Rework finalizers

This commit involves a number of changes to finalizers, all aimed towards
better consistency and reliability.

A big part of this has to do with v8::Inspector's ability to move objects
across IsolatedWorlds. There has been a few previous efforts on this, the most
significant being https://github.com/lightpanda-io/browser/pull/1901. To recap,
a Zig instance can map to 0-N v8::Objects. Where N is the total number of
IsolatedWorlds. Generally, IsolatedWorlds between origins are...isolated...but
the v8::Inspector isn't bound by this. So a Zig instance cannot be tied to a
Context/Identity/IsolatedWorld...it has to live until all references, possibly
from different IsolatedWorlds, are released (or the page is reset).

Finalizers could previously be managed via reference counting or explicitly
toggling the instance as weak/strong. Now, only reference counting is supported.
weak/strong can essentially be seen as an acquireRef (rc += 1) and
releaseRef (rc -= 1). Explicit setting did make some things easier, like not
having to worry so much about double-releasing (e.g. XHR abort being called
multiple times), but it was only used in a few places AND it simply doesn't work
with objects shared between IsolatedWorlds. It is never a boolean now, as 3
different IsolatedWorlds can each hold a reference.

Temps and Globals are tracked on the Session. Previously, they were tracked on
the Identity, but that makes no sense. If a Zig instance can outlive an Identity,
then any of its Temp references can too. This hasn't been a problem because we've
only seen MutationObserver and IntersectionObserver be used cross-origin,
but the right CDP script can make this crash with a use-after-free (e.g.
`MessageEvent.data` is released when the Identity is done, but `MessageEvent` is
still referenced by a different IsolateWorld).

Rather than deinit with a `comptime shutdown: bool`, there is now an explicit
`releaseRef` and `deinit`.

Bridge registration has been streamlined. Previously, types had to register
their finalizer AND acquireRef/releaseRef/deinit had to be declared on the entire
prototype chain, even if these methods just delegated to their proto. Finalizers
are now automatically enabled if a type has a `acquireRef` function. If a type
has an `acquireRef`, then it must have a `releaseRef` and a `deinit`. So if
there's custom cleanup to do in `deinit`, then you also have to define
`acquireRef` and `releaseRef` which will just delegate to the _proto.

Furthermore these finalizer methods can be defined anywhere on the chain.

Previously:

```zig
const KeywboardEvent = struct {
  _proto: *Event,
  ...

  pub fn deinit(self: *KeyboardEvent, session: *Session) void {
    self._proto.deinit(session);
  }

  pub fn releaseRef(self: *KeyboardEvent, session: *Session) void {
    self._proto.releaseRef(session);
  }
}
```

```zig
const KeyboardEvent = struct {
  _proto: *Event,
  ...
  // no deinit, releaseRef, acquireref
}
```

Since the `KeyboardEvent` doesn't participate in finalization directly, it
doesn't have to define anything. The bridge will detect the most specific place
they are defined and call them there.
This commit is contained in:
Karl Seguin
2026-03-28 14:23:03 +08:00
parent 67bd555e75
commit 01ecb296e5
53 changed files with 541 additions and 479 deletions

