sdk/emulator/qemu.git
9 years agoui/curses: Fix pageup/pagedown on -curses
OGAWA Hirofumi [Mon, 19 Oct 2015 12:24:07 +0000 (21:24 +0900)]
ui/curses: Fix pageup/pagedown on -curses

Current KEY_NPAGE/KEY_PPAGE handling is broken on -curses. Those uses
"GREY", but "KEY_MASK" masked out "GREY".

To fix, we have to use correct mask value - SCANCODE_KEYMASK.

Then, this adds support of "shift + pageup/pagedown". With this,
-curses mode can use scroll-up/down as usual like other display modes.

Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
9 years agoui/curses: Support line graphics chars on -curses mode
OGAWA Hirofumi [Mon, 19 Oct 2015 12:23:46 +0000 (21:23 +0900)]
ui/curses: Support line graphics chars on -curses mode

This converts vga code to curses code in console_write_bh().

With this changes, we can see line graphics (for example, dialog uses)
correctly.

Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
9 years agoui/curses: Fix monitor color with -curses when 256 colors
OGAWA Hirofumi [Mon, 19 Oct 2015 12:23:10 +0000 (21:23 +0900)]
ui/curses: Fix monitor color with -curses when 256 colors

If TERM=xterm-256color, COLOR_PAIRS==256 and monitor passes chtype
like 0x74xx. Then, the code uses uninitialized color pair. As result,
monitor uses black for both of fg and bg color, i.e. terminal is
filled by black.

To fix, this initialize above than 64 with default color (fg=white,bg=black).

FIXME: on 256 color, curses may be possible better vga color emulation.

Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
9 years agoMerge remote-tracking branch 'remotes/armbru/tags/pull-qapi-2015-11-02' into staging
Peter Maydell [Mon, 2 Nov 2015 11:11:39 +0000 (11:11 +0000)]
Merge remote-tracking branch 'remotes/armbru/tags/pull-qapi-2015-11-02' into staging

QAPI patches

# gpg: Signature made Mon 02 Nov 2015 09:07:23 GMT using RSA key ID EB918653
# gpg: Good signature from "Markus Armbruster <armbru@redhat.com>"
# gpg:                 aka "Markus Armbruster <armbru@pond.sub.org>"

* remotes/armbru/tags/pull-qapi-2015-11-02: (25 commits)
  qapi: Simplify gen_struct_field()
  qapi: Reserve 'u' member name
  qapi: Finish converting to new qapi union layout
  tpm: Convert to new qapi union layout
  memory: Convert to new qapi union layout
  input: Convert to new qapi union layout
  char: Convert to new qapi union layout
  net: Convert to new qapi union layout
  sockets: Convert to new qapi union layout
  block: Convert to new qapi union layout
  tests: Convert to new qapi union layout
  qapi-visit: Convert to new qapi union layout
  qapi: Start converting to new qapi union layout
  qapi-visit: Remove redundant functions for flat union base
  qapi: Unbox base members
  qapi: Prefer typesafe upcasts to qapi base classes
  qapi-types: Refactor base fields output
  qapi-visit: Split off visit_type_FOO_fields forward decl
  vnc: Hoist allocation of VncBasicInfo to callers
  qapi: Reserve 'q_*' and 'has_*' member names
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agoqapi: Simplify gen_struct_field()
Eric Blake [Mon, 26 Oct 2015 22:35:03 +0000 (16:35 -0600)]
qapi: Simplify gen_struct_field()

Rather than having all callers pass a name, type, and optional
flag, have them instead pass a QAPISchemaObjectTypeMember which
already has all that information.

No change to generated code.

Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-25-git-send-email-eblake@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agoqapi: Reserve 'u' member name
Eric Blake [Mon, 26 Oct 2015 22:35:02 +0000 (16:35 -0600)]
qapi: Reserve 'u' member name

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: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-24-git-send-email-eblake@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agoqapi: Finish converting to new qapi union layout
Eric Blake [Mon, 26 Oct 2015 22:35:01 +0000 (16:35 -0600)]
qapi: Finish converting to new qapi union layout

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.

This patch is the back end for a series that converts to a
saner qapi union layout.  Now that all clients have been
converted to use 'type' and 'obj->u.value', we can drop the
temporary parallel support for 'kind' and 'obj->value'.

Given a simple union qapi type:

{ 'union':'Foo', 'data': { 'a':'int', 'b':'bool' } }

this is the overall effect, when compared to the state before
this series of patches:

| struct Foo {
|-    FooKind kind;
|-    union { /* union tag is @kind */
|+    FooKind type;
|+    union { /* union tag is @type */
|         void *data;
|         int64_t a;
|         bool b;
|-    };
|+    } u;
| };

The testsuite still contains some examples of artificial restrictions
(see flat-union-clash-type.json, for example) that are no longer
technically necessary, now that there is no longer a collision between
enum tag values and non-variant member names; but fixing this will be
done in later patches, in part because some further changes are required
to keep QAPISchema*.check() from asserting.  Also, a later patch will
add a reservation for the member name 'u' to avoid a collision between a
user's non-variant names and our internal choice of C union name.

Note, however, that we do not rename the generated enum, which
is still 'FooKind'.  A further patch could generate implicit
enums as 'FooType', but while the generator already reserved
the '*Kind' namespace (commit 4dc2e69), there are already QMP
constructs with '*Type' naming, which means changing our
reservation namespace would have lots of churn to C code to
deal with a forced name change.

Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-23-git-send-email-eblake@redhat.com>
[Commit message tweaked]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agotpm: Convert to new qapi union layout
Eric Blake [Mon, 26 Oct 2015 22:35:00 +0000 (16:35 -0600)]
tpm: Convert to new qapi union layout

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 TPM-related code.

Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-22-git-send-email-eblake@redhat.com>
[Commit message tweaked slightly]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agomemory: Convert to new qapi union layout
Eric Blake [Mon, 26 Oct 2015 22:34:59 +0000 (16:34 -0600)]
memory: Convert to new qapi union layout

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 memory-related code.

Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-21-git-send-email-eblake@redhat.com>
[Commit message tweaked slightly]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agoinput: Convert to new qapi union layout
Eric Blake [Mon, 26 Oct 2015 22:34:58 +0000 (16:34 -0600)]
input: Convert to new qapi union layout

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 input-related code.

Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-20-git-send-email-eblake@redhat.com>
[Commit message tweaked slightly]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agochar: Convert to new qapi union layout
Eric Blake [Mon, 26 Oct 2015 22:34:57 +0000 (16:34 -0600)]
char: Convert to new qapi union layout

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 character-related
code.

Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-19-git-send-email-eblake@redhat.com>
[Commit message tweaked slightly]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agonet: Convert to new qapi union layout
Eric Blake [Mon, 26 Oct 2015 22:34:56 +0000 (16:34 -0600)]
net: Convert to new qapi union layout

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 net-related code.

Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-18-git-send-email-eblake@redhat.com>
[Commit message tweaked slightly]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agosockets: Convert to new qapi union layout
Eric Blake [Mon, 26 Oct 2015 22:34:55 +0000 (16:34 -0600)]
sockets: Convert to new qapi union layout

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 socket-related code.

Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-17-git-send-email-eblake@redhat.com>
[Commit message tweaked slightly]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agoblock: Convert to new qapi union layout
Eric Blake [Mon, 26 Oct 2015 22:34:54 +0000 (16:34 -0600)]
block: Convert to new qapi union layout

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 block-related code.

Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-16-git-send-email-eblake@redhat.com>
[Commit message tweaked slightly]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agotests: Convert to new qapi union layout
Eric Blake [Mon, 26 Oct 2015 22:34:53 +0000 (16:34 -0600)]
tests: Convert to new qapi union layout

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: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-15-git-send-email-eblake@redhat.com>
[Commit message tweaked slightly]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agoqapi-visit: Convert to new qapi union layout
Eric Blake [Mon, 26 Oct 2015 22:34:52 +0000 (16:34 -0600)]
qapi-visit: Convert to new qapi union layout

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 qapi-visit.py.

Generated code changes look like:

|@@ -4912,16 +4912,16 @@ void visit_type_MemoryDeviceInfo(Visitor
|     if (!*obj) {
|         goto out_obj;
|     }
|-    visit_type_MemoryDeviceInfoKind(v, &(*obj)->kind, "type", &err);
|+    visit_type_MemoryDeviceInfoKind(v, &(*obj)->type, "type", &err);
|     if (err) {
|         goto out_obj;
|     }
|-    if (!visit_start_union(v, !!(*obj)->data, &err) || err) {
|+    if (!visit_start_union(v, !!(*obj)->u.data, &err) || err) {
|         goto out_obj;
|     }
|-    switch ((*obj)->kind) {
|+    switch ((*obj)->type) {
|     case MEMORY_DEVICE_INFO_KIND_DIMM:
|-        visit_type_PCDIMMDeviceInfo(v, &(*obj)->dimm, "data", &err);
|+        visit_type_PCDIMMDeviceInfo(v, &(*obj)->u.dimm, "data", &err);
|         break;
|     default:
|         abort();
|@@ -4930,7 +4930,7 @@ out_obj:
|     error_propagate(errp, err);
|     err = NULL;
|     if (*obj) {
|-        visit_end_union(v, !!(*obj)->data, &err);
|+        visit_end_union(v, !!(*obj)->u.data, &err);
|     }
|     error_propagate(errp, err);
|     err = NULL;

Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-14-git-send-email-eblake@redhat.com>
[Commit message tweaked slightly]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agoqapi: Start converting to new qapi union layout
Eric Blake [Mon, 26 Oct 2015 22:34:51 +0000 (16:34 -0600)]
qapi: Start converting to new qapi union layout

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.

This patch is the front end for a series that converts to a
saner qapi union layout.  By the end of the series, we will no
longer have the type/kind mismatch, and all tag values will be
under a named union, which requires clients to access
'obj->u.value' instead of 'obj->value'.  But since the
conversion touches a number of files, it is easiest if we
temporarily support BOTH layouts simultaneously.

Given a simple union qapi type:

{ 'union':'Foo', 'data': { 'a':'int', 'b':'bool' } }

make the following changes in generated qapi-types.h:

| struct Foo {
|-    FooKind kind;
|-    union { /* union tag is @kind */
|+    union {
|+        FooKind kind;
|+        FooKind type;
|+    };
|+    union { /* union tag is @type */
|         void *data;
|         int64_t a;
|         bool b;
|+        union { /* union tag is @type */
|+            void *data;
|+            int64_t a;
|+            bool b;
|+        } u;
|     };
| };

Flat unions do not need the anonymous union for the tag member,
as we already fixed that to use the member name instead of 'kind'
back in commit 0f61af3e.

One additional change is needed in qapi.py: check_union() now
needs to check for collisions with 'type' in addition to those
with 'kind'.

Later, when the conversions are complete, we will remove the
duplication hacks, and also drop the check_union() restrictions.

Note, however, that we do not rename the generated enum, which
is still 'FooKind'.  A further patch could generate implicit
enums as 'FooType', but while the generator already reserved
the '*Kind' namespace (commit 4dc2e69), there are already QMP
constructs with '*Type' naming, which means changing our
reservation namespace would have lots of churn to C code to
deal with a forced name change.

Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-13-git-send-email-eblake@redhat.com>
[Commit message tweaked slightly]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agoqapi-visit: Remove redundant functions for flat union base
Eric Blake [Mon, 26 Oct 2015 22:34:50 +0000 (16:34 -0600)]
qapi-visit: Remove redundant functions for flat union base

The code for visiting the base class of a child struct created
visit_type_Base_fields() which covers all fields of Base; while
the code for visiting the base class of a flat union created
visit_type_Union_fields() covering all fields of the base
except the discriminator.  But since the base class includes
the discriminator of a flat union, we can just visit the entire
base, without needing a separate visit of the discriminator.
Not only is consistently visiting all fields easier to
understand, it lets us share code.

The generated code in qapi-visit.c loses several now-unused
visit_type_UNION_fields(), along with changes like:

|@@ -1654,11 +1557,7 @@ void visit_type_BlockdevOptions(Visitor
|     if (!*obj) {
|         goto out_obj;
|     }
|-    visit_type_BlockdevOptions_fields(v, obj, &err);
|-    if (err) {
|-        goto out_obj;
|-    }
|-    visit_type_BlockdevDriver(v, &(*obj)->driver, "driver", &err);
|+    visit_type_BlockdevOptionsBase_fields(v, (BlockdevOptionsBase **)obj, &err);
|     if (err) {
|         goto out_obj;
|     }

and forward declarations where needed.  Note that the cast of obj
to BASE ** is necessary to call visit_type_BASE_fields() (and we
can't use our upcast wrappers, because those work on pointers while
we have a pointer-to-pointer).

Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-12-git-send-email-eblake@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agoqapi: Unbox base members
Eric Blake [Mon, 26 Oct 2015 22:34:49 +0000 (16:34 -0600)]
qapi: Unbox base members

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 1e6c1616a "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: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-11-git-send-email-eblake@redhat.com>
[Commit message tweaked slightly]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agoqapi: Prefer typesafe upcasts to qapi base classes
Eric Blake [Mon, 26 Oct 2015 22:34:48 +0000 (16:34 -0600)]
qapi: Prefer typesafe upcasts to qapi base classes

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: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-10-git-send-email-eblake@redhat.com>
[Commit message tweaked]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agoqapi-types: Refactor base fields output
Eric Blake [Mon, 26 Oct 2015 22:34:47 +0000 (16:34 -0600)]
qapi-types: Refactor base fields output

Move code from gen_union() into gen_struct_fields() in order for
a later patch to share code when enumerating inherited fields
for struct types.

No change to generated code.

Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-9-git-send-email-eblake@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agoqapi-visit: Split off visit_type_FOO_fields forward decl
Eric Blake [Mon, 26 Oct 2015 22:34:46 +0000 (16:34 -0600)]
qapi-visit: Split off visit_type_FOO_fields forward decl

We generate a static visit_type_FOO_fields() for every type
FOO.  However, sometimes we need a forward declaration. Split
the code to generate the forward declaration out of
gen_visit_implicit_struct() into a new gen_visit_fields_decl(),
and also prepare for a forward declaration to be emitted
during gen_visit_struct(), so that a future patch can switch
from using visit_type_FOO_implicit() to the simpler
visit_type_FOO_fields() as part of unboxing the base class
of a struct.

No change to generated code.

Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-8-git-send-email-eblake@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agovnc: Hoist allocation of VncBasicInfo to callers
Eric Blake [Mon, 26 Oct 2015 22:34:45 +0000 (16:34 -0600)]
vnc: Hoist allocation of VncBasicInfo to callers

A future qapi patch will rework generated structs with a base
class to be unboxed.  In preparation for that, change the code
that allocates then populates an info struct to instead merely
populate the fields of an info field passed in as a parameter
(renaming vnc_basic_info_get* to vnc_init_basic_info*). Add
rudimentary Error handling at the lowest levels for cases
where the old code returned NULL; but rather than plumb Error
all the way through the stack, the callers drop the error and
return NULL as before.

Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-7-git-send-email-eblake@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agoqapi: Reserve 'q_*' and 'has_*' member names
Eric Blake [Mon, 26 Oct 2015 22:34:44 +0000 (16:34 -0600)]
qapi: Reserve 'q_*' and 'has_*' member names

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: Eric 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: Markus Armbruster <armbru@redhat.com>
9 years agoqapi: Reserve '*List' type names for list types
Eric Blake [Mon, 26 Oct 2015 22:34:43 +0000 (16:34 -0600)]
qapi: Reserve '*List' type names for list types

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 4dc2e69, which did the
same for names ending in 'Kind' versus implicit enum types for
qapi unions.

