From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ravikiran G Thirumalai Subject: Qdisc drops not being reported to user space Date: Thu, 17 Oct 2002 15:59:55 +0530 Sender: netdev-bounce@oss.sgi.com Message-ID: <20021017155955.E20237@in.ibm.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: linux-kernel@vger.kernel.org Return-path: To: netdev Content-Disposition: inline Errors-to: netdev-bounce@oss.sgi.com List-Id: netdev.vger.kernel.org Sometime back, I had posted a patch to report Qdisc drops under /proc/net/dev. I'd noticed that Qdisc.drops was not being reported for the default queuing discipline (pfifo_fast). I was told that my approach mucked SNMP data. I was also told that prio will be made the default queuing discipline and pfifo_fast will be done away with. Is this the case even now? will prio be made the default queuing disc in the 2.5 time frame? If not, any solution in sight to report Qdisc drops to userland? (I think there is a kernel patch and a patch for tc from Jamal). I have noticed Qdisc drops under webserver loads, and heavy tx loads (I used my proc reporting patch .. it did not break ifconfig atleast..). IMO It'd not be right if packets are being dropped by the kernel (default queuing discipline) and the unsuspecting admin has no means to find that out. IMHO It'd also not be fair to expect users to use CONFIG_NET_SCH_PRIO and add the prio qdisc before using a n/w interface just to be able to make out Qdisc drops. One more thing...IMHO, there must be some documentation somewhere .... tc man page (which doesn't exist as of now?) or atleast the readme which should indicate to the user that drops seen from ifconfig (packets dropped by the adapter) are different from packets dropped from the qdisc as shown by tc. Thanks, Kiran