View File

@@ -17,10 +17,11 @@
// along with this program. If not, see <https://www.gnu.org/licenses/>.
const std = @import("std");
const Session = @import("../Session.zig");
const log = @import("../../log.zig");
const string = @import("../../string.zig");
const Session = @import("../Session.zig");
const js = @import("js.zig");
const bridge = @import("bridge.zig");
const Caller = @import("Caller.zig");
@@ -213,7 +214,8 @@ pub fn mapZigInstanceToJs(self: *const Local, js_obj_handle: ?*const v8.Object,
.pointer => |ptr| {
const resolved = resolveValue(value);
const gop = try ctx.addIdentity(@intFromPtr(resolved.ptr));
const resolved_ptr_id = @intFromPtr(resolved.ptr);
const gop = try ctx.addIdentity(resolved_ptr_id);
if (gop.found_existing) {
// we've seen this instance before, return the same object
return (js.Object.Global{ .handle = gop.value_ptr.* }).local(self);
@@ -262,31 +264,27 @@ pub fn mapZigInstanceToJs(self: *const Local, js_obj_handle: ?*const v8.Object,
// dont' use js_obj.persist(), because we don't want to track this in
// context.global_objects, we want to track it in context.identity_map.
v8.v8__Global__New(isolate.handle, js_obj.handle, gop.value_ptr);
if (@hasDecl(JsApi.Meta, "finalizer")) {
// It would be great if resolved knew the resolved type, but I
// can't figure out how to make that work, since it depends on
// the [runtime] `value`.
// We need the resolved finalizer, which we have in resolved.
//
// The above if statement would be more clear as:
// if (resolved.finalizer_from_v8) |finalizer| {
// But that's a runtime check.
// Instead, we check if the base has finalizer. The assumption
// here is that if a resolve type has a finalizer, then the base
// should have a finalizer too.
const fc = try ctx.createFinalizerCallback(gop.value_ptr.*, resolved.ptr, resolved.finalizer_from_zig.?);
{
errdefer fc.deinit();
try ctx.identity.finalizer_callbacks.put(ctx.identity_arena, @intFromPtr(resolved.ptr), fc);
}
if (resolved.finalizer) |finalizer| {
const finalizer_ptr_id = finalizer.ptr_id;
finalizer.acquireRef(finalizer_ptr_id);
conditionallyReference(value);
if (@hasDecl(JsApi.Meta, "weak")) {
if (comptime IS_DEBUG) {
std.debug.assert(JsApi.Meta.weak == true);
}
v8.v8__Global__SetWeakFinalizer(gop.value_ptr, fc, resolved.finalizer_from_v8, v8.kParameter);
const session = ctx.session;
const finalizer_gop = try session.finalizer_callbacks.getOrPut(session.page_arena, finalizer_ptr_id);
if (finalizer_gop.found_existing == false) {
// This is the first context (and very likely only one) to
// see this Zig instance. We need to create the FinalizerCallback
// so that we can cleanup on page reset if v8 doesn't finalize.
errdefer _ = session.finalizer_callbacks.remove(finalizer_ptr_id);
finalizer_gop.value_ptr.* = try self.createFinalizerCallback(resolved_ptr_id, finalizer_ptr_id, finalizer.deinit);
}
const fc = finalizer_gop.value_ptr.*;
const identity_finalizer = try fc.arena.create(Session.FinalizerCallback.Identity);
identity_finalizer.* = .{
.fc = fc,
.identity = ctx.identity,
};
v8.v8__Global__SetWeakFinalizer(gop.value_ptr, identity_finalizer, finalizer.release, v8.kParameter);
}
return js_obj;
},
@@ -1121,12 +1119,19 @@ fn jsUnsignedIntToZig(comptime T: type, max: comptime_int, maybe: u32) !T {
// This function recursively walks the _type union field (if there is one) to
// get the most specific class_id possible.
const Resolved = struct {
weak: bool,
ptr: *anyopaque,
class_id: u16,
prototype_chain: []const @import("TaggedOpaque.zig").PrototypeChainEntry,
finalizer_from_v8: ?*const fn (handle: ?*const v8.WeakCallbackInfo) callconv(.c) void = null,
finalizer_from_zig: ?*const fn (ptr: *anyopaque, session: *Session) void = null,
finalizer: ?Finalizer,
const Finalizer = struct {
// Resolved.ptr is the most specific value in a chain (e.g. IFrame, not EventTarget, Node, ...)
// Finalizer.ptr_id is the most specific value in a chain that defines an acquireRef
ptr_id: usize,
deinit: *const fn (ptr_id: usize, session: *Session) void,
acquireRef: *const fn (ptr_id: usize) void,