Update the testsuite to match.

Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-5-git-send-email-eblake@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agoqapi: More robust conditions for when labels are needed
Eric Blake [Mon, 26 Oct 2015 22:34:42 +0000 (16:34 -0600)]
qapi: More robust conditions for when labels are needed

We were using regular expressions to see if ret included
any earlier text that emitted a 'goto out;' line, to decide
whether we needed to output an 'out:' label.  But this is
fragile, if the ret text can possibly combine more than one
generated function body, where the first function used a
goto but the second does not.  Change the code to just check
for the known conditions which cause an error check to be
needed.  Besides, it's slightly more efficient to use plain
checks than regular expression searching.

No change to generated code.

Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-4-git-send-email-eblake@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agoqapi: More idiomatic string operations
Eric Blake [Mon, 26 Oct 2015 22:34:41 +0000 (16:34 -0600)]
qapi: More idiomatic string operations

Rather than slicing the end of a string, we can use python's
endswith().  And rather than creating a set of characters,
we can search for a character within a string.

Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-3-git-send-email-eblake@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agotests/qapi-schema: Test for reserved names, empty struct
Eric Blake [Mon, 26 Oct 2015 22:34:40 +0000 (16:34 -0600)]
tests/qapi-schema: Test for reserved names, empty struct

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: Eric Blake <eblake@redhat.com>
Message-Id: <1445898903-12082-2-git-send-email-eblake@redhat.com>
[Commit message tweaked slightly]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agoqapi-schema: mark InetSocketAddress as mandatory again
Daniel P. Berrange [Thu, 22 Oct 2015 10:25:43 +0000 (11:25 +0100)]
qapi-schema: mark InetSocketAddress as mandatory again

Revert the qapi-schema.json change done in:

  commit 0983f5e6af76d5df8c6346cbdfff9d8305fb6da0
  Author: Daniel P. Berrange <berrange@redhat.com>
  Date:   Tue Sep 1 14:46:50 2015 +0100

    sockets: allow port to be NULL when listening on IP address

Switching "port" from mandatory to optional causes the QAPI
code generator to add a 'has_port' field to the InetSocketAddress
struct. No code that created InetSocketAddress objects was updated
to set 'has_port = true', which caused the non-NULL port strings
to be silently dropped when copying InetSocketAddress objects.

Reported-by: Knut Omang <knuto@ifi.uio.no>
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Message-Id: <1445509543-30679-1-git-send-email-berrange@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
9 years agoMerge remote-tracking branch 'remotes/rth/tags/pull-tile-20151030' into staging
Peter Maydell [Fri, 30 Oct 2015 21:59:48 +0000 (21:59 +0000)]
Merge remote-tracking branch 'remotes/rth/tags/pull-tile-20151030' into staging

Prefetch in y2 pipe

# gpg: Signature made Fri 30 Oct 2015 20:40:02 GMT using RSA key ID 4DD0279B
# gpg: Good signature from "Richard Henderson <rth7680@gmail.com>"
# gpg:                 aka "Richard Henderson <rth@redhat.com>"
# gpg:                 aka "Richard Henderson <rth@twiddle.net>"

* remotes/rth/tags/pull-tile-20151030:
  target-tilegx: Implement prefetch instructions in pipe y2

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agoMerge remote-tracking branch 'remotes/stefanha/tags/block-pull-request' into staging
Peter Maydell [Fri, 30 Oct 2015 19:47:47 +0000 (19:47 +0000)]
Merge remote-tracking branch 'remotes/stefanha/tags/block-pull-request' into staging

# gpg: Signature made Thu 29 Oct 2015 18:09:16 GMT using RSA key ID 81AB73C8
# gpg: Good signature from "Stefan Hajnoczi <stefanha@redhat.com>"
# gpg:                 aka "Stefan Hajnoczi <stefanha@gmail.com>"

* remotes/stefanha/tags/block-pull-request:
  block: Consider all child nodes in bdrv_requests_pending()
  target-arm: xlnx-zynqmp: Add sdhci support.
  sdhci: Split sdhci.h for public and internal device usage
  sd.h: Move sd.h to include/hw/sd/
  virtio: sync the dataplane vring state to the virtqueue before virtio_save
  gdb command: qemu handlers
  virtio-blk: switch off scsi-passthrough by default
  ppc/spapr: add 2.4 compat props
  s390x: include HW_COMPAT_* props
  qemu-gdb: add $qemu_coroutine_sp and $qemu_coroutine_pc
  qemu-gdb: extract parts of "qemu coroutine" implementation
  qemu-gdb: allow using glibc_pointer_guard() on core dumps

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agoMerge remote-tracking branch 'remotes/lalrae/tags/mips-20151030' into staging
Peter Maydell [Fri, 30 Oct 2015 16:30:25 +0000 (16:30 +0000)]
Merge remote-tracking branch 'remotes/lalrae/tags/mips-20151030' into staging

MIPS patches 2015-10-30

Changes:
* R6 CPU can be woken up by non-enabled interrupts
* PC fix in KVM
* Coprocessor 0 XContext calculation fix
* various MIPS R6 updates

# gpg: Signature made Fri 30 Oct 2015 14:51:56 GMT using RSA key ID 0B29DA6B
# gpg: Good signature from "Leon Alrae <leon.alrae@imgtec.com>"

* remotes/lalrae/tags/mips-20151030:
  target-mips: fix updating XContext on mmu exception
  target-mips: add SIGRIE instruction
  target-mips: Set Config5.XNP for R6 cores
  target-mips: add PC, XNP reg numbers to RDHWR
  hw/mips_malta: Fix KVM PC initialisation
  target-mips: Add enum for BREAK32
  target-mips: update writing to CP0.Status.KX/SX/UX in MIPS Release R6
  target-mips: implement the CPU wake-up on non-enabled interrupts in R6
  target-mips: move the test for enabled interrupts to a separate function

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-mips: fix updating XContext on mmu exception
Yongbok Kim [Thu, 29 Oct 2015 17:17:52 +0000 (17:17 +0000)]
target-mips: fix updating XContext on mmu exception

Correct updating XContext.Region field on mmu exceptions.
If Config3.CTXTC = 0 then the R field of XContext has to be updated
with the value of bits 63..62 of the virtual address upon a TLB
exception.
Also fixed the below line which overs 80 characters.

Signed-off-by: Yongbok Kim <yongbok.kim@imgtec.com>
Reviewed-by: James Hogan <james.hogan@imgtec.com>
Reviewed-by: Leon Alrae <leon.alrae@imgtec.com>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
9 years agotarget-mips: add SIGRIE instruction
Yongbok Kim [Thu, 29 Oct 2015 15:18:38 +0000 (15:18 +0000)]
target-mips: add SIGRIE instruction

Add SIGRIE (Signal Reserved Instruction Exception) for both MIPS and
microMIPS.
The instruction allows to use the 16-bit code field for software use.
This instruction is introduced by and required as of Release 6.

Signed-off-by: Yongbok Kim <yongbok.kim@imgtec.com>
Reviewed-by: Leon Alrae <leon.alrae@imgtec.com>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
9 years agotarget-mips: Set Config5.XNP for R6 cores
Yongbok Kim [Mon, 5 Oct 2015 13:45:45 +0000 (14:45 +0100)]
target-mips: Set Config5.XNP for R6 cores

Set Config5.XNP for R6 cores to indicate the extended LL/SC family
of instructions NOT present.

