OSDN Git Service

qmiga/qemu.git
5 years agoiotests.py: Add node_info()
Max Reitz [Fri, 1 Feb 2019 19:29:11 +0000 (20:29 +0100)]
iotests.py: Add node_info()

This function queries a node; since we cannot do that right now, it
executes query-named-block-nodes and returns the matching node's object.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20190201192935.18394-8-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
5 years agoiotests.py: Add filter_imgfmt()
Max Reitz [Fri, 1 Feb 2019 19:29:10 +0000 (20:29 +0100)]
iotests.py: Add filter_imgfmt()

Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-id: 20190201192935.18394-7-mreitz@redhat.com
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
5 years agoblock: Respect backing bs in bdrv_refresh_filename
Max Reitz [Fri, 1 Feb 2019 19:29:09 +0000 (20:29 +0100)]
block: Respect backing bs in bdrv_refresh_filename

Basically, bdrv_refresh_filename() should respect all children of a
BlockDriverState. However, generally those children are driver-specific,
so this function cannot handle the general case. On the other hand,
there are only few drivers which use other children than @file and
@backing (that being vmdk, quorum, and blkverify).

Most block drivers only use @file and/or @backing (if they use any
children at all). Both can be implemented directly in
bdrv_refresh_filename.

The user overriding the file's filename is already handled, however, the
user overriding the backing file is not. If this is done, opening the
BDS with the plain filename of its file will not be correct, so we may
not set bs->exact_filename in that case.

iotest 051 contains test cases for overriding the backing file, and so
its output changes with this patch applied.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Message-id: 20190201192935.18394-6-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
5 years agoblock: Add BDS.auto_backing_file
Max Reitz [Fri, 1 Feb 2019 19:29:08 +0000 (20:29 +0100)]
block: Add BDS.auto_backing_file

If the backing file is overridden, this most probably does change the
guest-visible data of a BDS.  Therefore, we will need to consider this
in bdrv_refresh_filename().

To see whether it has been overridden, we might want to compare
bs->backing_file and bs->backing->bs->filename.  However,
bs->backing_file is changed by bdrv_set_backing_hd() (which is just used
to change the backing child at runtime, without modifying the image
header), so bs->backing_file most of the time simply contains a copy of
bs->backing->bs->filename anyway, so it is useless for such a
comparison.

This patch adds an auto_backing_file BDS field which contains the
backing file path as indicated by the image header, which is not changed
by bdrv_set_backing_hd().

Because of bdrv_refresh_filename() magic, however, a BDS's filename may
differ from what has been specified during bdrv_open().  Then, the
comparison between bs->auto_backing_file and bs->backing->bs->filename
may fail even though bs->backing was opened from bs->auto_backing_file.
To mitigate this, we can copy the real BDS's filename (after the whole
bdrv_open() and bdrv_refresh_filename() process) into
bs->auto_backing_file, if we know the former has been opened based on
the latter.  This is only possible if no options modifying the backing
file's behavior have been specified, though.  To simplify things, this
patch only copies the filename from the backing file if no options have
been specified for it at all.

Furthermore, there are cases where an overlay is created by qemu which
already contains a BDS's filename (e.g. in blockdev-snapshot-sync).  We
do not need to worry about updating the overlay's bs->auto_backing_file
there, because we actually wrote a post-bdrv_refresh_filename() filename
into the image header.

So all in all, there will be false negatives where (as of a future
patch) bdrv_refresh_filename() will assume that the backing file differs
from what was specified in the image header, even though it really does
not.  However, these cases should be limited to where (1) the user
actually did override something in the backing chain (e.g. by specifying
options for the backing file), or (2) the user executed a QMP command to
change some node's backing file (e.g. change-backing-file or
block-commit with @backing-file given) where the given filename does not
happen to coincide with qemu's idea of the backing BDS's filename.

Then again, (1) really is limited to -drive.  With -blockdev or
blockdev-add, you have to adhere to the schema, so a user cannot give
partial "unimportant" options (e.g. by just setting backing.node-name
and leaving the rest to the image header).  Therefore, trying to fix
this would mean trying to fix something for -drive only.

To improve on (2), we would need a full infrastructure to "canonicalize"
an arbitrary filename (+ options), so it can be compared against
another.  That seems a bit over the top, considering that filenames
nowadays are there mostly for the user's entertainment.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Message-id: 20190201192935.18394-5-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
5 years agoblock: Skip implicit nodes for filename info
Max Reitz [Fri, 1 Feb 2019 19:29:07 +0000 (20:29 +0100)]
block: Skip implicit nodes for filename info

bdrv_refresh_filename() should simply skip all implicit nodes.  They are
supposed to be invisible to the user, so they should not appear in
filename information.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Message-id: 20190201192935.18394-4-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
5 years agoblock: Use children list in bdrv_refresh_filename
Max Reitz [Fri, 1 Feb 2019 19:29:06 +0000 (20:29 +0100)]
block: Use children list in bdrv_refresh_filename

bdrv_refresh_filename() should invoke itself recursively on all
children, not just on file.

With that change, we can remove the manual invocations in blkverify,
quorum, commit, mirror, and blklogwrites.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Message-id: 20190201192935.18394-3-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
5 years agoblock: Use bdrv_refresh_filename() to pull
Max Reitz [Fri, 1 Feb 2019 19:29:05 +0000 (20:29 +0100)]
block: Use bdrv_refresh_filename() to pull

Before this patch, bdrv_refresh_filename() is used in a pushing manner:
Whenever the BDS graph is modified, the parents of the modified edges
are supposed to be updated (recursively upwards).  However, that is
nonviable, considering that we want child changes not to concern
parents.

Also, in the long run we want a pull model anyway: Here, we would have a
bdrv_filename() function which returns a BDS's filename, freshly
constructed.

This patch is an intermediate step.  It adds bdrv_refresh_filename()
calls before every place a BDS.filename value is used.  The only
exceptions are protocol drivers that use their own filename, which
clearly would not profit from refreshing that filename before.

Also, bdrv_get_encrypted_filename() is removed along the way (as a user
of BDS.filename), since it is completely unused.

In turn, all of the calls to bdrv_refresh_filename() before this patch
are removed, because we no longer have to call this function on graph
changes.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-id: 20190201192935.18394-2-mreitz@redhat.com
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
5 years agoblock/nvme: Remove QEMU_PACKED from naturally aligned NVMeRegs struct
Thomas Huth [Mon, 25 Feb 2019 11:59:30 +0000 (12:59 +0100)]
block/nvme: Remove QEMU_PACKED from naturally aligned NVMeRegs struct

The QEMU_PACKED is causing a compiler warning/error with GCC 9:

  CC      block/nvme.o
block/nvme.c: In function ‘nvme_create_queue_pair’:
block/nvme.c:209:22: error: taking address of packed member of
 ‘struct <anonymous>’ may result in an unaligned pointer value
 [-Werror=address-of-packed-member]
  209 |     q->sq.doorbell = &s->regs->doorbells[idx * 2 * s->doorbell_scale];

All members of the struct are naturally aligned, so there should
not be the need for QEMU_PACKED here, and the following QEMU_BUILD_BUG_ON
also ensures that there is no padding. Thus simply remove the QEMU_PACKED
here.

Buglink: https://bugs.launchpad.net/qemu/+bug/1817525
Reported-by: Satheesh Rajendran <sathnaga@linux.vnet.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
5 years agoqcow2: Assert that L2 table offsets fit in the L1 table
Alberto Garcia [Fri, 8 Feb 2019 15:44:53 +0000 (17:44 +0200)]
qcow2: Assert that L2 table offsets fit in the L1 table

L1 table entries have a field to store the offset of an L2 table.
The rest of the bits of the entry are currently reserved except from
bit 63, which stores the COPIED flag.

The offset is always taken from the entry using L1E_OFFSET_MASK to
ensure that we only use the bits that belong to that field.

While that mask is used every time we read from the L1 table, it is
never used when we write to it. Due to the limits set elsewhere in the
code QEMU can never produce L2 table offsets that don't fit in that
field so any such offset when allocating an L2 table would indicate a
bug in QEMU.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
5 years agotests: add test-bdrv-graph-mod
Vladimir Sementsov-Ogievskiy [Sat, 23 Feb 2019 19:20:41 +0000 (22:20 +0300)]
tests: add test-bdrv-graph-mod

Add two tests of node graph modification.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
5 years agoblock: fix bdrv_check_perm for non-tree subgraph
Vladimir Sementsov-Ogievskiy [Sat, 23 Feb 2019 19:20:40 +0000 (22:20 +0300)]
block: fix bdrv_check_perm for non-tree subgraph

bdrv_check_perm in it's recursion checks each node in context of new
permissions for one parent, because of nature of DFS. It works well,
while children subgraph of top-most updated node is a tree, i.e. it
doesn't have any kind of loops. But if we have a loop (not oriented,
of course), i.e. we have two different ways from top-node to some
child-node, then bdrv_check_perm will do wrong thing:

  top
  | \
  |  |
  v  v
  A  B
  |  |
  v  v
  node

It will once check new permissions of node in context of new A
permissions and old B permissions and once visa-versa. It's a wrong way
and may lead to corruption of permission system. We may start with
no-permissions and all-shared for both A->node and B->node relations
and finish up with non shared write permission for both ways.

