Skip to content

Q
qemu

openeuler / qemu

通知 10
Star 0
Fork 0
Q
qemu

体验新版 GitCode，发现更多精彩内容 >>

切换分支/标签
  1. 14 11月, 2015 4 次提交
  3. 12 11月, 2015 8 次提交
  5. 11 11月, 2015 9 次提交
  7. 10 11月, 2015 4 次提交
    • E
      qapi: More tests of input arrays · 2533377c
      由 Eric Blake 提交于 
      Our testsuite had no coverage of empty arrays, nor of what
happens when the input does not match the expected type.
Useful to have, especially if we start changing the visitor
contracts.

I did not think it worth duplicating these additions to
test-qmp-input-strict; since all strict mode does is add
the ability to reject JSON input that has more keys than
what the visitor expects, yet the additions in this patch
error out earlier than that point regardless of whether
strict mode was requested.
Signed-off-by: NEric Blake <eblake@redhat.com>
Message-Id: <1446791754-23823-11-git-send-email-eblake@redhat.com>
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
      2533377c
    • E
      qapi: Test failure in middle of array parse · dd5ee2c2
      由 Eric Blake 提交于 
      Our generated list visitors have the same problem as has been
mentioned elsewhere (see commit 2f52e205): they allocate data
even on failure. An upcoming patch will correct things to
provide saner guarantees, but first we need to expose the
behavior in the testsuite to ensure we aren't introducing any
memory usage bugs.

There are more test cases throughout the test-qmp-input-* tests
that already deal with partial allocation; a later commit will
clean up all visit_type_FOO(), without marking all of the tests
with FIXME at this time.
Signed-off-by: NEric Blake <eblake@redhat.com>
Message-Id: <1446791754-23823-10-git-send-email-eblake@redhat.com>
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
      dd5ee2c2
    • E
      qapi: More tests of alternate output · 12fafd7c
      由 Eric Blake 提交于 
      The testsuite was only covering that we could output the 'int'
branch of an alternate (no additional allocation/cleanup required).
Add a test of the 'str' branch, to make sure that things still
work even when a branch involves allocation.

Update to modern style of g_new0() over g_malloc0() while
touching it.
Signed-off-by: NEric Blake <eblake@redhat.com>
Message-Id: <1446791754-23823-9-git-send-email-eblake@redhat.com>
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
      12fafd7c
    • E
      qapi: Simplify error cleanup in test-qmp-* · a12a5a1a
      由 Eric Blake 提交于 
      We have several tests that perform multiple sub-actions that are
expected to fail.  Asserting that an error occurred, then clearing
it up to prepare for the next action, turned into enough
boilerplate that it was sometimes forgotten (for example, a number
of tests added to test-qmp-input-visitor.c in d88f5fd1 leaked err).
Worse, if an error is not reset to NULL, we risk invalidating
later use of that error (passing a non-NULL err into a function
is generally a bad idea).  Encapsulate the boilerplate into a
single helper function error_free_or_abort(), and consistently
use it.

The new function is added into error.c for use everywhere,
although it is anticipated that testsuites will be the main
client.
Signed-off-by: NEric Blake <eblake@redhat.com>
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
      a12a5a1a
  9. 09 11月, 2015 5 次提交
    • E
      qapi: Simplify non-error testing in test-qmp-* · 3f66f764
      由 Eric Blake 提交于 
      By using &error_abort, we can avoid a local err variable in
situations where we expect success.  It also has the nice
effect that if the test breaks, the error message from
error_abort tends to be nicer than that of g_assert().

This patch has an additional bonus of fixing several call sites that
were passing &err to two different functions without checking it in
between.  In general that is unsafe practice; because if the first
function sets an error, the second function could abort() if it tries to
set a different error. We got away with it because we were asserting
that err was NULL through the entire chain, but switching to
&error_abort avoids the questionable practice up front.
Signed-off-by: NEric Blake <eblake@redhat.com>
Message-Id: <1446791754-23823-7-git-send-email-eblake@redhat.com>
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
      3f66f764
    • E
      qapi: Plug leaks in test-qmp-* · b18f1141
      由 Eric Blake 提交于 
      Make valgrind happy with the current state of the tests, so that
it is easier to see if future patches introduce new memory problems
without being drowned in noise.  Many of the leaks were due to
calling a second init without tearing down the data from an earlier
visit.  But since teardown is already idempotent, and we already
register teardown as part of input_visitor_test_add(), it is nicer
to just make init() safe to call multiple times than it is to have
to make all tests call teardown.

