Returning a partial object on error is an invitation for a careless
caller to leak memory. We already fixed things in an earlier
patch to guarantee NULL if visit_start fails ("qapi: Guarantee
NULL obj on input visitor callback error"), but that does not
help the case where visit_start succeeds but some other failure
happens before visit_end, such that we leak a partially constructed
object outside visit_type_FOO(). As no one outside the testsuite
was actually relying on these semantics, it is cleaner to just
document and guarantee that ALL pointer-based visit_type_FOO()
functions always leave a safe value in *obj during an input visitor
(either the new object on success, or NULL if an error is
encountered), so callers can now unconditionally use
qapi_free_FOO() to clean up regardless of whether an error occurred.
The decision is done by adding visit_is_input(), then updating the
generated code to check if additional cleanup is needed based on
the type of visitor in use.
Note that we still leave *obj unchanged after a scalar-based
visit_type_FOO(); I did not feel like auditing all uses of
visit_type_Enum() to see if the callers would tolerate a specific
sentinel value (not to mention having to decide whether it would
be better to use 0 or ENUM__MAX as that sentinel).
Backports commit 68ab47e4b4ecc1c4649362b8cc1e49794d1a6537 from qemu
The semantics of the list visit are somewhat baroque, with the
following pseudocode when FooList is used:
start()
for (prev = head; cur = next(prev); prev = &cur) {
visit(&cur->value)
}
Note that these semantics (advance before visit) requires that
the first call to next() return the list head, while all other
calls return the next element of the list; that is, every visitor
implementation is required to track extra state to decide whether
to return the input as-is, or to advance. It also requires an
argument of 'GenericList **' to next(), solely because the first
iteration might need to modify the caller's GenericList head, so
that all other calls have to do a layer of dereferencing.
Thankfully, we only have two uses of list visits in the entire
code base: one in spapr_drc (which completely avoids
visit_next_list(), feeding in integers from a different source
than uint8List), and one in qapi-visit.py. That is, all other
list visitors are generated in qapi-visit.c, and share the same
paradigm based on a qapi FooList type, so we can refactor how
lists are laid out with minimal churn among clients.
We can greatly simplify things by hoisting the special case
into the start() routine, and flipping the order in the loop
to visit before advance:
start(head)
for (tail = *head; tail; tail = next(tail)) {
visit(&tail->value)
}
With the simpler semantics, visitors have less state to track,
the argument to next() is reduced to 'GenericList *', and it
also becomes obvious whether an input visitor is allocating a
FooList during visit_start_list() (rather than the old way of
not knowing if an allocation happened until the first
visit_next_list()). As a minor drawback, we now allocate in
two functions instead of one, and have to pass the size to
both functions (unless we were to tweak the input visitors to
cache the size to start_list for reuse during next_list, but
that defeats the goal of less visitor state).
The signature of visit_start_list() is chosen to match
visit_start_struct(), with the new parameters after 'name'.
The spapr_drc case is a virtual visit, done by passing NULL for
list, similarly to how NULL is passed to visit_start_struct()
when a qapi type is not used in those visits. It was easy to
provide these semantics for qmp-output and dealloc visitors,
and a bit harder for qmp-input (several prerequisite patches
refactored things to make this patch straightforward). But it
turned out that the string and opts visitors munge enough other
state during visit_next_list() to make it easier to just
document and require a GenericList visit for now; an assertion
will remind us to adjust things if we need the semantics in the
future.
Several pre-requisite cleanup patches made the reshuffling of
the various visitors easier; particularly the qmp input visitor.
Backports commit d9f62dde1303286b24ac8ce88be27e2b9b9c5f46 from qemu
As shown in the previous commit, the string input visitor was
treating bogus input as an empty list rather than an error.
Fix parse_str() to set errp, then the callers to exit early if
an error was reported.
Meanwhile, fix the testsuite to use the generated
qapi_free_int16List() instead of rolling our own, and to
validate the fixed behavior, while at the same time documenting
one more change that we'd like to make in a later patch (a
failed visit_start_list should guarantee a NULL pointer,
regardless of what things were on input).
Backports commit 74f24cb6306d065045d0e2215a7d10533fa59c57 from qemu
As mentioned in previous patches, we want to call visit_end_struct()
functions unconditionally, so that visitors can release resources
tied up since the matching visit_start_struct() without also having
to worry about error priority if more than one error occurs.
Even though error_propagate() can be safely used to ignore a second
error during cleanup caused by a first error, it is simpler if the
cleanup cannot set an error. So, split out the error checking
portion (basically, input visitors checking for unvisited keys) into
a new function visit_check_struct(), which can be safely skipped if
any earlier errors are encountered, and leave the cleanup portion
(which never fails, but must be called unconditionally if
visit_start_struct() succeeded) in visit_end_struct().
