From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Dumazet Subject: Re: [PATCH net 00/15] netpoll: avoid capture effects for NAPI drivers Date: Mon, 24 Sep 2018 12:41:20 -0700 Message-ID: References: <20180921222752.101307-1-edumazet@google.com> <20180923.220440.434897583339366571.davem@davemloft.net> Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Cc: David Miller , netdev , michael.chan@broadcom.com, Ariel Elior , Eric Dumazet , Tariq Toukan , Saeed Mahameed , Jeff Kirsher , jakub.kicinski@netronome.com, Jay Vosburgh , Veaceslav Falico , Andy Gospodarek To: songliubraving@fb.com Return-path: Received: from mail-it1-f169.google.com ([209.85.166.169]:40131 "EHLO mail-it1-f169.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1725824AbeIYBpV (ORCPT ); Mon, 24 Sep 2018 21:45:21 -0400 Received: by mail-it1-f169.google.com with SMTP id h23-v6so12151883ita.5 for ; Mon, 24 Sep 2018 12:41:34 -0700 (PDT) In-Reply-To: Sender: netdev-owner@vger.kernel.org List-ID: On Mon, Sep 24, 2018 at 12:31 PM Song Liu wrote: > This triggers dev_watchdog() on a simple netperf TCP_RR on bnxt (I haven't > tested other drivers yet). > > I guess this is because NAPI_STATE_SCHED is set when poll_one_napi() calls > napi->poll(). And then cleared by napi->poll(). Which part of napi->poll() could possibly clear NAPI_STATE_SCHED when called by netpoll ? AFAIK, napi_complete_done() should exit early (before having a chance to clear NAPI_STATE_SCHED) because of : if (unlikely(n->state & (NAPIF_STATE_NPSVC | NAPIF_STATE_IN_BUSY_POLL))) return false; Thanks !