Stack buffer overflow (OOB write) in [@ soundtouch::FIRFilter::evaluateFilterMulti] with >16 audio channels reachable from web content
Categories
(Core :: Audio/Video: Playback, defect)
Tracking
()
| Tracking | Status | |
|---|---|---|
| firefox-esr115 | --- | unaffected |
| firefox-esr140 | --- | unaffected |
| firefox152 | + | fixed |
| firefox153 | + | fixed |
| firefox154 | + | fixed |
People
(Reporter: bugmon, Assigned: chunmin)
References
(Regression)
Details
(6 keywords, Whiteboard: [adv-main152.0.4+r])
Attachments
(8 files, 3 external files, 1 obsolete file)
The SoundTouch library vendored into Firefox (media/libsoundtouch, upstream revision d994965 per moz.yaml) contains a stack buffer overflow in soundtouch::FIRFilter::evaluateFilterMulti. The function declares a fixed-size local accumulator array 'LONG_SAMPLETYPE sums[16]' (float[16] in Firefox builds) but indexes it with the caller-supplied channel count, which may legally be as large as 32: the vendored revision includes the upstream change (commit ddf2866) that raised SOUNDTOUCH_MAX_CHANNELS from 16 to 32 in STTypes.h without resizing the array, and every channel-count validation site (FIFOSamplePipe::verifyNumberOfChannels, used by SoundTouch/RateTransposer/TDStretch/FIFOSampleBuffer::setChannels) accepts up to 32 channels. The in-function assert(numChannels <= SOUNDTOUCH_MAX_CHANNELS) checks against 32, so channel counts of 17-32 pass every check even in debug builds, and the three per-output-frame loops write sums[16..numChannels-1] out of bounds on the stack, with the accumulation loop storing attacker-influenced FIR-filtered audio sample values. Upstream silently fixed this after the vendored revision in commit bb49799 'Fix buffer overflow when processing >16 channels' (2026-03-29), changing sums[16] to sums[SOUNDTOUCH_MAX_CHANNELS]; the fix is not in the Firefox tree.
Web content reaches the bug without any user interaction beyond loading a page: a JS-constructed 24-channel WAV is played through an <audio> element (WaveDemuxer takes the channel count from the WAV fmt chunk unclamped; the ffvpx PCM decoder only rejects zero channels), mozCaptureStream() forces the DecodedStream/AudioDecoderInputTrack media sink (which runs on the MediaTrackGraph and works headless via the system-clock fallback driver), and setting preservesPitch=false with playbackRate=2.0 engages the SoundTouch time stretcher in rate-transposing mode: AudioDecoderInputTrack::EnsureTimeStretcher calls setChannels(24) (accepted, <=32) and setRate(2.0); each graph iteration then flows FillDataToTimeStretcher -> RLBoxSoundTouch::putSamples -> SoundTouch::putSamples -> RateTransposer::processSamples (anti-alias filter enabled by default) -> AAFilter::evaluate -> FIRFilter::evaluate(dest, src, n, 24) -> evaluateFilterMulti, overflowing the stack array. The sibling consumer AudioStream::GetTimeStretched (regular non-captured playback with cubeb audio output) passes the unclamped output channel count the same way and is equally affected.
In official little-endian Mozilla builds libsoundtouch is RLBox wasm-sandboxed, which contains the corruption inside the sandbox's linear memory (still a security bug under Firefox's defense-in-depth policy). The full native stack-corruption impact applies to supported shipping configurations that compile soundtouch natively: builds configured --without-wasm-sandboxed-libraries (e.g. Linux distro builds without a wasi sysroot) and all big-endian targets, where toolkit/moz.configure disables wasm sandboxing entirely. For demonstration, the library was compiled natively via a minimal build-config patch (equivalent to --without-wasm-sandboxed-libraries for this library, matching Mozilla's fuzzing/code-coverage configurations); the vulnerable code itself is unmodified vendored upstream source. The testcase crashes the content process with a fatal sanitizer out-of-bounds-index report at FIRFilter.cpp:174 (the first OOB store of the overflow, sums[16] of float[16]), independently reproduced and approved by the verifier.
