lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
Open Source and information security mailing list archives
| ||
|
Date: Thu, 30 Apr 2015 10:19:29 -0700 From: Eric Dumazet <eric.dumazet@...il.com> To: Dave Taht <dave.taht@...il.com> Cc: David Miller <davem@...emloft.net>, Kathleen Nichols <nichols@...lere.com>, Van Jacobson <vanj@...gle.com>, netdev <netdev@...r.kernel.org> Subject: Re: [PATCH net] codel: fix maxpacket/mtu confusion On Thu, 2015-04-30 at 10:01 -0700, Dave Taht wrote: > I long ago dropped the maxpacket check and the statistic/statistic > collection from various test versions of codel in favor of either a > constant (1500 byte) check or none at all. There is always some level > of buffering underneath codel. I have always wanted the big > GRO/TSO/GRO superpackets to get peeled apart also - the present system > worked OK with ECN in most cases, but seemed like it could get ugly > if we started dropping 64k worth of superpacket. The peeling part does not belong to codel or fq_codel, but eventually the parent qdisc, like TBF (peeling is already there), or HTB (no peeling yet) codel has binary choice : drop or not drop. Starting to doing X MSS drops out of GRO packets seems not needed to me if sender is elastic (properly reacts to drop/ecn_marks). It put extra logic in the wrong place. Again, if you have GRO super packets in codel/fq_codel, it means the sender is buggy enough to send big bursts. Just drop the big GRO packet, do not try to be gentle with the bad guy. -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@...r.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists