| Message ID | 20200607132737.1375-1-jamrial@gmail.com |
|---|---|
| State | New |
| Headers | show |
| Series | [FFmpeg-devel] avcodec/cbs_h2645: keep separate parameter set lists for reading and writing | expand |
| Context | Check | Description |
|---|---|---|
| andriy/default | pending | |
| andriy/make | success | Make finished |
| andriy/make_fate | success | Make fate finished |
On Sun, Jun 07, 2020 at 10:27:37AM -0300, James Almer wrote: > Similar logic as 4e2bef6a82. In scearios where an Access Unit is written right > after reading it using the same CBS context (hevc_metadata), the list of parsed > parameters sets used by the writer must not be the one that's the result of the > reader having already parsed the current Access Unit in question. > > Fixes: out of array access > Fixes: 23034/clusterfuzz-testcase-minimized-ffmpeg_BSF_HEVC_METADATA_fuzzer-5074645169733632.fuzz > > Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg > Signed-off-by: James Almer <jamrial@gmail.com> > --- > An alternative is forcing the usage of separate CBS contexts for reading and > writing. > > libavcodec/cbs_h264.h | 18 ++++++++--- > libavcodec/cbs_h2645.c | 72 ++++++++++++++++++++++++++++++++---------- > libavcodec/cbs_h265.h | 26 +++++++++++---- > 3 files changed, 89 insertions(+), 27 deletions(-) i think the change is probably ok and it fixes the issue what i feel uneasy about is if this is the sole way the security issue is fixed. let me try to explain what i mean by a simpler example: if you have a sprintf() that overwrites the buffer there are 3 ways to fix that A. You make the buffer big enough for what is written B. You make the amount written only as large as the buffer C. You check by using snprintf() Now like here A/B may represent a bugfix the problem with A/B is that security rests on potentially complex code So even when A/B is done, we also should do C This patch fixes the inconsistency on the write side be keeping more references to the parameter sets. For security one would have to proof that no crafted input to the reader fed into any available useer of the reader could result in an inconsistency to a writer following. Are you sure thats the case now and with future users of the code ? OTOH as dumb as a check in the writer may look, anyone can proof it fixes the specific inconsistency. What i suggest specifically is to also include or apply the simple check on top of this, or a equivalent / more generic check. So that security does not rest on alot of spread out code Thanks [...]
On 6/7/2020 11:45 AM, Michael Niedermayer wrote: > On Sun, Jun 07, 2020 at 10:27:37AM -0300, James Almer wrote: >> Similar logic as 4e2bef6a82. In scearios where an Access Unit is written right >> after reading it using the same CBS context (hevc_metadata), the list of parsed >> parameters sets used by the writer must not be the one that's the result of the >> reader having already parsed the current Access Unit in question. >> >> Fixes: out of array access >> Fixes: 23034/clusterfuzz-testcase-minimized-ffmpeg_BSF_HEVC_METADATA_fuzzer-5074645169733632.fuzz >> >> Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg >> Signed-off-by: James Almer <jamrial@gmail.com> >> --- >> An alternative is forcing the usage of separate CBS contexts for reading and >> writing. >> >> libavcodec/cbs_h264.h | 18 ++++++++--- >> libavcodec/cbs_h2645.c | 72 ++++++++++++++++++++++++++++++++---------- >> libavcodec/cbs_h265.h | 26 +++++++++++---- >> 3 files changed, 89 insertions(+), 27 deletions(-) > > i think the change is probably ok and it fixes the issue > what i feel uneasy about is if this is the sole way the security > issue is fixed. > > let me try to explain what i mean by a simpler example: > if you have a sprintf() that overwrites the buffer there are 3 ways > to fix that > A. You make the buffer big enough for what is written > B. You make the amount written only as large as the buffer > C. You check by using snprintf() > > Now like here A/B may represent a bugfix > the problem with A/B is that security rests on potentially complex code > So even when A/B is done, we also should do C > > This patch fixes the inconsistency on the write side be keeping more references > to the parameter sets. > For security one would have to proof that no crafted input to the reader > fed into any available useer of the reader could result in an inconsistency > to a writer following. > Are you sure thats the case now and with future users of the code ? > OTOH as dumb as a check in the writer may look, anyone can proof it fixes the > specific inconsistency. > > What i suggest specifically is to also include or apply the simple check > on top of this, or a equivalent / more generic check. So that security does not > rest on alot of spread out code > > Thanks Well, one possibility is that after this, the infer() warning could be replaced with an assert() instead. The CBS framework is not public, so crashing with an assert() would be acceptable as infer() failing in writing scenarios should be considered an internal bug (bitstream filters, or any CBS user for that matter, should ensure to not change fields in a way that would result in an invalid bitstream and thus failing infer() checks). The issue shouldn't be treated as "If inter_ref_pic_set_prediction_flag is 1 in this scenario, then we should stop to avoid out of array access", but as "We did something wrong because inter_ref_pic_set_prediction_flag was absolutely not meant to be 1 at this point". So using assert() after this patch sounds like a good solution and will help detect future bugs in the parsing code.