Build Info
- Branch: main
- Revision: 92d220f44b0d7165e2cb271406d62c313ec79d69
- Timestamp: 2026-06-06T16:14:48+00:00
Affected Code
File: media/libsoundtouch/src/FIRFilter.cpp, line 149-196
uint FIRFilter::evaluateFilterMulti(SAMPLETYPE *dest, const SAMPLETYPE *src, uint numSamples, uint numChannels)
{
int j, end;
assert(length != 0);
assert(src != nullptr);
assert(dest != nullptr);
assert(filterCoeffs != nullptr);
assert(numChannels <= SOUNDTOUCH_MAX_CHANNELS); // 24 <= 32: passes; no-op in NDEBUG builds anyway
// hint compiler autovectorization that loop length is divisible by 8
int ilength = length & -8;
end = numChannels * (numSamples - ilength);
#pragma omp parallel for
for (j = 0; j < end; j += numChannels)
{
const SAMPLETYPE *ptr;
LONG_SAMPLETYPE sums[16]; // <-- fixed 16 floats, but numChannels may be up to 32
uint c;
int i;
for (c = 0; c < numChannels; c ++)
{
sums[c] = 0; // <-- line 174: OOB stack WRITE for c >= 16
}
ptr = src + j;
for (i = 0; i < ilength; i ++)
{
SAMPLETYPE coef=filterCoeffs[i];
for (c = 0; c < numChannels; c ++)
{
sums[c] += ptr[0] * coef; // <-- OOB read-modify-write with attacker-influenced values
ptr ++;
}
}
for (c = 0; c < numChannels; c ++)
{
#ifdef SOUNDTOUCH_INTEGER_SAMPLES
sums[c] >>= resultDivFactor;
#endif // SOUNDTOUCH_INTEGER_SAMPLES
dest[j+c] = (SAMPLETYPE)sums[c]; // <-- OOB read
}
}
return numSamples - ilength;
}
File: media/libsoundtouch/src/STTypes.h, line 57
#define SOUNDTOUCH_MAX_CHANNELS 32 // raised from 16 upstream (commit ddf2866) without resizing sums[16]
File: media/libsoundtouch/src/FIFOSamplePipe.h, line 56-64
bool verifyNumberOfChannels(int nChannels) const
{
if ((nChannels > 0) && (nChannels <= SOUNDTOUCH_MAX_CHANNELS)) // accepts up to 32
{
return true;
}
ST_THROW_RT_ERROR("Error: Illegal number of channels"); // no-op: ST_NO_EXCEPTION_HANDLING + NDEBUG
return false;
}
File: dom/media/mediasink/AudioDecoderInputTrack.cpp, line 600-616, 655-662
void AudioDecoderInputTrack::EnsureTimeStretcher() {
...
mTimeStretcher->setSampleRate(Graph()->GraphRate());
mTimeStretcher->setChannels(GetChannelCountForTimeStretcher()); // unclamped decoded channel count
...
}
uint32_t AudioDecoderInputTrack::GetChannelCountForTimeStretcher() const {
// we always use the initial input channel count to initialize the time stretcher
return mInitialInputChannels; // = AudioInfo::mChannels from the media file (e.g. 24 from the WAV fmt chunk)
}
The fixed 16-element stack array is indexed by a channel count that may be up to 32. SOUNDTOUCH_MAX_CHANNELS was raised from 16 to 32 in the vendored revision without resizing the array; the upstream fix (bb49799) that changes sums[16] to sums[SOUNDTOUCH_MAX_CHANNELS] is not vendored.
Exploit Chain
- Attacker serves a web page that synthesizes a 24-channel, 48 kHz, 16-bit PCM WAV file in JavaScript (any channel count 17-32 works) and assigns it as the source of an <audio> element via a Blob URL.
- WaveDemuxer parses the WAV fmt chunk and propagates channels=24 into AudioInfo unclamped; the ffvpx PCM decoder accepts it (only channels==0 is rejected), so decoded AudioData carries 24 channels.
- The page calls audio.mozCaptureStream(), forcing the media decoder to use the DecodedStream sink, which feeds audio into the MediaTrackGraph through AudioDecoderInputTrack with mInitialInputChannels=24 (works headless via the MTG system-clock fallback driver).