Generated code in qapi-visit.c has diffs resembling:
|@@ -59,10 +59,12 @@ void visit_type_ACPIOSTInfo(Visitor *v,
| goto out_obj;
| }
| visit_type_ACPIOSTInfo_members(v, obj, &err);
|- error_propagate(errp, err);
|- err = NULL;
|+ if (err) {
|+ goto out_obj;
|+ }
|+ visit_check_struct(v, &err);
| out_obj:
|- visit_end_struct(v, &err);
|+ visit_end_struct(v);
| out:
and in qapi-event.c:
@@ -47,7 +47,10 @@ void qapi_event_send_acpi_device_ost(ACP
| goto out;
| }
| visit_type_q_obj_ACPI_DEVICE_OST_arg_members(v, ¶m, &err);
|- visit_end_struct(v, err ? NULL : &err);
|+ if (!err) {
|+ visit_check_struct(v, &err);
|+ }
|+ visit_end_struct(v);
| if (err) {
| goto out;
Backports commit 15c2f669e3fb2bc97f7b42d1871f595c0ac24af8 from qemu
Tighten assertions in the QMP output visitor, so that:
- qmp_output_get_qobject() can only be called after pairing a
visit_end_* for every visit_start_* (rather than allowing it on
a partially built object)
- qmp_output_get_qobject() cannot be called unless at least one
visit_type_* or visit_start/visit_end pair has occurred since
creation/reset (the accidental return of NULL fixed by commit
ab8bf1d7 would have been much easier to diagnose)
- ensure that we are encountering the expected object or list
type, to provide protection against mismatched push(struct)/
pop(list) or push(list)/pop(struct), similar to the qmp-input
protection added in commit bdd8e6b5.
- ensure that except for the root, 'name' is non-null inside a
dict, and NULL inside a list (this may need changing later if
we add "name.0" support for better error messages for a list,
but for now it makes sure all users are at least consistent)
Backports commit 56a6f02b8ce1fe41a2a9077593e46eca7d98267d from qemu
Implement the new type_null() callback for the qmp input and
output visitors. While we don't yet have a use for this in QAPI
input (the generator will need some tweaks first), some
potential usages have already been discussed on the list.
Meanwhile, the output visitor could already output explicit null
via type_any, but this gives us finer control.
At any rate, it's easy to test that we can round-trip an explicit
null through manual use of visit_type_null() wrapped by a virtual
visit_start_struct() walk, even if we can't do the visit in a
QAPI type. Repurpose the test_visitor_out_empty test,
particularly since a future patch will tighten semantics to
forbid use of qmp_output_get_qobject() without at least one
intervening visit_type_*.
Backports commit 3df016f185521f8dfa5bd89168722887156405c7 from qemu
Right now, qmp-output-visitor happens to produce a QNull result
if nothing is actually visited between the creation of the visitor
and the request for the resulting QObject. A stronger protocol
would require that a QMP output visit MUST visit something. But
to still be able to produce a JSON 'null' output, we need a new
visitor function that states our intentions. Yes, we could say
that such a visit must go through visit_type_any(), but that
feels clunky.
So this patch introduces the new visit_type_null() interface and
its no-op interface in the dealloc visitor, and stubs in the
qmp visitors (the next patch will finish the implementation).
For the visitors that will not implement the callback, document
the situation. The code in qapi-visit-core unconditionally
dereferences the callback pointer, so that a segfault will inform
a developer if they need to implement the callback for their
choice of visitor.
Note that JSON has a primitive null type, with the single value
null; likewise with the QNull type for QObject; but for QAPI,
we just have the 'null' value without a null type. We may
eventually want to add more support in QAPI for null (most likely,
we'd use it via an alternate type that permits 'null' or an
object); but we'll create that usage when we need it.
Backports commit 3bc97fd5924561d92f32758c67eaffd2e4e25038 from qemu
The visitor interface for mapping between QObject/QemuOpts/string
and QAPI is scandalously under-documented, making changes to visitor
core, individual visitors, and users of visitors difficult to
coordinate. Among other questions: when is it safe to pass NULL,
vs. when a string must be provided; which visitors implement which
callbacks; the difference between concrete and virtual visits.
Correct this by retrofitting proper contracts, and document where some
of the interface warts remain (for example, we may want to modify
visit_end_* to require the same 'obj' as the visit_start counterpart,
so the dealloc visitor can be simplified). Later patches in this
series will tackle some, but not all, of these warts.
Add assertions to (partially) enforce the contract. Some of these
were only made possible by recent cleanup commits.
Backports commit adfb264c9ed04bfc694921b72173be8e29e90024 from qemu
After recent changes, the only remaining use of
visit_start_implicit_struct() is for allocating the space needed
when visiting an alternate. Since the term 'implicit struct' is
hard to explain, rename the function to its current usage. While
at it, we can merge the functionality of visit_get_next_type()
into the same function, making it more like visit_start_struct().
Generated code is now slightly smaller:
| {
| Error *err = NULL;
|
|- visit_start_implicit_struct(v, (void**) obj, sizeof(BlockdevRef), &err);
|+ visit_start_alternate(v, name, (GenericAlternate **)obj, sizeof(**obj),
|+ true, &err);
| if (err) {
| goto out;
| }
|- visit_get_next_type(v, name, &(*obj)->type, true, &err);
|- if (err) {
|- goto out_obj;
|- }
| switch ((*obj)->type) {
| case QTYPE_QDICT:
| visit_start_struct(v, name, NULL, 0, &err);
...
| }
|-out_obj:
|- visit_end_implicit_struct(v);
|+ visit_end_alternate(v);
| out:
| error_propagate(errp, err);
| }
Backports commit dbf11922622685934bfb41e7cf2be9bd4a0405c0 from qemu
In the QMP input visitor, visiting a list traverses two objects:
the QAPI GenericList of the caller (which gets advanced in
visit_next_list() regardless of this patch), and the QList input
that we are converting to QAPI. For consistency with QDict
visits, we want to consume elements from the input QList during
the visit_type_FOO() for the list element; that is, we want ALL
the code for consuming an input to live in qmp_input_get_object(),
rather than having it split according to whether we are visiting
a dict or a list. Making qmp_input_get_object() the common point
of consumption will make it easier for a later patch to refactor
visit_start_list() to cover the GenericList * head of a QAPI list,
and in turn will get rid of the 'first' flag (which lived in
qmp_input_next_list() pre-patch, and is hoisted to StackObject
by this patch).