Another common leak was forgetting to clean up an error object,
after testing that an error was raised.

Another leak was in test_visitor_in_struct_nested(), failing to
clean the base member of UserDefTwo.  Cleaning that up left
check_and_free_str() as dead code (since using the qapi_free_*
takes care of recursion, and we don't want double frees).

A final leak was in test_visitor_out_any(), which was reassigning
the qobj local variable to a subset of the overall structure
needing freeing; it did not result in a use-after-free, but
was not cleaning up all the qdict.

test-qmp-event and test-qmp-commands were already clean.
Signed-off-by: NEric Blake <eblake@redhat.com>
Message-Id: <1446791754-23823-6-git-send-email-eblake@redhat.com>
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
      b18f1141
    • E
      qapi: Share test_init code in test-qmp-input* · 0920a171
      由 Eric Blake 提交于 
      Rather than duplicate the body of two functions just to
decide between qobject_from_jsonv() and qobject_from_json(),
exploit the fact that qobject_from_jsonv() intentionally
takes 'va_list *' instead of the more common 'va_list', and
that qobject_from_json() just calls qobject_from_jsonv(,NULL).
For each file, our two existing init functions then become
thin wrappers around a new internal function, and future
updates to initialization don't have to be duplicated.
Suggested-by: NMarkus Armbruster <armbru@redhat.com>
Signed-off-by: NEric Blake <eblake@redhat.com>
Message-Id: <1446791754-23823-5-git-send-email-eblake@redhat.com>
[Two old comment typos fixed]
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
      0920a171
    • E
      qapi: Strengthen test of TestStructList · bd20588d
      由 Eric Blake 提交于 
      Make each list element different, to ensure that order is
preserved, and use the generated free function instead of
hand-rolling our own to ensure (under valgrind) that the
list is properly cleaned.
Suggested-by: NMarkus Armbruster <armbru@redhat.com>
Signed-off-by: NEric Blake <eblake@redhat.com>
Message-Id: <1446791754-23823-3-git-send-email-eblake@redhat.com>
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
      bd20588d
    • E
      qapi: Use generated TestStruct machinery in tests · 748053c9
      由 Eric Blake 提交于 
      Commit d88f5fd1 and friends first introduced the various test-qmp-*
tests in 2011, with duplicated hand-rolled TestStruct machinery,
to make sure the qapi visitor interface was tested.  Later, commit
4f193e34 in 2013 added a .json file for further testing use by the
files, but without consolidating any of the existing hand-rolled
visitors.  And with four copies, subtle differences have crept in,
between the tests themselves (mainly whitespace differences, but
also a question of whether to use NULL or "TestStruct" when
calling visit_start_struct()) and from what the generator produces
(the hand-rolled versions did not cater to partially-allocated
objects, because they did not have a deallocation usage).

Of course, just because the visitor interface is tested does not
mean it is a sane interface; and future patches will be changing
some of the visitor contracts.  Rather than having to duplicate
the cleanup work in each copy of the TestStruct visitor, and keep
each hand-rolled copy in sync with what the generator supplies, we
might as well just test what the generator should give us in the
first place.
Signed-off-by: NEric Blake <eblake@redhat.com>
Message-Id: <1446791754-23823-2-git-send-email-eblake@redhat.com>
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
      748053c9
  11. 06 11月, 2015 2 次提交
  13. 02 11月, 2015 7 次提交
    • E
      qapi: Reserve 'u' member name · 5e59baf9
      由 Eric Blake 提交于 
      Now that we have separated union tag values from colliding with
non-variant C names, by naming the union 'u', we should reserve
this name for our use.  Note that we want to forbid 'u' even in
a struct with no variants, because it is possible for a future
qemu release to extend QMP in a backwards-compatible manner while
converting from a struct to a flat union.  Fortunately, no
existing clients were using this member name.  If we ever find
the need for QMP to have a member 'u', we could at that time
relax things, perhaps by having c_name() munge the QMP member to
'q_u'.