The following commit will add a test, which shows this bug.

To fix this situation, let's really set BdrvChild permissions during
bdrv_check_perm procedure. And we are happy here, as check-perm is
already written in transaction manner, so we just need to restore
backed-up permissions in _abort.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
5 years agoblock: improve should_update_child
Vladimir Sementsov-Ogievskiy [Sat, 23 Feb 2019 19:20:39 +0000 (22:20 +0300)]
block: improve should_update_child

As it already said in the comment, we don't want to create loops in
parent->child relations. So, when we try to append @to to @c, we should
check that @c is not in @to children subtree, and we should check it
recursively, not only the first level. The patch provides BFS-based
search, to check the relations.

This is needed for further fleecing-hook filter usage: we need to
append it to source, when the hook is already a parent of target, and
source may be in a backing chain of target (fleecing-scheme). So, on
appending, the hook should not became a child (direct or through
children subtree) of the target.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
5 years agoaio-posix: Assert that aio_poll() is always called in home thread
Kevin Wolf [Thu, 14 Feb 2019 12:13:36 +0000 (13:13 +0100)]
aio-posix: Assert that aio_poll() is always called in home thread

aio_poll() has an existing assertion that the function is only called
from the AioContext's home thread if blocking is allowed.

This is not enough, some handlers make assumptions about the thread they
run in. Extend the assertion to non-blocking calls, too.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
5 years agoblock: Use normal drain for bdrv_set_aio_context()
Kevin Wolf [Fri, 8 Feb 2019 15:53:37 +0000 (16:53 +0100)]
block: Use normal drain for bdrv_set_aio_context()

Now that bdrv_set_aio_context() works inside drained sections, it can
also use the real drain function instead of open coding something
similar.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
5 years agotest-bdrv-drain: AioContext switch in drained section
Kevin Wolf [Fri, 8 Feb 2019 12:18:49 +0000 (13:18 +0100)]
test-bdrv-drain: AioContext switch in drained section

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
5 years agoblock: Fix AioContext switch for drained node
Kevin Wolf [Fri, 8 Feb 2019 15:51:17 +0000 (16:51 +0100)]
block: Fix AioContext switch for drained node

When a drained node changes its AioContext, we need to move its
aio_disable_external() to the new context, too.

Without this fix, drain_end will try to reenable the new context, which
has never been disabled, so an assertion failure is triggered.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
5 years agoblock: Don't poll in bdrv_set_aio_context()
Kevin Wolf [Fri, 15 Feb 2019 12:13:12 +0000 (13:13 +0100)]
block: Don't poll in bdrv_set_aio_context()

The explicit aio_poll() call in bdrv_set_aio_context() was added in
commit c2b6428d388 as a workaround for bdrv_drain() failing to achieve
to actually quiesce everything (specifically the NBD client code to
switch AioContext).

Now that the NBD client has been fixed to complete this operation during
bdrv_drain(), we don't need the workaround any more.

It was wrong anyway: aio_poll() must always be run in the home thread of
the AioContext.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
5 years agonbd: Increase bs->in_flight during AioContext switch
Kevin Wolf [Mon, 18 Feb 2019 14:33:08 +0000 (15:33 +0100)]
nbd: Increase bs->in_flight during AioContext switch

bdrv_drain() must not leave connection_co scheduled, so bs->in_flight
needs to be increased while the coroutine is waiting to be scheduled
in the new AioContext after nbd_client_attach_aio_context().

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
5 years agonbd: Use low-level QIOChannel API in nbd_read_eof()
Kevin Wolf [Mon, 18 Feb 2019 13:56:01 +0000 (14:56 +0100)]
nbd: Use low-level QIOChannel API in nbd_read_eof()

Instead of using the convenience wrapper qio_channel_read_all_eof(), use
the lower level QIOChannel API. This means duplicating some code, but
we'll need this because this coroutine yield is special: We want it to
be interruptible so that nbd_client_attach_aio_context() can correctly
reenter the coroutine.

This moves the bdrv_dec/inc_in_flight() pair into nbd_read_eof(), so
that connection_co will always sit in this exact qio_channel_yield()
call when bdrv_drain() returns.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
5 years agonbd: Move nbd_read_eof() to nbd/client.c
Kevin Wolf [Mon, 18 Feb 2019 13:38:15 +0000 (14:38 +0100)]
nbd: Move nbd_read_eof() to nbd/client.c

The only caller of nbd_read_eof() is nbd_receive_reply(), so it doesn't
have to live in the header file, but can move next to its caller.

Also add the missing coroutine_fn to the function and its caller.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
5 years agoio: Remove redundant read/write_coroutine assignments
Kevin Wolf [Wed, 20 Feb 2019 17:00:07 +0000 (18:00 +0100)]
io: Remove redundant read/write_coroutine assignments

qio_channel_yield() now updates ioc->read_write/coroutine and calls
qio_channel_set_aio_fd_handlers(), so the code in the handlers has
become redundant and can be removed.

This does not make a difference in intermediate states because
aio_co_wake() really enters the coroutine immediately here: These
handlers are never run in coroutine context, and we're in the right
AioContext because qio_channel_attach_aio_context() asserts that the
handlers are inactive.

To make these conditions more obvious, assert the right AioContext.

Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
5 years agoio: Make qio_channel_yield() interruptible
Kevin Wolf [Mon, 18 Feb 2019 13:09:32 +0000 (14:09 +0100)]
io: Make qio_channel_yield() interruptible

Similar to how qemu_co_sleep_ns() allows preemption from an external
coroutine entry, allow reentering qio_channel_yield() early.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
5 years agonbd: Restrict connection_co reentrance
Kevin Wolf [Fri, 15 Feb 2019 15:53:51 +0000 (16:53 +0100)]
nbd: Restrict connection_co reentrance

nbd_client_attach_aio_context() schedules connection_co in the new
AioContext and this way reenters it in any arbitrary place that has
yielded. We can restrict this a bit to the function call where the
coroutine actually sits waiting when it's idle.

This doesn't solve any bug yet, but it shows where in the code we need
to support this random reentrance and where we don't have to care.

Add FIXME comments for the existing bugs that the rest of this series
will fix.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
5 years agovirtio-blk: Increase in_flight for request restart BH
Kevin Wolf [Thu, 14 Feb 2019 17:51:03 +0000 (18:51 +0100)]
virtio-blk: Increase in_flight for request restart BH

virtio_blk_dma_restart_bh() submits new requests, so in order to make
sure that these requests are not started inside a drained section of the
attached BlockBackend, we need to make sure that draining the
BlockBackend waits for the BH to be executed.

This BH is still questionable because its scheduled in the main thread
instead of the configured iothread. Leave a FIXME comment for this.

But with this fix, enabling the data plane at least waits for these
requests (in bdrv_set_aio_context()) instead of changing the AioContext
under their feet and making them run in the wrong thread, causing
crashes and failures (e.g. due to missing locking).

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
5 years agoblock-backend: Make blk_inc/dec_in_flight public
Kevin Wolf [Thu, 14 Feb 2019 17:42:44 +0000 (18:42 +0100)]
block-backend: Make blk_inc/dec_in_flight public

For some users of BlockBackends, just increasing the in_flight counter
is easier than implementing separate handlers in BlockDevOps. Make the
helper functions for this public.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
5 years agoqemu-img: fix error reporting for -object
Daniel P. Berrangé [Tue, 19 Feb 2019 11:46:09 +0000 (11:46 +0000)]
qemu-img: fix error reporting for -object

Error reporting for user_creatable_add_opts_foreach was changed so that
it no longer called 'error_report_err' in:

  commit 7e1e0c11127bde81cff260fc6859690435c509d6
  Author: Markus Armbruster <armbru@redhat.com>
  Date:   Wed Oct 17 10:26:43 2018 +0200

    qom: Clean up error reporting in user_creatable_add_opts_foreach()

Some callers were updated to pass in "&error_fatal" but all the ones in
qemu-img were left passing NULL. As a result all errors went to
/dev/null instead of being reported to the user.

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
5 years agocommit: Replace commit_top_bs on failure after deleting the block job
Alberto Garcia [Fri, 15 Feb 2019 13:49:32 +0000 (15:49 +0200)]
commit: Replace commit_top_bs on failure after deleting the block job

If there's an error in commit_start() then the block job must be
deleted before replacing commit_top_bs, otherwise it will fail because
of lack of permissions. This happens since the permission system was
introduced in 8dfba2797761d8a43744e4e6571c8175e448a478.

Fortunately this bug doesn't seem to be possible to reproduce at the
moment without changing the code.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
5 years agoblock: don't set the same context
Denis Plotnikov [Fri, 15 Feb 2019 13:03:25 +0000 (16:03 +0300)]
block: don't set the same context

Adds a fast path on aio context setting preventing
unnecessary context setting routine.
Also, it prevents issues with cyclic walk of child
bds-es appeared because of registering aio walking
notifiers:

Call stack:

0  __GI_raise
1  __GI_abort
2  __assert_fail_base
3  __GI___assert_fail
4  bdrv_detach_aio_context (bs=0x55f54d65c000)      <<<
5  bdrv_detach_aio_context (bs=0x55f54fc8a800)
6  bdrv_set_aio_context (bs=0x55f54fc8a800, ...)
7  block_job_attached_aio_context
8  bdrv_attach_aio_context (bs=0x55f54d65c000, ...) <<<
9  bdrv_set_aio_context (bs=0x55f54d65c000)
10 blk_set_aio_context
11 virtio_blk_data_plane_stop
12 virtio_bus_stop_ioeventfd
13 virtio_vmstate_change
14 vm_state_notify (running=0, state=RUN_STATE_SHUTDOWN)
15 do_vm_stop (state=RUN_STATE_SHUTDOWN, send_stop=true)
16 vm_stop (state=RUN_STATE_SHUTDOWN)
17 main_loop_should_exit
18 main_loop
19 main

This can happen because of "new" context attachment to VM disk bds.
When attaching a new context the corresponding aio context handler is
called for each of aio_notifiers registered on the VM disk bds context.
Among those handlers, there is the block_job_attached_aio_context handler
which sets a new aio context for the block job bds. When doing so,
the old context is detached from all the block job bds children and one of
them is the VM disk bds, serving as backing store for the blockjob bds,
although the VM disk bds is actually the initializer of that process.
Since the VM disk bds is protected with walking_aio_notifiers flag
from double processing in recursive calls, the assert fires.

Signed-off-by: Denis Plotnikov <dplotnikov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
5 years agoqcow2-snapshot: remove redundant find_snapshot_by_id_and_name call
Daniel Henrique Barboza [Wed, 7 Nov 2018 13:10:00 +0000 (11:10 -0200)]
qcow2-snapshot: remove redundant find_snapshot_by_id_and_name call

In qcow2_snapshot_create there is the following code block:

    /* Generate an ID */
    find_new_snapshot_id(bs, sn_info->id_str, sizeof(sn_info->id_str));

    /* Check that the ID is unique */
    if (find_snapshot_by_id_and_name(bs, sn_info->id_str, NULL) >= 0) {
        return -EEXIST;
    }

find_new_snapshot_id cycles through all snapshots, getting the id_str
as an unsigned long int, calculating the max id_max value of all the
existing id_strs and writing in the id_str pointer id_max + 1:

    for(i = 0; i < s->nb_snapshots; i++) {
        sn = s->snapshots + i;
        id = strtoul(sn->id_str, NULL, 10);
        if (id > id_max)
            id_max = id;
    }
    snprintf(id_str, id_str_size, "%lu", id_max + 1);

Here, sn_info->id_str will have the unique value id_max + 1. Right
after that, find_snapshot_by_id_and_name is called with
id = sn_info->id_str and name = NULL. This will cause the function
to execute the following:

    } else if (id) {
        for (i = 0; i < s->nb_snapshots; i++) {
            if (!strcmp(s->snapshots[i].id_str, id)) {
                return i;
            }
        }
    }

In short, we're searching the existing snapshots to see if sn_info->id_str
matches any existing id, right after we set in the previous line a
sn_info->id_str value that is already unique.

The first code block goes way back to commit 585f8587ad, a 2006 commit from
Fabrice Bellard that simply says "new qcow2 disk image format". No more
info is provided about this logic in any subsequent commits that moved
this code block around.

I can't say about the original design, but the current logic is redundant.
bdrv_snapshot_create is called in aio_context lock, forbidding any
concurrent call to accidentally create a new snapshot between
the find_new_snapshot_id and find_snapshot_by_id_and_name calls. What
we're ending up doing is to cycle through the snapshots two times
for no viable reason.

This patch eliminates the redundancy by removing the 'id is unique'
check that calls find_snapshot_by_id_and_name.

Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
5 years agoblock/snapshot: remove bdrv_snapshot_delete_by_id_or_name
Daniel Henrique Barboza [Wed, 7 Nov 2018 13:09:59 +0000 (11:09 -0200)]
block/snapshot: remove bdrv_snapshot_delete_by_id_or_name

After the previous patch, the only instance of this function left
is inside qemu-img.c.

qemu-img is using it inside the 'img_snapshot' function to delete
snapshots in the SNAPSHOT_DELETE case, based on a "snapshot_name"
string that refers to the tag, not ID, of the QEMUSnapshotInfo struct.
This can be verified by checking the SNAPSHOT_CREATE case that
comes shortly before SNAPSHOT_DELETE. In that case, the same
"snapshot_name" variable is being strcpy to the 'name' field
of the QEMUSnapshotInfo struct sn:

pstrcpy(sn.name, sizeof(sn.name), snapshot_name);

Based on that, it is unlikely that "snapshot_name" might contain
an "id" in SNAPSHOT_DELETE.

This patch changes SNAPSHOT_DELETE to use snapshot_find() and
snapshot_delete() instead of bdrv_snapshot_delete_by_id_or_name.
After that, there is no instances left of bdrv_snapshot_delete_by_id_or_name
in the code, so it is safe to remove it entirely.

Suggested-by: Murilo Opsfelder Araujo <muriloo@linux.ibm.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
5 years agoblock/snapshot.c: eliminate use of ID input in snapshot operations
Daniel Henrique Barboza [Wed, 7 Nov 2018 13:09:58 +0000 (11:09 -0200)]
block/snapshot.c: eliminate use of ID input in snapshot operations

At this moment, QEMU attempts to create/load/delete snapshots
by using either an ID (id_str) or a name. The problem is that the code
isn't consistent of whether the entered argument is an ID or a name,
causing unexpected behaviors.

For example, when creating snapshots via savevm <arg>, what happens is that
"arg" is treated as both name and id_str. In a guest without snapshots, create
a single snapshot via savevm:

(qemu) savevm 0
(qemu) info snapshots
List of snapshots present on all disks:
ID        TAG                 VM SIZE                DATE       VM CLOCK
--        0                      741M 2018-07-31 13:39:56   00:41:25.313

A snapshot with name "0" is created. ID is hidden from the user, but the
ID is a non-zero integer that starts at "1". Thus, this snapshot has
id_str=1, TAG="0". Creating a second snapshot with arg = 1, the first one
is deleted:

(qemu) savevm 1
(qemu) info snapshots
List of snapshots present on all disks:
ID        TAG                 VM SIZE                DATE       VM CLOCK
--        1                      741M 2018-07-31 13:42:14   00:41:55.252

What happened?

- when creating the second snapshot, a verification is done inside
bdrv_all_delete_snapshot to delete any existing snapshots that matches an
string argument. Here, the code calls bdrv_all_delete_snapshot("1", ...);

- bdrv_all_delete_snapshot calls bdrv_snapshot_find(..., "1") for each
BlockDriverState of the guest. And this is where things goes tilting:
bdrv_snapshot_find does a search by both id_str and name. It finds
out that there is a snapshot that has id_str = 1, stores a reference
to the snapshot in the sn_info pointer and then returns match found;

- since a match was found, a call to bdrv_snapshot_delete_by_id_or_name() is
made. This function ignores the pointer written by bdrv_snapshot_find. Instead,
it deletes the snapshot using bdrv_snapshot_delete() calling it first with
id_str = 1. If it fails to delete, then it calls it again with name = 1.

- after all that, QEMU creates the new snapshot, that has id_str = 1 and
name = 1. The user is left wondering that happened with the first snapshot
created. Similar bugs can be triggered when using loadvm and delvm.

Before contemplating discarding the use of ID input in these operations,
I've searched the code of what would be the implications. My findings
are:

- the RBD and Sheepdog drivers don't care. Both uses the 'name' field as
key in their logic, making id_str = name when appropriate.
replay-snapshot.c does not make any special use of id_str;

- qcow2 uses id_str as an unique identifier but it is automatically
calculated, not being influenced by user input. Other than that, there are
no distinguish operations made only with id_str;

- in blockdev.c, the delete operation uses a match of both id_str AND
name. Given that id_str is either a copy of 'name' or auto-generated,
we're fine here.

This gives motivation to not consider ID as a valid user input in HMP
commands - sticking with 'name' input only is more consistent. To
accomplish that, the following changes were made in this patch:

- bdrv_snapshot_find() does not match for id_str anymore, only 'name'. The
function is called in save_snapshot(), load_snapshot(), bdrv_all_delete_snapshot()
and bdrv_all_find_snapshot(). This change makes the search function more
predictable and does not change the behavior of any underlying code that uses
these affected functions, which are related to HMP (which is fine) and the
main loop inside vl.c (which doesn't care about it anyways);

- bdrv_all_delete_snapshot() does not call bdrv_snapshot_delete_by_id_or_name
anymore. Instead, it uses the pointer returned by bdrv_snapshot_find to
erase the snapshot with the exact match of id_str an name. This function
is called in save_snapshot and hmp_delvm, thus this change  produces the
intended effect;

- documentation changes to reflect the new behavior. I consider this to
be an API fix instead of an API change - the user was already creating
snapshots using 'name', but now he/she will also enjoy a consistent
behavior.

Ideally we would get rid of the id_str field entirely, but this would have
repercussions on existing snapshots. Another day perhaps.

Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Acked-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
5 years agoMAINTAINERS: Remove myself as block maintainer
Jeff Cody [Wed, 26 Sep 2018 18:05:33 +0000 (14:05 -0400)]
MAINTAINERS: Remove myself as block maintainer

I'll not be involved in day-to-day qemu development.  Remove myself as
maintainer from the remainder of the network block drivers, and revert
them to the general block layer maintainership.

