[prev in list] [next in list] [prev in thread] [next in thread] 

List:       gpsd-dev
Subject:    Re: [gpsd-dev] C Client API Issues
From:       "Gary E. Miller" <gem () rellim ! com>
Date:       2016-07-21 19:35:47
Message-ID: 20160721123547.4e6ce6a4 () spidey ! rellim ! com
[Download RAW message or body]


Yo Robert!

Good analysis.  Patches anyone?

On Thu, 21 Jul 2016 19:19:45 +0000
Robert Norris <rw_norris@hotmail.com> wrote:

> > From: Hal Murray <hmurray@megapathdsl.net>  
>   
> >rw_norris@hotmail.com said:  
> > >gpsd needs a flag so close is a noop if it isn't open.
> >> Wouldn't such a flag would need to go into the gps_data_t
> >> structure?   
> 
> >I was being sloppy when I said "flag".   I don't care how it's
> >implemented as long as duplicate calls to close don't cause any
> >problems.  
> 
> Even though I was on about initialization, currently calling
> gps_close() twice will crash the client with a double free.
> 
> privdata should be explicitly set to NULL after the free in
> gps_socket_close().
> 
> There also looks to be issues with gps_shm_* functions too, such as:
> 
> gps_shm_open() doesn't set gpsdata->privdata to NULL when attaching
> to shared memory fails.
> 
> Thus if you tried to then use gps_shm_read() [which you shouldn't do
> anyway] it would probably crash.
> 
> gps_shm_close() never checks if gpsdata->privdata is NULL, so again
> it would crash the client if used after an unsuccessful open.
> 
> It should be easy to make these more robust against API incorrect use.
> 
> 
> 
>     
> 




RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97703
	gem@rellim.com  Tel:+1 541 382 8588

[Attachment #3 (application/pgp-signature)]

[prev in list] [next in list] [prev in thread] [next in thread]