Note that we cannot forbid 'u' everywhere (by adding the
rejection code to check_name()), because the existing QKeyCode
enum already uses it; therefore we only reserve it as a struct
type member name.
Signed-off-by: NEric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-24-git-send-email-eblake@redhat.com>
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
      5e59baf9
    • E
      tests: Convert to new qapi union layout · c363acef
      由 Eric Blake 提交于 
      We have two issues with our qapi union layout:
1) Even though the QMP wire format spells the tag 'type', the
C code spells it 'kind', requiring some hacks in the generator.
2) The C struct uses an anonymous union, which places all tag
values in the same namespace as all non-variant members. This
leads to spurious collisions if a tag value matches a non-variant
member's name.

Make the conversion to the new layout for testsuite code.
Signed-off-by: NEric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-15-git-send-email-eblake@redhat.com>
[Commit message tweaked slightly]
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
      c363acef
    • E
      qapi: Unbox base members · ddf21908
      由 Eric Blake 提交于 
      Rather than storing a base class as a pointer to a box, just
store the fields of that base class in the same order, so that
a child struct can be directly cast to its parent.  This gives
less malloc overhead, less pointer dereferencing, and even less
generated code.  Compare to the earlier commit 1e6c1616 "qapi:
Generate a nicer struct for flat unions" (although that patch
had fewer places to change, as less of qemu was directly using
qapi structs for flat unions).  It also allows us to turn on
automatic type-safe wrappers for upcasting to the base class
of a struct.

Changes to the generated code look like this in qapi-types.h:

| struct SpiceChannel {
|-    SpiceBasicInfo *base;
|+    /* Members inherited from SpiceBasicInfo: */
|+    char *host;
|+    char *port;
|+    NetworkAddressFamily family;
|+    /* Own members: */
|     int64_t connection_id;

as well as additional upcast functions like qapi_SpiceChannel_base().
Meanwhile, changes to qapi-visit.c look like:

| static void visit_type_SpiceChannel_fields(Visitor *v, SpiceChannel **obj, Error **errp)
| {
|     Error *err = NULL;
|
|-    visit_type_implicit_SpiceBasicInfo(v, &(*obj)->base, &err);
|+    visit_type_SpiceBasicInfo_fields(v, (SpiceBasicInfo **)obj, &err);
|     if (err) {

(the cast is necessary, since our upcast wrappers only deal with a
single pointer, not pointer-to-pointer); plus the wholesale
elimination of some now-unused visit_type_implicit_FOO() functions.

Without boxing, the corner case of one empty struct having
another empty struct as its base type now requires inserting a
dummy member (previously, the 'Base *base' member sufficed).

And now that we no longer consume a 'base' member in the generated
C struct, we can delete the former negative struct-base-clash-base
test.
Signed-off-by: NEric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-11-git-send-email-eblake@redhat.com>
[Commit message tweaked slightly]
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
      ddf21908
    • E
      qapi: Prefer typesafe upcasts to qapi base classes · 30594fe1
      由 Eric Blake 提交于 
      A previous patch (commit 1e6c1616) made it possible to
directly cast from a qapi flat union type to its base type.
However, it requires the use of a C cast, which turns off
compiler type-safety checks.  Fortunately, no such casts
exist, just yet.

Regardless, add inline type-safe wrappers named
qapi_FOO_base() for any union type FOO that has a base,
which can be used for a safer upcast, and enhance the
testsuite to cover the new functionality.

A future patch will extend the upcast support to structs,
where such conversions do exist already.

Note that C makes const-correct upcasts annoying because
it lacks overloads; these functions cast away const so that
they can accept user pointers whether const or not, and the
result in turn can be assigned to normal or const pointers.
Alternatively, this could have been done with macros, but
type-safe macros are hairy, and not worthwhile here.