This patch is therefore altering the post-condition use of 'entry',
while keeping what gets visited unchanged, from:
start_list next_list type_ELT ... next_list type_ELT next_list end_list
visits 1st elt last elt
entry NULL 1st elt 1st elt last elt last elt NULL gone
where type_ELT() returns (entry ? entry : 1st elt) and next_list() steps
entry
to this usage:
start_list next_list type_ELT ... next_list type_ELT next_list end_list
visits 1st elt last elt
entry 1st elt 1nd elt 2nd elt last elt NULL NULL gone
where type_ELT() steps entry and returns the old entry, and next_list()
leaves entry alone.
Backports commit fcf3cb21783b2dae3358fdbe7001cb2f74e0cedf from qemu
Don't embed the root of the visit into the stack of current
containers being visited. That way, we no longer get confused
on whether the first visit of a dictionary is to the dictionary
itself or to one of the members of the dictionary, based on
whether the caller passed name=NULL; and makes the QMP Input
visitor like other visitors where the value of 'name' is now
ignored on the root visit. (We may someday want to revisit
the rules on what 'name' should be on a top-level visit,
rather than just ignoring it; but that would be the topic of
another patch).
An audit of all qmp_input_visitor_new() call sites shows that
there were only two places where callers had previously been
visiting to a QDict with a non-NULL name to bypass a call to
visit_start_struct(), and those were fixed in prior patches.
Backports commit ce140b176920b5b65184020735a3c65ed3e9aeda from qemu
Commit e8316d7 mistakenly passed consume=true within
qmp_input_optional() when checking if an optional member was
present, but the mistake was silently ignored since the code
happily let us extract a member more than once. Fix
qmp_input_optional() to not consume anything, then tighten up
the input visitor to ensure that a member is consumed exactly
once (all generated code follows this pattern; and the new
assert will catch any hand-written code that tries to visit
the same key more than once).
Backports commit e5826a2fd727f0be54a81083f31fe02a275465cd from qemu
Rather than having two separate ways to create a QMP input
visitor, where the safer approach has the more verbose name,
it is better to consolidate things into a single function
where the caller must explicitly choose whether to be strict
or to ignore excess input. This patch is the strictly
mechanical conversion; the next patch will then audit which
uses can be made stricter.
Backports commit fc471c18d5d2ec713d5a019f9530398675494bc8 from qemu
Management of the top of stack was a bit verbose; creating a
temporary variable and adding some comments makes the existing
code more legible before the next few patches improve things.
No semantic changes other than asserting that we are always
visiting a QObject, and not a NULL value. In particular, the
check for 'name && qobject_type(qobj) == QTYPE_QDICT)' is a
bit overkill (a dict visit should always have a name); a later
patch revisits that, while this patch is only changing one
layer of indentation due to dropping 'if (qobj)'.
Backports commit b471d012e5d7bec1d2272738141e121b5581fcdf from qemu
Our existing input visitors were not very consistent on errors in a
function taking 'TYPE **obj'. These are start_struct(),
start_alternate(), type_str(), and type_any(). next_list() is
similar, but can't fail (see commit 08f9541). While all of them set
'*obj' to allocated storage on success, it was not obvious whether
'*obj' was guaranteed safe on failure, or whether it was left
uninitialized. But a future patch wants to guarantee that
visit_type_FOO() does not leak a partially-constructed obj back to
the caller; it is easier to implement this if we can reliably state
that input visitors assign '*obj' regardless of success or failure,
and that on failure *obj is NULL. Add assertions to enforce
consistency in the final setting of err vs. *obj.
The opts-visitor start_struct() doesn't set an error, but it
also was doing a weird check for 0 size; all callers pass in
non-zero size if obj is non-NULL.
The testsuite has at least one spot where we no longer need
to pre-initialize a variable prior to a visit; valgrind confirms
that the test is still fine with the cleanup.
A later patch will document the design constraint implemented
here.
Backports commit e58d695e6c3a5cfa0aa2fc91b87ade017ef28b05 from qemu
By sticking the next pointer first, we don't need a union with
64-bit padding for smaller types. On 32-bit platforms, this
can reduce the size of uint8List from 16 bytes (or 12, depending
on whether 64-bit ints can tolerate 4-byte alignment) down to 8.
It has no effect on 64-bit platforms (where alignment still
dictates a 16-byte struct); but fewer anonymous unions is still
a win in my book.
It requires visit_next_list() to gain a size parameter, to know
what size element to allocate; comparable to the size parameter
of visit_start_struct().
I debated about going one step further, to allow for fewer casts,
by doing:
typedef GenericList GenericList;
struct GenericList {
GenericList *next;
};
struct FooList {
GenericList base;
Foo *value;
};
so that you convert to 'GenericList *' by '&foolist->base', and
back by 'container_of(generic, GenericList, base)' (as opposed to
the existing '(GenericList *)foolist' and '(FooList *)generic').
