Skip to content

[bigendian] Fix crash during PThread initialization #23700

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
sbc100 merged 3 commits into from
Feb 25, 2025
Merged

[bigendian] Fix crash during PThread initialization #23700

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
331431c
fix: Crash during PThread initialization on big-endian machine
slavek-kucera Feb 19, 2025
1d77a71
Merge branch 'main' into fix_bigendian_pthread_init
kripken Feb 25, 2025
97adaf4
Merge branch 'main' into fix_bigendian_pthread_init
sbc100 Feb 25, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
76 changes: 76 additions & 0 deletions src/lib/liblittle_endian_heap.js
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,6 +34,82 @@ var LibraryLittleEndianHeap = {

$LE_HEAP_LOAD_F64: (byteOffset) =>
HEAP_DATA_VIEW.getFloat64(byteOffset, true),

$LE_ATOMICS_NATIVE_BYTE_ORDER__postset: `
LE_ATOMICS_NATIVE_BYTE_ORDER = (new Int8Array(new Int16Array([1]).buffer)[0] === 1)
? [ /* little endian */
(x => x),
(x => x),
undefined,
(x => x),
]
: [ /* big endian */
(x => x),
(x => (((x & 0xff00) << 8) | ((x & 0xff) << 24)) >> 16),
undefined,
(x => ((x >> 24) & 0xff) | ((x >> 8) & 0xff00) | ((x & 0xff00) << 8) | ((x & 0xff) << 24)),
];
function LE_HEAP_UPDATE() {
HEAPU16.unsigned = (x => x & 0xffff);
HEAPU32.unsigned = (x => x >>> 0);
}
`,
$LE_ATOMICS_NATIVE_BYTE_ORDER: [],

$LE_ATOMICS_ADD: (heap, offset, value) => {
const order = LE_ATOMICS_NATIVE_BYTE_ORDER[heap.BYTES_PER_ELEMENT - 1];
const res = order(Atomics.add(heap, offset, order(value)));
return heap.unsigned ? heap.unsigned(res) : res;
},
$LE_ATOMICS_AND: (heap, offset, value) => {
const order = LE_ATOMICS_NATIVE_BYTE_ORDER[heap.BYTES_PER_ELEMENT - 1];
const res = order(Atomics.and(heap, offset, order(value)));
return heap.unsigned ? heap.unsigned(res) : res;
},
$LE_ATOMICS_COMPAREEXCHANGE: (heap, offset, expected, replacement) => {
const order = LE_ATOMICS_NATIVE_BYTE_ORDER[heap.BYTES_PER_ELEMENT - 1];
const res = order(Atomics.compareExchange(heap, offset, order(expected), order(replacement)));
return heap.unsigned ? heap.unsigned(res) : res;
},
$LE_ATOMICS_EXCHANGE: (heap, offset, value) => {
const order = LE_ATOMICS_NATIVE_BYTE_ORDER[heap.BYTES_PER_ELEMENT - 1];
const res = order(Atomics.exchange(heap, offset, order(value)));
return heap.unsigned ? heap.unsigned(res) : res;
},
$LE_ATOMICS_ISLOCKFREE: (size) => Atomics.isLockFree(size),
$LE_ATOMICS_LOAD: (heap, offset) => {
const order = LE_ATOMICS_NATIVE_BYTE_ORDER[heap.BYTES_PER_ELEMENT - 1];
const res = order(Atomics.load(heap, offset));
return heap.unsigned ? heap.unsigned(res) : res;
},
$LE_ATOMICS_NOTIFY: (heap, offset, count) => Atomics.notify(heap, offset, count),
$LE_ATOMICS_OR: (heap, offset, value) => {
const order = LE_ATOMICS_NATIVE_BYTE_ORDER[heap.BYTES_PER_ELEMENT - 1];
const res = order(Atomics.or(heap, offset, order(value)));
return heap.unsigned ? heap.unsigned(res) : res;
},
$LE_ATOMICS_STORE: (heap, offset, value) => {
const order = LE_ATOMICS_NATIVE_BYTE_ORDER[heap.BYTES_PER_ELEMENT - 1];
Atomics.store(heap, offset, order(value));
},
$LE_ATOMICS_SUB: (heap, offset, value) => {
const order = LE_ATOMICS_NATIVE_BYTE_ORDER[heap.BYTES_PER_ELEMENT - 1];
const res = order(Atomics.sub(heap, offset, order(value)));
return heap.unsigned ? heap.unsigned(res) : res;
},
$LE_ATOMICS_WAIT: (heap, offset, value, timeout) => {
const order = LE_ATOMICS_NATIVE_BYTE_ORDER[heap.BYTES_PER_ELEMENT - 1];
return Atomics.wait(heap, offset, order(value), timeout);
},
$LE_ATOMICS_WAITASYNC: (heap, offset, value, timeout) => {
const order = LE_ATOMICS_NATIVE_BYTE_ORDER[heap.BYTES_PER_ELEMENT - 1];
return Atomics.waitAsync(heap, offset, order(value), timeout);
},
$LE_ATOMICS_XOR: (heap, offset, value) => {
const order = LE_ATOMICS_NATIVE_BYTE_ORDER[heap.BYTES_PER_ELEMENT - 1];
const res = order(Atomics.xor(heap, offset, order(value)));
return heap.unsigned ? heap.unsigned(res) : res;
},
}

addToLibrary(LibraryLittleEndianHeap);
7 changes: 4 additions & 3 deletions src/runtime_shared.js
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -66,9 +66,6 @@ var wasmOffsetConverter;

function updateMemoryViews() {
var b = wasmMemory.buffer;
#if SUPPORT_BIG_ENDIAN
{{{ maybeExportHeap('HEAP_DATA_VIEW') }}} HEAP_DATA_VIEW = new DataView(b);
#endif
{{{ maybeExportHeap('HEAP8') }}}HEAP8 = new Int8Array(b);
{{{ maybeExportHeap('HEAP16') }}}HEAP16 = new Int16Array(b);
{{{ maybeExportHeap('HEAPU8') }}}HEAPU8 = new Uint8Array(b);
Expand All @@ -81,6 +78,10 @@ function updateMemoryViews() {
{{{ maybeExportHeap('HEAP64') }}}HEAP64 = new BigInt64Array(b);
{{{ maybeExportHeap('HEAPU64') }}}HEAPU64 = new BigUint64Array(b);
#endif
#if SUPPORT_BIG_ENDIAN
{{{ maybeExportHeap('HEAP_DATA_VIEW') }}} HEAP_DATA_VIEW = new DataView(b);
LE_HEAP_UPDATE();
#endif
}

#if ENVIRONMENT_MAY_BE_NODE && MIN_NODE_VERSION < 160000
Expand Down
20 changes: 20 additions & 0 deletions test/js_optimizer/LittleEndianHeap-output.js
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,3 +33,23 @@ LE_HEAP_STORE_F64(x * 8, a);
HEAP[x];

HeAp[x];

LE_ATOMICS_ADD(heap, offset, value);

LE_ATOMICS_AND(heap, offset, value);

LE_ATOMICS_COMPAREEXCHANGE(heap, offset, expected, replacement);

LE_ATOMICS_EXCHANGE(heap, offset, value);

LE_ATOMICS_LOAD(heap, offset);

LE_ATOMICS_OR(heap, offset, value);

LE_ATOMICS_SUB(heap, offset, value);

LE_ATOMICS_WAIT(heap, offset, value, timeout);

LE_ATOMICS_WAITASYNC(heap, offset, value, timeout);