Move 'sheepdog' to the 'Odd Fixes' support level.

For VHDX, added my personal email address as a maintainer, as I can
answer questions or send the occassional bug fix.  Leaving it as
'Supported', instead of 'Odd Fixes', because I think the rest of the
block layer maintainers and developers will upkeep it as well, if
needed.

Signed-off-by: Jeff Cody <jcody@redhat.com>
Acked-by: Max Reitz <mreitz@redhat.com>
Message-Id: <63e205cb84c8f0a10c1bc6d5d6856d72ceb56e41.1537984851.git.jcody@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
5 years agoMAINTAINERS: Replace myself with John Snow for block jobs
Jeff Cody [Wed, 26 Sep 2018 18:05:32 +0000 (14:05 -0400)]
MAINTAINERS: Replace myself with John Snow for block jobs

I'll not be involved with day-to-day qemu development, and John
Snow is a block jobs wizard.  Have him take over block job
maintainership duties.

Signed-off-by: Jeff Cody <jcody@redhat.com>
Acked-by: John Snow <jsnow@redhat.com>
Message-Id: <d56d7c6592e7d68aa72764e9616878394bffbc14.1537984851.git.jcody@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
5 years agoMerge remote-tracking branch 'remotes/kraxel/tags/vga-20190222-pull-request' into...
Peter Maydell [Mon, 25 Feb 2019 12:49:07 +0000 (12:49 +0000)]
Merge remote-tracking branch 'remotes/kraxel/tags/vga-20190222-pull-request' into staging

vga: bugfixes and edid support for virtio-vga

# gpg: Signature made Fri 22 Feb 2019 08:24:25 GMT
# gpg:                using RSA key 4CB6D8EED3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <kraxel@redhat.com>" [full]
# gpg:                 aka "Gerd Hoffmann <gerd@kraxel.org>" [full]
# gpg:                 aka "Gerd Hoffmann (private) <kraxel@gmail.com>" [full]
# Primary key fingerprint: A032 8CFF B93A 17A7 9901  FE7D 4CB6 D8EE D3E8 7138

* remotes/kraxel/tags/vga-20190222-pull-request:
  display/virtio: add edid support.
  virtio-gpu: remove useless 'waiting' field
  virtio-gpu: block both 2d and 3d rendering
  virtio-gpu: remove unused config_size
  virtio-gpu: remove unused qdev

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
5 years agoMerge remote-tracking branch 'remotes/kraxel/tags/ui-20190222-pull-request' into...
Peter Maydell [Mon, 25 Feb 2019 09:05:41 +0000 (09:05 +0000)]
Merge remote-tracking branch 'remotes/kraxel/tags/ui-20190222-pull-request' into staging

ui: add support for -display spice-app
ui: gtk+sdl bugfixes.

# gpg: Signature made Fri 22 Feb 2019 07:53:13 GMT
# gpg:                using RSA key 4CB6D8EED3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <kraxel@redhat.com>" [full]
# gpg:                 aka "Gerd Hoffmann <gerd@kraxel.org>" [full]
# gpg:                 aka "Gerd Hoffmann (private) <kraxel@gmail.com>" [full]
# Primary key fingerprint: A032 8CFF B93A 17A7 9901  FE7D 4CB6 D8EE D3E8 7138

* remotes/kraxel/tags/ui-20190222-pull-request:
  display: add -display spice-app launching a Spice client
  spice: use a default name for the server
  qapi: document DisplayType enum
  build-sys: add gio-2.0 check
  char: register spice ports after spice started
  char: move SpiceChardev and open_spice_port() to spice.h header
  spice: do not stop spice if VM is paused
  spice: merge options lists
  spice: avoid spice runtime assert
  char/spice: discard write() if backend is disconnected
  char/spice: trigger HUP event
  ui/gtk: Fix the license information
  sdl2: drop qemu_input_event_send_key_qcode call
  spice: set device address and device display ID in QXL interface
  kbd-state: don't block auto-repeat events

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
5 years agoMerge remote-tracking branch 'remotes/awilliam/tags/vfio-updates-20190221.0' into...
Peter Maydell [Fri, 22 Feb 2019 15:48:04 +0000 (15:48 +0000)]
Merge remote-tracking branch 'remotes/awilliam/tags/vfio-updates-20190221.0' into staging

VFIO updates 2019-02-21

 - Workaround kernel overflow bug in vfio type1 DMA unmap
   (Alex Williamson)

 - Refactor vfio container initialization (Eric Auger)

# gpg: Signature made Fri 22 Feb 2019 05:21:07 GMT
# gpg:                using RSA key 239B9B6E3BB08B22
# gpg: Good signature from "Alex Williamson <alex.williamson@redhat.com>" [full]
# gpg:                 aka "Alex Williamson <alex@shazbot.org>" [full]
# gpg:                 aka "Alex Williamson <alwillia@redhat.com>" [full]
# gpg:                 aka "Alex Williamson <alex.l.williamson@gmail.com>" [full]
# Primary key fingerprint: 42F6 C04E 540B D1A9 9E7B  8A90 239B 9B6E 3BB0 8B22

* remotes/awilliam/tags/vfio-updates-20190221.0:
  hw/vfio/common: Refactor container initialization
  vfio/common: Work around kernel overflow bug in DMA unmap

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
5 years agoMerge remote-tracking branch 'remotes/rth/tags/pull-hppa-20190221' into staging
Peter Maydell [Fri, 22 Feb 2019 13:53:12 +0000 (13:53 +0000)]
Merge remote-tracking branch 'remotes/rth/tags/pull-hppa-20190221' into staging

Fix dino pci config access.

# gpg: Signature made Thu 21 Feb 2019 19:03:26 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-20190221:
  hw/hppa/dino: mask out lower 2 bits of PCI config addr

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
5 years agoMerge remote-tracking branch 'remotes/rth/tags/pull-tcg-20190221' into staging
Peter Maydell [Fri, 22 Feb 2019 13:04:42 +0000 (13:04 +0000)]
Merge remote-tracking branch 'remotes/rth/tags/pull-tcg-20190221' into staging

Allow const void * as argument to helpers.
Remove obsolete TODO file.

# gpg: Signature made Thu 21 Feb 2019 18:59:11 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-20190221:
  include/exec/helper-head.h: support "const void *" in helper calls
  tcg: Remove TODO file

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
5 years agoMerge remote-tracking branch 'remotes/amarkovic/tags/mips-queue-feb-21-2019-v2' into...
Peter Maydell [Fri, 22 Feb 2019 11:26:17 +0000 (11:26 +0000)]
Merge remote-tracking branch 'remotes/amarkovic/tags/mips-queue-feb-21-2019-v2' into staging

MIPS queue for February 21st, 2019, v2

# gpg: Signature made Thu 21 Feb 2019 18:37:04 GMT
# gpg:                using RSA key D4972A8967F75A65
# gpg: Good signature from "Aleksandar Markovic <amarkovic@wavecomp.com>" [unknown]
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg:          There is no indication that the signature belongs to the owner.
# Primary key fingerprint: 8526 FBF1 5DA3 811F 4A01  DD75 D497 2A89 67F7 5A65

* remotes/amarkovic/tags/mips-queue-feb-21-2019-v2:
  target/mips: fulong2e: Dynamically generate SPD EEPROM data
  target/mips: fulong2e: Fix bios flash size
  hw/pci-host/bonito.c: Add PCI mem region mapped at the correct address
  target/mips: implement QMP query-cpu-definitions command
  tests/tcg: target/mips: Add wrappers for MSA integer compare instructions
  tests/tcg: target/mips: Change directory name 'bit-counting' to 'bit-count'
  tests/tcg: target/mips: Correct path to headers in some test source files
  hw/misc: mips_itu: Fix 32/64 bit issue in a line involving shift operator

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
5 years agodisplay: add -display spice-app launching a Spice client
Marc-André Lureau [Thu, 21 Feb 2019 11:07:03 +0000 (12:07 +0100)]
display: add -display spice-app launching a Spice client

Add a new display backend that will configure Spice to allow a remote
client to control QEMU in a similar fashion as other QEMU display
backend/UI like GTK.

For this to work, it will set up Spice server with a unix socket, and
register a VC chardev that will be exposed as Spice ports. A QMP
monitor is also exposed as a Spice port, this allows the remote client
fuller qemu control and state handling.

- doesn't handle VC set_echo() - this doesn't seem a strong
  requirement, very few front-end use it
- spice options can be tweaked with other -spice arguments
- Windows support shouldn't be hard to do, but will probably use a TCP
  port instead
- we may want to watch the child process to quit automatically if it
  crashed

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Tested-by: Victor Toso <victortoso@redhat.com>
Message-id: 20190221110703.5775-12-marcandre.lureau@redhat.com

[ kraxel: squash incremental fix ]

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
5 years agodisplay/virtio: add edid support.
Gerd Hoffmann [Thu, 21 Feb 2019 08:10:54 +0000 (09:10 +0100)]
display/virtio: add edid support.