This patch just adds upcasts.  None of our code needed to
downcast from a base qapi class to a child.  Also, in the
case of grandchildren (such as BlockdevOptionsQcow2), the
caller will need to call two functions to get to the inner
base (although it wouldn't be too hard to generate a
qapi_FOO_base_base() if desired).  If a user changes qapi
to alter the base class hierarchy, such as going from
'A -> C' to 'A -> B -> C', it will change the type of
'qapi_C_base()', and the compiler will point out the places
that are affected by the new base.

One alternative was proposed, but was deemed too ugly to use
in practice: the generators could output redundant
information using anonymous types:
| struct Child {
|     union {
|         struct {
|             Type1 parent_member1;
|             Type2 parent_member2;
|         };
|         Parent base;
|     };
| };
With that ugly proposal, for a given qapi type, obj->member
and obj->base.member would refer to the same storage; allowing
convenience in working with members without needing 'base.'
allowing typesafe upcast without needing a C cast by accessing
'&obj->base', and allowing downcasts from the parent back to
the child possible through container_of(obj, Child, base).
Signed-off-by: NEric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-10-git-send-email-eblake@redhat.com>
[Commit message tweaked]
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
      30594fe1
    • E
      qapi: Reserve 'q_*' and 'has_*' member names · 9fb081e0
      由 Eric Blake 提交于 
      c_name() produces names starting with 'q_' when protecting a
dictionary member name that would fail to directly compile, but
in doing so can cause clashes with any member name already
beginning with 'q-' or 'q_'.  Likewise, we create a C name 'has_'
for any optional member that can clash with any member name
beginning with 'has-' or 'has_'.

Technically, rather than blindly reserving the namespace,
we could try to complain about user names only when an actual
collision occurs, or even teach c_name() how to munge names
to avoid collisions.  But it is not trivial, especially when
collisions can occur across multiple types (such as via
inheritance or flat unions).  Besides, no existing .json
files are trying to use these names.  So it's easier to just
outright forbid the potential for collision.  We can always
relax things in the future if a real need arises for QMP to
express member names that have been forbidden here.

'has_' only has to be reserved for struct/union member names,
while 'q_' is reserved everywhere (matching the fact that
only members can be optional, while we use c_name() for munging
both members and entities).  Note that we could relax 'q_'
restrictions on entities independently from member names; for
example, c_name('qmp_' + 'unix') would result in a different
function name than our current 'qmp_' + c_name('unix').

Update and add tests to cover the new error messages.
Signed-off-by: NEric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-6-git-send-email-eblake@redhat.com>
[Consistently pass protect=False to c_name(); commit message tweaked
slightly]
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
      9fb081e0
    • E
      qapi: Reserve '*List' type names for list types · 255960dd
      由 Eric Blake 提交于 
      Type names ending in 'List' can clash with qapi list types in
generated C.  We don't currently use such names. It is easier to
outlaw them now than to worry about how to resolve such a clash
in the future. For precedence, see commit 4dc2e690, which did the
same for names ending in 'Kind' versus implicit enum types for
qapi unions.

Update the testsuite to match.
Signed-off-by: NEric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-5-git-send-email-eblake@redhat.com>
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
      255960dd
    • E
      tests/qapi-schema: Test for reserved names, empty struct · 19767083
      由 Eric Blake 提交于 
      Add some testsuite coverage to ensure future patches are on
the right track:

Our current C representation of qapi arrays is done by appending
'List' to the element name; but we are not preventing the
creation of an object type with the same name.  Add
reserved-type-list.json to test this.  Then rename
enum-union-clash.json to reserved-type-kind.json to cover the
reservation that we DO detect, and shorten it to match the fact
that the name is reserved even if there is no clash.

We are failing to detect a collision between a dictionary member
and the implicit 'has_*' flag for another optional member. The
easiest fix would be for a future patch to reserve the entire
"has[-_]" namespace for member names (the collision is also
possible for branch names within flat unions, but only as long as
branch names can collide with (non-variant) members; however,
since future patches are about to remove that, it is not worth
testing here). Add reserved-member-has.json to test this.

A similar collision exists between a dictionary member where
c_name() munges what might otherwise be a reserved name to start
with 'q_', and another member explicitly starts with "q[-_]".
Again, the easiest solution for a future patch will be reserving
the entire namespace, but here for commands as well as members.
Add reserved-member-q.json and reserved-command-q.json to test
this; separate tests since arguably our munging of command 'unix'
to 'qmp_q_unix()' could be done without a q_, which is different
than the munging of a member 'unix' to 'foo.q_unix'.

Finally, our testsuite does not have any compilation coverage
of struct inheritance with empty qapi structs.  Update
qapi-schema-test.json to test this.

Note that there is currently no technical reason to forbid type
name patterns from member names, or member name patterns from
types, since the two are not in the same namespace in C and
won't collide; but it's not worth adding positive tests of these
corner cases at this time, especially while there is other churn
pending in patches that rearrange which collisions actually
happen.
Signed-off-by: NEric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-2-git-send-email-eblake@redhat.com>
[Commit message tweaked slightly]
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
      19767083
  15. 29 10月, 2015 1 次提交
    • V
      tests/vhost-user-bridge: add vhost-user bridge application · 3595e2eb
      由 Victor Kaplansky 提交于 
      The test existing in QEMU for vhost-user feature is good for
testing the management protocol, but does not allow actual
traffic. This patch proposes Vhost-User Bridge application, which
can serve the QEMU community as a comprehensive test by running
real internet traffic by means of vhost-user interface.

Essentially the Vhost-User Bridge is a very basic vhost-user
backend for QEMU. It runs as a standalone user-level process.
For packet processing Vhost-User Bridge uses an additional QEMU
instance with a backend configured by "-net socket" as a shared
VLAN.  This way another QEMU virtual machine can effectively
serve as a shared bus by means of UDP communication.

For a more simple setup, the another QEMU instance running the
SLiRP backend can be the same QEMU instance running vhost-user
client.

This Vhost-User Bridge implementation is very preliminary.  It is
missing many features. I has been studying vhost-user protocol
internals, so I've written vhost-user-bridge bit by bit as I
progressed through the protocol.  Most probably its internal
architecture will change significantly.

To run Vhost-User Bridge application:

1. Build vhost-user-bridge with a regular procedure. This will
create a vhost-user-bridge executable under tests directory:

    $ configure; make tests/vhost-user-bridge

2. Ensure the machine has hugepages enabled in kernel with
command line like:

    default_hugepagesz=2M hugepagesz=2M hugepages=2048

3. Run Vhost-User Bridge with:

    $ tests/vhost-user-bridge

The above will run vhost-user server listening for connections
on UNIX domain socket /tmp/vubr.sock, and will try to connect
by UDP to VLAN bridge to localhost:5555, while listening on
localhost:4444

Run qemu with a virtio-net backed by vhost-user:

    $ qemu \
        -enable-kvm -m 512 -smp 2 \
        -object memory-backend-file,id=mem,size=512M,mem-path=/dev/hugepages,share=on \
        -numa node,memdev=mem -mem-prealloc \
        -chardev socket,id=char0,path=/tmp/vubr.sock \
        -netdev type=vhost-user,id=mynet1,chardev=char0,vhostforce \
        -device virtio-net-pci,netdev=mynet1 \
        -net none \
        -net socket,vlan=0,udp=localhost:4444,localaddr=localhost:5555 \
        -net user,vlan=0 \
        disk.img

vhost-user-bridge was tested very lightly: it's able to bringup a
linux on client VM with the virtio-net driver, and execute transmits
and receives to the internet. I tested with "wget redhat.com",
"dig redhat.com".

PS. I've consulted DPDK's code for vhost-user during Vhost-User
Bridge implementation.
Signed-off-by: NVictor Kaplansky <victork@redhat.com>
Reviewed-by: NMichael S. Tsirkin <mst@redhat.com>
Signed-off-by: NMichael S. Tsirkin <mst@redhat.com>
      3595e2eb