But doing that would require hoisting the declaration of
GenericList prior to inclusion of qapi-types.h, rather than its
current spot in visitor.h; it also makes iteration a bit more
verbose through 'foolist->base.next' instead of 'foolist->next'.
Note that for lists of objects, the 'value' payload is still
hidden behind a boxed pointer. Someday, it would be nice to do:
struct FooList {
FooList *next;
Foo value;
};
for one less level of malloc for each list element. This patch
is a step in that direction (now that 'next' is no longer at a
fixed non-zero offset within the struct, we can store more than
just a pointer's-worth of data as the value payload), but the
actual conversion would be a task for another series, as it will
touch a lot of code.
Backports commit e65d89bf1a4484e0db0f3dc820a8b209f2fb1e8b from qemu
We have three classes of QAPI visitors: input, output, and dealloc.
Currently, all implementations of these visitors have one thing in
common based on their visitor type: the implementation used for the
visit_type_enum() callback. But since we plan to add more such
common behavior, in relation to documenting and further refining
the semantics, it makes more sense to have the visitor
implementations advertise which class they belong to, so the common
qapi-visit-core code can use that information in multiple places.
A later patch will better document the types of visitors directly
in visitor.h.
For this patch, knowing the class of a visitor implementation lets
us make input_type_enum() and output_type_enum() become static
functions, by replacing the callback function Visitor.type_enum()
with the simpler enum member Visitor.type. Share a common
assertion in qapi-visit-core as part of the refactoring.
Move comments in opts-visitor.c to match the refactored layout.
Backports commit 983f52d4b3f86fb9dc9f8b142132feb5a8723016 from qemu
Commit 57cb38b included qapi/error.h into qemu/osdep.h to get the
Error typedef. Since then, we've moved to include qemu/osdep.h
everywhere. Its file comment explains: "To avoid getting into
possible circular include dependencies, this file should not include
any other QEMU headers, with the exceptions of config-host.h,
compiler.h, os-posix.h and os-win32.h, all of which are doing a
similar job to this file and are under similar constraints."
qapi/error.h doesn't do a similar job, and it doesn't adhere to
similar constraints: it includes qapi-types.h. That's in excess of
100KiB of crap most .c files don't actually need.
Add the typedef to qemu/typedefs.h, and include that instead of
qapi/error.h. Include qapi/error.h in .c files that need it and don't
get it now. Include qapi-types.h in qom/object.h for uint16List.
Update scripts/clean-includes accordingly. Update it further to match
reality: replace config.h by config-target.h, add sysemu/os-posix.h,
sysemu/os-win32.h. Update the list of includes in the qemu/osdep.h
comment quoted above similarly.
This reduces the number of objects depending on qapi/error.h from "all
of them" to less than a third. Unfortunately, the number depending on
qapi-types.h shrinks only a little. More work is needed for that one.
Backports commit da34e65cb4025728566d6504a99916f6e7e1dd6a from qemu
There's no reason to do two malloc's for a flat union; let's just
inline the branch struct directly into the C union branch of the
flat union.
Surprisingly, fewer clients were actually using explicit references
to the branch types in comparison to the number of flat unions
thus modified.
This lets us reduce the hack in qapi-types:gen_variants() added in
the previous patch; we no longer need to distinguish between
alternates and flat unions.
The change to unboxed structs means that u.data (added in commit
cee2dedb) is now coincident with random fields of each branch of
the flat union, whereas beforehand it was only coincident with
pointers (since all branches of a flat union have to be objects).
Note that this was already the case for simple unions - but there
we got lucky. Remember, visit_start_union() blindly returns true
for all visitors except for the dealloc visitor, where it returns
the value !!obj->u.data, and that this result then controls
whether to proceed with the visit to the variant. Pre-patch,
this meant that flat unions were testing whether the boxed pointer
was still NULL, and thereby skipping visit_end_implicit_struct()
and avoiding a NULL dereference if the pointer had not been
allocated. The same was true for simple unions where the current
branch had pointer type, except there we bypassed visit_type_FOO().
But for simple unions where the current branch had scalar type, the
contents of that scalar meant that the decision to call
visit_type_FOO() was data-dependent - the reason we got lucky there
is that visit_type_FOO() for all scalar types in the dealloc visitor
is a no-op (only the pointer variants had anything to free), so it
did not matter whether the dealloc visit was skipped. But with this
patch, we would risk leaking memory if we could skip a call to
visit_type_FOO_fields() based solely on a data-dependent decision.
But notice: in the dealloc visitor, visit_type_FOO() already handles
a NULL obj - it was only the visit_type_implicit_FOO() that was
failing to check for NULL. And now that we have refactored things to
have the branch be part of the parent struct, we no longer have a
separate pointer that can be NULL in the first place. So we can just
delete the call to visit_start_union() altogether, and blindly visit
the branch type; there is no change in behavior except to the dealloc
visitor, where we now unconditionally visit the branch, but where that
visit is now always safe (for a flat union, we can no longer
dereference NULL, and for a simple union, visit_type_FOO() was already
safely handling NULL on pointer types).