Signed-off-by: Yongbok Kim <yongbok.kim@imgtec.com>
Reviewed-by: Leon Alrae <leon.alrae@imgtec.com>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
9 years agotarget-mips: add PC, XNP reg numbers to RDHWR
Yongbok Kim [Thu, 29 Oct 2015 15:18:39 +0000 (15:18 +0000)]
target-mips: add PC, XNP reg numbers to RDHWR

Add Performance Counter (4) and XNP (5) register numbers to RDHWR.
Add check_hwrena() to simplify access control checkings.
Add RDHWR support to microMIPS R6.

Signed-off-by: Yongbok Kim <yongbok.kim@imgtec.com>
Reviewed-by: Leon Alrae <leon.alrae@imgtec.com>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
9 years agohw/mips_malta: Fix KVM PC initialisation
James Hogan [Mon, 12 Oct 2015 16:54:39 +0000 (17:54 +0100)]
hw/mips_malta: Fix KVM PC initialisation

Commit 71c199c81d29 ("mips_malta: provide ememsize env variable to
kernels") changed the meaning of loaderparams.ram_size to be the whole
of RAM rather than just the low part below where the boot code is placed
for KVM, but it didn't update the PC initialisation for KVM to use
ram_low_size. Fix that now.

Fixes: 71c199c81d29 ("mips_malta: provide ememsize env variable to kernels")
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paul Burton <paul.burton@imgtec.com>
Cc: Leon Alrae <leon.alrae@imgtec.com>
Cc: Aurelien Jarno <aurelien@aurel32.net>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Reviewed-by: Leon Alrae <leon.alrae@imgtec.com>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
9 years agoMerge remote-tracking branch 'remotes/armbru/tags/pull-monitor-2015-10-30' into staging
Peter Maydell [Fri, 30 Oct 2015 09:41:14 +0000 (09:41 +0000)]
Merge remote-tracking branch 'remotes/armbru/tags/pull-monitor-2015-10-30' into staging

QMP and QObject patches

# gpg: Signature made Fri 30 Oct 2015 08:06:26 GMT using RSA key ID EB918653
# gpg: Good signature from "Markus Armbruster <armbru@redhat.com>"
# gpg:                 aka "Markus Armbruster <armbru@pond.sub.org>"

* remotes/armbru/tags/pull-monitor-2015-10-30:
  docs: Document QMP event rate limiting
  monitor: Throttle event VSERPORT_CHANGE separately by "id"
  monitor: Turn monitor_qapi_event_state[] into a hash table
  glib: add compatibility interface for g_hash_table_add()
  monitor: Split MonitorQAPIEventConf off MonitorQAPIEventState
  monitor: Switch from timer_new() to timer_new_ns()
  monitor: Simplify event throttling
  monitor: Reduce casting of QAPI event QDict
  qstring: Make conversion from QObject * accept null
  qlist: Make conversion from QObject * accept null
  qfloat qint: Make conversion from QObject * accept null
  qdict: Make conversion from QObject * accept null
  qbool: Make conversion from QObject * accept null
  qobject: Drop QObject_HEAD

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agodocs: Document QMP event rate limiting
Markus Armbruster [Thu, 15 Oct 2015 15:08:36 +0000 (17:08 +0200)]
docs: Document QMP event rate limiting

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <1444921716-9511-8-git-send-email-armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
9 years agomonitor: Throttle event VSERPORT_CHANGE separately by "id"
Markus Armbruster [Thu, 15 Oct 2015 15:08:35 +0000 (17:08 +0200)]
monitor: Throttle event VSERPORT_CHANGE separately by "id"

VSERPORT_CHANGE is emitted when the guest opens or closes a
virtio-serial port.  The event's member "id" identifies the port.

When several events arrive quickly, throttling drops all but the last
of them.  Because of that, a QMP client must assume that *any* port
may have changed state when it receives a VSERPORT_CHANGE event and
throttling may have happened.

Make the event more useful by throttling it for each port separately.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <1444921716-9511-7-git-send-email-armbru@redhat.com>

9 years agomonitor: Turn monitor_qapi_event_state[] into a hash table
Markus Armbruster [Thu, 15 Oct 2015 15:08:34 +0000 (17:08 +0200)]
monitor: Turn monitor_qapi_event_state[] into a hash table

In preparation of finer grained throttling.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <1444921716-9511-6-git-send-email-armbru@redhat.com>

9 years agoglib: add compatibility interface for g_hash_table_add()
Markus Armbruster [Tue, 27 Oct 2015 14:44:00 +0000 (15:44 +0100)]
glib: add compatibility interface for g_hash_table_add()

The next commit will use it.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
9 years agoblock: Consider all child nodes in bdrv_requests_pending()
Kevin Wolf [Wed, 28 Oct 2015 10:46:51 +0000 (11:46 +0100)]
block: Consider all child nodes in bdrv_requests_pending()

The function manually recursed into bs->file and bs->backing to check
whether there were any requests pending, but it ignored other children.

There's no need to special case file and backing here, so just replace
these two explicit recursions by a loop recursing for all child nodes.

Reported-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Jeff Cody <jcody@redhat.com>
Message-id: 1446029211-27148-1-git-send-email-kwolf@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agotarget-arm: xlnx-zynqmp: Add sdhci support.
Sai Pavan Boddu [Thu, 8 Oct 2015 13:21:03 +0000 (18:51 +0530)]
target-arm: xlnx-zynqmp: Add sdhci support.

Add two SYSBUS_SDHCI devices for xlnx-zynqmp

Signed-off-by: Sai Pavan Boddu <saipava@xilinx.com>
Reviewed-by: Peter Crosthwaite <crosthwaite.peter@gmail.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agosdhci: Split sdhci.h for public and internal device usage
Sai Pavan Boddu [Thu, 8 Oct 2015 13:21:02 +0000 (18:51 +0530)]
sdhci: Split sdhci.h for public and internal device usage

Split sdhci.h into pubilc version (i.e include/hw/sd/sdhci.h) and
internal version (i.e hw/sd/sdhci-interna.h) based on register
declarations and object declaration.

Signed-off-by: Sai Pavan Boddu <saipava@xilinx.com>
Reviewed-by: Alistair Francis <alistair.francis@xilinx.com>
Reviewed-by: Peter Crosthwaite <crosthwaite.peter@gmail.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agosd.h: Move sd.h to include/hw/sd/
Sai Pavan Boddu [Thu, 8 Oct 2015 13:21:01 +0000 (18:51 +0530)]
sd.h: Move sd.h to include/hw/sd/

Create a sd directory under include/hw/ and move sd.h to
include/hw/sd/

Signed-off-by: Sai Pavan Boddu <saipava@xilinx.com>
Reviewed-by: Alistair Francis <alistair.francis@xilinx.com>
Reviewed-by: Peter Crosthwaite <crosthwaite.peter@gmail.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agovirtio: sync the dataplane vring state to the virtqueue before virtio_save
Pavel Butsykin [Mon, 26 Oct 2015 11:42:57 +0000 (14:42 +0300)]
virtio: sync the dataplane vring state to the virtqueue before virtio_save

When creating snapshot with the dataplane enabled, the snapshot file gets
not the actual state of virtqueue, because the current state is stored in
VirtIOBlockDataPlane. Therefore, before saving snapshot need to sync
the dataplane vring state to the virtqueue. The dataplane will resume its
work at the next notify virtqueue.

When snapshot loads with loadvm we get a message:
VQ 0 size 0x80 Guest index 0x15f5 inconsistent with Host index 0x0:
    delta 0x15f5
error while loading state for instance 0x0 of device
    '0000:00:08.0/virtio-blk'
Error -1 while loading VM state

to reproduce the error I used the following hmp commands:
savevm snap1
loadvm snap1

qemu parameters:
--enable-kvm -smp 4 -m 1024 -drive file=/var/lib/libvirt/images/centos6.4.qcow2,if=none,id=drive-virtio-disk0,format=qcow2,cache=none,aio=native -device virtio-blk-pci,scsi=off,bus=pci.0,addr=0x8,drive=drive-virtio-disk0,id=virtio-disk0 -set device.virtio-disk0.x-data-plane=on

Signed-off-by: Pavel Butsykin <pbutsykin@virtuozzo.com>
Signed-off-by: Denis V. Lunev <den@openvz.org>
Message-id: 1445859777-2982-1-git-send-email-den@openvz.org
CC: Stefan Hajnoczi <stefanha@redhat.com>
CC: "Michael S. Tsirkin" <mst@redhat.com>
CC: Kevin Wolf <kwolf@redhat.com>
CC: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agogdb command: qemu handlers
Dr. David Alan Gilbert [Tue, 27 Oct 2015 13:09:45 +0000 (13:09 +0000)]
gdb command: qemu handlers

A new gdb commands are added:

  qemu handlers

     That dumps an AioContext list (by default qemu_aio_context)
     possibly including a backtrace for cases it knows about
     (with the verbose option).  Intended to help find why something
     is hanging waiting for IO.

  Use 'qemu handlers --verbose iohandler_ctx'  to find out why
your incoming migration is stuck.

Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Message-id: 1445951385-11924-1-git-send-email-dgilbert@redhat.com

V2:
  Merge into one command with optional handlers arg, and only do
    backtrace in verbose mode

 (gdb) qemu handlers
 ----
 {pfd = {fd = 6, events = 25, revents = 0}, io_read = 0x55869656ffd0
 <event_notifier_dummy_cb>, io_write = 0x0, deleted = 0, opaque =
 0x558698c4ce08, node = {le_next = 0x0, le_prev = 0x558698c4cdc0}}

 (gdb) qemu handlers iohandler_ctx
 ----
 {pfd = {fd = 9, events = 25, revents = 0}, io_read = 0x558696581380
 <fd_coroutine_enter>, io_write = 0x0, deleted = 0, opaque =
 0x558698dc99d0, node = {le_next = 0x558698c4cca0, le_prev =
 0x558698c4c1d0}}
 ----
 {pfd = {fd = 4, events = 25, revents = 0}, io_read = 0x55869657b330
 <sigfd_handler>, io_write = 0x0, deleted = 0, opaque = 0x4, node =
 {le_next = 0x558698c4c260, le_prev = 0x558699f72508}}
 ----
 {pfd = {fd = 5, events = 25, revents = 0}, io_read = 0x55869656ffd0
 <event_notifier_dummy_cb>, io_write = 0x0, deleted = 0, opaque =
 0x558698c4c218, node = {le_next = 0x0, le_prev = 0x558698c4ccc8}}
 ----
 (gdb) qemu handlers --verbose iohandler_ctx
 ----
 {pfd = {fd = 9, events = 25, revents = 0}, io_read = 0x558696581380
 <fd_coroutine_enter>, io_write = 0x0, deleted = 0, opaque =
 0x558698dc99d0, node = {le_next = 0x558698c4cca0, le_prev =
 0x558698c4c1d0}}
 #0  0x0000558696581820 in qemu_coroutine_switch
 (from_=from_@entry=0x558698cb3cf0, to_=to_@entry=0x7f421c37eac8,
 action=action@entry=COROUTINE_YIELD) at
 /home/dgilbert/git/qemu/coroutine-ucontext.c:177
 #1  0x0000558696580c00 in qemu_coroutine_yield () at
 /home/dgilbert/git/qemu/qemu-coroutine.c:145
 #2  0x00005586965814f5 in yield_until_fd_readable (fd=9) at
 /home/dgilbert/git/qemu/qemu-coroutine-io.c:90
 #3  0x0000558696523937 in socket_get_buffer (opaque=0x55869a3dc620,
 buf=0x558698c505a0 "", pos=<optimized out>, size=32768) at
 /home/dgilbert/git/qemu/migration/qemu-file-unix.c:101
 #4  0x0000558696521fac in qemu_fill_buffer (f=0x558698c50570) at
 /home/dgilbert/git/qemu/migration/qemu-file.c:227
 #5  0x0000558696522989 in qemu_peek_byte (f=0x558698c50570, offset=0)
     at /home/dgilbert/git/qemu/migration/qemu-file.c:507
 #6  0x0000558696522bf4 in qemu_get_be32 (f=0x558698c50570) at
 /home/dgilbert/git/qemu/migration/qemu-file.c:520
 #7  0x0000558696522bf4 in qemu_get_be32 (f=f@entry=0x558698c50570)
     at /home/dgilbert/git/qemu/migration/qemu-file.c:604
 #8  0x0000558696347e5c in qemu_loadvm_state (f=f@entry=0x558698c50570)
     at /home/dgilbert/git/qemu/migration/savevm.c:1821
 #9  0x000055869651de8c in process_incoming_migration_co
 (opaque=0x558698c50570)
     at /home/dgilbert/git/qemu/migration/migration.c:336
 #10 0x000055869658188a in coroutine_trampoline (i0=<optimized out>,
 i1=<optimized out>)
     at /home/dgilbert/git/qemu/coroutine-ucontext.c:80
 #11 0x00007f420f05df10 in __start_context () at /lib64/libc.so.6
 #12 0x00007ffc40815f50 in  ()
 #13 0x0000000000000000 in  ()

  ----
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agovirtio-blk: switch off scsi-passthrough by default
Cornelia Huck [Fri, 16 Oct 2015 10:25:54 +0000 (12:25 +0200)]
virtio-blk: switch off scsi-passthrough by default

Devices that are compliant with virtio-1 do not support scsi
passthrough any more (and it has not been a recommended setup
anyway for quite some time). To avoid having to switch it off
explicitly in newer qemus that turn on virtio-1 by default, let's
switch the default to scsi=false for 2.5.

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Message-id: 1444991154-79217-4-git-send-email-cornelia.huck@de.ibm.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agoppc/spapr: add 2.4 compat props
Cornelia Huck [Fri, 16 Oct 2015 10:25:53 +0000 (12:25 +0200)]
ppc/spapr: add 2.4 compat props

HW_COMPAT_2_4 will become non-empty: prepare for it.

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Message-id: 1444991154-79217-3-git-send-email-cornelia.huck@de.ibm.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agos390x: include HW_COMPAT_* props
Cornelia Huck [Fri, 16 Oct 2015 10:25:52 +0000 (12:25 +0200)]
s390x: include HW_COMPAT_* props

We want to inherit generic hw compat as well.

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Message-id: 1444991154-79217-2-git-send-email-cornelia.huck@de.ibm.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agoqemu-gdb: add $qemu_coroutine_sp and $qemu_coroutine_pc
Paolo Bonzini [Mon, 12 Oct 2015 08:02:54 +0000 (10:02 +0200)]
qemu-gdb: add $qemu_coroutine_sp and $qemu_coroutine_pc

These can be useful to manually get a stack trace of a coroutine inside
a core dump.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 1444636974-19950-4-git-send-email-pbonzini@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agoqemu-gdb: extract parts of "qemu coroutine" implementation
Paolo Bonzini [Mon, 12 Oct 2015 08:02:53 +0000 (10:02 +0200)]
qemu-gdb: extract parts of "qemu coroutine" implementation

Provide useful Python functions to reach and decipher a jmpbuf.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 1444636974-19950-3-git-send-email-pbonzini@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agoqemu-gdb: allow using glibc_pointer_guard() on core dumps
Paolo Bonzini [Mon, 12 Oct 2015 08:02:52 +0000 (10:02 +0200)]
qemu-gdb: allow using glibc_pointer_guard() on core dumps

get_fs_base() cannot be run on a core dump, because it uses the arch_prctl
system call.  The fs base is the value that is returned by pthread_self(),
and it would be nice to just glean it from the "info threads" output:

* 1    Thread 0x7f16a3fff700 (LWP 33642) pthread_cond_wait@@GLIBC_2.3.2 ()
              ^^^^^^^^^^^^^^

but unfortunately the gdb API does not provide that.  Instead, we can
look for the "arg" argument of the start_thread function if glibc debug
information are available.  If not, fall back to the old mechanism.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 1444636974-19950-2-git-send-email-pbonzini@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agotarget-mips: Add enum for BREAK32
Yongbok Kim [Fri, 2 Oct 2015 16:50:50 +0000 (17:50 +0100)]
target-mips: Add enum for BREAK32

Add enum for BREAK32

Signed-off-by: Yongbok Kim <yongbok.kim@imgtec.com>
Reviewed-by: Leon Alrae <leon.alrae@imgtec.com>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
9 years agotarget-mips: update writing to CP0.Status.KX/SX/UX in MIPS Release R6
Leon Alrae [Mon, 14 Sep 2015 12:51:31 +0000 (13:51 +0100)]
target-mips: update writing to CP0.Status.KX/SX/UX in MIPS Release R6

Implement the relationship between CP0.Status.KX, SX and UX. It should not
be possible to set UX bit if SX is 0, the same applies for setting SX if
KX is 0.

Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
9 years agotarget-mips: implement the CPU wake-up on non-enabled interrupts in R6
Leon Alrae [Mon, 14 Sep 2015 12:58:24 +0000 (13:58 +0100)]
target-mips: implement the CPU wake-up on non-enabled interrupts in R6

In Release 6, the behaviour of WAIT has been modified to make it a
requirement that a processor that has disabled operation as a result of
executing a WAIT will resume operation on arrival of an interrupt even if
interrupts are not enabled.

Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
9 years agotarget-mips: move the test for enabled interrupts to a separate function
Leon Alrae [Mon, 14 Sep 2015 12:58:23 +0000 (13:58 +0100)]
target-mips: move the test for enabled interrupts to a separate function

Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
9 years agomonitor: Split MonitorQAPIEventConf off MonitorQAPIEventState
Markus Armbruster [Thu, 15 Oct 2015 15:08:33 +0000 (17:08 +0200)]
monitor: Split MonitorQAPIEventConf off MonitorQAPIEventState

In preparation of turning monitor_qapi_event_state[] into a hash table
for finer grained throttling.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <1444921716-9511-5-git-send-email-armbru@redhat.com>

9 years agomonitor: Switch from timer_new() to timer_new_ns()
Markus Armbruster [Thu, 15 Oct 2015 15:08:32 +0000 (17:08 +0200)]
monitor: Switch from timer_new() to timer_new_ns()

We don't actually care for the scale, so we can just as well use the
simpler interface.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <1444921716-9511-4-git-send-email-armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
9 years agomonitor: Simplify event throttling
Markus Armbruster [Thu, 15 Oct 2015 15:08:31 +0000 (17:08 +0200)]
monitor: Simplify event throttling

The event throttling state machine is hard to understand.  I'm not
sure it's entirely correct.  Rewrite it in a more straightforward
manner:

State 1: No event sent recently (less than evconf->rate ns ago)

    Invariant: evstate->timer is not pending, evstate->qdict is null

    On event: send event, arm timer, goto state 2

State 2: Event sent recently, no additional event being delayed

    Invariant: evstate->timer is pending, evstate->qdict is null

    On event: store it in evstate->qdict, goto state 3

    On timer: goto state 1

State 3: Event sent recently, additional event being delayed

    Invariant: evstate->timer is pending, evstate->qdict is non-null

    On event: store it in evstate->qdict, goto state 3

    On timer: send evstate->qdict, clear evstate->qdict,
              arm timer, goto state 2

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <1444921716-9511-3-git-send-email-armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
9 years agomonitor: Reduce casting of QAPI event QDict
Markus Armbruster [Thu, 15 Oct 2015 15:08:30 +0000 (17:08 +0200)]
monitor: Reduce casting of QAPI event QDict

Make the variables holding the event QDict instead of QObject.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <1444921716-9511-2-git-send-email-armbru@redhat.com>

9 years agoqstring: Make conversion from QObject * accept null
Markus Armbruster [Thu, 15 Oct 2015 14:15:37 +0000 (16:15 +0200)]
qstring: Make conversion from QObject * accept null

qobject_to_qstring() crashes on null, which is a trap for the unwary.
Return null instead, and simplify a few callers.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <1444918537-18107-7-git-send-email-armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Luiz Capitulino <lcapitulino@redhat.com>
9 years agoqlist: Make conversion from QObject * accept null
Markus Armbruster [Thu, 15 Oct 2015 14:15:36 +0000 (16:15 +0200)]
qlist: Make conversion from QObject * accept null

qobject_to_qlist() crashes on null, which is a trap for the unwary.
Return null instead.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <1444918537-18107-6-git-send-email-armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Luiz Capitulino <lcapitulino@redhat.com>
9 years agoqfloat qint: Make conversion from QObject * accept null
Markus Armbruster [Thu, 15 Oct 2015 14:15:35 +0000 (16:15 +0200)]
qfloat qint: Make conversion from QObject * accept null

qobject_to_qfloat() and qobject_to_qint() crash on null, which is a
trap for the unwary.  Return null instead, and simplify a few callers.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <1444918537-18107-5-git-send-email-armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Luiz Capitulino <lcapitulino@redhat.com>
9 years agoqdict: Make conversion from QObject * accept null
Markus Armbruster [Thu, 15 Oct 2015 14:15:34 +0000 (16:15 +0200)]
qdict: Make conversion from QObject * accept null

qobject_to_qdict() crashes on null, which is a trap for the unwary.
Return null instead, and simplify a few callers.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <1444918537-18107-4-git-send-email-armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Luiz Capitulino <lcapitulino@redhat.com>
9 years agoqbool: Make conversion from QObject * accept null
Markus Armbruster [Thu, 15 Oct 2015 14:15:33 +0000 (16:15 +0200)]
qbool: Make conversion from QObject * accept null

qobject_to_qbool() crashes on null, which is a trap for the unwary.
Return null instead, and simplify a few callers.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <1444918537-18107-3-git-send-email-armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Luiz Capitulino <lcapitulino@redhat.com>
9 years agoqobject: Drop QObject_HEAD
Markus Armbruster [Thu, 15 Oct 2015 14:15:32 +0000 (16:15 +0200)]
qobject: Drop QObject_HEAD

QObject_HEAD is a macro expanding into the common part of structs that
are sub-types of QObject.  It's always been just QObject base, and
unlikely to change.  Drop the macro, because the code is clearer with
out it.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <1444918537-18107-2-git-send-email-armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Luiz Capitulino <lcapitulino@redhat.com>
9 years agoMerge remote-tracking branch 'remotes/mst/tags/for_upstream' into staging
Peter Maydell [Thu, 29 Oct 2015 09:49:52 +0000 (09:49 +0000)]
Merge remote-tracking branch 'remotes/mst/tags/for_upstream' into staging

virtio, pc, memory: fixes+features for 2.5

New features:
    This enables hotplug for multifunction devices.
    Patches are very small, so I think it's OK to merge
    at this stage.

    There's also some new infrastructure for vhost-user testing
    not enabled yet so it's harmless to merge.

