On Fri, Apr 08, 2022 at 09:33:42AM -0700, Jakub Kicinski wrote: > On Fri, 8 Apr 2022 14:35:22 +0200 Petr Malat wrote: > > On Mon, Mar 07, 2022 at 01:33:21PM -0800, Jakub Kicinski wrote: > > > On Mon, 7 Mar 2022 20:59:29 +0100 Petr Malat wrote: > > > > Function sctp_do_peeloff() wrongly initializes daddr of the original > > > > socket instead of the peeled off one, which makes getpeername() return > > > > zeroes instead of the primary address. Initialize the new socket > > > > instead. > > > > > > Could you add a Fixes tag? > > > > Hi Jakub, > > have you got some time to review the updated version with "Fixes" tag > > added? > > Thanks for adding the tag. Long story short if you got no replies from > SCTP folks on the posting to linux-sctp@ repost with netdev@ added. > That way it'll get into general networking patchwork / patch processing > queue. Yup. Also, the patch itself always need to be posted to netdev@ anyway, because we can't take patches directly from linux-sctp@. Well, unless you mean to not officially submit the patch at a given time for some reason, that is. > > > The issue has been in the kernel for a while, because my app is using > > peer addresses for storing sockets in a hash table and the hash table can > > handle collisions, thus I haven't noticed it's broken until I dumped > > the hash table while working on another problem.