  Author: Peter Xu <peterx@redhat.com>
  Date:   Tue Mar 6 13:33:17 2018 +0800

    chardev: use chardev's gcontext for async connect

Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20190211182442.8542-9-berrange@redhat.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

chardev: ensure qemu_chr_parse_compat reports missing driver error

If no valid char driver was identified the qemu_chr_parse_compat method
was silent, leaving callers no clue what failed.

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20190211182442.8542-8-berrange@redhat.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

chardev: remove many local variables in qemu_chr_parse_socket

Now that all validation is separated off into a separate method,
we can directly populate the ChardevSocket struct from the
QemuOpts values, avoiding many local variables.

Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20190211182442.8542-7-berrange@redhat.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

chardev: forbid 'wait' option with client sockets

The 'wait'/'nowait' parameter is used to tell server sockets whether to
block until a client is accepted during initialization. Client chardevs
have always silently ignored this option. Various tests were mistakenly
passing this option for their client chardevs.

Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20190211182442.8542-6-berrange@redhat.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

chardev: forbid 'reconnect' option with server sockets

The 'reconnect' option is used to give the sleep time, in seconds,
before a client socket attempts to re-establish a connection to the
server. It does not make sense to set this for server sockets, as they
will always accept a new client connection immediately after the
previous one went away.

Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20190211182442.8542-5-berrange@redhat.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

chardev: fix validation of options for QMP created chardevs

The TLS creds option is not valid with certain address types. The user
config was only checked for errors when parsing legacy QemuOpts, thus
the user could pass unsupported values via QMP.

Pull all code for validating options out into a new method
qmp_chardev_validate_socket, that is called from the main
qmp_chardev_open_socket method. This adds a missing check for rejecting
TLS creds with the vsock address type.

Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20190211182442.8542-4-berrange@redhat.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

io: add qio_task_wait_thread to join with a background thread

Add the ability for a caller to wait for completion of the
background thread to synchronously dispatch its result, without
needing to wait for the main loop to run the idle callback.

This method needs very careful usage to avoid a dangerous
race condition with the free'ing of the task. The completion
callback is normally invoked from an idle callback registered
with the main loop context. The qio_task_wait_thread method
must only be called if the completion callback has not yet
run. The only safe way to achieve this is to run the
qio_task_wait_thread method from the thread that executes
the main loop.

It is generally a bad idea to use this method since it will
block execution of the main loop, however, the design of
the character devices and its usage from vhostuser already
requires blocking execution.

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20190211182442.8542-3-berrange@redhat.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

io: store reference to thread information in the QIOTask struct

Currently the struct QIOTaskThreadData is only needed by the worker
thread, but a subsequent patch will need to access it from another
context.

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20190211182442.8542-2-berrange@redhat.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

tests/test-char: add muxed chardev testing for open/close

Validate that frontend callbacks for CHR_EVENT_OPENED/CHR_EVENT_CLOSED
events are being issued when expected and in strictly pairing order.

Signed-off-by: Artem Pisarenko <artem.k.pisarenko@gmail.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <ac67ff2d27dd51a0075d5d634355c9e4f7bb53de.1541507990.git.artem.k.pisarenko@gmail.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

chardev: fix mess in OPENED/CLOSED events when muxed

When chardev is multiplexed (mux=on) there are a lot of cases where
CHR_EVENT_OPENED/CHR_EVENT_CLOSED events pairing (expected from
frontend side) is broken. There are either generation of multiple
repeated or extra CHR_EVENT_OPENED events, or CHR_EVENT_CLOSED just
isn't generated at all.
This is mostly because 'qemu_chr_fe_set_handlers()' function makes its
own (and often wrong) implicit decision on updated frontend state and
invokes 'fd_event' callback with 'CHR_EVENT_OPENED'. And even worse,
it doesn't do symmetric action in opposite direction, as someone may
expect (i.e. it doesn't invoke previously set 'fd_event' with
'CHR_EVENT_CLOSED'). Muxed chardev uses trick by calling this function
again to replace callback handlers with its own ones, but it doesn't
account for such side effect.
Fix that using extended version of this function with added argument
for disabling side effect and keep original function for compatibility
with lots of frontends already using this interface and being
"tolerant" to its side effects.
One more source of event duplication is just line of code in
char-mux.c, which does far more than comment above says (obvious fix).

Signed-off-by: Artem Pisarenko <artem.k.pisarenko@gmail.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <7dde6abbd21682857f8294644013173c0b9949b3.1541507990.git.artem.k.pisarenko@gmail.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

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

Pull request

# gpg: Signature made Tue 12 Feb 2019 03:58:58 GMT
# gpg:                using RSA key 9CA4ABB381AB73C8
# gpg: Good signature from "Stefan Hajnoczi <stefanha@redhat.com>" [full]
# gpg:                 aka "Stefan Hajnoczi <stefanha@gmail.com>" [full]
# Primary key fingerprint: 8695 A8BF D3F9 7CDA AC35  775A 9CA4 ABB3 81AB 73C8

* remotes/stefanha/tags/block-pull-request:
  virtio-blk: cleanup using VirtIOBlock *s and VirtIODevice *vdev
  qemugdb/coroutine: fix arch_prctl has unknown return type
  iothread: fix iothread hang when stop too soon