Unfortunately, simple unions are not as easy to switch to unboxed
layout; because we are special-casing the hidden implicit type with
a single 'data' member, we really DO need to keep calling another
layer of visit_start_struct(), with a second malloc; although there
are some cleanups planned for simple unions in later patches.
visit_start_union() and gen_visit_implicit_struct() are now unused.
Drop them.
Note that after this patch, the only remaining use of
visit_start_implicit_struct() is for alternate types; the next patch
will do further cleanup based on that fact.
Backports commit 544a3731591f5d53e15f22de00ce5ac758d490b3 from qemu
The previous commit documented an inconsistency in how we are
using the stack of qmp-output-visitor. Normally, pushing a
single top-level object puts the object on the stack twice:
once as the root, and once as the current container being
appended to; but popping that struct only pops once. However,
qmp_ouput_add() was trying to either set up the added object
as the new root (works if you parse two top-level scalars in a
row: the second replaces the first as the root) or as a member
of the current container (works as long as you have an open
container on the stack; but if you have popped the first
top-level container, it then resolves to the root and still
tries to add into that existing container).
Fix the stupidity by not tracking two separate things in the
stack. Drop the now-useless qmp_output_first() and
qmp_output_last() while at it.
Saved for a later patch: we still are rather sloppy in that
qmp_output_get_object() can be called in the middle of a parse,
rather than requiring that a visit is complete.
Backports commit 455ba08afde784466420449d01c6458f88349d55 from qemu
Commit 6c2f9a15 ensured that we would not return NULL when the
caller used an output visitor but had nothing to visit. But
in doing so, it added a FIXME about a reference count leak
that could abort qemu in the (unlikely) case of SIZE_MAX such
visits (more plausible on 32-bit). (Although that commit
suggested we might fix it in time for 2.5, we ran out of time;
fortunately, it is unlikely enough to bite that it was not
worth worrying about during the 2.5 release.)
This fixes things by documenting the internal contracts, and
explaining why the internal function can return NULL and only
the public facing interface needs to worry about qnull(),
thus avoiding over-referencing the qnull_ global object.
It does not, however, fix the stupidity of the stack mixing
up two separate pieces of information; add a FIXME to explain
that issue, which will be fixed shortly in a future patch.
Backports commit a86156401559cb4401cf9ecc704faeab6fc8bb19 from qemu
The only way that qmp_input_pop() will set errp is if a dictionary
was the most recent thing pushed. Since we don't have any
push(struct)/pop(list) or push(list)/pop(struct) mismatches (such
a mismatch is a programming bug), we therefore cannot set errp
inside qmp_input_end_list(). Make this obvious by
using &error_abort. A later patch will then remove the errp
parameter of qmp_input_pop(), but that will first require the
larger task of splitting visit_end_struct().
Backports commit bdd8e6b5d8a9def83d491a3f41c10424fc366258 from qemu
visit_start_struct() and visit_type_enum() had a 'kind' argument
that was usually set to either the stringized version of the
corresponding qapi type name, or to NULL (although some clients
didn't even get that right). But nothing ever used the argument.
It's even hard to argue that it would be useful in a debugger,
as a stack backtrace also tells which type is being visited.
Therefore, drop the 'kind' argument as dead.
Backports commit 337283dffbb5ad5860ed00408a5fd0665c21be07 from qemu
As explained in the previous patches, matching argument order of
'name, &value' to JSON's "name":value makes sense. However,
while the last two patches were easy with Coccinelle, I ended up
doing this one all by hand. Now all the visitor callbacks match
the main interface.
The compiler is able to enforce that all clients match the changed
interface in visitor-impl.h, even where two pointers are being
swapped, because only one of the two pointers is const (if that
were not the case, then C's looseness on treating 'char *' like
'void *' would have made review a bit harder).
Backports commit 0b2a0d6bb2446060944061e53e87d0c7addede79 from qemu
Commit 4e27e819 introduced optional visitor callbacks for all
sorts of int types, but no visitor has supplied any of the
callbacks for sizes less than 64 bits. In other words, the
generic implementation based on using type_[u]int64() followed
by bounds-checking works just fine. In the interest of
simplicity, it's easier to make the visitor callback interface
not have to worry about the other sizes.
Adding some helper functions minimizes the boilerplate required
to correct FIXMEs added earlier with regards to questionable
reuse of errp, particularly now that we can guarantee from a
single file audit that value is unchanged if an error is set.
Backports commit 04e070d217b4414f1f91aa8ad25fc0ae7ca0be93 from qemu
JSON uses "name":value, but many of our visitor interfaces were
called with visit_type_FOO(v, &value, name, errp). This can be
a bit confusing to have to mentally swap the parameter order to
match JSON order. It's particularly bad for visit_start_struct(),
where the 'name' parameter is smack in the middle of the
otherwise-related group of 'obj, kind, size' parameters! It's
time to do a global swap of the parameter ordering, so that the
'name' parameter is always immediately after the Visitor argument.
Additional reason in favor of the swap: the existing include/qjson.h
prefers listing 'name' first in json_prop_*(), and I have plans to
unify that file with the qapi visitors; listing 'name' first in
qapi will minimize churn to the (admittedly few) qjson.h clients.
Later patches will then fix docs, object.h, visitor-impl.h, and
those clients to match.
