From mboxrd@z Thu Jan 1 00:00:00 1970 From: Enrico Mioso Subject: Re: [PATCH V2] cdc_ncm: Add support for moving NDP to end of NCM frame Date: Tue, 7 Jul 2015 22:50:41 +0200 (CEST) Message-ID: References: <1435787748-30393-1-git-send-email-mrkiko.rs@gmail.com> <1436177296.8225.25.camel@suse.com> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII; format=flowed Cc: linux-usb@vger.kernel.org, netdev@vger.kernel.org To: Oliver Neukum Return-path: Received: from mail-wi0-f178.google.com ([209.85.212.178]:33144 "EHLO mail-wi0-f178.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932713AbbGGUus (ORCPT ); Tue, 7 Jul 2015 16:50:48 -0400 In-Reply-To: <1436177296.8225.25.camel@suse.com> Sender: netdev-owner@vger.kernel.org List-ID: Hi Oliver, hello to who is reading this message. i was re-reading the code and the oops, without understanding what's the problem. Still: what impressed me is the fact that at some point you see NULL ptr dereference in unrelated code (fbcon). Is it possible that at some point the memory portion (172 bytes if device is affected by NCM errata, and mine is), that the portion of memory to which ctx->delayed_ndp16 points to is somehow moved / thrown away? It doesn't make sense, because otherwise even accesses to the ctx variable would give problems. And they don't. Looking around then, I see kzalloc() / kmalloc (kzalloc =kmalloc | __GFP_ZERO) are used to allocate any size of memory (with the only requirement for it to be small). In rndis_host.c 1025 bytes (not 1024) are allocated, so I am excluding any kind of alignment problem here. Thank you, Enrico