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

Merge remote-tracking branch 'remotes/ericb/tags/pull-nbd-2019-02-11' into staging

nbd patches for 2019-02-11

- Add qcow2 bitmap details to 'qemu-img info'

# gpg: Signature made Mon 11 Feb 2019 20:38:32 GMT
# gpg:                using RSA key A7A16B4A2527436A
# gpg: Good signature from "Eric Blake <eblake@redhat.com>" [full]
# gpg:                 aka "Eric Blake (Free Software Programmer) <ebb9@byu.net>" [full]
# gpg:                 aka "[jpeg image of size 6874]" [full]
# Primary key fingerprint: 71C2 CC22 B1C4 6029 27D2  F3AA A7A1 6B4A 2527 436A

* remotes/ericb/tags/pull-nbd-2019-02-11:
  qcow2: list of bitmaps new test 242
  qcow2: Add list of bitmaps to ImageInfoSpecificQCow2
  bdrv_query_image_info Error parameter added
  nbd/server: Kill pointless shadowed variable

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

virtio-blk: cleanup using VirtIOBlock *s and VirtIODevice *vdev

In several part we still using req->dev or VIRTIO_DEVICE(req->dev)
when we have already defined s and vdev pointers:
    VirtIOBlock *s = req->dev;
    VirtIODevice *vdev = VIRTIO_DEVICE(s);

Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Liam Merwick <liam.merwick@oracle.com>
Message-id: 20190208142347.214815-1-sgarzare@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

qemugdb/coroutine: fix arch_prctl has unknown return type

qemu coroutine command results in following error output:

Python Exception <class 'gdb.error'> 'arch_prctl' has unknown return
type; cast the call to its declared return type: Error occurred in
Python command: 'arch_prctl' has unknown return type; cast the call to
its declared return type

Fix it by giving it what it wants: arch_prctl return type.

Information on the topic:
   https://sourceware.org/gdb/onlinedocs/gdb/Calling.html

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-id: 20190206151425.105871-1-vsementsov@virtuozzo.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

iothread: fix iothread hang when stop too soon

Lukas reported an hard to reproduce QMP iothread hang on s390 that
QEMU might hang at pthread_join() of the QMP monitor iothread before
quitting:

  Thread 1
  #0  0x000003ffad10932c in pthread_join
  #1  0x0000000109e95750 in qemu_thread_join
      at /home/thuth/devel/qemu/util/qemu-thread-posix.c:570
  #2  0x0000000109c95a1c in iothread_stop
  #3  0x0000000109bb0874 in monitor_cleanup
  #4  0x0000000109b55042 in main

While the iothread is still in the main loop:

  Thread 4
  #0  0x000003ffad0010e4 in ??
  #1  0x000003ffad553958 in g_main_context_iterate.isra.19
  #2  0x000003ffad553d90 in g_main_loop_run
  #3  0x0000000109c9585a in iothread_run
      at /home/thuth/devel/qemu/iothread.c:74
  #4  0x0000000109e94752 in qemu_thread_start
      at /home/thuth/devel/qemu/util/qemu-thread-posix.c:502
  #5  0x000003ffad10825a in start_thread
  #6  0x000003ffad00dcf2 in thread_start

IMHO it's because there's a race between the main thread and iothread
when stopping the thread in following sequence:

    main thread                       iothread
    ===========                       ==============
                                      aio_poll()
    iothread_get_g_main_context
      set iothread->worker_context
    iothread_stop
      schedule iothread_stop_bh
                                        execute iothread_stop_bh [1]
                                          set iothread->running=false
                                          (since main_loop==NULL so
                                           skip to quit main loop.
                                           Note: although main_loop is
                                           NULL but worker_context is
                                           not!)
                                      atomic_read(&iothread->worker_context) [2]
                                        create main_loop object
                                        g_main_loop_run() [3]
    pthread_join() [4]

We can see that when execute iothread_stop_bh() at [1] it's possible
that main_loop is still NULL because it's only created until the first
check of the worker_context later at [2].  Then the iothread will hang
in the main loop [3] and it'll starve the main thread too [4].

Here the simple solution should be that we check again the "running"
variable before check against worker_context.

CC: Thomas Huth <thuth@redhat.com>
CC: Dr. David Alan Gilbert <dgilbert@redhat.com>
CC: Stefan Hajnoczi <stefanha@redhat.com>
CC: Lukáš Doktor <ldoktor@redhat.com>
CC: Markus Armbruster <armbru@redhat.com>
CC: Eric Blake <eblake@redhat.com>
CC: Paolo Bonzini <pbonzini@redhat.com>
Reported-by: Lukáš Doktor <ldoktor@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Tested-by: Thomas Huth <thuth@redhat.com>
Message-id: 20190129051432.22023-1-peterx@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

qcow2: list of bitmaps new test 242

A new test file 242 added to the qemu-iotests set. It checks
the format of qcow2 specific information for the new added
section that lists details of bitmaps.