- The page sets audio.preservesPitch=false and audio.playbackRate=2.0, then calls audio.play(); AudioDecoderInputTrack creates the SoundTouch time stretcher with setChannels(24) — accepted because 24 <= SOUNDTOUCH_MAX_CHANNELS (32) — and configures rate mode via setTempo(1.0)/setRate(2.0).
- On each MediaTrackGraph iteration, FillDataToTimeStretcher interleaves the 24-channel audio and calls RLBoxSoundTouch::putSamples -> SoundTouch::putSamples; since rate > 1.0, samples are routed to RateTransposer::processSamples, which applies the anti-alias FIR filter (enabled by default): AAFilter::evaluate -> FIRFilter::evaluate(dest, src, numSamples, 24).
- numChannels is neither 1 nor 2, so evaluateFilterMulti runs; once >= 64 frames (the filter length) are buffered, the per-frame loops index sums[0..23] into the 16-element stack array: sums[16..23] are written out of bounds (32 bytes past the array), and the accumulation loop repeatedly stores attacker-influenced FIR-filtered sample values into stack memory beyond the array for every output frame.
- Impact: repeated controlled-value out-of-bounds stack writes in the content process media thread. In native-soundtouch configurations (distro builds without wasm sandboxing, big-endian targets) this corrupts adjacent stack memory (locals/spills/canaries) and is a basis for further exploitation; in wasm-RLBox builds the corruption lands inside the sandbox heap, corrupting the audio pipeline state (defense-in-depth violation).
Steps to Reproduce
- Apply source_patch.diff to the Firefox source tree (git apply source_patch.diff). It only excludes soundtouch from the default RLBox wasm-sandboxed library set (equivalent to --without-wasm-sandboxed-libraries for this library; matches Linux distro / big-endian / Mozilla fuzzing+coverage configurations) so the native, sanitizer-instrumented code is used. The vulnerable code is unmodified.
- Build Firefox with the ASAN/UBSan fuzzing mozconfig: mcp__clauditor__build_firefox with mozconfig /app/mozconfigs/mozconfig.linux.asan.fuzzing (build_dir: /firefox/obj-firefox-asan).
- Load testcase.html via mcp__clauditor__browser_evaluator against <build_dir>/dist/bin/firefox with prefs {"media.autoplay.default": 0, "media.autoplay.blocking_policy": 0, "media.forcestereo.enabled": false, "accessibility.monoaudio.enable": false, "fuzzing.enabled": true} (timeout 45 s). The autoplay prefs merely substitute for a user clicking play; the others match shipping defaults on Windows/macOS/PulseAudio-Linux.
- Within a few seconds of playback the content process aborts with a fatal sanitizer out-of-bounds-index report at media/libsoundtouch/src/FIRFilter.cpp:174 ('index 16 out of bounds for type LONG_SAMPLETYPE[16]') in soundtouch::FIRFilter::evaluateFilterMulti, on the MediaTrackGraph GraphRunner thread.
Security Impact
- Severity: High
- Attacker capability: A remote attacker hosting a web page obtains repeated out-of-bounds stack writes (and reads) in the Firefox content process, fully triggerable with no user interaction beyond visiting the page. The overwritten values are FIR-accumulated sums of attacker-supplied audio samples multiplied by known filter coefficients, giving the attacker substantial control over the bytes written past the array; the overflow extent is selectable (4 to 64 bytes past the buffer via the file's channel count, 17-32) and repeats every output frame on every graph iteration, providing many corruption attempts per second. In configurations where soundtouch is compiled natively (builds without wasm sandboxing such as typical Linux distro builds lacking a wasi sysroot, and all big-endian platforms), this is a classic stack buffer overflow suitable as a foundation for content-process code execution. In default Mozilla wasm-RLBox builds the writes are contained in the sandbox linear memory, where they corrupt the audio pipeline's data structures (sandbox-internal corruption; treated as a security-sensitive defect under Mozilla's defense-in-depth policy).
- Preconditions: Victim loads an attacker-controlled page that plays a 17-32 channel audio resource with preservesPitch=false and playbackRate != 1.0 (autoplay permission or any user gesture that starts playback). Full native memory-corruption impact requires a Firefox build where libsoundtouch is not wasm-sandboxed (--without-wasm-sandboxed-libraries / no wasi sysroot / big-endian); in default wasm-RLBox desktop builds the corruption is confined to the RLBox sandbox.