release: *const fn (handle: ?*const v8.WeakCallbackInfo) callconv(.c) void,
};
};
pub fn resolveValue(value: anytype) Resolved {
const T = bridge.Struct(@TypeOf(value));
@@ -1153,27 +1158,85 @@ pub fn resolveValue(value: anytype) Resolved {
unreachable;
}
fn resolveT(comptime T: type, value: *anyopaque) Resolved {
fn resolveT(comptime T: type, value: *T) Resolved {
const Meta = T.JsApi.Meta;
return .{
.ptr = value,
.class_id = Meta.class_id,
.prototype_chain = &Meta.prototype_chain,
.weak = if (@hasDecl(Meta, "weak")) Meta.weak else false,
.finalizer_from_v8 = if (@hasDecl(Meta, "finalizer")) Meta.finalizer.from_v8 else null,
.finalizer_from_zig = if (@hasDecl(Meta, "finalizer")) Meta.finalizer.from_zig else null,
.finalizer = blk: {
const FT = (comptime findFinalizerType(T)) orelse break :blk null;
const getFinalizerPtr = comptime finalizerPtrGetter(T, FT);
const finalizer_ptr = getFinalizerPtr(value);
const Wrap = struct {
fn deinit(ptr_id: usize, session: *Session) void {
FT.deinit(@ptrFromInt(ptr_id), session);
}
fn acquireRef(ptr_id: usize) void {
FT.acquireRef(@ptrFromInt(ptr_id));
}
fn release(handle: ?*const v8.WeakCallbackInfo) callconv(.c) void {
const ptr = v8.v8__WeakCallbackInfo__GetParameter(handle.?).?;
const identity_finalizer: *Session.FinalizerCallback.Identity = @ptrCast(@alignCast(ptr));
const fc = identity_finalizer.fc;
if (identity_finalizer.identity.identity_map.fetchRemove(fc.resolved_ptr_id)) |kv| {
var global = kv.value;
v8.v8__Global__Reset(&global);
}
FT.releaseRef(@ptrFromInt(fc.finalizer_ptr_id), fc.session);
}
};
break :blk .{
.ptr_id = @intFromPtr(finalizer_ptr),
.deinit = Wrap.deinit,
.acquireRef = Wrap.acquireRef,
.release = Wrap.release,
};
},
};
}
fn conditionallyReference(value: anytype) void {
const T = bridge.Struct(@TypeOf(value));
if (@hasDecl(T, "acquireRef")) {
value.acquireRef();
return;
// Start at the "resolved" type (the most specific) and work our way up the
// prototype chain looking for the type that defines acquireRef
fn findFinalizerType(comptime T: type) ?type {
const S = bridge.Struct(T);
if (@hasDecl(S, "acquireRef")) {
return S;
}
if (@hasField(T, "_proto")) {
conditionallyReference(value._proto);
if (@hasField(S, "_proto")) {
const ProtoPtr = std.meta.fieldInfo(S, ._proto).type;
const ProtoChild = @typeInfo(ProtoPtr).pointer.child;
return findFinalizerType(ProtoChild);
}
return null;
}
// Generate a function that follows the _proto pointer chain to get to the finalizer type
fn finalizerPtrGetter(comptime T: type, comptime FT: type) *const fn (*T) *FT {
const S = bridge.Struct(T);
if (S == FT) {
return struct {
fn get(v: *T) *FT {
return v;
}
}.get;
}
if (@hasField(S, "_proto")) {
const ProtoPtr = std.meta.fieldInfo(S, ._proto).type;
const ProtoChild = @typeInfo(ProtoPtr).pointer.child;
const childGetter = comptime finalizerPtrGetter(ProtoChild, FT);
return struct {
fn get(v: *T) *FT {
return childGetter(v._proto);
}
}.get;
}
@compileError("Cannot find path from " ++ @typeName(T) ++ " to " ++ @typeName(FT));
}
pub fn stackTrace(self: *const Local) !?[]const u8 {
@@ -1381,6 +1444,34 @@ pub fn debugContextId(self: *const Local) i32 {
return v8.v8__Context__DebugContextId(self.handle);
}
fn createFinalizerCallback(
self: *const Local,
// Key in identity map
// The most specific value (KeyboardEvent, not Event)
resolved_ptr_id: usize,
// The most specific value where finalizers are defined
// What actually gets acquired / released / deinit
finalizer_ptr_id: usize,
deinit: *const fn (ptr_id: usize, session: *Session) void,
) !*Session.FinalizerCallback {
const session = self.ctx.session;
const arena = try session.getArena(.{ .debug = "FinalizerCallback" });
errdefer session.releaseArena(arena);
const fc = try arena.create(Session.FinalizerCallback);
fc.* = .{
.arena = arena,
.session = session,
._deinit = deinit,
.resolved_ptr_id = resolved_ptr_id,
.finalizer_ptr_id = finalizer_ptr_id,
};
return fc;
}
// Encapsulates a Local and a HandleScope. When we're going from V8->Zig
// we easily get both a Local and a HandleScope via Caller.init.
// But when we're going from Zig -> V8, things are more complicated.