I've reverted the "gap between DIMMs" workaround, as it seems too risky, and
applied my own patch in virtio, but not in dataplane code.  This means that
dataplane is broken for some complex DIMM configurations for now.  Waiting for
Stefan to review the dataplane fix.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
# gpg: Signature made Thu 29 Oct 2015 09:36:16 GMT using RSA key ID D28D5469
# gpg: Good signature from "Michael S. Tsirkin <mst@kernel.org>"
# gpg:                 aka "Michael S. Tsirkin <mst@redhat.com>"

* remotes/mst/tags/for_upstream:
  enable multi-function hot-add
  remove function during multi-function hot-add
  tests/vhost-user-bridge: add vhost-user bridge application
  Revert "memhp: extend address auto assignment to support gaps"
  Revert "pc: memhp: force gaps between DIMM's GPA"
  virtio: drop virtqueue_map_sg
  virtio-scsi: convert to virtqueue_map
  virtio-serial: convert to virtio_map
  virtio-blk: convert to virtqueue_map
  virtio: switch to virtio_map
  virtio: introduce virtio_map
  mmap-alloc: fix error handling
  pc: memhp: do not emit inserting event for coldplugged DIMMs
  vhost-user-test: fix up rhel6 build
  vhost-user: cleanup msg size math
  vhost-user: cleanup struct size math

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agoenable multi-function hot-add
Cao jin [Wed, 28 Oct 2015 06:20:31 +0000 (14:20 +0800)]
enable multi-function hot-add

Enable PCIe device multi-function hot-add, just ensure function 0 is added
last, then driver will get the notification to scan the slot.

Signed-off-by: Cao jin <caoj.fnst@cn.fujitsu.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
9 years agoremove function during multi-function hot-add
Cao jin [Wed, 28 Oct 2015 06:20:30 +0000 (14:20 +0800)]
remove function during multi-function hot-add

In case user want to cancel the hot-add operation, should roll back,
device_del the added function that still don`t work.

Signed-off-by: Cao jin <caoj.fnst@cn.fujitsu.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
9 years agotests/vhost-user-bridge: add vhost-user bridge application
Victor Kaplansky [Wed, 28 Oct 2015 12:53:07 +0000 (14:53 +0200)]
tests/vhost-user-bridge: add vhost-user bridge application

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: Victor Kaplansky <victork@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
9 years agoRevert "memhp: extend address auto assignment to support gaps"
Michael S. Tsirkin [Wed, 28 Oct 2015 16:55:06 +0000 (18:55 +0200)]
Revert "memhp: extend address auto assignment to support gaps"

This reverts commit df0acded19ec4b826aa095cfc19d341bd66fafd3.

There's no point to it now that the only user has been reverted.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
9 years agoRevert "pc: memhp: force gaps between DIMM's GPA"
Michael S. Tsirkin [Wed, 28 Oct 2015 16:54:05 +0000 (18:54 +0200)]
Revert "pc: memhp: force gaps between DIMM's GPA"

This reverts commit aa8580cddf011e8cedcf87f7a0fdea7549fc4704.

As described in
http://article.gmane.org/gmane.comp.emulators.qemu/371432
that commit causes linux guests to crash on memory hot-unplug.

The original problem it's trying to solve has now
been addressed within virtio.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
9 years agovirtio: drop virtqueue_map_sg
Michael S. Tsirkin [Tue, 27 Oct 2015 08:22:59 +0000 (10:22 +0200)]
virtio: drop virtqueue_map_sg

Deprecated in favor of virtqueue_map.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
9 years agovirtio-scsi: convert to virtqueue_map
Michael S. Tsirkin [Tue, 27 Oct 2015 08:22:13 +0000 (10:22 +0200)]
virtio-scsi: convert to virtqueue_map

Note: virtqueue_map already validates input
so virtio-scsi does not have to.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
9 years agovirtio-serial: convert to virtio_map
Michael S. Tsirkin [Tue, 27 Oct 2015 08:19:43 +0000 (10:19 +0200)]
virtio-serial: convert to virtio_map

This also fixes a minor bug:
-                virtqueue_map_sg(port->elem.out_sg, port->elem.out_addr,
-                                 port->elem.out_num, 1);
is wrong: out_sg is not written so should not be marked dirty.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
9 years agovirtio-blk: convert to virtqueue_map
Michael S. Tsirkin [Tue, 27 Oct 2015 08:18:24 +0000 (10:18 +0200)]
virtio-blk: convert to virtqueue_map

Drop deprecated use of virtqueue_map_sg.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
9 years agovirtio: switch to virtio_map
Michael S. Tsirkin [Tue, 27 Oct 2015 08:09:16 +0000 (10:09 +0200)]
virtio: switch to virtio_map

Drop use of the deprecated virtio_map_sg in virtio core.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
9 years agovirtio: introduce virtio_map
Michael S. Tsirkin [Tue, 27 Oct 2015 08:01:44 +0000 (10:01 +0200)]
virtio: introduce virtio_map

virtio_map_sg currently fails if one of the entries it's mapping is
contigious in GPA but not HVA address space.  Introduce virtio_map which
handles this by splitting sg entries.

This new API generally turns out to be a good idea since it's harder to
misuse: at least in one case the existing one was used incorrectly.

This will still fail if there's no space left in the sg, but luckily max
queue size in use is currently 256, while max sg size is 1024, so we
should be OK even is all entries happen to cross a single DIMM boundary.

Won't work well with very small DIMM sizes, unfortunately:
e.g. this will fail with 4K DIMMs where a single
request might span a large number of DIMMs.

Let's hope these are uncommon - at least we are not breaking things.

Note: virtio-scsi calls virtio_map_sg on data loaded from network, and
validates input, asserting on failure.  Copy the validating code here -
it will be dropped from virtio-scsi in a follow-up patch.

Reported-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
9 years agommap-alloc: fix error handling
Michael S. Tsirkin [Sun, 25 Oct 2015 15:07:45 +0000 (17:07 +0200)]
mmap-alloc: fix error handling

Existing callers are checking for MAP_FAILED,
so we should return that on error.

Reported-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
9 years agopc: memhp: do not emit inserting event for coldplugged DIMMs
Igor Mammedov [Fri, 23 Oct 2015 12:55:26 +0000 (14:55 +0200)]
pc: memhp: do not emit inserting event for coldplugged DIMMs

currently acpi_memory_plug_cb() sets is_inserting for
cold- and hot-plugged DIMMs as result ASL MHPD.MSCN()
method issues device check even for every coldplugged
DIMM. There isn't much harm in it but if we try to
unplug such DIMM, OSPM will issue device check
intstead of device eject event. So OSPM won't eject
memory module as expected and it will try to eject it
only when another memory device is hot-(un)plugged.

As a fix do not set 'is_inserting' event and do not
issue SCI for cold-plugged DIMMs as they are
enumerated and activated by OSPM during guest's boot.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
9 years agovhost-user-test: fix up rhel6 build
Michael S. Tsirkin [Thu, 22 Oct 2015 19:28:37 +0000 (22:28 +0300)]
vhost-user-test: fix up rhel6 build

Build on RHEL6 fails:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=42875

Apparently unnamed unions couldn't use C99  named field initializers.
Let's just name the payload union field.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
9 years agovhost-user: cleanup msg size math
Michael S. Tsirkin [Thu, 22 Oct 2015 19:33:39 +0000 (22:33 +0300)]
vhost-user: cleanup msg size math

We are sending msg fields, use sizeof on these
and not on local variables which happen to
have a matching type.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
9 years agovhost-user: cleanup struct size math
Michael S. Tsirkin [Thu, 22 Oct 2015 19:31:28 +0000 (22:31 +0300)]
vhost-user: cleanup struct size math

We are using local msg structures everywhere, use them
for sizeof as well.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
9 years agoMerge remote-tracking branch 'remotes/rth/tags/pull-tcg-20151028' into staging
Peter Maydell [Wed, 28 Oct 2015 20:10:22 +0000 (20:10 +0000)]
Merge remote-tracking branch 'remotes/rth/tags/pull-tcg-20151028' into staging

Breakpoint fixes

# gpg: Signature made Wed 28 Oct 2015 17:58:52 GMT using RSA key ID 4DD0279B
# gpg: Good signature from "Richard Henderson <rth7680@gmail.com>"
# gpg:                 aka "Richard Henderson <rth@redhat.com>"
# gpg:                 aka "Richard Henderson <rth@twiddle.net>"