Done by first patching scripts/qapi*.py by hand to make generated
files do what I want, then by running the following Coccinelle
script to affect the rest of the code base:
$ spatch --sp-file script `git grep -l '\bvisit_' -- '**/*.[ch]'`
I then had to apply some touchups (Coccinelle insisted on TAB
indentation in visitor.h, and botched the signature of
visit_type_enum() by rewriting 'const char *const strings[]' to
the syntactically invalid 'const char*const[] strings'). The
movement of parameters is sufficient to provoke compiler errors
if any callers were missed.
// Part 1: Swap declaration order
@@
type TV, TErr, TObj, T1, T2;
identifier OBJ, ARG1, ARG2;
@@
void visit_start_struct
-(TV v, TObj OBJ, T1 ARG1, const char *name, T2 ARG2, TErr errp)
+(TV v, const char *name, TObj OBJ, T1 ARG1, T2 ARG2, TErr errp)
{ ... }
@@
type bool, TV, T1;
identifier ARG1;
@@
bool visit_optional
-(TV v, T1 ARG1, const char *name)
+(TV v, const char *name, T1 ARG1)
{ ... }
@@
type TV, TErr, TObj, T1;
identifier OBJ, ARG1;
@@
void visit_get_next_type
-(TV v, TObj OBJ, T1 ARG1, const char *name, TErr errp)
+(TV v, const char *name, TObj OBJ, T1 ARG1, TErr errp)
{ ... }
@@
type TV, TErr, TObj, T1, T2;
identifier OBJ, ARG1, ARG2;
@@
void visit_type_enum
-(TV v, TObj OBJ, T1 ARG1, T2 ARG2, const char *name, TErr errp)
+(TV v, const char *name, TObj OBJ, T1 ARG1, T2 ARG2, TErr errp)
{ ... }
@@
type TV, TErr, TObj;
identifier OBJ;
identifier VISIT_TYPE =~ "^visit_type_";
@@
void VISIT_TYPE
-(TV v, TObj OBJ, const char *name, TErr errp)
+(TV v, const char *name, TObj OBJ, TErr errp)
{ ... }
// Part 2: swap caller order
@@
expression V, NAME, OBJ, ARG1, ARG2, ERR;
identifier VISIT_TYPE =~ "^visit_type_";
@@
(
-visit_start_struct(V, OBJ, ARG1, NAME, ARG2, ERR)
+visit_start_struct(V, NAME, OBJ, ARG1, ARG2, ERR)
|
-visit_optional(V, ARG1, NAME)
+visit_optional(V, NAME, ARG1)
|
-visit_get_next_type(V, OBJ, ARG1, NAME, ERR)
+visit_get_next_type(V, NAME, OBJ, ARG1, ERR)
|
-visit_type_enum(V, OBJ, ARG1, ARG2, NAME, ERR)
+visit_type_enum(V, NAME, OBJ, ARG1, ARG2, ERR)
|
-VISIT_TYPE(V, OBJ, NAME, ERR)
+VISIT_TYPE(V, NAME, OBJ, ERR)
)
Backports commit 51e72bc1dd6ace6e91d675f41a1f09bd00ab8043 from qemu
The generated code can call visit_end_union() without having called
visit_start_union(). Example:
if (!*obj) {
goto out_obj;
}
visit_type_CpuInfoBase_fields(v, (CpuInfoBase **)obj, &err);
if (err) {
goto out_obj; // if we go from here...
}
if (!visit_start_union(v, !!(*obj)->u.data, &err) || err) {
goto out_obj;
}
switch ((*obj)->arch) {
[...]
}
out_obj:
// ... then *obj is true, and ...
error_propagate(errp, err);
err = NULL;
if (*obj) {
// we end up here
visit_end_union(v, !!(*obj)->u.data, &err);
}
error_propagate(errp, err);
Harmless only because no visitor implements end_union(). Clean it up
anyway, by deleting the function as useless.
Messed up since we have visit_end_union (commit cee2ded).
Backports commit 7c91aabd8964cfdf637f302c579c95401f21ce92 from qemu
For less code, reflect the determined boolean value of an optional
visit back to the caller instead of making the caller read the
boolean after the fact.
The resulting generated code has the following diff:
|- visit_optional(v, &has_fdset_id, "fdset-id");
|- if (has_fdset_id) {
|+ if (visit_optional(v, &has_fdset_id, "fdset-id")) {
| visit_type_int(v, &fdset_id, "fdset-id", &err);
| if (err) {
| goto out;
| }
| }
Backports commit 29637a6ee913df8fcdf371426ee48956b945b618 from qemu
None of the visitor callbacks would set an error when testing
if an optional field was present; make this part of the interface
contract by eliminating the errp argument.
The resulting generated code has a nice diff:
|- visit_optional(v, &has_fdset_id, "fdset-id", &err);
|- if (err) {
|- goto out;
|- }
|+ visit_optional(v, &has_fdset_id, "fdset-id");
| if (has_fdset_id) {
| visit_type_int(v, &fdset_id, "fdset-id", &err);
| if (err) {
| goto out;
| }
| }
Backports commit 5cdc8831a795fb8452d7e34f644202fd724e122a from qemu
The QMP input visitor allows integral values to be assigned by
promotion to a QTYPE_QFLOAT. However, when parsing an alternate,
we did not take this into account, such that an alternate that
accepts 'number' and some other type, but not 'int', would reject
integral values.