On 6/7/2020 12:20 PM, James Almer wrote: > On 6/7/2020 11:45 AM, Michael Niedermayer wrote: >> On Sun, Jun 07, 2020 at 10:27:37AM -0300, James Almer wrote: >>> Similar logic as 4e2bef6a82. In scearios where an Access Unit is written right >>> after reading it using the same CBS context (hevc_metadata), the list of parsed >>> parameters sets used by the writer must not be the one that's the result of the >>> reader having already parsed the current Access Unit in question. >>> >>> Fixes: out of array access >>> Fixes: 23034/clusterfuzz-testcase-minimized-ffmpeg_BSF_HEVC_METADATA_fuzzer-5074645169733632.fuzz >>> >>> Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg >>> Signed-off-by: James Almer <jamrial@gmail.com> >>> --- >>> An alternative is forcing the usage of separate CBS contexts for reading and >>> writing. >>> >>> libavcodec/cbs_h264.h | 18 ++++++++--- >>> libavcodec/cbs_h2645.c | 72 ++++++++++++++++++++++++++++++++---------- >>> libavcodec/cbs_h265.h | 26 +++++++++++---- >>> 3 files changed, 89 insertions(+), 27 deletions(-) >> >> i think the change is probably ok and it fixes the issue >> what i feel uneasy about is if this is the sole way the security >> issue is fixed. >> >> let me try to explain what i mean by a simpler example: >> if you have a sprintf() that overwrites the buffer there are 3 ways >> to fix that >> A. You make the buffer big enough for what is written >> B. You make the amount written only as large as the buffer >> C. You check by using snprintf() >> >> Now like here A/B may represent a bugfix >> the problem with A/B is that security rests on potentially complex code >> So even when A/B is done, we also should do C >> >> This patch fixes the inconsistency on the write side be keeping more references >> to the parameter sets. >> For security one would have to proof that no crafted input to the reader >> fed into any available useer of the reader could result in an inconsistency >> to a writer following. >> Are you sure thats the case now and with future users of the code ? >> OTOH as dumb as a check in the writer may look, anyone can proof it fixes the >> specific inconsistency. >> >> What i suggest specifically is to also include or apply the simple check >> on top of this, or a equivalent / more generic check. So that security does not >> rest on alot of spread out code >> >> Thanks > > Well, one possibility is that after this, the infer() warning could be > replaced with an assert() instead. The CBS framework is not public, so > crashing with an assert() would be acceptable as infer() failing in > writing scenarios should be considered an internal bug (bitstream > filters, or any CBS user for that matter, should ensure to not change > fields in a way that would result in an invalid bitstream and thus > failing infer() checks). > > The issue shouldn't be treated as "If inter_ref_pic_set_prediction_flag > is 1 in this scenario, then we should stop to avoid out of array > access", but as "We did something wrong because > inter_ref_pic_set_prediction_flag was absolutely not meant to be 1 at > this point". So using assert() after this patch sounds like a good > solution and will help detect future bugs in the parsing code. It could also be a generic return AVERROR_INVALIDDATA as you suggested, to be fair. All the standard writing helpers abort gracefully that way, so infer() could do the same. Or the other helpers could be changed to assert().