Signed-off-by: Andrey Shinkevich <andrey.shinkevich@virtuozzo.com>
Message-Id: <1549638368-530182-4-git-send-email-andrey.shinkevich@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: pep8 compliance, avoid trailing blank line]
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Eric Blake <eblake@redhat.com>

qcow2: Add list of bitmaps to ImageInfoSpecificQCow2

In the 'Format specific information' section of the 'qemu-img info'
command output, the supplemental information about existing QCOW2
bitmaps will be shown, such as a bitmap name, flags and granularity:

image: /vz/vmprivate/VM1/harddisk.hdd
file format: qcow2
virtual size: 64G (68719476736 bytes)
disk size: 3.0M
cluster_size: 1048576
Format specific information:
    compat: 1.1
    lazy refcounts: true
    bitmaps:
        [0]:
            flags:
                [0]: in-use
                [1]: auto
            name: back-up1
            granularity: 65536
        [1]:
            flags:
                [0]: in-use
                [1]: auto
            name: back-up2
            granularity: 65536
    refcount bits: 16
    corrupt: false

Signed-off-by: Andrey Shinkevich <andrey.shinkevich@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <1549638368-530182-3-git-send-email-andrey.shinkevich@virtuozzo.com>
Signed-off-by: Eric Blake <eblake@redhat.com>

bdrv_query_image_info Error parameter added

Inform a user in case qcow2_get_specific_info fails to obtain
QCOW2 image specific information. This patch is preliminary to
the one "qcow2: Add list of bitmaps to ImageInfoSpecificQCow2".

Signed-off-by: Andrey Shinkevich <andrey.shinkevich@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <1549638368-530182-2-git-send-email-andrey.shinkevich@virtuozzo.com>
Signed-off-by: Eric Blake <eblake@redhat.com>

nbd/server: Kill pointless shadowed variable

lgtm.com pointed out that commit 678ba275 introduced a shadowed
declaration of local variable 'bs'; thankfully, the inner 'bs'
obtained by 'blk_bs(blk)' matches the outer one given that we had
'blk_insert_bs(blk, bs, errp)' a few lines earlier, and there are
no later uses of 'bs' beyond the scope of the 'if (bitmap)' to
care if we change the value stored in 'bs' while traveling the
backing chain to find a bitmap.  So simply get rid of the extra
declaration.

Reported-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <20190207191357.6665-1-eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>

Merge remote-tracking branch 'remotes/rth/tags/pull-tcg-20190211' into staging

Fix dynamic tlb resize
Fix x86 host vector saturation
Diagnose missing tcg labels

# gpg: Signature made Mon 11 Feb 2019 16:57:52 GMT
# gpg:                using RSA key 64DF38E8AF7E215F
# 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/tags/pull-tcg-20190211:
  cputlb: update TLB entry/index after tlb_fill
  exec-all: document that tlb_fill can trigger a TLB resize
  tcg/i386: fix unsigned vector saturating arithmetic
  tcg: Diagnose referenced labels that have not been emitted

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

cputlb: update TLB entry/index after tlb_fill

We are failing to take into account that tlb_fill() can cause a
TLB resize, which renders prior TLB entry pointers/indices stale.
Fix it by re-doing the TLB entry lookups immediately after tlb_fill.

Fixes: 86e1eff8bc ("tcg: introduce dynamic TLB sizing", 2019-01-28)
Reported-by: Max Filippov <jcmvbkbc@gmail.com>
Tested-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Message-Id: <20190209162745.12668-3-cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

exec-all: document that tlb_fill can trigger a TLB resize

Signed-off-by: Emilio G. Cota <cota@braap.org>
Message-Id: <20190209162745.12668-2-cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/i386: fix unsigned vector saturating arithmetic

Due to a cut/paste error in the original implementation, the unsigned
vector saturating arithmetic was erroneously being calculated as signed
vector saturating arithmetic.

Fixes: 8ffafbcec2 ("tcg/i386: Implement vector saturating arithmetic")
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Message-Id: <20190207224258.426-1-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg: Diagnose referenced labels that have not been emitted

Currently, a jump to a label that is not defined anywhere will
be emitted not be relocated.  This results in a jump to a random
jump target.  With tcg debugging, print a diagnostic to the -d op
file and abort.

This could help debug or detect errors like
c2d9644e6d ("target/arm: Fix crash on conditional instruction in an IT block")

Reported-by: Roman Kapl <code@rkapl.cz>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

Merge remote-tracking branch 'remotes/stsquad/tags/pull-testing-next-110219-1' into staging

Testing updates:

  - .travis.yml tweaks and optimisations
  - .cirrus.yml enabled for FreeBSD CI
  - docker.py clean-ups for binfmt_misc
  - more control of vm-test builds

# gpg: Signature made Mon 11 Feb 2019 13:03:14 GMT
# gpg:                using RSA key F715F7CD46F94435F4F588658E520D61289519AE
# gpg: Good signature from "Alex Bennée (Master Work Key) <alex.bennee@linaro.org>" [full]
# Primary key fingerprint: 6685 AE99 E751 67BC AFC8  DF35 FBD0 DB09 5A9E 2A44
#      Subkey fingerprint: F715 F7CD 46F9 4435 F4F5  8865 8E52 0D61 2895 19AE