ASAN Report
../../../../media/libsoundtouch/src/FIRFilter.cpp:174:13: runtime error: index 16 out of bounds for type 'LONG_SAMPLETYPE[16]' (aka 'float[16]')
#0 0x7ad8a8a95eff in soundtouch::FIRFilter::evaluateFilterMulti(float*, float const*, unsigned int, unsigned int) /firefox/obj-firefox-asan/media/libsoundtouch/src/./../../../../media/libsoundtouch/src/FIRFilter.cpp:174:21
#1 0x7ad8a8a93bfe in soundtouch::FIRFilter::evaluate(float*, float const*, unsigned int, unsigned int) /firefox/obj-firefox-asan/media/libsoundtouch/src/./../../../../media/libsoundtouch/src/FIRFilter.cpp:268:16
#2 0x7ad8a8a93bfe in soundtouch::AAFilter::evaluate(soundtouch::FIFOSampleBuffer&, soundtouch::FIFOSampleBuffer&) const /firefox/obj-firefox-asan/media/libsoundtouch/src/./../../../../media/libsoundtouch/src/AAFilter.cpp:211:20
#3 0x7ad8a8a9c87d in soundtouch::RateTransposer::processSamples(float const*, unsigned int) /firefox/obj-firefox-asan/media/libsoundtouch/src/./../../../../media/libsoundtouch/src/RateTransposer.cpp:168:20
#4 0x7ad8a8a9ed7d in soundtouch::FIFOSamplePipe::moveSamples(soundtouch::FIFOSamplePipe&) /firefox/media/libsoundtouch/src/FIFOSamplePipe.h:93:9
#5 0x7ad8a8a9ed7d in soundtouch::SoundTouch::putSamples(float const*, unsigned int) /firefox/obj-firefox-asan/media/libsoundtouch/src/./../../../../media/libsoundtouch/src/SoundTouch.cpp:0:0
#6 0x7ad8a8a8f210 in auto rlbox::rlbox_noop_sandbox::impl_invoke_with_func_ptr<...> /firefox/obj-firefox-asan/dist/include/mozilla/rlbox/rlbox_noop_sandbox.hpp:187:12
#7 0x7ad8a8a8f210 in auto rlbox::rlbox_sandbox<rlbox::rlbox_noop_sandbox>::INTERNAL_invoke_with_func_ptr<...> /firefox/obj-firefox-asan/dist/include/mozilla/rlbox/rlbox_sandbox.hpp:809:22
#8 0x7ad8a8a8f210 in mozilla::RLBoxSoundTouch::putSamples(float const*, unsigned int) /firefox/media/libsoundtouch/src/RLBoxSoundTouch.cpp:100:12
#9 0x76d88543c9a9 in mozilla::AudioDecoderInputTrack::FillDataToTimeStretcher(long)::$_0::operator()(mozilla::AudioChunk*) const /firefox/obj-firefox-asan/dom/media/mediasink/./../../../../dom/media/mediasink/AudioDecoderInputTrack.cpp:468:21
#10 0x76d88543c9a9 in void mozilla::AudioSegment::IterateOnChunks<...> /firefox/dom/media/AudioSegment.h:460:11
#11 0x76d88543c9a9 in mozilla::AudioDecoderInputTrack::FillDataToTimeStretcher(long) /firefox/obj-firefox-asan/dom/media/mediasink/./../../../../dom/media/mediasink/AudioDecoderInputTrack.cpp:437:17
#12 0x76d88543ae62 in _ZZN7mozilla22AudioDecoderInputTrack32AppendTimeStretchedDataToSegmentElRNS_12AudioSegmentEENK3$_0clIjEEDaT_ /firefox/obj-firefox-asan/dom/media/mediasink/./../../../../dom/media/mediasink/AudioDecoderInputTrack.cpp:420:26
#13 0x76d88543ae62 in auto rlbox::tainted_base_impl<...>::copy_and_verify<...> /firefox/obj-firefox-asan/dist/include/mozilla/rlbox/rlbox.hpp:494:14
#14 0x76d88543ae62 in mozilla::AudioDecoderInputTrack::AppendTimeStretchedDataToSegment(long, mozilla::AudioSegment&) /firefox/obj-firefox-asan/dom/media/mediasink/./../../../../dom/media/mediasink/AudioDecoderInputTrack.cpp:415:32
