zqz/platform-external-webrtc
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Integer overflow bug in low_cut_filter.

Browse Source 
A multiplication result doesn't fit in an int32_t type. This change
rewrites the code to avoid the overflowing multiplication.

Here y[0], y[1] are int16 numbers containing the (truncated) topmost
18 and (scaled Q2 to use the full int16) the least significant 13
bits of a 32-bit value. The change makes y[1] to be calculated 
directly instead of using y[0] as an intermediate value. 

TESTED=this change passes the bit exactness tests, and has also been 
running on the audio_processing fuzzer with a CHECK comparing the
old and new value.

Bug: chromium:747202
Change-Id: Iafc69eb7391d494afdadf65f5b7f399a57bbe9a8
Reviewed-on: https://chromium-review.googlesource.com/580907
Reviewed-by: Minyue Li <minyue@webrtc.org>
Commit-Queue: Alex Loiko <aleloi@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#19120}
This commit is contained in:
Alex Loiko
2017-07-24 12:25:52 +02:00
committed by Commit Bot
parent fe43df155c
commit e029d99f19
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
webrtc/modules/audio_processing/low_cut_filter.cc
View File

@ -58,8 +58,8 @@ class LowCutFilter::BiquadFilter {
y[2] = y[0]; y[2] = y[0];
y[3] = y[1]; y[3] = y[1];
y[0] = static_cast<int16_t>(tmp_int32 >> 13); y[0] = static_cast<int16_t>(tmp_int32 >> 13);
y[1] = static_cast<int16_t>(
(tmp_int32 - (static_cast<int32_t>(y[0]) * ( 1 << 13))) * 4); y[1] = static_cast<int16_t>((tmp_int32 & 0x00001FFF) * 4);
// Rounding in Q12, i.e. add 2^11. // Rounding in Q12, i.e. add 2^11.
tmp_int32 += 2048; tmp_int32 += 2048;