* remotes/stsquad/tags/pull-testing-next-110219-1:
  tests/vm: Be verbose while extracting compressed images
  docs/devel/testing: Add -a option to usermod command on docker setup
  scripts/qemu.py: allow arches use KVM for their 32bit cousins
  tests/vm: expose BUILD_TARGET, TARGET_LIST and EXTRA_CONFIGURE_OPTS
  tests/vm: add --build-target option
  tests/vm: call make check directly for netbsd/freebsd/ubuntu.i386
  tests/vm: move images to $HOME/.cache/qemu-vm/images
  tests: PEP8 cleanup of docker.py, mostly white space
  tests: docker.py be even smarter with persistent binfmt_misc
  tests: make docker.py check for persistent configs
  tests: make docker.py update use configured binfmt path
  docker: add debian-buster-arm64-cross
  archive-source.sh: Clone the submodules locally
  MAINTAINERS: Add an entry for scripts/archive-source.sh
  .travis.yml: fold --disable-tcg into alternate coroutine builds
  .travis.yml: separate tools and docs into another entry
  .travis.yml: stop requesting libffi & gettext from homebrew
  .cirrus.yml: basic compile and test for FreeBSD

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

tests/vm: Be verbose while extracting compressed images

Depending of the host hardware, copying and extracting VM images can
take up to few minutes. Add verbosity to avoid the user to worry about
VMs hanging.

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190129175403.18017-2-philmd@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

docs/devel/testing: Add -a option to usermod command on docker setup

The option -G of usermod command will remove user from other groups
not listed, i.e.: $USER will belong only to group 'docker' after
following the documentation as is.

From usermod(8) manual page:

    If the user is currently a member of a group which is not listed,
    the user will be removed from the group. This behaviour can be
    changed via the -a option, which appends the user to the current
    supplementary group list.

This patch improves the situation by adding the -a option to the
usermod command, which will just append user to the supplementary
group list.

Cc: qemu-trivial@nongnu.org
Signed-off-by: Murilo Opsfelder Araujo <muriloo@linux.ibm.com>
Message-Id: <20190207184346.6840-1-muriloo@linux.ibm.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

scripts/qemu.py: allow arches use KVM for their 32bit cousins

A lot of architectures can run their 32 bit cousins on KVM so the
kvm_available function needs to be a little less restricting when
deciding if KVM is available.

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

tests/vm: expose BUILD_TARGET, TARGET_LIST and EXTRA_CONFIGURE_OPTS

Now the underlying basevm support passes these along we can expose
some additional variables to our Makefile to allow more customised
tweaking of the build. For example:

  make vm-build-freebsd TARGET_LIST=aarch64-softmmu \
    EXTRA_CONFIGURE_OPTS="--disable-tools --disable-docs" \
    BUILD_TARGET=check-softfloat

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

tests/vm: add --build-target option

This allows us to invoke the build with a custom target (for the VMs
that use the {target} format string specifier). Currently OpenBSD is
still hardwired due to problems running check.

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

tests/vm: call make check directly for netbsd/freebsd/ubuntu.i386

The "make check" target calls check-qtest which has the appropriate
system binaries as dependencies so we shouldn't need to do two steps
of make invocation. Doing it in two steps was a hangover from when our
make check couldn't run tests in parallel.

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

tests/vm: move images to $HOME/.cache/qemu-vm/images

It's easier to move around the images then, by replacing the
subdirectory with a symlink.  Allows to share the images between
multiple qemu checkouts for example.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

tests: PEP8 cleanup of docker.py, mostly white space

My editor keeps putting squiggly lines under a bunch of the python
lines to remind me how non-PEP8 compliant it is. Clean that up so it's
easier to spot new errors.

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

tests: docker.py be even smarter with persistent binfmt_misc

If we have a persistent mapping we don't need the QEMU binary copied
into the container as the kernel has already opened the file and will
pass the fd in. However the support libraries will still need to be
there.

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

tests: make docker.py check for persistent configs

binfmt_misc configured with the "F" flag opens the interpreter at
config time. This means it can use an already open file-descriptor to
run QEMU so there is no point trying to copy the binary into a
container.

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

tests: make docker.py update use configured binfmt path

When copying a QEMU binary into a linux-user docker image we should
check what the current configured binfmt_misc path is rather than
just assuming "/usr/bin/qemu-bin". Obviously if the user changes the
configuration afterwards they will break their images again.

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

docker: add debian-buster-arm64-cross

We can't build QEMU with this but we can use this image to build newer
arm64 testcases which need more up to date tools.

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

archive-source.sh: Clone the submodules locally

We cloned the QEMU repository from the local storage. Since the
submodules are also available there, clone them too. This is
quicker and reduce network use.

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
[AJB: incorporated review suggestions from danpb]
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

MAINTAINERS: Add an entry for scripts/archive-source.sh

The scripts/archive-source.sh is used by the VM tests, it makes
sense to add it in the "Build and test automation" section.

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

.travis.yml: fold --disable-tcg into alternate coroutine builds

The alternate coroutine builds are really only of interest to people
running KVM (although I think you could use them for TCG if you really
tried). As they tend to run long lets kill two birds with one stone
and fold the --disable-tcg build into them.