#15 0x76d88543a677 in mozilla::AudioDecoderInputTrack::AppendBufferedDataToOutput(long) /firefox/obj-firefox-asan/dom/media/mediasink/./../../../../dom/media/mediasink/AudioDecoderInputTrack.cpp:368:9
#16 0x76d885439717 in mozilla::AudioDecoderInputTrack::ProcessInput(long, long, unsigned int) /firefox/obj-firefox-asan/dom/media/mediasink/./../../../../dom/media/mediasink/AudioDecoderInputTrack.cpp:321:23
#17 0x76d884ccda81 in mozilla::MediaTrackGraphImpl::Process(mozilla::MixerCallbackReceiver*) /firefox/obj-firefox-asan/dom/media/./../../../dom/media/MediaTrackGraph.cpp:1466:15
#18 0x76d884ccf832 in mozilla::MediaTrackGraphImpl::OneIterationImpl(long, mozilla::MixerCallbackReceiver*) /firefox/obj-firefox-asan/dom/media/./../../../dom/media/MediaTrackGraph.cpp:1624:3
#19 0x76d884972d5d in mozilla::GraphRunner::Run() /firefox/obj-firefox-asan/dom/media/./../../../dom/media/GraphRunner.cpp:143:32
#20 0x76d87d48b37a in nsThread::ProcessNextEvent(bool, bool*) /firefox/obj-firefox-asan/xpcom/threads/./../../../xpcom/threads/nsThread.cpp:1173:16
#21 0x76d87d494ae9 in NS_ProcessNextEvent(nsIThread*, bool) /firefox/obj-firefox-asan/xpcom/threads/./../../../xpcom/threads/nsThreadUtils.cpp:472:10
#22 0x76d87d780f21 in mozilla::ipc::MessagePumpForNonMainThreads::Run(base::MessagePump::Delegate*) /firefox/obj-firefox-asan/ipc/glue/./../../../ipc/glue/MessagePump.cpp:297:20
#23 0x76d87d5bc944 in MessageLoop::RunInternal() /firefox/obj-firefox-asan/ipc/chromium/./../../../ipc/chromium/src/base/message_loop.cc:371:10
#24 0x76d87d5bc944 in MessageLoop::RunHandler() /firefox/obj-firefox-asan/ipc/chromium/./../../../ipc/chromium/src/base/message_loop.cc:364:3
#25 0x76d87d5bc944 in MessageLoop::Run() /firefox/obj-firefox-asan/ipc/chromium/./../../../ipc/chromium/src/base/message_loop.cc:346:3
#26 0x76d87d483194 in nsThread::ThreadFunc(void*) /firefox/obj-firefox-asan/xpcom/threads/./../../../xpcom/threads/nsThread.cpp:374:10
#27 0x7ad8a8e998cf in _pt_root /firefox/obj-firefox-asan/config/external/nspr/pr/./../../../../../nsprpub/pr/src/pthreads/ptthread.c:190:3
#28 0x5c9b235fe946 in asan_thread_start(void*) _asan_rtl_:28
#29 0x7ad8a9337aa3 in pthread_condattr_setpshared ??:?
#30 0x7ad8a93c4a63 in clone ??:0:0
SUMMARY: UndefinedBehaviorSanitizer: out-of-bounds-index ../../../../media/libsoundtouch/src/FIRFilter.cpp:174:13
| Reporter | ||
Comment 1•5 months ago
|
||
| Reporter | ||
Comment 2•5 months ago
|
||
| Reporter | ||
Comment 3•5 months ago
|
||
| Reporter | ||
Comment 4•5 months ago
|
||
| Reporter | ||
Comment 5•5 months ago
|
||
| Reporter | ||
Comment 6•5 months ago
|
||
| Reporter | ||
Comment 7•5 months ago
|
||
Updated•5 months ago
|
| Assignee | ||
Comment 9•5 months ago
|
||
Plays a 24-channel WAV with preservesPitch=false and playbackRate=2.0 (and
captures the output to drive the MediaTrackGraph), which routes the audio
through SoundTouch's rate transposer into FIRFilter::evaluateFilterMulti.
