Return-path: Received: from mail-io0-f177.google.com ([209.85.223.177]:33816 "EHLO mail-io0-f177.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752975AbcA0X13 (ORCPT ); Wed, 27 Jan 2016 18:27:29 -0500 MIME-Version: 1.0 In-Reply-To: <20160127154618.GA5717@localhost> References: <20160126111730.GA6765@localhost> <20160127154618.GA5717@localhost> From: Julian Calaby Date: Thu, 28 Jan 2016 10:27:09 +1100 Message-ID: (sfid-20160128_002736_648758_DED52DA5) Subject: Re: [PATCH] net/mac80211/agg-rx.c: fix use of uninitialised values To: Chris Bainbridge Cc: "linux-kernel@vger.kernel.org" , Johannes Berg , linux-wireless , aryabinin@virtuozzo.com, Julia Lawall , kernel-janitors@vger.kernel.org, Joe Perches Content-Type: text/plain; charset=UTF-8 Sender: linux-wireless-owner@vger.kernel.org List-ID: Hi Chris, On Thu, Jan 28, 2016 at 2:46 AM, Chris Bainbridge wrote: > Use kzalloc instead of kmalloc for struct tid_ampdu_rx. Fixes: > > [ 7.976605] UBSAN: Undefined behaviour in net/mac80211/rx.c:932:29 > [ 7.976608] load of value 2 is not a valid value for type '_Bool' > [ 7.976611] CPU: 3 PID: 1134 Comm: kworker/u16:7 Not tainted 4.5.0-rc1+ #265 > [ 7.976613] Hardware name: Apple Inc. MacBookPro10,2/Mac-AFD8A9D944EA4843, BIOS MBP102.88Z.0106.B0A.1509130955 09/13/2015 > [ 7.976616] Workqueue: phy0 rt2x00usb_work_rxdone > [ 7.976619] 0000000000000004 ffff880254a7ba50 ffffffff8181d866 0000000000000007 > [ 7.976622] ffff880254a7ba78 ffff880254a7ba68 ffffffff8188422d ffffffff8379b500 > [ 7.976626] ffff880254a7bab8 ffffffff81884747 0000000000000202 0000000348620032 > [ 7.976629] Call Trace: > [ 7.976633] [] dump_stack+0x45/0x5f > [ 7.976637] [] ubsan_epilogue+0xd/0x40 > [ 7.976642] [] __ubsan_handle_load_invalid_value+0x67/0x70 > [ 7.976646] [] ieee80211_sta_reorder_release.isra.16+0x5ed/0x730 > [ 7.976650] [] ieee80211_prepare_and_rx_handle+0xd04/0x1c00 > [ 7.976654] [] ? usb_hcd_map_urb_for_dma+0x65e/0x960 > [ 7.976659] [] __ieee80211_rx_handle_packet+0x1f3/0x750 > [ 7.976663] [] ieee80211_rx_napi+0x447/0x990 > [ 7.976667] [] rt2x00lib_rxdone+0x305/0xbd0 > [ 7.976670] [] ? dequeue_task_fair+0x64f/0x1de0 > [ 7.976674] [] ? sched_clock_cpu+0xe6/0x150 > [ 7.976678] [] rt2x00usb_work_rxdone+0x7c/0x140 > [ 7.976682] [] process_one_work+0x226/0x860 > [ 7.976686] [] worker_thread+0x5c/0x680 > [ 7.976690] [] ? process_one_work+0x860/0x860 > [ 7.976693] [] kthread+0xf6/0x150 > [ 7.976697] [] ? kthread_worker_fn+0x310/0x310 > [ 7.976700] [] ret_from_fork+0x3f/0x70 > [ 7.976703] [] ? kthread_worker_fn+0x310/0x310 > > Link: https://lkml.org/lkml/2016/1/26/230 > Signed-off-by: Chris Bainbridge > --- > net/mac80211/agg-rx.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/net/mac80211/agg-rx.c b/net/mac80211/agg-rx.c > index 10ad4ac1fa0b..bde3344cbdd0 100644 > --- a/net/mac80211/agg-rx.c > +++ b/net/mac80211/agg-rx.c > @@ -291,7 +291,7 @@ void __ieee80211_start_rx_ba_session(struct sta_info *sta, > } > > /* prepare A-MPDU MLME for Rx aggregation */ > - tid_agg_rx = kmalloc(sizeof(struct tid_ampdu_rx), GFP_KERNEL); > + tid_agg_rx = kzalloc(sizeof(struct tid_ampdu_rx), GFP_KERNEL); This looks like a "big hammer" solution to this problem. My reading of this is that the events leading up to UBSAN's warning are: 1. In __ieee80211_start_rx_ba_session() tid_agg_rx is kmalloc'd and has "random" data 2. All of tid_ampdu_rx's fields except the boolean "removed" are initialised 3. Stuff happens 4. We get to "set_release_timer" in ieee80211_sta_reorder_release 5. the random data means that "removed" has an incorrect value for a boolean 6. UBSAN barfs as above I believe that false is stored as 0 internally (and true as 1) so kzalloc is a correct solution, but it's the heavy weight solution as all but one of the fields in the struct are initialised immediately after, so we're essentially zeroing the entire struct to ensure that one field is set to zero. I'd prefer to just set ->removed to false right after we set ->auto_seq as that should be faster, however I don't know if __ieee80211_start_rx_ba_session() is a fast path so I don't know if this is saving anything. Either way, this looks sane to me. Reviewed-by: Julian Calaby On another note, this is an error that should be pretty easy to spot. Could any of the automated tools find cases where a struct containing a bool variable is kmalloc'd and returned without assigning all the bools? Thanks, -- Julian Calaby Email: julian.calaby@gmail.com Profile: http://www.google.com/profiles/julian.calaby/