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

.travis.yml: separate tools and docs into another entry

Re-building the tools and documents by default is a little wasteful as
they are not really affected by the main build options. Split tools
and documents into their own task with a minimal softmmu and
linux-user target list just to check they don't interact badly.

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

.travis.yml: stop requesting libffi & gettext from homebrew

The default package set installed on macOS builders from Travis already
includes libffi and gettext as shown by log messages:

  Skipping install of libffi formula. It is already up-to-date.
  Using libffi
  Skipping install of gettext formula. It is already up-to-date.
  Using gettext

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

.cirrus.yml: basic compile and test for FreeBSD

Signed-off-by: Ed Maste <emaste@freebsd.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

Merge remote-tracking branch 'remotes/thibault/tags/samuel-thibault' into staging

More work towards libslirp

Marc-André Lureau (27):
  slirp: generalize guestfwd with a callback based approach
  net/slirp: simplify checking for cmd: prefix
  net/slirp: free forwarding rules on cleanup
  net/slirp: fix leaks on forwarding rule registration error
  slirp: add callbacks for timer
  slirp: replace trace functions with DEBUG calls
  slirp: replace QEMU_PACKED with SLIRP_PACKED
  slirp: replace most qemu socket utilities with slirp own version
  slirp: replace qemu_set_nonblock()
  slirp: add unregister_poll_fd() callback
  slirp: replace qemu_notify_event() with a callback
  slirp: move QEMU state saving to a separate unit
  slirp: do not include qemu headers in libslirp.h public API header
  slirp: improve windows headers inclusion
  slirp: add slirp own version of pstrcpy
  slirp: remove qemu timer.h dependency
  slirp: remove now useless QEMU headers inclusions
  slirp: replace net/eth.h inclusion with own defines
  slirp: replace qemu qtailq with slirp own copy
  slirp: replace remaining qemu headers dependency
  slirp: prefer c99 types over BSD kind
  slirp: improve send_packet() callback
  slirp: replace global polling with per-instance & notifier
  slirp: remove slirp_instances list
  slirp: use polling callbacks, drop glib requirement
  slirp: pass opaque to all callbacks
  slirp: API is extern C

Peter Maydell (2):
  slirp: Avoid marking naturally packed structs as QEMU_PACKED
  slirp: Don't mark struct ipq or struct ipasfrag as packed

Samuel Thibault (3):
  slirp: Avoid unaligned 16bit memory access
  slirp: replace QEMU_BUILD_BUG_ON with G_STATIC_ASSERT
  slirp: Move g_spawn_async_with_fds_qemu compatibility to slirp/

# gpg: Signature made Thu 07 Feb 2019 14:02:41 GMT
# gpg:                using RSA key E61DBB15D4172BDEC97E92D9DB550E89F0FA54F3
# gpg: Good signature from "Samuel Thibault <samuel.thibault@aquilenet.fr>" [unknown]
# gpg:                 aka "Samuel Thibault <sthibault@debian.org>" [marginal]
# gpg:                 aka "Samuel Thibault <samuel.thibault@gnu.org>" [unknown]
# gpg:                 aka "Samuel Thibault <samuel.thibault@inria.fr>" [marginal]
# gpg:                 aka "Samuel Thibault <samuel.thibault@labri.fr>" [marginal]
# gpg:                 aka "Samuel Thibault <samuel.thibault@ens-lyon.org>" [marginal]
# gpg:                 aka "Samuel Thibault <samuel.thibault@u-bordeaux.fr>" [unknown]
# gpg: WARNING: This key is not certified with sufficiently trusted signatures!
# gpg:          It is not certain that the signature belongs to the owner.
# Primary key fingerprint: 900C B024 B679 31D4 0F82  304B D017 8C76 7D06 9EE6
#      Subkey fingerprint: E61D BB15 D417 2BDE C97E  92D9 DB55 0E89 F0FA 54F3

* remotes/thibault/tags/samuel-thibault: (32 commits)
  slirp: API is extern C
  slirp: pass opaque to all callbacks
  slirp: use polling callbacks, drop glib requirement
  slirp: remove slirp_instances list
  slirp: replace global polling with per-instance & notifier
  slirp: improve send_packet() callback
  slirp: prefer c99 types over BSD kind
  slirp: replace remaining qemu headers dependency
  slirp: Move g_spawn_async_with_fds_qemu compatibility to slirp/
  slirp: replace QEMU_BUILD_BUG_ON with G_STATIC_ASSERT
  slirp: replace qemu qtailq with slirp own copy
  slirp: replace net/eth.h inclusion with own defines
  slirp: remove now useless QEMU headers inclusions
  slirp: remove qemu timer.h dependency
  slirp: add slirp own version of pstrcpy
  slirp: improve windows headers inclusion
  slirp: do not include qemu headers in libslirp.h public API header
  slirp: move QEMU state saving to a separate unit
  slirp: replace qemu_notify_event() with a callback
  slirp: add unregister_poll_fd() callback
  ...