* remotes/rth/tags/pull-tcg-20151028:
  target-*: Advance pc after recognizing a breakpoint

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-*: Advance pc after recognizing a breakpoint
Richard Henderson [Tue, 13 Oct 2015 22:07:49 +0000 (22:07 +0000)]
target-*: Advance pc after recognizing a breakpoint

Some targets already had this within their logic, but make sure
it's present for all targets.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
9 years agoMerge remote-tracking branch 'remotes/ehabkost/tags/x86-pull-request' into staging
Peter Maydell [Wed, 28 Oct 2015 15:08:36 +0000 (15:08 +0000)]
Merge remote-tracking branch 'remotes/ehabkost/tags/x86-pull-request' into staging

target-i386: finally enable "check" mode by default

# gpg: Signature made Wed 28 Oct 2015 14:13:10 GMT using RSA key ID 984DC5A6
# gpg: Good signature from "Eduardo Habkost <ehabkost@redhat.com>"

* remotes/ehabkost/tags/x86-pull-request:
  target-i386: Enable "check" mode by default
  target-i386: Don't left shift negative constant

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agoMerge remote-tracking branch 'remotes/mcayland/tags/qemu-openbios-signed' into staging
Peter Maydell [Wed, 28 Oct 2015 14:02:27 +0000 (14:02 +0000)]
Merge remote-tracking branch 'remotes/mcayland/tags/qemu-openbios-signed' into staging

Update OpenBIOS images

# gpg: Signature made Wed 28 Oct 2015 00:02:46 GMT using RSA key ID AE0F321F
# gpg: Good signature from "Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>"

* remotes/mcayland/tags/qemu-openbios-signed:
  Update OpenBIOS images

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agoUpdate OpenBIOS images
Mark Cave-Ayland [Wed, 28 Oct 2015 00:01:28 +0000 (00:01 +0000)]
Update OpenBIOS images

Update OpenBIOS images to SVN r1353 built from submodule.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
9 years agotarget-i386: Enable "check" mode by default
Eduardo Habkost [Wed, 26 Aug 2015 16:25:44 +0000 (13:25 -0300)]
target-i386: Enable "check" mode by default

Current default behavior of QEMU is to silently disable features that
are not supported by the host when a CPU model is requested in the
command-line. This means that in addition to risking breaking guest ABI
by default, we are silent about it.

I would like to enable "enforce" by default, but this can easily break
existing production systems because of the way libvirt makes assumptions
about CPU models today (this will change in the future, once QEMU
provide a proper interface for checking if a CPU model is runnable).

But there's no reason we should be silent about it. So, change
target-i386 to enable "check" mode by default so at least we have some
warning printed to stderr (and hopefully logged somewhere) when QEMU
disables a feature that is not supported by the host system.

Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
9 years agotarget-i386: Don't left shift negative constant
Eduardo Habkost [Tue, 29 Sep 2015 20:34:23 +0000 (17:34 -0300)]
target-i386: Don't left shift negative constant

Left shift of negative values is undefined behavior. Detected by clang:
  qemu/target-i386/translate.c:2423:26: runtime error:
    left shift of negative value -8

This changes the code to reverse the sign after the left shift.

Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
9 years agoMerge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20151027-1' into...
Peter Maydell [Tue, 27 Oct 2015 16:17:55 +0000 (16:17 +0000)]
Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20151027-1' into staging

target-arm queue:
 * more EL2 preparation: handling for stage 2 translations
 * standardize debug macros in i.MX devices
 * improve error message in a corner case for virt board
 * disable live migration of KVM GIC if the kernel can't handle it
 * add SPSR_(ABT|UND|IRQ|FIQ) registers
 * handle non-executable page-straddling Thumb instructions
 * fix a "no 64-bit EL2" assumption in arm_excp_unmasked()

# gpg: Signature made Tue 27 Oct 2015 16:03:31 GMT using RSA key ID 14360CDE
# gpg: Good signature from "Peter Maydell <peter.maydell@linaro.org>"
# gpg:                 aka "Peter Maydell <pmaydell@gmail.com>"
# gpg:                 aka "Peter Maydell <pmaydell@chiark.greenend.org.uk>"

* remotes/pmaydell/tags/pull-target-arm-20151027-1: (27 commits)
  target-arm: Add support for S1 + S2 MMU translations
  target-arm: Route S2 MMU faults to EL2
  target-arm: Add S2 translation to 32bit S1 PTWs
  target-arm: Add S2 translation to 64bit S1 PTWs
  target-arm: Add ARMMMUFaultInfo
  target-arm: Avoid inline for get_phys_addr
  target-arm: Add support for S2 page-table protection bits
  target-arm: Add computation of starting level for S2 PTW
  target-arm: lpae: Rename granule_sz to stride
  target-arm: lpae: Replace tsz with computed inputsize
  target-arm: Add support for AArch32 S2 negative t0sz
  target-arm: lpae: Move declaration of t0sz and t1sz
  target-arm: lpae: Make t0sz and t1sz signed integers
  target-arm: Add HPFAR_EL2
  i.MX: Standardize i.MX GPT debug
  i.MX: Standardize i.MX EPIT debug
  i.MX: Standardize i.MX FEC debug
  i.MX: Standardize i.MX CCM debug
  i.MX: Standardize i.MX AVIC debug
  i.MX: Standardize i.MX I2C debug
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: Add support for S1 + S2 MMU translations
Edgar E. Iglesias [Mon, 26 Oct 2015 13:02:07 +0000 (14:02 +0100)]
target-arm: Add support for S1 + S2 MMU translations

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1445864527-14520-15-git-send-email-edgar.iglesias@gmail.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: Route S2 MMU faults to EL2
Edgar E. Iglesias [Mon, 26 Oct 2015 13:02:06 +0000 (14:02 +0100)]
target-arm: Route S2 MMU faults to EL2

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1445864527-14520-14-git-send-email-edgar.iglesias@gmail.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: Add S2 translation to 32bit S1 PTWs
Edgar E. Iglesias [Mon, 26 Oct 2015 13:02:05 +0000 (14:02 +0100)]
target-arm: Add S2 translation to 32bit S1 PTWs

Add support for applying S2 translation to 32bit S1
page-table walks.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1445864527-14520-13-git-send-email-edgar.iglesias@gmail.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: Add S2 translation to 64bit S1 PTWs
Edgar E. Iglesias [Mon, 26 Oct 2015 13:02:04 +0000 (14:02 +0100)]
target-arm: Add S2 translation to 64bit S1 PTWs

Add support for applying S2 translation to 64bit S1
page-table walks.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1445864527-14520-12-git-send-email-edgar.iglesias@gmail.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: Add ARMMMUFaultInfo
Edgar E. Iglesias [Mon, 26 Oct 2015 13:02:03 +0000 (14:02 +0100)]
target-arm: Add ARMMMUFaultInfo

Introduce ARMMMUFaultInfo to propagate MMU Fault information
across the MMU translation code path. This is in preparation for
adding Stage-2 translation.

No functional changes.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1445864527-14520-11-git-send-email-edgar.iglesias@gmail.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: Avoid inline for get_phys_addr
Edgar E. Iglesias [Mon, 26 Oct 2015 13:02:02 +0000 (14:02 +0100)]
target-arm: Avoid inline for get_phys_addr

Avoid inline for get_phys_addr() to prepare for future recursive use.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1445864527-14520-10-git-send-email-edgar.iglesias@gmail.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: Add support for S2 page-table protection bits
Edgar E. Iglesias [Mon, 26 Oct 2015 13:02:01 +0000 (14:02 +0100)]
target-arm: Add support for S2 page-table protection bits

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1445864527-14520-9-git-send-email-edgar.iglesias@gmail.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>