This patch adds EDID support to the family of virtio-gpu devices.  It is
turned off by default, use the new edid property to enable it.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-id: 20190221081054.13853-1-kraxel@redhat.com

5 years agovirtio-gpu: remove useless 'waiting' field
Marc-André Lureau [Thu, 21 Feb 2019 11:43:30 +0000 (12:43 +0100)]
virtio-gpu: remove useless 'waiting' field

Let's check renderer_blocked instead directly.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Christophe Fergeau <cfergeau@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-id: 20190221114330.17968-5-marcandre.lureau@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
5 years agovirtio-gpu: block both 2d and 3d rendering
Marc-André Lureau [Thu, 21 Feb 2019 11:43:29 +0000 (12:43 +0100)]
virtio-gpu: block both 2d and 3d rendering

Now that 2d commands are translated to 3d rendering, qemu must stop
sending 3d updates (from 2d) to Spice as well.

Fixes:
https://bugzilla.redhat.com/show_bug.cgi?id=1674324

Cc: cfergeau@redhat.com
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Christophe Fergeau <cfergeau@redhat.com>
Tested-by: Christophe Fergeau <cfergeau@redhat.com>
Message-id: 20190221114330.17968-4-marcandre.lureau@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
5 years agovirtio-gpu: remove unused config_size
Marc-André Lureau [Thu, 21 Feb 2019 11:43:28 +0000 (12:43 +0100)]
virtio-gpu: remove unused config_size

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Christophe Fergeau <cfergeau@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-id: 20190221114330.17968-3-marcandre.lureau@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
5 years agovirtio-gpu: remove unused qdev
Marc-André Lureau [Thu, 21 Feb 2019 11:43:27 +0000 (12:43 +0100)]
virtio-gpu: remove unused qdev

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Christophe Fergeau <cfergeau@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-id: 20190221114330.17968-2-marcandre.lureau@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
5 years agohw/vfio/common: Refactor container initialization
Eric Auger [Fri, 22 Feb 2019 04:07:03 +0000 (21:07 -0700)]
hw/vfio/common: Refactor container initialization

We introduce the vfio_init_container_type() helper.
It computes the highest usable iommu type and then
set the container and the iommu type.

Its usage in vfio_connect_container() makes the code
ready for addition of new iommu types.

Signed-off-by: Eric Auger <eric.auger@redhat.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
5 years agovfio/common: Work around kernel overflow bug in DMA unmap
Alex Williamson [Fri, 22 Feb 2019 04:07:03 +0000 (21:07 -0700)]
vfio/common: Work around kernel overflow bug in DMA unmap

A kernel bug was introduced in v4.15 via commit 71a7d3d78e3c which
adds a test for address space wrap-around in the vfio DMA unmap path.
Unfortunately due to overflow, the kernel detects an unmap of the last
page in the 64-bit address space as a wrap-around.  In QEMU, a Q35
guest with VT-d emulation and guest IOMMU enabled will attempt to make
such an unmap request during VM system reset, triggering an error:

  qemu-kvm: VFIO_UNMAP_DMA: -22
  qemu-kvm: vfio_dma_unmap(0x561f059948f0, 0xfef00000, 0xffffffff01100000) = -22 (Invalid argument)

Here the IOVA start address (0xfef00000) and the size parameter
(0xffffffff01100000) add to exactly 2^64, triggering the bug.  A
kernel fix is queued for the Linux v5.0 release to address this.

This patch implements a workaround to retry the unmap, excluding the
final page of the range when we detect an unmap failing which matches
the requirements for this issue.  This is expected to be a safe and
complete workaround as the VT-d address space does not extend to the
full 64-bit space and therefore the last page should never be mapped.

This workaround can be removed once all kernels with this bug are
sufficiently deprecated.

Link: https://bugzilla.redhat.com/show_bug.cgi?id=1662291
Reported-by: Pei Zhang <pezhang@redhat.com>
Debugged-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
5 years agoMerge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20190221' into...
Peter Maydell [Thu, 21 Feb 2019 18:58:35 +0000 (18:58 +0000)]
Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20190221' into staging

target-arm queue:
 * Model the Arm "Musca" development boards: "musca-a" and "musca-b1"
 * Implement the ARMv8.3-JSConv extension
 * v8M MPU should use background region as default, not always
 * Stop unintentional sign extension in pmu_init

# gpg: Signature made Thu 21 Feb 2019 18:56:32 GMT
# gpg:                using RSA key E1A5C593CD419DE28E8315CF3C2525ED14360CDE
# gpg:                issuer "peter.maydell@linaro.org"
# gpg: Good signature from "Peter Maydell <peter.maydell@linaro.org>" [ultimate]
# gpg:                 aka "Peter Maydell <pmaydell@gmail.com>" [ultimate]
# gpg:                 aka "Peter Maydell <pmaydell@chiark.greenend.org.uk>" [ultimate]
# Primary key fingerprint: E1A5 C593 CD41 9DE2 8E83  15CF 3C25 25ED 1436 0CDE

* remotes/pmaydell/tags/pull-target-arm-20190221: (21 commits)
  hw/arm/armsse: Make 0x5... alias region work for per-CPU devices
  hw/arm/musca: Wire up PL011 UARTs
  hw/arm/musca: Wire up PL031 RTC
  hw/arm/musca: Add MPCs
  hw/arm/musca: Add PPCs
  hw/arm/musca.c: Implement models of the Musca-A and -B1 boards
  hw/arm/armsse: Allow boards to specify init-svtor
  hw/arm/armsse: Document SRAM_ADDR_WIDTH property in header comment
  hw/char/pl011: Use '0x' prefix when logging hex numbers
  hw/char/pl011: Support all interrupt lines
  hw/char/pl011: Allow use as an embedded-struct device
  hw/timer/pl031: Convert to using trace events
  hw/timer/pl031: Allow use as an embedded-struct device
  hw/misc/tz-ppc: Support having unused ports in the middle of the range
  target/arm: Implement ARMv8.3-JSConv
  target/arm: Rearrange Floating-point data-processing (2 regs)
  target/arm: Split out vfp_helper.c
  target/arm: Restructure disas_fp_int_conv
  target/arm: Stop unintentional sign extension in pmu_init
  target/arm: v8M MPU should use background region as default, not always
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
5 years agotarget/mips: fulong2e: Dynamically generate SPD EEPROM data
BALATON Zoltan [Thu, 21 Feb 2019 12:34:07 +0000 (13:34 +0100)]
target/mips: fulong2e: Dynamically generate SPD EEPROM data

The machine comes with 256M memory module by default but it's
upgradable so it could have different memory size. There was a TODO
comment to replace static SPD EEPROM data with dynamically generated
one to support this. Now that we have a function for that, it's easy
to do. Although this would allow larger RAM sizes, the peculiar memory
map of the machine may need some special handling to map it as low and
high memory. Because I don't know what the correct place would be for
highmem, I've left memory size fixed at 256M for now and TODO is moved
there instead.

Signed-off-by: BALATON Zoltan <balaton@eik.bme.hu>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
5 years agotarget/mips: fulong2e: Fix bios flash size
BALATON Zoltan [Thu, 21 Feb 2019 12:29:15 +0000 (13:29 +0100)]
target/mips: fulong2e: Fix bios flash size

According to both the specifications on linux-mips.org referenced in a
comment at the beginning of the file and the flash chip part number
the bios size should be 512k not 1M.

Signed-off-by: BALATON Zoltan <balaton@eik.bme.hu>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
5 years agohw/pci-host/bonito.c: Add PCI mem region mapped at the correct address
BALATON Zoltan [Thu, 21 Feb 2019 12:25:00 +0000 (13:25 +0100)]
hw/pci-host/bonito.c: Add PCI mem region mapped at the correct address

Stop using system memory as PCI memory otherwise devices such as VGA
that have regions mapped to PCI memory clash with RAM. Use a separate
memory region for PCI memory and map it to the correct address in
system memory which allows PCI mem regions to show at the correct
address where clients expect them.

Signed-off-by: BALATON Zoltan <balaton@eik.bme.hu>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
5 years agotarget/mips: implement QMP query-cpu-definitions command
Pavel Dovgalyuk [Tue, 19 Feb 2019 17:02:55 +0000 (18:02 +0100)]
target/mips: implement QMP query-cpu-definitions command

This patch enables QMP-based querying of the available CPU types for
MIPS and MIPS64 platforms.

Signed-off-by: Pavel Dovgalyuk <pavel.dovgaluk@ispras.ru>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
5 years agotests/tcg: target/mips: Add wrappers for MSA integer compare instructions
Aleksandar Markovic [Wed, 13 Feb 2019 17:04:12 +0000 (18:04 +0100)]
tests/tcg: target/mips: Add wrappers for MSA integer compare instructions

Add wrappers for MSA integer compare instructions.

Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Aleksandar Rikalo <arikalo@wavecomp.com>
5 years agotests/tcg: target/mips: Change directory name 'bit-counting' to 'bit-count'
Aleksandar Markovic [Mon, 18 Feb 2019 06:57:47 +0000 (07:57 +0100)]
tests/tcg: target/mips: Change directory name 'bit-counting' to 'bit-count'

Change directory name 'bit-counting' to 'bit-count'. This is just for
cosmetic and consistency sake. This was the only subdirectory in MSA
test directory that uses ending 'ing'.

Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
5 years agotests/tcg: target/mips: Correct path to headers in some test source files
Aleksandar Markovic [Mon, 18 Feb 2019 06:55:56 +0000 (07:55 +0100)]
tests/tcg: target/mips: Correct path to headers in some test source files

Correct path to headers in tests/tcg/mips/user/ase/msa/bit-counting/*
source files.

Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Aleksandar Rikalo <arikalo@wavecomp.com>
5 years agohw/misc: mips_itu: Fix 32/64 bit issue in a line involving shift operator
Aleksandar Markovic [Mon, 18 Feb 2019 07:32:48 +0000 (08:32 +0100)]
hw/misc: mips_itu: Fix 32/64 bit issue in a line involving shift operator

Fix 32/64 bit issue in a line involving shift operator. "1 << ..."
calculation of size is done as a 32-bit signed integer which may
then be unintentionally sign-extended into the 64-bit result. The
problem was discovered by Coverity (CID 1398648). Using "1ULL"
instead of "1" on the LHS of the shift fixes this problem.

Reported-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
5 years agoinclude/exec/helper-head.h: support "const void *" in helper calls
David Hildenbrand [Thu, 21 Feb 2019 09:34:59 +0000 (10:34 +0100)]
include/exec/helper-head.h: support "const void *" in helper calls

Especially when dealing with out-of-line gvec helpers, it is often
helpful to specify some vector pointers as constant. E.g. when
we have two inputs and one output, marking the two inputs as consts
pointers helps to avoid bugs.

Const pointers can be specified via "cptr", however behave in TCG just
like ordinary pointers. We can specify helpers like:

DEF_HELPER_FLAGS_4(gvec_vbperm, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32)

void HELPER(gvec_vbperm)(void *v1, const void *v2, const void *v3,
                         uint32_t desc)

And make sure that here, only v1 will be written (as long as const is
not casted away, of course).

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20190221093459.22547-1-david@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
5 years agotcg: Remove TODO file
Richard Henderson [Wed, 20 Feb 2019 17:29:06 +0000 (09:29 -0800)]
tcg: Remove TODO file

The last update to this file was 9 years ago.  In the meantime,
4 of the 6 ideas have actually been completed.  The lat two do
not actually make sense anymore.

Suggested-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
5 years agohw/arm/armsse: Make 0x5... alias region work for per-CPU devices
Peter Maydell [Thu, 21 Feb 2019 18:17:48 +0000 (18:17 +0000)]
hw/arm/armsse: Make 0x5... alias region work for per-CPU devices

The region 0x40010000 .. 0x4001ffff and its secure-only alias
at 0x50010000... are for per-CPU devices. We implement this by
giving each CPU its own container memory region, where the
per-CPU devices live. Unfortunately, the alias region which
makes devices mapped at 0x4... addresses also appear at 0x5...
is only implemented in the overall "all CPUs" container. The
effect of this bug is that the CPU_IDENTITY register block appears
only at 0x4001f000, but not at the 0x5001f000 alias where it should
also appear. Guests (like very recent Arm Trusted Firmware-M)
which try to access it at 0x5001f000 will crash.

Fix this by moving the handling for this alias from the "all CPUs"
container to the per-CPU container. (We leave the aliases for
0x1... and 0x3... in the overall container, because there are
no per-CPU devices there.)

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20190215180500.6906-1-peter.maydell@linaro.org
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
5 years agohw/arm/musca: Wire up PL011 UARTs
Peter Maydell [Thu, 21 Feb 2019 18:17:47 +0000 (18:17 +0000)]
hw/arm/musca: Wire up PL011 UARTs

Wire up the two PL011 UARTs in the Musca board.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
5 years agohw/arm/musca: Wire up PL031 RTC
Peter Maydell [Thu, 21 Feb 2019 18:17:47 +0000 (18:17 +0000)]
hw/arm/musca: Wire up PL031 RTC

Wire up the PL031 RTC for the Musca board.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
5 years agohw/arm/musca: Add MPCs
Peter Maydell [Thu, 21 Feb 2019 18:17:47 +0000 (18:17 +0000)]
hw/arm/musca: Add MPCs

The Musca board puts its SRAM and flash behind TrustZone
Memory Protection Controllers (MPCs). Each MPC sits between
the CPU and the RAM/flash, and also has a set of memory mapped
control registers. Wire up the MPCs, and the memory behind them.
For the moment we implement the flash as simple ROM, which
cannot be reprogrammed by the guest.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
5 years agohw/arm/musca: Add PPCs
Peter Maydell [Thu, 21 Feb 2019 18:17:47 +0000 (18:17 +0000)]
hw/arm/musca: Add PPCs

Many of the devices on the Musca board live behind TrustZone
Peripheral Protection Controllers (PPCs); add models of the
PPCs, using a similar scheme to the MPS2 board models.
This commit wires up the PPCs with "unimplemented device"
stubs behind them in the correct places in the address map.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
5 years agohw/arm/musca.c: Implement models of the Musca-A and -B1 boards
Peter Maydell [Thu, 21 Feb 2019 18:17:47 +0000 (18:17 +0000)]
hw/arm/musca.c: Implement models of the Musca-A and -B1 boards

The Musca-A and Musca-B1 development boards are based on the
SSE-200 subsystem for embedded. Implement an initial skeleton
model of these boards, which are similar but not identical.

This commit creates the board model with the SSE and the IRQ
splitters to wire IRQs up to its two CPUs. As yet there
are no devices and no memory: these will be added later.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
5 years agohw/arm/armsse: Allow boards to specify init-svtor
Peter Maydell [Thu, 21 Feb 2019 18:17:47 +0000 (18:17 +0000)]
hw/arm/armsse: Allow boards to specify init-svtor

The Musca boards have DAPLink firmware that sets the initial
secure VTOR value (the location of the vector table) differently
depending on the boot mode (from flash, from RAM, etc). Export
the init-svtor as a QOM property of the ARMSSE object so that
the board can change it.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
5 years agohw/arm/armsse: Document SRAM_ADDR_WIDTH property in header comment
Peter Maydell [Thu, 21 Feb 2019 18:17:47 +0000 (18:17 +0000)]
hw/arm/armsse: Document SRAM_ADDR_WIDTH property in header comment

In commit 4b635cf7a95e501211 we added a QOM property to the ARMSSE
object, but forgot to add it to the documentation comment in the
header. Correct the omission.

Fixes: 4b635cf7a95e501211 ("hw/arm/armsse: Make SRAM bank size configurable")
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
5 years agohw/char/pl011: Use '0x' prefix when logging hex numbers
Peter Maydell [Thu, 21 Feb 2019 18:17:46 +0000 (18:17 +0000)]
hw/char/pl011: Use '0x' prefix when logging hex numbers

The pl011 logs when the guest makes a bad access. It prints
the address offset in hex but confusingly omits the '0x'
prefix; add it.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
5 years agohw/char/pl011: Support all interrupt lines
Peter Maydell [Thu, 21 Feb 2019 18:17:46 +0000 (18:17 +0000)]
hw/char/pl011: Support all interrupt lines

The PL011 UART has six interrupt lines:
 * RX (receive data)
 * TX (transmit data)
 * RT (receive timeout)
 * MS (modem status)
 * E (errors)
 * combined (logical OR of all the above)

So far we have only emulated the combined interrupt line;
add support for the others, so that boards that wire them
up to different interrupt controller inputs can do so.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
5 years agohw/char/pl011: Allow use as an embedded-struct device
Peter Maydell [Thu, 21 Feb 2019 18:17:46 +0000 (18:17 +0000)]
hw/char/pl011: Allow use as an embedded-struct device

Create a new include file for the pl011's device struct,
type macros, etc, so that it can be instantiated using
the "embedded struct" coding style.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
5 years agohw/timer/pl031: Convert to using trace events
Peter Maydell [Thu, 21 Feb 2019 18:17:46 +0000 (18:17 +0000)]
hw/timer/pl031: Convert to using trace events

Convert the debug printing in the PL031 device to use trace events,
and augment it to cover the interesting parts of device operation.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
5 years agohw/timer/pl031: Allow use as an embedded-struct device
Peter Maydell [Thu, 21 Feb 2019 18:17:46 +0000 (18:17 +0000)]
hw/timer/pl031: Allow use as an embedded-struct device

Create a new include file for the pl031's device struct,
type macros, etc, so that it can be instantiated using
the "embedded struct" coding style.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
5 years agohw/misc/tz-ppc: Support having unused ports in the middle of the range
Peter Maydell [Thu, 21 Feb 2019 18:17:46 +0000 (18:17 +0000)]
hw/misc/tz-ppc: Support having unused ports in the middle of the range

The Peripheral Protection Controller's handling of unused ports
is that if there is nothing connected to the port's downstream
then it does not create the sysbus MMIO region for the upstream
end of the port. This results in odd behaviour when there is
an unused port in the middle of the range: since sysbus MMIO
regions are implicitly consecutively allocated, any used ports
above the unused ones end up with sysbus MMIO region numbers
that don't match the port number.

