Open tolikheha opened 6 years ago
Mailing list
The main communication channel for the project is our mailing list: https://list.openfastpath.org/mailman/listinfo/openfastpath
To post a message to all the list members you need to be a subscriber. Email: openfastpath@list.openfastpath.org
Patches
Patches shall be submitted via the mailinglist in diff format. Please keep patches small, atomic (ie isolated to one fix or addition) and well-explained. Consider that the project uses Linux kernel coding style, which is verified through checkpatch.pl.
Hi Thank you for the patch but @OpenFastPath patches are accepted why mailing lists. Please use git format-patch and send using git send-email to send the patch for review on the mailing list.
Hi,
Anyone who wants to contribute to the OpenFastPath Foundation needs to sign the contribution agreement. Please fill and sign the Contribution License Agreement, then send a scanned PDF of it to me: http://www.openfastpath.org/downloads/OpenFastPath_ContributionAgreement.pdf
More at: http://www.openfastpath.org/index.php/about-us/
Kind Regards, Sorin Vultureanu OFP maintainer
email: sorin.vultureanu@enea.com
Data corruption from #93 still relevant and situation worse than it looks.
There are few moment:
First,
ofp_sbappendstream_locked
can drop packet. So tcp controll block shouldn't be updated beforeofp_sbappendstream_locked
called and return value checked. Currently it doesn't return anything, butint ofp_sockbuf_put_last
can return -1 in case of error and why it not passed here is a big mistery.Next, this implementation will call
ofp_tcp_reass
even if there are space only for 1 packet. But this function can try to add a lot more packets than 1. And it will try to do this withofp_sbappendstream_locked
, which will drop all that packets with succesfull updating of tcp controll block. So check of socket buffer should be added toofp_tcp_reass
too.