Re: [Dnsmasq-discuss] [PATCH] Fix buffer overflow in TCP requests

2020-07-06 Thread Simon Kelley
On 30/06/2020 20:22, Frank wrote: > Resending this because I realized I sent it to Simon rather than the list: > > Hi Simon, > > This bug is fairly easy to reproduce. It can take 10 mins or more to > reproduce a crash so I suggest checking the length you get in > cache_recv_insert. If the

Re: [Dnsmasq-discuss] [PATCH] Fix buffer overflow in TCP requests

2020-06-30 Thread Frank
Resending this because I realized I sent it to Simon rather than the list: Hi Simon, This bug is fairly easy to reproduce. It can take 10 mins or more to reproduce a crash so I suggest checking the length you get in cache_recv_insert. If the domains being used are > 4 characters, this will catch

Re: [Dnsmasq-discuss] [PATCH] Fix buffer overflow in TCP requests

2020-06-28 Thread Simon Kelley
That's a nasty bug, and could explain what pi-hole users are seeing. If I understand things correctly, this bug will only manifest itself when the write() or read() syscalls return EINTR ir EAGAIN, which is possible, but not common, hence the bugs wasn't detected earlier. Frank, did you find a

Re: [Dnsmasq-discuss] [PATCH] Fix buffer overflow in TCP requests

2020-06-18 Thread Dominik
Hey Frank, dear list-members, thanks for your proposed fix. The Pi-hole team adopted dnsmasq v2.81 early on and we're seeing reports for mysterious crashes scattered all over the dnsmasq code since the very first days of releasing our latest version. Crashes reports show several locations, e.g.,