1408584 - Assertion failure: aInitialCapacity != 0, at mozilla/BufferList.h:131 with clonebuffer

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect, P1)

Description

[Christian Holler (:decoder)]

The following testcase crashes on mozilla-central revision 196dadb2fe50 (build with --enable-posix-nspr-emulation --enable-valgrind --enable-gczeal --disable-tests --enable-stdcxx-compat --disable-profiling --enable-debug --enable-optimize, run with --fuzzing-safe min.js):

function byteArray(str) {}
var mutated = byteArray(serialize(new Date(NaN)).clonebuffer);
var a = [1 / 0, -1 / 0, ];
for (var i = 0; i < a.length; i++) {
    var n = a[i];
    var nbuf = serialize(n);
    nbuf.clonebuffer = String.fromCharCode.apply(null, mutated);
}


Backtrace:

received signal SIGSEGV, Segmentation fault.
#0  0x00000000008e161e in mozilla::BufferList<js::SystemAllocPolicy>::Init (aInitialCapacity=<optimized out>, aInitialSize=<optimized out>, this=0x7ffff4297b60) at mozilla/BufferList.h:131
#1  CloneBufferObject::setCloneBuffer_impl (cx=cx@entry=0x7ffff6955000, args=...) at js/src/builtin/TestingFunctions.cpp:2757
#2  0x00000000008e176b in JS::CallNonGenericMethod<&CloneBufferObject::is, &CloneBufferObject::setCloneBuffer_impl> (args=..., cx=0x7ffff6955000) at js/CallNonGenericMethod.h:100
#3  CloneBufferObject::setCloneBuffer (cx=0x7ffff6955000, argc=<optimized out>, vp=<optimized out>) at js/src/builtin/TestingFunctions.cpp:2775
#4  0x000000000056153d in js::CallJSNative (cx=0x7ffff6955000, native=0x8e16d0 <CloneBufferObject::setCloneBuffer(JSContext*, unsigned int, JS::Value*)>, args=...) at js/src/jscntxtinlines.h:291
#5  0x0000000000555c9f in js::InternalCallOrConstruct (cx=cx@entry=0x7ffff6955000, args=..., construct=construct@entry=js::NO_CONSTRUCT) at js/src/vm/Interpreter.cpp:482
#6  0x000000000055607d in InternalCall (cx=cx@entry=0x7ffff6955000, args=...) at js/src/vm/Interpreter.cpp:531
#7  0x00000000005561e0 in js::Call (cx=cx@entry=0x7ffff6955000, fval=..., fval@entry=..., thisv=..., thisv@entry=..., args=..., rval=...) at js/src/vm/Interpreter.cpp:550
#8  0x0000000000556762 in js::CallSetter (cx=0x7ffff6955000, thisv=..., thisv@entry=..., setter=setter@entry=..., v=v@entry=...) at js/src/vm/Interpreter.cpp:679
#9  0x0000000000bddafe in SetExistingProperty (cx=0x7ffff6955000, obj=..., obj@entry=..., id=..., id@entry=..., v=v@entry=..., receiver=receiver@entry=..., pobj=..., pobj@entry=..., prop=..., result=...) at js/src/vm/NativeObject.cpp:2732
#10 0x0000000000bfca2f in js::NativeSetProperty<(js::QualifiedBool)1> (cx=cx@entry=0x7ffff6955000, obj=..., id=id@entry=..., value=..., value@entry=..., receiver=..., receiver@entry=..., result=...) at js/src/vm/NativeObject.cpp:2768
#11 0x000000000055e43c in js::SetProperty (cx=0x7ffff6955000, obj=..., id=..., v=..., receiver=..., result=...) at js/src/vm/NativeObject.h:1615
#12 0x000000000054a7e7 in SetPropertyOperation (rval=..., id=..., lval=..., op=<optimized out>, cx=0x7ffff6955000) at js/src/vm/Interpreter.cpp:269
#13 Interpret (cx=0x7ffff6955000, state=...) at js/src/vm/Interpreter.cpp:2873
[...]
#23 main (argc=<optimized out>, argv=<optimized out>, envp=<optimized out>) at js/src/shell/js.cpp:8892
rax	0x0	0
rbx	0x7fffffffd1a0	140737488343456
rcx	0x7ffff6c28a2d	140737333332525
rdx	0x0	0
rsi	0x7ffff6ef7770	140737336276848
rdi	0x7ffff6ef6540	140737336272192
rbp	0x7fffffffd190	140737488343440
rsp	0x7fffffffd110	140737488343312
r8	0x7ffff6ef7770	140737336276848
r9	0x7ffff7fe4740	140737354024768
r10	0x58	88
r11	0x7ffff6b9f750	140737332770640
r12	0x7ffff4297b60	140737289747296
r13	0x7fffffffd140	140737488343360
r14	0x7ffff69560e8	140737330372840
r15	0x0	0
rip	0x8e161e <CloneBufferObject::setCloneBuffer_impl(JSContext*, JS::CallArgs const&)+1150>
=> 0x8e161e <CloneBufferObject::setCloneBuffer_impl(JSContext*, JS::CallArgs const&)+1150>:	movl   $0x0,(%eax)
   0x8e1625 <CloneBufferObject::setCloneBuffer_impl(JSContext*, JS::CallArgs const&)+1157>:	ud2

Comment 1

[Fuzzing Team]

JSBugMon: Bisection requested, result:
autoBisect shows this is probably related to the following changeset:

The first bad revision is:
changeset:   https://hg.mozilla.org/mozilla-central/rev/8c07eaec94c4
user:        Kan-Ru Chen
date:        Fri Apr 22 18:04:20 2016 +0800
summary:     Bug 1264642 - Part 4. Use BufferList to replace raw buffers in StructuredClone. r=baku r=billm r=jorendorff

This iteration took 225.980 seconds to run.

Comment 2

[Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)]

Kan-Ru, is bug 1264642 a likely regressor?

Comment 3

[Steve Fink [:sfink] [:s:]]

This looks more like I introduced a bug when adding the ability to use binary data as a structured clone buffer. The minimal test case I came up with, fwiw, is

  serialize().clonebuffer = String.fromCharCode()

I think I moved the discard of the previous buffer to after where it is needed.

Comment 4

[Steve Fink [:sfink] [:s:]]

That was in bug 1400466. That's my guess as to the regressor.

Comment 5

[Steve Fink [:sfink] [:s:]]

Oh, it's simpler than that. serialize().clonebuffer = '' would also work. Init just doesn't want an empty buffer, it seems. I'll just extend the bad length check to cover it.

Comment 6

[Steve Fink [:sfink] [:s:]]

Attachment: Disallow empty clonebuffer

Really I should remove the string interface entirely, but this particular bug is shared with the ArrayBuffer interface.

Comment 7

[Steve Fink [:sfink] [:s:]]

and this means the autodetected bug 1264642 probably is the regressor, though the bug was really in the test function.

Comment 8

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/a2f0768ff0b7
Disallow empty clonebuffer, r=kanru

Comment 9

[Sebastian Hengst [:aryx] (needinfo me if it's about an intermittent or backout)]

https://hg.mozilla.org/mozilla-central/rev/a2f0768ff0b7