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

Merge remote-tracking branch 'remotes/vivier2/tags/linux-user-for-4.0-pull-request' into staging

- add new netlink type from linux v4.18 and v4.19
- fix coverity warning (CID 1390634)
- fix ioctl(SIOCGIFCONF) crash

# gpg: Signature made Thu 07 Feb 2019 13:12:53 GMT
# gpg:                using RSA key F30C38BD3F2FBE3C
# gpg: Good signature from "Laurent Vivier <lvivier@redhat.com>" [full]
# gpg:                 aka "Laurent Vivier <laurent@vivier.eu>" [full]
# gpg:                 aka "Laurent Vivier (Red Hat) <lvivier@redhat.com>" [full]
# Primary key fingerprint: CD2F 75DD C8E3 A4DC 2E4F  5173 F30C 38BD 3F2F BE3C

* remotes/vivier2/tags/linux-user-for-4.0-pull-request:
  linux-user: add new netlink types
  linux-user: Check sscanf return value in open_net_route()
  Fix linux-user crashes in ioctl(SIOCGIFCONF) when ifc_buf is NULL.

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

Merge remote-tracking branch 'remotes/mcayland/tags/qemu-sparc-20190206' into staging

qemu-sparc queue

# gpg: Signature made Wed 06 Feb 2019 21:25:18 GMT
# gpg:                using RSA key CC621AB98E82200D915CC9C45BC2C56FAE0F321F
# gpg:                issuer "mark.cave-ayland@ilande.co.uk"
# gpg: Good signature from "Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>" [full]
# Primary key fingerprint: CC62 1AB9 8E82 200D 915C  C9C4 5BC2 C56F AE0F 321F

* remotes/mcayland/tags/qemu-sparc-20190206:
  sun4m: pass initrd size to OpenBIOS via fw_cfg interface
  sun4u: add power_mem_read routine
  hw/sparc64: Create VGA device only if it has really been requested

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

Merge remote-tracking branch 'remotes/vivier2/tags/trivial-patches-pull-request' into staging

Trivial patches:
* Update copyright
* Fix LGPL in target/moxie
* configure portability fix
* Drop useless inclusion of "hw/i386/pc.h"
* Mark the cpu-cluster device with user_creatable = false
* tsc210x: Fix building with no verbosity

# gpg: Signature made Wed 06 Feb 2019 15:27:35 GMT
# gpg:                using RSA key F30C38BD3F2FBE3C
# gpg: Good signature from "Laurent Vivier <lvivier@redhat.com>" [full]
# gpg:                 aka "Laurent Vivier <laurent@vivier.eu>" [full]
# gpg:                 aka "Laurent Vivier (Red Hat) <lvivier@redhat.com>" [full]
# Primary key fingerprint: CD2F 75DD C8E3 A4DC 2E4F  5173 F30C 38BD 3F2F BE3C

* remotes/vivier2/tags/trivial-patches-pull-request:
  hw/input/tsc210x: Fix building with no verbosity
  hw/cpu/cluster: Mark the cpu-cluster device with user_creatable = false
  hw/unicore32/puv3: Drop useless inclusion of "hw/i386/pc.h"
  hw/sparc64/sun4u: Drop useless inclusion of "hw/i386/pc.h"
  configure: Avoid non-portable 'test -o/-a'
  target/moxie: Fix LGPL information in the file headers
  qemu-common.h: Update copyright string for 2019

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

Merge remote-tracking branch 'remotes/elmarco/tags/dump-pull-request' into staging

Pull request

# gpg: Signature made Wed 06 Feb 2019 14:57:54 GMT
# gpg:                using RSA key DAE8E10975969CE5
# gpg: Good signature from "Marc-André Lureau <marcandre.lureau@redhat.com>" [full]
# gpg:                 aka "Marc-André Lureau <marcandre.lureau@gmail.com>" [full]
# Primary key fingerprint: 87A9 BD93 3F87 C606 D276  F62D DAE8 E109 7596 9CE5

* remotes/elmarco/tags/dump-pull-request:
  dump: Set correct vaddr for ELF dump

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

slirp: API is extern C

Make it possible to use headers easily with C++ projects.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: pass opaque to all callbacks

This is friendlier for FFI bindings.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: use polling callbacks, drop glib requirement

It would be legitimate to use libslirp without glib. Let's
add_poll/get_revents pair of callbacks to provide the same
functionality.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: remove slirp_instances list

Now that polling is done per-instance, we don't need a global list of
slirp instances.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: replace global polling with per-instance & notifier

Remove hard-coded dependency on slirp in main-loop, and use a "poll"
notifier instead. The notifier is registered per slirp instance.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: improve send_packet() callback

Use a more descriptive name for the callback.

Reuse the SlirpWriteCb type. Wrap it to check that all data has been written.

Return a ssize_t for potential error handling and data-loss reporting.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: prefer c99 types over BSD kind

Replace:
- u_char -> uint8_t
- u_short -> uint16_t
- u_long -> uint32_t
- u_int -> unsigned
- caddr_t -> char *

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: replace remaining qemu headers dependency

Except for the migration code which is gated by WITH_QEMU, only
include our own headers, so libslirp can be built standalone.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: Move g_spawn_async_with_fds_qemu compatibility to slirp/

Only slirp actually needs it, and will need it along in libslirp.