On Sun, Jun 07, 2020 at 12:45:15PM -0300, James Almer wrote: > On 6/7/2020 12:20 PM, James Almer wrote: > > On 6/7/2020 11:45 AM, Michael Niedermayer wrote: > >> On Sun, Jun 07, 2020 at 10:27:37AM -0300, James Almer wrote: > >>> Similar logic as 4e2bef6a82. In scearios where an Access Unit is written right > >>> after reading it using the same CBS context (hevc_metadata), the list of parsed > >>> parameters sets used by the writer must not be the one that's the result of the > >>> reader having already parsed the current Access Unit in question. > >>> > >>> Fixes: out of array access > >>> Fixes: 23034/clusterfuzz-testcase-minimized-ffmpeg_BSF_HEVC_METADATA_fuzzer-5074645169733632.fuzz > >>> > >>> Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg > >>> Signed-off-by: James Almer <jamrial@gmail.com> > >>> --- > >>> An alternative is forcing the usage of separate CBS contexts for reading and > >>> writing. > >>> > >>> libavcodec/cbs_h264.h | 18 ++++++++--- > >>> libavcodec/cbs_h2645.c | 72 ++++++++++++++++++++++++++++++++---------- > >>> libavcodec/cbs_h265.h | 26 +++++++++++---- > >>> 3 files changed, 89 insertions(+), 27 deletions(-) > >> > >> i think the change is probably ok and it fixes the issue > >> what i feel uneasy about is if this is the sole way the security > >> issue is fixed. > >> > >> let me try to explain what i mean by a simpler example: > >> if you have a sprintf() that overwrites the buffer there are 3 ways > >> to fix that > >> A. You make the buffer big enough for what is written > >> B. You make the amount written only as large as the buffer > >> C. You check by using snprintf() > >> > >> Now like here A/B may represent a bugfix > >> the problem with A/B is that security rests on potentially complex code > >> So even when A/B is done, we also should do C > >> > >> This patch fixes the inconsistency on the write side be keeping more references > >> to the parameter sets. > >> For security one would have to proof that no crafted input to the reader > >> fed into any available useer of the reader could result in an inconsistency > >> to a writer following. > >> Are you sure thats the case now and with future users of the code ? > >> OTOH as dumb as a check in the writer may look, anyone can proof it fixes the > >> specific inconsistency. > >> > >> What i suggest specifically is to also include or apply the simple check > >> on top of this, or a equivalent / more generic check. So that security does not > >> rest on alot of spread out code > >> > >> Thanks > > > > Well, one possibility is that after this, the infer() warning could be > > replaced with an assert() instead. The CBS framework is not public, so > > crashing with an assert() would be acceptable as infer() failing in > > writing scenarios should be considered an internal bug (bitstream > > filters, or any CBS user for that matter, should ensure to not change > > fields in a way that would result in an invalid bitstream and thus > > failing infer() checks). > > > > The issue shouldn't be treated as "If inter_ref_pic_set_prediction_flag > > is 1 in this scenario, then we should stop to avoid out of array > > access", but as "We did something wrong because > > inter_ref_pic_set_prediction_flag was absolutely not meant to be 1 at > > this point". So using assert() after this patch sounds like a good > > solution and will help detect future bugs in the parsing code. > > It could also be a generic return AVERROR_INVALIDDATA as you suggested, > to be fair. All the standard writing helpers abort gracefully that way, > so infer() could do the same. > > Or the other helpers could be changed to assert(). whats the status of this ? has this issue been fixed in some other way i missed ? will this get applied ? should i apply my not so pretty fix for 23034 ? what will be done about releases ? can this be backported ? PS: please just make sure 23034/ is mentioned in the commit message so whatever fixes it can be kept track of (i know its already mentioned this is more intended as a remainder for other alternative fixes) thx [...]