Avoid this numbering mismatch by creating dummy MMIO regions
for the unused ports. This doesn't change anything for our
existing boards, which don't have any gaps in the middle of
the port ranges they use; but it will be needed for the Musca
board.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
5 years agotarget/arm: Implement ARMv8.3-JSConv
Richard Henderson [Thu, 21 Feb 2019 18:17:46 +0000 (18:17 +0000)]
target/arm: Implement ARMv8.3-JSConv

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20190215192302.27855-5-richard.henderson@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
[PMM: fixed a couple of comment typos]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
5 years agotarget/arm: Rearrange Floating-point data-processing (2 regs)
Richard Henderson [Thu, 21 Feb 2019 18:17:45 +0000 (18:17 +0000)]
target/arm: Rearrange Floating-point data-processing (2 regs)

There are lots of special cases within these insns.  Split the
major argument decode/loading/saving into no_output (compares),
rd_is_dp, and rm_is_dp.

We still need to special case argument load for compare (rd as
input, rm as zero) and vcvt fixed (rd as input+output), but lots
of special cases do disappear.

Now that we have a full switch at the beginning, hoist the ISA
checks from the code generation.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20190215192302.27855-4-richard.henderson@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
5 years agotarget/arm: Split out vfp_helper.c
Richard Henderson [Thu, 21 Feb 2019 18:17:45 +0000 (18:17 +0000)]
target/arm: Split out vfp_helper.c

Move all of the fp helpers out of helper.c into a new file.
This is code movement only.  Since helper.c has no copyright
header, take the one from cpu.h for the new file.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20190215192302.27855-3-richard.henderson@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
5 years agotarget/arm: Restructure disas_fp_int_conv
Richard Henderson [Thu, 21 Feb 2019 18:17:45 +0000 (18:17 +0000)]
target/arm: Restructure disas_fp_int_conv

For opcodes 0-5, move some if conditions into the structure
of a switch statement.  For opcodes 6 & 7, decode everything
at once with a second switch.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20190215192302.27855-2-richard.henderson@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
5 years agotarget/arm: Stop unintentional sign extension in pmu_init
Aaron Lindsay OS [Thu, 21 Feb 2019 18:17:45 +0000 (18:17 +0000)]
target/arm: Stop unintentional sign extension in pmu_init

This was introduced by
    commit bf8d09694ccc07487cd73d7562081fdaec3370c8
    target/arm: Don't clear supported PMU events when initializing PMCEID1
and identified by Coverity (CID 1398645).

Signed-off-by: Aaron Lindsay <aaron@os.amperecomputing.com>
Reported-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20190219144621.450-1-aaron@os.amperecomputing.com
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
5 years agotarget/arm: v8M MPU should use background region as default, not always
Peter Maydell [Thu, 21 Feb 2019 18:17:45 +0000 (18:17 +0000)]
target/arm: v8M MPU should use background region as default, not always

The "background region" for a v8M MPU is a default which will be used
(if enabled, and if the access is privileged) if the access does
not match any specific MPU region. We were incorrectly using it
always (by putting the condition at the wrong nesting level). This
meant that we would always return the default background permissions
rather than the correct permissions for a specific region, and also
that we would not return the right information in response to a
TT instruction.

Move the check for the background region to the same place in the
logic as the equivalent v8M MPUCheck() pseudocode puts it.
This in turn means we must adjust the condition we use to detect
matches in multiple regions to avoid false-positives.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20190214113408.10214-1-peter.maydell@linaro.org

5 years agohw/arm/armsse: Fix memory leak in error-exit path
Peter Maydell [Thu, 21 Feb 2019 18:17:45 +0000 (18:17 +0000)]
hw/arm/armsse: Fix memory leak in error-exit path

Coverity points out (CID 1398632, CID 1398650) that we
leak a couple of allocated strings in the error-exit
code path for setting up the MHUs in the ARMSSE.
Fix this bug by moving the allocate-and-free of each
string to be closer to the use, so we do the free before
doing the error-exit check.

Fixes: f8574705f62b38a ("hw/arm/armsse: Add unimplemented-device stubs for MHUs")
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-id: 20190215113707.24553-1-peter.maydell@linaro.org

5 years agohw/hppa/dino: mask out lower 2 bits of PCI config addr
Sven Schnelle [Mon, 18 Feb 2019 18:33:14 +0000 (19:33 +0100)]
hw/hppa/dino: mask out lower 2 bits of PCI config addr

some versions of HP-UX 10.20 seems to rely on the fact that DINO
strips out the lower 2 bits of the PCI configuration address.
Also update the binary SeaBIOS distributed to the latest version
from Helge's repository, which is required with that change.

Signed-off-by: Sven Schnelle <svens@stackframe.org>
Message-Id: <20190218183314.20157-1-svens@stackframe.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
5 years agoMerge remote-tracking branch 'remotes/jnsnow/tags/bitmaps-pull-request' into staging
Peter Maydell [Thu, 21 Feb 2019 13:09:33 +0000 (13:09 +0000)]
Merge remote-tracking branch 'remotes/jnsnow/tags/bitmaps-pull-request' into staging

Pull request

# gpg: Signature made Wed 20 Feb 2019 18:01:00 GMT
# gpg:                using RSA key F9B7ABDBBCACDF95BE76CBD07DEF8106AAFC390E
# gpg: Good signature from "John Snow (John Huston) <jsnow@redhat.com>" [full]
# Primary key fingerprint: FAEB 9711 A12C F475 812F  18F2 88A9 064D 1835 61EB
#      Subkey fingerprint: F9B7 ABDB BCAC DF95 BE76  CBD0 7DEF 8106 AAFC 390E

* remotes/jnsnow/tags/bitmaps-pull-request:
  blockdev: acquire aio_context for bitmap add/remove
  block/dirty-bitmap: Documentation and Comment fixups
  dirty-bitmap: Expose persistent flag to 'query-block'

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
5 years agospice: use a default name for the server
Marc-André Lureau [Thu, 21 Feb 2019 11:07:02 +0000 (12:07 +0100)]
spice: use a default name for the server

If no -name is given, let's use a friendly "QEMU version" server
name. This is sometime exposed on spice client side, for example on
remote-viewer title.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Tested-by: Victor Toso <victortoso@redhat.com>
Message-id: 20190221110703.5775-11-marcandre.lureau@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
5 years agoqapi: document DisplayType enum
Marc-André Lureau [Thu, 21 Feb 2019 11:07:01 +0000 (12:07 +0100)]
qapi: document DisplayType enum

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Tested-by: Victor Toso <victortoso@redhat.com>
Message-id: 20190221110703.5775-10-marcandre.lureau@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
5 years agobuild-sys: add gio-2.0 check
Marc-André Lureau [Thu, 21 Feb 2019 11:07:00 +0000 (12:07 +0100)]
build-sys: add gio-2.0 check

GIO is required for the "-display spice-app" backend.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Tested-by: Victor Toso <victortoso@redhat.com>
Message-id: 20190221110703.5775-9-marcandre.lureau@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
5 years agochar: register spice ports after spice started
Marc-André Lureau [Thu, 21 Feb 2019 11:06:59 +0000 (12:06 +0100)]
char: register spice ports after spice started

Spice port registration is delayed until the server is started. But
ports created after are not being registered. If the server is already
started, do vmc_register_interface() to register it from
qemu_chr_open_spice_port().

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Tested-by: Victor Toso <victortoso@redhat.com>
Message-id: 20190221110703.5775-8-marcandre.lureau@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
5 years agochar: move SpiceChardev and open_spice_port() to spice.h header
Marc-André Lureau [Thu, 21 Feb 2019 11:06:58 +0000 (12:06 +0100)]
char: move SpiceChardev and open_spice_port() to spice.h header

This will allow easier subclassing of SpiceChardev, in upcoming
"display: add -display spice-app launching external application"
patch.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Tested-by: Victor Toso <victortoso@redhat.com>
Message-id: 20190221110703.5775-7-marcandre.lureau@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
5 years agospice: do not stop spice if VM is paused
Marc-André Lureau [Thu, 21 Feb 2019 11:06:57 +0000 (12:06 +0100)]
spice: do not stop spice if VM is paused

spice_server_vm_start/stop() was added to help migration state (commit
f5bb039c6d97ef3e664094eab3c9a4dc1824ed73).

However, a paused VM could keep running the spice server. This will
allow a Spice client to keep sending commands to a spice chardev. This
allows to stop/cont a VM from a Spice monitor port. Character
devices (vdagent/usb/smartcard/..) should not read from Spice when the
VM is paused.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Tested-by: Victor Toso <victortoso@redhat.com>
Message-id: 20190221110703.5775-6-marcandre.lureau@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
5 years agospice: merge options lists
Marc-André Lureau [Thu, 21 Feb 2019 11:06:56 +0000 (12:06 +0100)]
spice: merge options lists

Passing several -spice options to qemu command line, or calling
several time qemu_opts_set() will ignore all but the first option
list. Since the spice server is a singleton, it makes sense to merge
all the options, the last value being the one taken into account.

This changes the behaviour from, for ex:
$ qemu... -spice port=5900 -spice port=5901 -> port: 5900
to:
$ qemu... -spice port=5900 -spice port=5901 -> port: 5901

(if necessary we could instead produce an error when an option is
given twice, although this makes handling default values and such more
complicated)

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Tested-by: Victor Toso <victortoso@redhat.com>
Message-id: 20190221110703.5775-5-marcandre.lureau@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
5 years agospice: avoid spice runtime assert
Marc-André Lureau [Thu, 21 Feb 2019 11:06:55 +0000 (12:06 +0100)]
spice: avoid spice runtime assert

The Spice server doesn't like to be started or stopped twice . It
aborts with:

(process:6191): Spice-ERROR **: 19:29:35.912: red-worker.c:623:handle_dev_start: assertion `!worker->running' failed

It's easy to avoid that situation since qemu spice_display_is_running
tracks the server state.