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>

slirp: replace QEMU_BUILD_BUG_ON with G_STATIC_ASSERT

to remove another dependency on qemu.

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>

slirp: replace qemu qtailq with slirp own copy

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: replace net/eth.h inclusion with own defines

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: remove now useless QEMU headers inclusions

Some of those could have been squashed earlier, but it is easier to do
it all here.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: remove qemu timer.h dependency

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: add slirp own version of pstrcpy

Remove a dependency on qemu util.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: improve windows headers inclusion

Our API usage requires Vista, set WIN32_LEAN_AND_MEAN to fix a number
of issues (winsock2.h include order for ex, which is better to include
first for legacy reasons).

While at it, group redundants #ifndef _WIN32 blocks.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: do not include qemu headers in libslirp.h public API header

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: move QEMU state saving to a separate unit

Make state saving optional: this will allow to build SLIRP without
QEMU. (eventually, the vmstate helpers will be extracted, so an
external project & process could save its state)

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: replace qemu_notify_event() with a callback

Introduce a SlirpCb callback to kick the main io-thread.

Add an intermediary sodrop() function that will call SlirpCb.notify
callback when sbdrop() returns true.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: add unregister_poll_fd() callback

Add a counter-part to register_poll_fd() for completeness.

(so far, register_poll_fd() is called only on struct socket fd)

Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: replace qemu_set_nonblock()

Replace qemu_set_nonblock() with slirp_set_nonblock()

qemu_set_nonblock() does some event registration with the main
loop. Add a new callback register_poll_fd() for that reason.

Always build the fd-register stub, to avoid #if WIN32.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: replace most qemu socket utilities with slirp own version

qemu_set_nonblock() is slightly more problematic and will be dealt
with in a separate patch.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: replace QEMU_PACKED with SLIRP_PACKED

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: replace trace functions with DEBUG calls

Remove a dependency on QEMU. Use the existing logging facilities.
Set SLIRP_DEBUG=tftp to get tftp log.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: add callbacks for timer

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

net/slirp: fix leaks on forwarding rule registration error

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

net/slirp: free forwarding rules on cleanup

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

net/slirp: simplify checking for cmd: prefix

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: generalize guestfwd with a callback based approach

Instead of calling into QEMU chardev directly, and mixing it with
slirp_add_exec() handling, add a new function slirp_add_guestfwd()
which takes a write callback.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: Don't mark struct ipq or struct ipasfrag as packed

There is no reason to mark the struct ipq and struct ipasfrag as
packed: they are naturally aligned anyway, and are not representing
any on-the-wire packet format.  Indeed they vary in size depending on
the size of pointers on the host system, because the 'struct qlink'
members include 'void *' fields.

Dropping the 'packed' annotation fixes clang -Waddress-of-packed-member
warnings and probably lets the compiler generate better code too.

The only thing we do care about in the layout of the struct is
that the frag_link matches up with the ipf_link of the struct
ipasfrag, as documented in the comment on that struct; assert
at build time that this is the case.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: Avoid marking naturally packed structs as QEMU_PACKED

Various ipv6 structs in the slirp headers are marked QEMU_PACKED,
but they are actually naturally aligned and will have no padding
in them. Instead of marking them with the 'packed' attribute,
assert at compile time that they are the size we expect. This
allows us to take the address of fields within the structs
without risking undefined behaviour, and suppresses clang
-Waddress-of-packed-member warnings.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

slirp: Avoid unaligned 16bit memory access

pkt parameter may be unaligned, so we must access it byte-wise.

This fixes sparc64 host SIGBUS during pxe boot.

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

linux-user: add new netlink types

Add QEMU_IFLA_MIN_MTU, QEMU_IFLA_MAX_MTU (from linux v4.19)
    QEMU_IFLA_BRPORT_ISOLATED (from linux v4.18) and
    QEMU_IFLA_BRPORT_BACKUP_PORT (from linux v4.19).

These new types fix this error flow with sudo:
...
Unknown host QEMU_IFLA type: 50
Unknown host QEMU_IFLA type: 51
Unknown QEMU_IFLA_BRPORT type 33
...

Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20190206193211.6683-1-laurent@vivier.eu>

linux-user: Check sscanf return value in open_net_route()

Coverity warns (CID 1390634) that open_net_route() is not
checking the return value from sscanf(), which means that
it might then use values that aren't initialized.

Errors here should in general not happen since we're passing
an assumed-good /proc/net/route from the host kernel, but
if we do fail to parse a line then just skip it in the output
we pass to the guest.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20190205174207.9278-1-peter.maydell@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

Fix linux-user crashes in ioctl(SIOCGIFCONF) when ifc_buf is NULL.

Summary:
This is to fix bug https://bugs.launchpad.net/qemu/+bug/1796754.
It is valid for ifc_buf to be NULL according to
http://man7.org/linux/man-pages/man7/netdevice.7.html.