James Almer: > Similar logic as 4e2bef6a82. In scearios where an Access Unit is written right > after reading it using the same CBS context (hevc_metadata), the list of parsed > parameters sets used by the writer must not be the one that's the result of the > reader having already parsed the current Access Unit in question. > > Fixes: out of array access > Fixes: 23034/clusterfuzz-testcase-minimized-ffmpeg_BSF_HEVC_METADATA_fuzzer-5074645169733632.fuzz > > Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg > Signed-off-by: James Almer <jamrial@gmail.com> > --- > An alternative is forcing the usage of separate CBS contexts for reading and > writing. > Am I the only one who thinks this would be cleaner, as it would directly show to the user that there are two separate states involved? It would also lead to a smaller diff. - Andreas
On 7/1/2020 6:40 AM, Michael Niedermayer wrote: > On Sun, Jun 07, 2020 at 12:45:15PM -0300, James Almer wrote: >> On 6/7/2020 12:20 PM, James Almer wrote: >>> On 6/7/2020 11:45 AM, Michael Niedermayer wrote: >>>> On Sun, Jun 07, 2020 at 10:27:37AM -0300, James Almer wrote: >>>>> Similar logic as 4e2bef6a82. In scearios where an Access Unit is written right >>>>> after reading it using the same CBS context (hevc_metadata), the list of parsed >>>>> parameters sets used by the writer must not be the one that's the result of the >>>>> reader having already parsed the current Access Unit in question. >>>>> >>>>> Fixes: out of array access >>>>> Fixes: 23034/clusterfuzz-testcase-minimized-ffmpeg_BSF_HEVC_METADATA_fuzzer-5074645169733632.fuzz >>>>> >>>>> Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg >>>>> Signed-off-by: James Almer <jamrial@gmail.com> >>>>> --- >>>>> An alternative is forcing the usage of separate CBS contexts for reading and >>>>> writing. >>>>> >>>>> libavcodec/cbs_h264.h | 18 ++++++++--- >>>>> libavcodec/cbs_h2645.c | 72 ++++++++++++++++++++++++++++++++---------- >>>>> libavcodec/cbs_h265.h | 26 +++++++++++---- >>>>> 3 files changed, 89 insertions(+), 27 deletions(-) >>>> >>>> i think the change is probably ok and it fixes the issue >>>> what i feel uneasy about is if this is the sole way the security >>>> issue is fixed. >>>> >>>> let me try to explain what i mean by a simpler example: >>>> if you have a sprintf() that overwrites the buffer there are 3 ways >>>> to fix that >>>> A. You make the buffer big enough for what is written >>>> B. You make the amount written only as large as the buffer >>>> C. You check by using snprintf() >>>> >>>> Now like here A/B may represent a bugfix >>>> the problem with A/B is that security rests on potentially complex code >>>> So even when A/B is done, we also should do C >>>> >>>> This patch fixes the inconsistency on the write side be keeping more references >>>> to the parameter sets. >>>> For security one would have to proof that no crafted input to the reader >>>> fed into any available useer of the reader could result in an inconsistency >>>> to a writer following. >>>> Are you sure thats the case now and with future users of the code ? >>>> OTOH as dumb as a check in the writer may look, anyone can proof it fixes the >>>> specific inconsistency. >>>> >>>> What i suggest specifically is to also include or apply the simple check >>>> on top of this, or a equivalent / more generic check. So that security does not >>>> rest on alot of spread out code >>>> >>>> Thanks >>> >>> Well, one possibility is that after this, the infer() warning could be >>> replaced with an assert() instead. The CBS framework is not public, so >>> crashing with an assert() would be acceptable as infer() failing in >>> writing scenarios should be considered an internal bug (bitstream >>> filters, or any CBS user for that matter, should ensure to not change >>> fields in a way that would result in an invalid bitstream and thus >>> failing infer() checks). >>> >>> The issue shouldn't be treated as "If inter_ref_pic_set_prediction_flag >>> is 1 in this scenario, then we should stop to avoid out of array >>> access", but as "We did something wrong because >>> inter_ref_pic_set_prediction_flag was absolutely not meant to be 1 at >>> this point". So using assert() after this patch sounds like a good >>> solution and will help detect future bugs in the parsing code. >> >> It could also be a generic return AVERROR_INVALIDDATA as you suggested, >> to be fair. All the standard writing helpers abort gracefully that way, >> so infer() could do the same. >> >> Or the other helpers could be changed to assert(). > > whats the status of this ? > has this issue been fixed in some other way i missed ? I pushed the first two patches and backported them, so the issue should be fixed by the first. > will this get applied ? I delayed applying this one waiting for more opinions, especially Mark's, since it's kind of ugly. > should i apply my not so pretty fix for 23034 ? > what will be done about releases ? can this be backported ? Already answered above, but maybe confirm it's fixed just to be sure. > > PS: please just make sure 23034/ is mentioned in the commit message so > whatever fixes it can be kept track of (i know its already mentioned > this is more intended as a remainder for other alternative fixes) Ah, guess i should have mentioned that ossfuzz issue in ef13fafe22 (and e6ab99f324by extension) seeing i didn't push this one alongside it... > > thx > > [...] > > > _______________________________________________ > ffmpeg-devel mailing list > ffmpeg-devel@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-devel > > To unsubscribe, visit link above, or email > ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe". >