media.forcestereo.enabled=false keeps the >16 channels from being downmixed
before reaching the time-stretcher.
| Assignee | ||
Comment 10•5 months ago
|
||
upstream report with a standalone test and fix. Will report to upstream tomorrow
| Assignee | ||
Updated•5 months ago
|
| Assignee | ||
Comment 11•5 months ago
|
||
evaluateFilterMulti declared a fixed LONG_SAMPLETYPE sums[16] but indexes it for
each of numChannels, which setChannels permits up to SOUNDTOUCH_MAX_CHANNELS (32).
Audio with 17-32 channels through the rate transposer therefore wrote past the
end of the stack array. Size the accumulator to SOUNDTOUCH_MAX_CHANNELS, matching
the validated ceiling. Carried as a local vendor patch (firfilter-max-channels.patch)
until the next revendor picks up the equivalent upstream change.
| Assignee | ||
Comment 12•5 months ago
|
||
I've contacted to the soundtouch developers. Waiting for replies now. Ideally it should be fixed on upstream and then imported to Firefox.
| Assignee | ||
Comment 13•5 months ago
|
||
I found the fix is already in the upstream, and I've uploaded libsoundtouch via bug 2048343. Sorry for the delay. My bad.
| Assignee | ||
Comment 14•5 months ago
|
||
Comment on attachment 9595701 [details]
(secure)
Security Approval Request
-
How easily could an exploit be constructed based on the patch?: Easy. The one-line diff (a fixed-size array renamed to use the library's defined maximum) directly reveals that the old size was smaller than the maximum allowed channel count; an attacker reading the diff immediately knows which channel counts (greater than the former array size) trigger the defect. The affected code path is reachable with standard web APIs — an audio element with a multi-channel source,
preservesPitch=false, and a non-unityplaybackRate— requiring no special timing, heap shaping, or privileged access. -
Do comments in the patch, the check-in comment, or tests included in the patch paint a bulls-eye on the security problem?: No. The patch has been audited and security-revealing comments and commit body detail have been removed. The regression test for this fix will land separately, approximately one month after the fix ships.
-
Which branches (beta, release, and/or ESR) are affected by this flaw, and do the release status flags reflect this affected/unaffected state correctly?: Introduced in Firefox 150 by Bug 2018973. Affects Nightly (153), Beta (153), and Release (150–152). ESR 140, ESR 128, and ESR 115 are not affected — those branches carry the library at a prior version where the channel ceiling matched the accumulator size. The status flags have not yet been updated; they should be set to affected for Firefox 150–152 and unaffected for ESR 140, ESR 128, and ESR 115.
-
If not all supported branches, which bug introduced the flaw?: Bug 2018973 (Update SoundTouch to version 2.4.0). Bug 1176300 originally introduced the fixed-size accumulator
sums[16]with preconditionassert(numChannels < 16). Althoughassertis compiled out in release builds, the flaw could not be reached regardless: the API-level gateverifyNumberOfChannelsindependently limits the channel count to<= SOUNDTOUCH_MAX_CHANNELS, which was 16 at the time, andsums[16]holds exactly 16 elements (indices 0–15). Bug 2018973 raisedSOUNDTOUCH_MAX_CHANNELSto 32 and updated the precondition toassert(numChannels <= SOUNDTOUCH_MAX_CHANNELS), but left the accumulator atsums[16]; channel counts 17–32 now pass the gate, reach the loop, and write past the end of the array. -
Do you have backports for the affected branches?: Yes. The fix is a one-line local vendor patch (firfilter-max-channels.patch) already prepared and verified; it applies identically to beta and release since the affected file has not diverged between those branches and trunk. Bug 2048343 will carry the equivalent upstream change on Nightly via a full revendor, so this local patch is the uplift vehicle only.
-
If not, how different, hard to create, and risky will they be?: N/A — backport is already prepared.