With this patch, we now have the following desirable table:
alternate has case selected for
'int' 'number' QTYPE_QINT QTYPE_QFLOAT
no no error error
no yes 'number' 'number'
yes no 'int' error
yes yes 'int' 'number'
While it is unlikely that we will ever use 'number' in an
alternate other than in the testsuite, it never hurts to be
more precise in what we allow.
Backports commit d00341af384665d259af475b14c96bb8414df415 from qemu
Previously, working with alternates required two lookup arrays
and some indirection: for type Foo, we created Foo_qtypes[]
which maps each qtype to a value of the generated FooKind enum,
then look up that value in FooKind_lookup[] like we do for other
union types.
This has a couple of subtle bugs. First, the generator was
creating a call with a parameter '(int *) &(*obj)->type' where
type is an enum type; this is unsafe if the compiler chooses
to store the enum type in a different size than int, where
assigning through the wrong size pointer can corrupt data or
cause a SIGBUS.
Related bug, not not fixed in this patch: qapi-visit.py's
gen_visit_enum() generates a cast of its enum * argument to
int *. Marked FIXME.
Second, since the values of the FooKind enum start at zero, all
entries of the Foo_qtypes[] array that were not explicitly
initialized will map to the same branch of the union as the
first member of the alternate, rather than triggering a desired
failure in visit_get_next_type(). Fortunately, the bug seldom
bites; the very next thing the input visitor does is try to
parse the incoming JSON with the wrong parser, which normally
fails; the output visitor is not used with a C struct in that
state, and the dealloc visitor has nothing to clean up (so
there is no leak).
However, the second bug IS observable in one case: parsing an
integer causes unusual behavior in an alternate that contains
at least a 'number' member but no 'int' member, because the
'number' parser accepts QTYPE_QINT in addition to the expected
QTYPE_QFLOAT (that is, since 'int' is not a member, the type
QTYPE_QINT accidentally maps to FooKind 0; if this enum value
is the 'number' branch the integer parses successfully, but if
the 'number' branch is not first, some other branch tries to
parse the integer and rejects it). A later patch will worry
about fixing alternates to always parse all inputs that a
non-alternate 'number' would accept, for now this is still
marked FIXME in the updated test-qmp-input-visitor.c, to
merely point out that new undesired behavior of 'ans' matches
the existing undesired behavior of 'asn'.
This patch fixes the default-initialization bug by deleting the
indirection, and modifying get_next_type() to directly assign a
QTypeCode parameter. This in turn fixes the type-casting bug,
as we are no longer casting a pointer to enum to a questionable
size. There is no longer a need to generate an implicit FooKind
enum associated with the alternate type (since the QMP wire
format never uses the stringized counterparts of the C union
member names). Since the updated visit_get_next_type() does not
know which qtypes are expected, the generated visitor is
modified to generate an error statement if an unexpected type is
encountered.
Callers now have to know the QTYPE_* mapping when looking at the
discriminator; but so far, only the testsuite was even using the
C struct of an alternate types. I considered the possibility of
keeping the internal enum FooKind, but initialized differently
than most generated arrays, as in:
typedef enum FooKind {
FOO_KIND_A = QTYPE_QDICT,
FOO_KIND_B = QTYPE_QINT,
} FooKind;
to create nicer aliases for knowing when to use foo->a or foo->b
when inspecting foo->type; but it turned out to add too much
complexity, especially without a client.
There is a user-visible side effect to this change, but I
consider it to be an improvement. Previously,
the invalid QMP command:
{"execute":"blockdev-add", "arguments":{"options":
{"driver":"raw", "id":"a", "file":true}}}
failed with:
{"error": {"class": "GenericError",
"desc": "Invalid parameter type for 'file', expected: QDict"}}
(visit_get_next_type() succeeded, and the error comes from the
visit_type_BlockdevOptions() expecting {}; there is no mention of
the fact that a string would also work). Now it fails with:
{"error": {"class": "GenericError",
"desc": "Invalid parameter type for 'file', expected: BlockdevRef"}}
(the error when the next type doesn't match any expected types for
the overall alternate).
Backports commit 0426d53c6530606bf7641b83f2b755fe61c280ee from qemu
The qapi enum ErrorClass is unusual that it uses 'CamelCase' names,
contrary to our documented convention of preferring 'lower-case'.
However, this enum is entrenched in the API; we cannot change
what strings QMP outputs. Meanwhile, we want to simplify how
c_enum_const() is used to generate enum constants, by moving away
from the heuristics of camel_to_upper() to a more straightforward
c_name(N).upper() - but doing so will rename all of the ErrorClass
constants and cause churn to all client files, where the new names
are aesthetically less pleasing (ERROR_CLASS_DEVICENOTFOUND looks
like we can't make up our minds on whether to break between words).
So as always in computer science, solve the problem by some more
indirection: rename the qapi type to QapiErrorClass, and add a
new enum ErrorClass in error.h whose members are aliases of the
qapi type, but with the spelling expected elsewhere in the tree.
Then, when c_enum_const() changes the munging, we only have to
adjust the one alias spot.
Backports commit f22a28b898322c01b0463a8b7ec551d72bc61a5b from qemu
It's first class, because unlike '**', it actually works, i.e. doesn't
require 'gen': false.
'**' will go away next.