On Wed, Jul 01, 2020 at 10:22:48AM -0300, James Almer wrote: > On 7/1/2020 6:40 AM, Michael Niedermayer wrote: > > On Sun, Jun 07, 2020 at 12:45:15PM -0300, James Almer wrote: > >> On 6/7/2020 12:20 PM, James Almer wrote: > >>> On 6/7/2020 11:45 AM, Michael Niedermayer wrote: > >>>> On Sun, Jun 07, 2020 at 10:27:37AM -0300, James Almer wrote: > >>>>> Similar logic as 4e2bef6a82. In scearios where an Access Unit is written right > >>>>> after reading it using the same CBS context (hevc_metadata), the list of parsed > >>>>> parameters sets used by the writer must not be the one that's the result of the > >>>>> reader having already parsed the current Access Unit in question. > >>>>> > >>>>> Fixes: out of array access > >>>>> Fixes: 23034/clusterfuzz-testcase-minimized-ffmpeg_BSF_HEVC_METADATA_fuzzer-5074645169733632.fuzz > >>>>> > >>>>> Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg > >>>>> Signed-off-by: James Almer <jamrial@gmail.com> > >>>>> --- > >>>>> An alternative is forcing the usage of separate CBS contexts for reading and > >>>>> writing. > >>>>> > >>>>> libavcodec/cbs_h264.h | 18 ++++++++--- > >>>>> libavcodec/cbs_h2645.c | 72 ++++++++++++++++++++++++++++++++---------- > >>>>> libavcodec/cbs_h265.h | 26 +++++++++++---- > >>>>> 3 files changed, 89 insertions(+), 27 deletions(-) > >>>> > >>>> i think the change is probably ok and it fixes the issue > >>>> what i feel uneasy about is if this is the sole way the security > >>>> issue is fixed. > >>>> > >>>> let me try to explain what i mean by a simpler example: > >>>> if you have a sprintf() that overwrites the buffer there are 3 ways > >>>> to fix that > >>>> A. You make the buffer big enough for what is written > >>>> B. You make the amount written only as large as the buffer > >>>> C. You check by using snprintf() > >>>> > >>>> Now like here A/B may represent a bugfix > >>>> the problem with A/B is that security rests on potentially complex code > >>>> So even when A/B is done, we also should do C > >>>> > >>>> This patch fixes the inconsistency on the write side be keeping more references > >>>> to the parameter sets. > >>>> For security one would have to proof that no crafted input to the reader > >>>> fed into any available useer of the reader could result in an inconsistency > >>>> to a writer following. > >>>> Are you sure thats the case now and with future users of the code ? > >>>> OTOH as dumb as a check in the writer may look, anyone can proof it fixes the > >>>> specific inconsistency. > >>>> > >>>> What i suggest specifically is to also include or apply the simple check > >>>> on top of this, or a equivalent / more generic check. So that security does not > >>>> rest on alot of spread out code > >>>> > >>>> Thanks > >>> > >>> Well, one possibility is that after this, the infer() warning could be > >>> replaced with an assert() instead. The CBS framework is not public, so > >>> crashing with an assert() would be acceptable as infer() failing in > >>> writing scenarios should be considered an internal bug (bitstream > >>> filters, or any CBS user for that matter, should ensure to not change > >>> fields in a way that would result in an invalid bitstream and thus > >>> failing infer() checks). > >>> > >>> The issue shouldn't be treated as "If inter_ref_pic_set_prediction_flag > >>> is 1 in this scenario, then we should stop to avoid out of array > >>> access", but as "We did something wrong because > >>> inter_ref_pic_set_prediction_flag was absolutely not meant to be 1 at > >>> this point". So using assert() after this patch sounds like a good > >>> solution and will help detect future bugs in the parsing code. > >> > >> It could also be a generic return AVERROR_INVALIDDATA as you suggested, > >> to be fair. All the standard writing helpers abort gracefully that way, > >> so infer() could do the same. > >> > >> Or the other helpers could be changed to assert(). > > > > whats the status of this ? > > has this issue been fixed in some other way i missed ? > > I pushed the first two patches and backported them, so the issue should > be fixed by the first. > > > will this get applied ? > > I delayed applying this one waiting for more opinions, especially > Mark's, since it's kind of ugly. > > > should i apply my not so pretty fix for 23034 ? > > what will be done about releases ? can this be backported ? > > Already answered above, but maybe confirm it's fixed just to be sure. 23034 no longer reproduces on master locally thx [...]