-
How likely is this patch to cause regressions; how much testing does it need?: Low. The change is a single line in a third-party library (array declaration in a hot FIR loop), with no behavioral effect for audio with 16 or fewer channels. Correctness for the previously out-of-bounds channel counts is verified by a local crashtest (FAIL → PASS on a native-soundtouch ASan build). The fix compiles cleanly in both default wasm-sandboxed and native configurations.
-
Is the patch ready to land after security approval is given?: Yes.
-
Is Android affected?: Yes.
media/libsoundtouch/is compiled for all platforms including Android/GeckoView, and the audio time-stretcher path is shared across platforms.
Drafted with the assistance of Claude Code — reviewed and approved by the patch author.
Updated•5 months ago
|
Updated•5 months ago
|
Comment 15•5 months ago
|
||
The bug has a release status flag that shows some version of Firefox is affected, thus it will be considered confirmed.
Comment 16•5 months ago
|
||
firefox-beta Uplift Approval Request
- User impact if declined/Reason for urgency: sec-high stack buffer overflow (out-of-bounds write, csectype-bounds) reachable from web content with no interaction beyond loading a page. A page that plays a 17-32 channel audio resource with
preservesPitch=falseand a non-unityplaybackRatedrives the SoundTouch time-stretcher's FIR filter to write attacker-influenced audio sample values past a fixed 16-element stack array. In builds where libsoundtouch is not wasm-sandboxed (e.g. Linux distro builds without a wasi sysroot, big-endian targets) this is a content-process stack memory-corruption primitive; in default RLBox builds the writes land inside the sandbox, corrupting audio-pipeline state (defense-in-depth violation). - Code covered by automated testing?: no
- Fix verified in Nightly?: yes
- Needs manual QE testing?: no
- Steps to reproduce for manual QE testing: N/A
- Risk associated with taking this patch: low
- Explanation of risk level: The fix is a one-line change to a vendored third-party library that enlarges a fixed-size stack accumulator to the already-validated maximum, matching the upstream fix. It has no behavioural effect for the common case (<=16 channels) and only widens the buffer the higher channel counts already require. Scope is minimal and structural, applied via a small vendored local patch.
- String changes made/needed?: None
- Is Android affected?: yes
| Assignee | ||
Comment 17•5 months ago
|
||
| Assignee | ||
Comment 18•5 months ago
|
||
Comment 19•5 months ago
|
||
firefox-release Uplift Approval Request
- User impact if declined/Reason for urgency: sec-high stack buffer overflow (out-of-bounds write, csectype-bounds) reachable from web content with no interaction beyond loading a page. A page that plays a 17-32 channel audio resource with
preservesPitch=falseand a non-unityplaybackRatedrives the SoundTouch time-stretcher's FIR filter to write attacker-influenced audio sample values past a fixed 16-element stack array. In builds where libsoundtouch is not wasm-sandboxed (e.g. Linux distro builds without a wasi sysroot, big-endian targets) this is a content-process stack memory-corruption primitive; in default RLBox builds the writes land inside the sandbox, corrupting audio-pipeline state (defense-in-depth violation). - Code covered by automated testing?: no
- Fix verified in Nightly?: yes
- Needs manual QE testing?: no
- Steps to reproduce for manual QE testing: N/A
- Risk associated with taking this patch: low
- Explanation of risk level: The fix is a one-line change to a vendored third-party library that enlarges a fixed-size stack accumulator to the already-validated maximum, matching the upstream fix. It has no behavioural effect for the common case (<=16 channels) and only widens the buffer the higher channel counts already require. Scope is minimal and structural, applied via a small vendored local patch.
- String changes made/needed?: None
- Is Android affected?: yes
Updated•4 months ago
|
| Assignee | ||
Updated•4 months ago
|
| Assignee | ||
Updated•4 months ago
|
Updated•4 months ago
|
Updated•4 months ago
|
Comment 20•4 months ago
|
||
| uplift | ||
Updated•4 months ago
|
Updated•4 months ago
|
Updated•4 months ago
|
Comment 21•4 months ago
|
||
| uplift | ||
Updated•4 months ago
|
Updated•3 months ago
|
Comment 22•2 months ago
|
||
Comment 23•2 months ago
|
||
| bugherder | ||
Updated•2 months ago
|
Updated•2 months ago
|
Description
•