Backports commit 28770e057f265a4e70bcbdfc2447cce7b5f2dc19 from qemu
The enum string table parameters in various QOM/QAPI methods
are declared 'const char *strings[]'. This results in const
warnings if passed a variable that was declared as
static const char * const strings[] = { .... };
Add the extra const annotation to the parameters, since
neither the string elements, nor the array itself should
ever be modified.
Backports commit 2e4450ff432daef524cb3557fca68a3b7b5c7823 from qemu
No backend was setting an error when ending the visit of a list or
implicit struct, or when moving to the next list node. Make the
callers a bit easier to follow by making this a part of the contract,
and removing the errp argument - callers can then unconditionally end
an object as part of cleanup without having to think about whether a
second error is dominated by a first, because there is no second
error.
A later patch will then tackle the larger task of splitting
visit_end_struct(), which can indeed set an error.
Backports commit 08f9541dec51700abef0c37994213164ca4e4fc9 from qemu
The only way that qmp_input_pop() will set errp is if a dictionary
was the most recent thing pushed. Since we don't have any
push(struct)/pop(list) or push(list)/pop(struct) mismatches (such
a mismatch is a programming bug), we therefore cannot set errp
inside qmp_input_end_list(). Make this obvious by
using &error_abort. A later patch will then remove the errp
parameter of qmp_input_pop(), but that will first require the
larger task of splitting visit_end_struct().
Backports commit bdd8e6b5d8a9def83d491a3f41c10424fc366258 from qemu
Our qapi visitor contract supports multiple integer visitors,
but left the type_uint64 visitor as optional (falling back on
type_int64); which in turn can lead to awkward behavior with
numbers larger than INT64_MAX (the user has to be aware of
twos complement, and deal with negatives).
This patch does not address the disparity in handling large
values as negatives. It merely moves the fallback from uint64
to int64 from the visitor core to the visitors, where the issue
can actually be fixed, by implementing the missing type_uint64()
callbacks on top of the respective type_int64() callbacks, and
with a FIXME comment explaining why that's wrong.
With that done, we now have a type_uint64() callback in every
driver, so we can make it mandatory from the core. And although
the type_int64() callback can cover the entire valid range of
type_uint{8,16,32} on valid user input, using type_uint64() to
avoid mixed signedness makes more sense.
Backports commit f755dea79dc81b0d6a8f6414e0672e165e28d8ba from qemu
The qapi builtin type 'int' is basically shorthand for the type
'int64'. In fact, since no visitor was providing the optional
type_int64() callback, visit_type_int64() was just always falling
back to type_int(), cementing the equivalence between the types.
However, some visitors are providing a type_uint64() callback.
For purposes of code consistency, it is nicer if all visitors
use the paired type_int64/type_uint64 names rather than the
mismatched type_int/type_uint64. So this patch just renames
the signed int callbacks in place, dropping the type_int()
callback as redundant, and a later patch will focus on the
unsigned int callbacks.
Add some FIXMEs to questionable reuse of errp in code touched
by the rename, while at it (the reuse works as long as the
callbacks don't modify value when setting an error, but it's not
a good example to set) - a later patch will then fix those.
No change in functionality here, although further cleanups are
in the pipeline.
Backports commit 4c40314a35816de635e7170eaacdc0c35be83a8a from qemu
The macro DO_UPCAST() is incorrectly named: it converts from a
parent class to a derived class (which is a downcast). Better,
and more consistent with some of the other qapi visitors, is
to use the container_of() macro through a to_FOO() helper. Names
like 'to_ov()' may be a bit short, but for a static helper it
doesn't hurt too much, and matches existing practice in files
like qmp-input-visitor.c.
Our current definition of container_of() is weaker than
DO_UPCAST(), in that it does not require the derived class to
have Visitor as its first member, but this does not hurt our
usage patterns in qapi visitors.
Backports commit d7bea75d35a44023efc9d481d3a1a2600677b2ef from qemu
Clean up includes so that osdep.h is included first and headers
which it implies are not included manually.
This commit was created with scripts/clean-includes.
Backports commit cbf21151906c935d4276268b59429c58546462ae from qemu
qobject_to_qstring() crashes on null, which is a trap for the unwary.
Return null instead, and simplify a few callers.
Backports commit 7f0278435df1fa845b3bd9556942f89296d4246b from qemu
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.
Backports commit fcf73f66a67f5e58c18216f8c8651e38cf4d90af from qemu
qobject_to_qbool() crashes on null, which is a trap for the unwary.
Return null instead, and simplify a few callers.
Backports commit 14b6160099f0caf5dc9d62e637b007bc5d719a96 from qemu
These macros expand into error class enumeration constant, comma,
string. Unclean. Has been that way since commit 13f59ae.
The error class is always ERROR_CLASS_GENERIC_ERROR since the previous
commit.
* Prepend every use of a QERR_ macro by ERROR_CLASS_GENERIC_ERROR, and
delete it from the QERR_ macro. No change after preprocessing.
* Rewrite error_set(ERROR_CLASS_GENERIC_ERROR, ...) into
error_setg(...). Again, no change after preprocessing.
Backports commit c6bd8c706a799eb0fece99f468aaa22b818036f3 from qemu
We require a C99 compiler, so let's use 'bool' instead of 'int'
when dealing with boolean values. There are few enough clients
to fix them all in one pass.
Backports commit fc48ffc39ed1060856475e4320d5896f26c945e8 from qemu