diff --git a/libavcodec/cbs_h264.h b/libavcodec/cbs_h264.h index 9f7c2a0d30..9d104787d9 100644 --- a/libavcodec/cbs_h264.h +++ b/libavcodec/cbs_h264.h @@ -448,10 +448,20 @@ typedef struct CodedBitstreamH264Context { // All currently available parameter sets. These are updated when // any parameter set NAL unit is read/written with this context. - AVBufferRef *sps_ref[H264_MAX_SPS_COUNT]; - AVBufferRef *pps_ref[H264_MAX_PPS_COUNT]; - H264RawSPS *sps[H264_MAX_SPS_COUNT]; - H264RawPPS *pps[H264_MAX_PPS_COUNT]; + AVBufferRef **sps_ref; + AVBufferRef **pps_ref; + H264RawSPS **sps; + H264RawPPS **pps; + + AVBufferRef *read_sps_ref[H264_MAX_SPS_COUNT]; + AVBufferRef *read_pps_ref[H264_MAX_PPS_COUNT]; + H264RawSPS *read_sps[H264_MAX_SPS_COUNT]; + H264RawPPS *read_pps[H264_MAX_PPS_COUNT]; + + AVBufferRef *write_sps_ref[H264_MAX_SPS_COUNT]; + AVBufferRef *write_pps_ref[H264_MAX_PPS_COUNT]; + H264RawSPS *write_sps[H264_MAX_SPS_COUNT]; + H264RawPPS *write_pps[H264_MAX_PPS_COUNT]; // The currently active parameter sets. These are updated when any // NAL unit refers to the relevant parameter set. These pointers diff --git a/libavcodec/cbs_h2645.c b/libavcodec/cbs_h2645.c index b432921ecc..69ed890c63 100644 --- a/libavcodec/cbs_h2645.c +++ b/libavcodec/cbs_h2645.c @@ -758,14 +758,14 @@ static int cbs_h2645_split_fragment(CodedBitstreamContext *ctx, return 0; } -#define cbs_h2645_replace_ps(h26n, ps_name, ps_var, id_element) \ +#define cbs_h2645_replace_ps(h26n, Hn, ps_name, ps_var, id_element) \ static int cbs_h26 ## h26n ## _replace_ ## ps_var(CodedBitstreamContext *ctx, \ CodedBitstreamUnit *unit) \ { \ CodedBitstreamH26 ## h26n ## Context *priv = ctx->priv_data; \ H26 ## h26n ## Raw ## ps_name *ps_var = unit->content; \ unsigned int id = ps_var->id_element; \ - if (id >= FF_ARRAY_ELEMS(priv->ps_var)) { \ + if (id >= Hn ## _MAX_ ## ps_name ## _COUNT) { \ av_log(ctx->log_ctx, AV_LOG_ERROR, "Invalid " #ps_name \ " id : %d.\n", id); \ return AVERROR_INVALIDDATA; \ @@ -785,18 +785,24 @@ static int cbs_h26 ## h26n ## _replace_ ## ps_var(CodedBitstreamContext *ctx, \ return 0; \ } -cbs_h2645_replace_ps(4, SPS, sps, seq_parameter_set_id) -cbs_h2645_replace_ps(4, PPS, pps, pic_parameter_set_id) -cbs_h2645_replace_ps(5, VPS, vps, vps_video_parameter_set_id) -cbs_h2645_replace_ps(5, SPS, sps, sps_seq_parameter_set_id) -cbs_h2645_replace_ps(5, PPS, pps, pps_pic_parameter_set_id) +cbs_h2645_replace_ps(4, H264, SPS, sps, seq_parameter_set_id) +cbs_h2645_replace_ps(4, H264, PPS, pps, pic_parameter_set_id) +cbs_h2645_replace_ps(5, HEVC, VPS, vps, vps_video_parameter_set_id) +cbs_h2645_replace_ps(5, HEVC, SPS, sps, sps_seq_parameter_set_id) +cbs_h2645_replace_ps(5, HEVC, PPS, pps, pps_pic_parameter_set_id) static int cbs_h264_read_nal_unit(CodedBitstreamContext *ctx, CodedBitstreamUnit *unit) { + CodedBitstreamH264Context *priv = ctx->priv_data; GetBitContext gbc; int err; + priv->sps_ref = (AVBufferRef **)priv->read_sps_ref; + priv->pps_ref = (AVBufferRef **)priv->read_pps_ref; + priv->sps = (H264RawSPS **)priv->read_sps; + priv->pps = (H264RawPPS **)priv->read_pps; + err = init_get_bits(&gbc, unit->data, 8 * unit->data_size); if (err < 0) return err; @@ -953,9 +959,17 @@ static int cbs_h264_read_nal_unit(CodedBitstreamContext *ctx, static int cbs_h265_read_nal_unit(CodedBitstreamContext *ctx, CodedBitstreamUnit *unit) { + CodedBitstreamH265Context *priv = ctx->priv_data; GetBitContext gbc; int err; + priv->vps_ref = (AVBufferRef **)priv->read_vps_ref; + priv->sps_ref = (AVBufferRef **)priv->read_sps_ref; + priv->pps_ref = (AVBufferRef **)priv->read_pps_ref; + priv->vps = (H265RawVPS **)priv->read_vps; + priv->sps = (H265RawSPS **)priv->read_sps; + priv->pps = (H265RawPPS **)priv->read_pps; + err = init_get_bits(&gbc, unit->data, 8 * unit->data_size); if (err < 0) return err; @@ -1164,8 +1178,14 @@ static int cbs_h264_write_nal_unit(CodedBitstreamContext *ctx, CodedBitstreamUnit *unit, PutBitContext *pbc) { + CodedBitstreamH264Context *priv = ctx->priv_data; int err; + priv->sps_ref = (AVBufferRef **)priv->write_sps_ref; + priv->pps_ref = (AVBufferRef **)priv->write_pps_ref; + priv->sps = (H264RawSPS **)priv->write_sps; + priv->pps = (H264RawPPS **)priv->write_pps; + switch (unit->type) { case H264_NAL_SPS: { @@ -1281,8 +1301,16 @@ static int cbs_h265_write_nal_unit(CodedBitstreamContext *ctx, CodedBitstreamUnit *unit, PutBitContext *pbc) { + CodedBitstreamH265Context *priv = ctx->priv_data; int err; + priv->vps_ref = (AVBufferRef **)priv->write_vps_ref; + priv->sps_ref = (AVBufferRef **)priv->write_sps_ref; + priv->pps_ref = (AVBufferRef **)priv->write_pps_ref; + priv->vps = (H265RawVPS **)priv->write_vps; + priv->sps = (H265RawSPS **)priv->write_sps; + priv->pps = (H265RawPPS **)priv->write_pps; + switch (unit->type) { case HEVC_NAL_VPS: { @@ -1483,10 +1511,14 @@ static void cbs_h264_close(CodedBitstreamContext *ctx) ff_h2645_packet_uninit(&h264->common.read_packet); - for (i = 0; i < FF_ARRAY_ELEMS(h264->sps); i++) - av_buffer_unref(&h264->sps_ref[i]); - for (i = 0; i < FF_ARRAY_ELEMS(h264->pps); i++) - av_buffer_unref(&h264->pps_ref[i]); + for (i = 0; i < H264_MAX_SPS_COUNT; i++) { + av_buffer_unref(&h264->read_sps_ref[i]); + av_buffer_unref(&h264->write_sps_ref[i]); + } + for (i = 0; i < H264_MAX_PPS_COUNT; i++) { + av_buffer_unref(&h264->read_pps_ref[i]); + av_buffer_unref(&h264->write_pps_ref[i]); + } } static void cbs_h265_close(CodedBitstreamContext *ctx) @@ -1496,12 +1528,18 @@ static void cbs_h265_close(CodedBitstreamContext *ctx) ff_h2645_packet_uninit(&h265->common.read_packet); - for (i = 0; i < FF_ARRAY_ELEMS(h265->vps); i++) - av_buffer_unref(&h265->vps_ref[i]); - for (i = 0; i < FF_ARRAY_ELEMS(h265->sps); i++) - av_buffer_unref(&h265->sps_ref[i]); - for (i = 0; i < FF_ARRAY_ELEMS(h265->pps); i++) - av_buffer_unref(&h265->pps_ref[i]); + for (i = 0; i < HEVC_MAX_VPS_COUNT; i++) { + av_buffer_unref(&h265->read_vps_ref[i]); + av_buffer_unref(&h265->write_vps_ref[i]); + } + for (i = 0; i < HEVC_MAX_SPS_COUNT; i++) { + av_buffer_unref(&h265->read_sps_ref[i]); + av_buffer_unref(&h265->write_sps_ref[i]); + } + for (i = 0; i < HEVC_MAX_PPS_COUNT; i++) { + av_buffer_unref(&h265->read_pps_ref[i]); + av_buffer_unref(&h265->write_pps_ref[i]); + } } const CodedBitstreamType ff_cbs_type_h264 = { diff --git a/libavcodec/cbs_h265.h b/libavcodec/cbs_h265.h index 73897f77a4..ab27f77f15 100644 --- a/libavcodec/cbs_h265.h +++ b/libavcodec/cbs_h265.h @@ -731,12 +731,26 @@ typedef struct CodedBitstreamH265Context { // All currently available parameter sets. These are updated when // any parameter set NAL unit is read/written with this context. - AVBufferRef *vps_ref[HEVC_MAX_VPS_COUNT]; - AVBufferRef *sps_ref[HEVC_MAX_SPS_COUNT]; - AVBufferRef *pps_ref[HEVC_MAX_PPS_COUNT]; - H265RawVPS *vps[HEVC_MAX_VPS_COUNT]; - H265RawSPS *sps[HEVC_MAX_SPS_COUNT]; - H265RawPPS *pps[HEVC_MAX_PPS_COUNT]; + AVBufferRef **vps_ref; + AVBufferRef **sps_ref; + AVBufferRef **pps_ref; + H265RawVPS **vps; + H265RawSPS **sps; + H265RawPPS **pps; + + AVBufferRef *read_vps_ref[HEVC_MAX_VPS_COUNT]; + AVBufferRef *read_sps_ref[HEVC_MAX_SPS_COUNT]; + AVBufferRef *read_pps_ref[HEVC_MAX_PPS_COUNT]; + H265RawVPS *read_vps[HEVC_MAX_VPS_COUNT]; + H265RawSPS *read_sps[HEVC_MAX_SPS_COUNT]; + H265RawPPS *read_pps[HEVC_MAX_PPS_COUNT]; + + AVBufferRef *write_vps_ref[HEVC_MAX_VPS_COUNT]; + AVBufferRef *write_sps_ref[HEVC_MAX_SPS_COUNT]; + AVBufferRef *write_pps_ref[HEVC_MAX_PPS_COUNT]; + H265RawVPS *write_vps[HEVC_MAX_VPS_COUNT]; + H265RawSPS *write_sps[HEVC_MAX_SPS_COUNT]; + H265RawPPS *write_pps[HEVC_MAX_PPS_COUNT]; // The currently active parameter sets. These are updated when any // NAL unit refers to the relevant parameter set. These pointers
Similar logic as 4e2bef6a82. In scearios where an Access Unit is written right after reading it using the same CBS context (hevc_metadata), the list of parsed parameters sets used by the writer must not be the one that's the result of the reader having already parsed the current Access Unit in question. Fixes: out of array access Fixes: 23034/clusterfuzz-testcase-minimized-ffmpeg_BSF_HEVC_METADATA_fuzzer-5074645169733632.fuzz Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: James Almer <jamrial@gmail.com> --- An alternative is forcing the usage of separate CBS contexts for reading and writing. libavcodec/cbs_h264.h | 18 ++++++++--- libavcodec/cbs_h2645.c | 72 ++++++++++++++++++++++++++++++++---------- libavcodec/cbs_h265.h | 26 +++++++++++---- 3 files changed, 89 insertions(+), 27 deletions(-)