LE_ATOMICS_XOR(heap, offset, value);
10 changes: 10 additions & 0 deletions test/js_optimizer/LittleEndianHeap.js
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,3 +16,13 @@ a = HEAPF64[x]; // HEAPF64
HEAPF64[x] = a;
HEAP[x]; // should not be changed
HeAp[x];
Atomics.add(heap, offset, value);
Atomics.and(heap, offset, value);
Atomics.compareExchange(heap, offset, expected, replacement);
Atomics.exchange(heap, offset, value);
Atomics.load(heap, offset);
Atomics.or(heap, offset, value);
Atomics.sub(heap, offset, value);
Atomics.wait(heap, offset, value, timeout);
Atomics.waitAsync(heap, offset, value, timeout);
Atomics.xor(heap, offset, value);
41 changes: 39 additions & 2 deletions tools/acorn-optimizer.mjs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1106,10 +1106,21 @@ function littleEndianHeap(ast) {
recursiveWalk(ast, {
FunctionDeclaration: (node, c) => {
// do not recurse into LE_HEAP_STORE, LE_HEAP_LOAD functions
if (!(node.id.type === 'Identifier' && node.id.name.startsWith('LE_HEAP'))) {
if (
!(
node.id.type === 'Identifier' &&
(node.id.name.startsWith('LE_HEAP') || node.id.name.startsWith('LE_ATOMICS_'))
)
) {
c(node.body);
}
},
VariableDeclarator: (node, c) => {
if (!(node.id.type === 'Identifier' && node.id.name.startsWith('LE_ATOMICS_'))) {
c(node.id);
if (node.init) c(node.init);
}
},
AssignmentExpression: (node, c) => {
const target = node.left;
const value = node.right;
Expand Down Expand Up @@ -1160,6 +1171,27 @@ function littleEndianHeap(ast) {
}
}
},
CallExpression: (node, c) => {
if (node.arguments) {
for (var a of node.arguments) c(a);
}
if (
// Atomics.X(args) -> LE_ATOMICS_X(args)
node.callee.type === 'MemberExpression' &&
node.callee.object.type === 'Identifier' &&
node.callee.object.name === 'Atomics' &&
node.callee.property.type === 'Identifier' &&
!node.computed
) {
makeCallExpression(
node,
'LE_ATOMICS_' + node.callee.property.name.toUpperCase(),
node.arguments,
);
} else {
c(node.callee);
}
},
MemberExpression: (node, c) => {
c(node.property);
if (!isHEAPAccess(node)) {
Expand Down Expand Up @@ -1217,7 +1249,12 @@ function growableHeap(ast) {
recursiveWalk(ast, {
FunctionDeclaration(node, c) {
// Do not recurse into to GROWABLE_HEAP_ helper functions themselves.
if (!(node.id.type === 'Identifier' && node.id.name.startsWith('GROWABLE_HEAP_'))) {
if (
!(
node.id.type === 'Identifier' &&
(node.id.name.startsWith('GROWABLE_HEAP_') || node.id.name === 'LE_HEAP_UPDATE')
)
) {
c(node.body);
}
},
Expand Down
16 changes: 15 additions & 1 deletion tools/link.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1122,7 +1122,21 @@ def phase_linker_setup(options, linker_args): # noqa: C901, PLR0912, PLR0915
'$LE_HEAP_LOAD_U32',
'$LE_HEAP_LOAD_I32',
'$LE_HEAP_LOAD_F32',
'$LE_HEAP_LOAD_F64'
'$LE_HEAP_LOAD_F64',
'$LE_ATOMICS_NATIVE_BYTE_ORDER',
'$LE_ATOMICS_ADD',
'$LE_ATOMICS_AND',
'$LE_ATOMICS_COMPAREEXCHANGE',
'$LE_ATOMICS_EXCHANGE',
'$LE_ATOMICS_ISLOCKFREE',
'$LE_ATOMICS_LOAD',
'$LE_ATOMICS_NOTIFY',
'$LE_ATOMICS_OR',
'$LE_ATOMICS_STORE',
'$LE_ATOMICS_SUB',
'$LE_ATOMICS_WAIT',
'$LE_ATOMICS_WAITASYNC',
'$LE_ATOMICS_XOR',

@sbc100 sbc100 Feb 20, 2025

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

While I agree it would be good to add all of these, I also thing we might want to only add the ones that we actually use (and also test).

@slavek-kucera slavek-kucera Feb 20, 2025

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I thought that the whole point of this approach was to actually implement comprehensive support like was done to the non-atomic memory accesses done from JS code.

@sbc100 sbc100 Feb 20, 2025

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, I suppose you right, seems reasonable to leave them all in then.

But lets skip 64-bit atomics and just fail in that case, to keep things simple. I don't think anyone should be doing that.

@slavek-kucera slavek-kucera Feb 21, 2025

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I've tested it on the following program:

#include <atomic>
#include <print>
#include <emscripten.h>

int main(){
        std::atomic<unsigned long long> x = 0x0123456789ABCDEF;

        std::println("x: {:X}", x.load());

        EM_ASM({ Atomics.compareExchange(HEAP64, ($0)/8, BigInt("0x0123456789ABCDEF"), BigInt("0xFEDCBA9876543210")); }, &x);

        std::println("x: {:X}", x.load());

}

built with

em++ -o main.js -s USE_PTHREADS=1 -s PROXY_TO_PTHREAD=1 -s PTHREAD_POOL_SIZE=8 -s SUPPORT_BIG_ENDIAN=1 -pthread -s EXIT_RUNTIME=1 -s ALLOW_MEMORY_GROWTH=1 -std=c++23 main.cpp

And I do get the expected result on both LE and BE:

x: 123456789ABCDEF
x: FEDCBA9876543210

]

if settings.RUNTIME_DEBUG or settings.ASSERTIONS or settings.STACK_OVERFLOW_CHECK or settings.PTHREADS_PROFILING or settings.GL_ASSERTIONS:
Expand Down