[asterisk-dev] [Code Review]: Fix a large variety of errors caused by negative return codes

[Matt Jordan]

> On April 11, 2012, 2:39 p.m., Paul Belanger wrote:
> > /branches/1.8/formats/format_g719.c, lines 118-129
> > <https://reviewboard.asterisk.org/r/1863/diff/1/?file=27154#file27154line118>
> >
> >     It looks like this code is duplicated in a lot of places.  Wouldn't it be cleaner to create ast_ftruncate() or something?

Maybe, but I don't think its justifiable as a bug fix for 1.8.

Each format in Asterisk is responsible for the implementation of the ast_format virtual table.  There aren't 'default' functions provided for any of the various function in the table, with the rule that if a function is not defined, then the format doesn't support that operation or it returns -1 if defined.  Hence, all information regarding those functions is defined locally in each format file.  The idea of having 'default' functions that could be pointed to in a format's vtable is a concept that doesn't currently exist, and would entail a new API for 1.8 - which is a large change that is not justifiable for just this issue (in my opinion).


- Matt


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviewboard.asterisk.org/r/1863/#review5968
-----------------------------------------------------------


On April 11, 2012, 2:26 p.m., Matt Jordan wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviewboard.asterisk.org/r/1863/
> -----------------------------------------------------------
> 
> (Updated April 11, 2012, 2:26 p.m.)
> 
> 
> Review request for Asterisk Developers.
> 
> 
> Summary
> -------
> 
> Through a static analysis tool, a large number of errors were found that dealt with using potential negative return values in harmful ways.  The following is a summary of the changes for the various affected files:
> 
> * app_voicemail
>   - a negative result from lseek is later passed into mmap as the size of the memory map
>   - a negative result from read closes the various file descriptors and unlinks the output file, but the next check against the return value only checks to see if its non-zero
> * chan_agent - ast_channel_unlock returns a negative value if the item passed to it is NULL or an invalid ao2 object.  strerror cannot accept a negative value.
> * chan_dahdi - if dahdi_get_index returns a negative value, we index directly into an array
> * format_* - various errors not checking return values of ftello and fseek, and passing the results into functions that expect non-negative values
> * func_env - various places where the result of ftello was directly used as input to functions that expect non-negative values
> * asterisk
>   - the request to read returns a negative value.  This causes us to enter into the retry logic.  If we attempt to reconnect and succeed, we would normally proceed in the for(;;) loop and attempt to index into buf using the negative return value from read.  Instead, if we do reconnect, we immediately return to the beginning of the for(;;) loop and attempt a new read.
>   - similar situation - if a read fails, don't attempt to index into a buffer using the return value
> * frame - if we can't determine a preferred codec using the provided values, don't attempt to use an index value that never got set
> * manager 
>   - various failures of mkstemp, lseek were not checked for and could be provided to methods that don't handle negative numbers
>   - passing a negative result from lseek into mmap as the size of the memory map
> * translate - powerof can return a negative result if no bits are set, which would then be used as an index into an array
> * res_agi - if read returns an error, we treated it as if bytes were read from the pipe
> * res_musiconhold - if we fail to spawn spawn_mp3 returns a negative number, we wait a bit and attempt again later.  However, the return 'file descriptor' srcfd is later passed into read.
> * res_rtp_asterisk - ast_rtp_codecs_payload_code can return a negative value if no codecs are found that match between instance1 and the specified payload.  If that's the case, the bridge should be broken, as there are no compatible formats between the two different endpoints (and we shouldn't index into an array using the return value)
> 
> 
> Diffs
> -----
> 
>   /branches/1.8/main/translate.c 361945 
>   /branches/1.8/main/manager.c 361945 
>   /branches/1.8/main/frame.c 361945 
>   /branches/1.8/main/asterisk.c 361945 
>   /branches/1.8/funcs/func_env.c 361945 
>   /branches/1.8/formats/format_wav_gsm.c 361945 
>   /branches/1.8/formats/format_wav.c 361945 
>   /branches/1.8/formats/format_vox.c 361945 
>   /branches/1.8/formats/format_sln16.c 361945 
>   /branches/1.8/formats/format_sln.c 361945 
>   /branches/1.8/formats/format_siren7.c 361945 
>   /branches/1.8/formats/format_siren14.c 361945 
>   /branches/1.8/formats/format_pcm.c 361945 
>   /branches/1.8/formats/format_h263.c 361945 
>   /branches/1.8/formats/format_h264.c 361945 
>   /branches/1.8/formats/format_ilbc.c 361945 
>   /branches/1.8/formats/format_g729.c 361945 
>   /branches/1.8/formats/format_gsm.c 361945 
>   /branches/1.8/formats/format_g719.c 361945 
>   /branches/1.8/formats/format_g723.c 361945 
>   /branches/1.8/channels/chan_agent.c 361945 
>   /branches/1.8/channels/chan_dahdi.c 361945 
>   /branches/1.8/apps/app_voicemail.c 361945 
>   /branches/1.8/res/res_agi.c 361945 
>   /branches/1.8/res/res_musiconhold.c 361945 
>   /branches/1.8/res/res_rtp_asterisk.c 361945 
> 
> Diff: https://reviewboard.asterisk.org/r/1863/diff
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Matt
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-dev/attachments/20120411/aa880080/attachment.htm>