After the commit "spice: do not stop spice if VM is paused", it will
be possible to pause and resume the VM, and this will call
qemu_spice_display_start() twice. The easiest is to add a check for
spice_display_is_running with this patch to avoid the assert.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Tested-by: Victor Toso <victortoso@redhat.com>
Message-id: 20190221110703.5775-4-marcandre.lureau@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
5 years agochar/spice: discard write() if backend is disconnected
Marc-André Lureau [Thu, 21 Feb 2019 11:06:54 +0000 (12:06 +0100)]
char/spice: discard write() if backend is disconnected

Most chardev backend handle write() as discarded data if underlying
system is disconnected. For unknown historical reasons, the Spice
backend has "reliable" write: it will wait until the client end is
reconnected to do further successful write().

To decide whether it make sense to wait until the client is
reconnected (or queue the writes), let's review Spice chardev usage
and handling of a disconnected client:

 * spice vdagent
   The agents reopen the virtio port on disconnect. In qemu side,
   virtio_serial_close() will also discard pending data.

 * usb redirection
   A disconnect creates a device disconnection.

 * smartcard emulation
   Data is discarded in passthru_apdu_from_guest().

   (Spice doesn't explicitly open the smartcard char device until
   upcoming 0.14.2, commit 69a5cfc74131ec0459f2eb5a231139f5a69a8037)

 * spice webdavd
   The daemon will restart the service, and reopen the virtio port.

 * spice ports (serial console, qemu monitor..)
   Depends on the associated device or usage.

   - serial, may be throttled or discarded on write, depending on
     device

   - QMP/HMP monitor have some CLOSED event handling, but want to
     flush the write, which will finish when a new client connects.

On disconnect/reconnect, the client starts with fresh sessions. If it
is a seamless migration, the client disconnects after the source
migrated. The handling of source disconnect in qemu is thus irrelevant
for the Spice session migration.

For all these use cases, it is better to discard writes when the
client is disconnected, and require the vm-side device/agent to behave
correctly on CHR_EVENT_CLOSED, to stop reading and writing from
the spice chardev.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Tested-by: Victor Toso <victortoso@redhat.com>
Message-id: 20190221110703.5775-3-marcandre.lureau@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
5 years agochar/spice: trigger HUP event
Marc-André Lureau [Thu, 21 Feb 2019 11:06:53 +0000 (12:06 +0100)]
char/spice: trigger HUP event

Inform the front-end of disconnected state (spice client
disconnected).

This will wakeup the source handler immediately, so it can detect the
disconnection asap.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Tested-by: Victor Toso <victortoso@redhat.com>
Message-id: 20190221110703.5775-2-marcandre.lureau@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
5 years agoui/gtk: Fix the license information
Thomas Huth [Thu, 21 Feb 2019 06:51:42 +0000 (07:51 +0100)]
ui/gtk: Fix the license information

The license information in this file is very messy. A short note at
the beginning says GPL first, but the long boilerplate code then
talks about "GNU Lesser General Public License version 2.0". First,
there is no such version of the "GNU Lesser GPL", it only started with
version 2.1. In version 2.0, it was still called "GNU Library GPL"
instead. Second, you can easily get the license of this file wrong
if you only quickly glance at the long boilerplate code.

Anyway, looking at the text of the LGPL (see COPYING.LIB in the top
directory), the license clearly states in section "3." that one should
rather replace the license information with the GPL information in
such a case of a mixture instead. Thus let's clean up the confusing
statements and use the proper GPL text only.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-id: 1550731902-28842-1-git-send-email-thuth@redhat.com

[ kraxel: s/v2/v2+/ as requested by Daniel ]

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
5 years agosdl2: drop qemu_input_event_send_key_qcode call
Gerd Hoffmann [Fri, 8 Feb 2019 07:27:44 +0000 (08:27 +0100)]
sdl2: drop qemu_input_event_send_key_qcode call

qkbd_state_key_event() does that for us.

Fixes: 07333e1ca3 kbd-state: use state tracker for sdl2
Reported-by: BALATON Zoltan <balaton@eik.bme.hu>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Tested-by: BALATON Zoltan <balaton@eik.bme.hu>
Message-id: 20190208072744.10687-1-kraxel@redhat.com

5 years agoMerge remote-tracking branch 'remotes/kraxel/tags/usb-20190220-pull-request' into...
Peter Maydell [Thu, 21 Feb 2019 09:41:11 +0000 (09:41 +0000)]
Merge remote-tracking branch 'remotes/kraxel/tags/usb-20190220-pull-request' into staging

usb: usb_ep_get() fixes

# gpg: Signature made Wed 20 Feb 2019 11:13:32 GMT
# gpg:                using RSA key 4CB6D8EED3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <kraxel@redhat.com>" [full]
# gpg:                 aka "Gerd Hoffmann <gerd@kraxel.org>" [full]
# gpg:                 aka "Gerd Hoffmann (private) <kraxel@gmail.com>" [full]
# Primary key fingerprint: A032 8CFF B93A 17A7 9901  FE7D 4CB6 D8EE D3E8 7138

* remotes/kraxel/tags/usb-20190220-pull-request:
  usb: remove unnecessary NULL device check from usb_ep_get()
  usb: add device checks before redirector calls to usb_ep_get()
  usb: check device is not NULL before calling usb_ep_get()
  uhci: check device is not NULL before calling usb_ep_get()
  ohci: check device is not NULL before calling usb_ep_get()
  ehci: check device is not NULL before calling usb_ep_get()
  xhci: check device is not NULL before calling usb_ep_get()
  xhci: add asserts to help with static code analysis
  usb: rearrange usb_ep_get()

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
5 years agospice: set device address and device display ID in QXL interface
Lukáš Hrázký [Fri, 15 Feb 2019 15:09:19 +0000 (16:09 +0100)]
spice: set device address and device display ID in QXL interface

Calls the new SPICE QXL interface function spice_qxl_set_device_info to
set the hardware address of the graphics device represented by the QXL
interface (e.g. a PCI path) and the device display IDs (the IDs of the
device's monitors that belong to this QXL interface).

Also stops using the deprecated spice_qxl_set_max_monitors, the new
interface function replaces it.

Signed-off-by: Lukáš Hrázký <lhrazky@redhat.com>
Message-Id: <20190215150919.8263-1-lhrazky@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
5 years agokbd-state: don't block auto-repeat events
Gerd Hoffmann [Wed, 20 Feb 2019 10:02:35 +0000 (11:02 +0100)]
kbd-state: don't block auto-repeat events

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-id: 20190220100235.20914-1-kraxel@redhat.com

5 years agousb: remove unnecessary NULL device check from usb_ep_get()
Liam Merwick [Wed, 6 Feb 2019 13:36:56 +0000 (13:36 +0000)]
usb: remove unnecessary NULL device check from usb_ep_get()

No caller of usb_ep_get() calls it with a NULL device (previous commits
have addressed the few remaining cases which didn't explicitly check).
Replace check for 'dev == NULL' with an assert instead.

Signed-off-by: Liam Merwick <liam.merwick@oracle.com>
Message-id: 1549460216-25808-10-git-send-email-liam.merwick@oracle.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
5 years agousb: add device checks before redirector calls to usb_ep_get()
Liam Merwick [Wed, 6 Feb 2019 13:36:55 +0000 (13:36 +0000)]
usb: add device checks before redirector calls to usb_ep_get()

Add an assert and an explicit check before the two callers to
usb_ep_get() in the USB redirector code to ensure the device
passed in is not NULL.

Signed-off-by: Liam Merwick <liam.merwick@oracle.com>
Message-id: 1549460216-25808-9-git-send-email-liam.merwick@oracle.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
5 years agousb: check device is not NULL before calling usb_ep_get()
Liam Merwick [Wed, 6 Feb 2019 13:36:54 +0000 (13:36 +0000)]
usb: check device is not NULL before calling usb_ep_get()

In musb_packet(), the call to usb_find_device() can return NULL
if it doesn't find a device matching 'addr' so explicitly check
the return value before passing it to usb_ep_get().  This then
allows the subsequent calculation of 'id' to be streamlined.

Signed-off-by: Liam Merwick <liam.merwick@oracle.com>
Message-id: 1549460216-25808-8-git-send-email-liam.merwick@oracle.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
5 years agouhci: check device is not NULL before calling usb_ep_get()
Liam Merwick [Wed, 6 Feb 2019 13:36:53 +0000 (13:36 +0000)]
uhci: check device is not NULL before calling usb_ep_get()

In uhci_handle_td(), the call to ehci_find_device() can return NULL
if it doesn't find a device matching 'addr' so explicitly check
the return value before passing it to usb_ep_get().

Signed-off-by: Liam Merwick <liam.merwick@oracle.com>
Message-id: 1549460216-25808-7-git-send-email-liam.merwick@oracle.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>