Signed-off-by: Kan Li <likan_999.student@sina.com>
Reviewed-by: Laurent Vivier <lvivier@redhat.com>
Message-Id: <20181024201303.114-1-likan_999.student@sina.com>
[lv: fix errors reported by checkpatch.pl]
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

Merge remote-tracking branch 'remotes/rth/tags/pull-hppa-20190206' into staging

Queued target/hppa patches

# gpg: Signature made Wed 06 Feb 2019 10:50:06 GMT
# gpg:                using RSA key 64DF38E8AF7E215F
# 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/tags/pull-hppa-20190206:
  target/hppa: fix PSW Q bit behaviour to match hardware
  target/hppa: fix setting registers via gdb
  target/hppa: use tb_cflags() to access tb->cflags

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

Merge remote-tracking branch 'remotes/rth/tags/pull-tcg-20190206' into staging

Queued accel/tcg patches

# gpg: Signature made Wed 06 Feb 2019 03:42:52 GMT
# gpg:                using RSA key 64DF38E8AF7E215F
# 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/tags/pull-tcg-20190206:
  accel/tcg: Consider cluster index in tb_lookup__cpu_state()
  tcg: add early clober modifier in atomic16_cmpxchg on aarch64

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

sun4m: pass initrd size to OpenBIOS via fw_cfg interface

This is to enable OpenBIOS to claim the initrd memory as in-use before attempting
to boot the kernel.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>

sun4u: add power_mem_read routine

Define skeleton 'power_mem_read' routine. Avoid NULL dereference.

Reported-by: Fakhri Zulkifli <mohdfakhrizulkifli@gmail.com>
Signed-off-by: Prasad J Pandit <pjp@fedoraproject.org>
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>

hw/sparc64: Create VGA device only if it has really been requested

The sun4u/sun4v machine currently always creates a VGA device, even if
the user started QEMU with "-nodefaults" or "-vga none". That's likely
not what the users expect in this case, so add a check whether the VGA
adapter has really been requested.

Signed-off-by: Thomas Huth <huth@tuxfamily.org>
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>

hw/input/tsc210x: Fix building with no verbosity

When building with TSC_VERBOSE not defined, we get:
      CC      arm-softmmu/hw/input/tsc210x.o
    hw/input/tsc210x.c: In function ‘tsc2102_data_register_write’:
    hw/input/tsc210x.c:554:5: error: label at end of compound statement
         default:
         ^~~~~~~
    hw/input/tsc210x.c: In function ‘tsc2102_control_register_write’:
    hw/input/tsc210x.c:638:5: error: label at end of compound statement
         bad_reg:
         ^~~~~~~
    hw/input/tsc210x.c: In function ‘tsc2102_audio_register_write’:
    hw/input/tsc210x.c:766:5: error: label at end of compound statement
         default:
         ^~~~~~~
    make[1]: *** [rules.mak:69: hw/input/tsc210x.o] Error 1

Fix this by replacing the culprit fprintf(stderr) calls by a more
recent API: qemu_log_mask(LOG_GUEST_ERROR). Other fprintf() calls
are left untouched.

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Message-Id: <20190204204517.23698-1-philmd@redhat.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

hw/cpu/cluster: Mark the cpu-cluster device with user_creatable = false

The device can not be instantiated by the user and QEMU currently
aborts when you try to use it:

$ x86_64-softmmu/qemu-system-x86_64 -device cpu-cluster
qemu-system-x86_64: hw/cpu/cluster.c:73: cpu_cluster_realize:
 Assertion `cbdata.cpu_count > 0' failed.
Aborted (core dumped)

Since this is an internal device only, mark it with user_creatable = false.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Luc Michel <luc.michel@greensocs.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-Id: <1549371525-29899-1-git-send-email-thuth@redhat.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

hw/unicore32/puv3: Drop useless inclusion of "hw/i386/pc.h"

In 47973a2dbf we split the last generic chipset out of the PC
board, but forgot to remove the include of "hw/i386/pc.h".
Since it is now unused, remove it.

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20190204210433.26088-3-philmd@redhat.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

hw/sparc64/sun4u: Drop useless inclusion of "hw/i386/pc.h"

In 47973a2dbf we split the last generic chipset out of the PC
board, but forgot to remove the include of "hw/i386/pc.h".
Since it is now unused, remove it.

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Message-Id: <20190204210433.26088-2-philmd@redhat.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

configure: Avoid non-portable 'test -o/-a'

POSIX says that it is better to use &&/|| and two separate test
invocations than it is to try and use -a and -o (in fact, there
are some tests that are inherently ambiguous to parse if the
user passes in corner-case input like "(").

Since we cannot guarantee which shell runs configure, we cannot
rely on -o/-a always following bash's parser rules.

Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190205023937.18245-1-eblake@redhat.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

dump: Set correct vaddr for ELF dump

vaddr needs to be equal to the paddr since the dump file represents the
physical memory image.

Without setting vaddr correctly, GDB would load all the different memory
regions on top of each other to vaddr 0, thus making GDB showing the wrong
memory data for a given address.

Signed-off-by: Jon Doron <arilou@gmail.com>
Message-Id: <20190109082203.27142-1-arilou@gmail.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Tested-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Acked-by: Laszlo Ersek <lersek@redhat.com>