From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from imap1.codethink.co.uk ([176.9.8.82]:60443 "EHLO imap1.codethink.co.uk" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1727627AbeIEXKs (ORCPT ); Wed, 5 Sep 2018 19:10:48 -0400 Message-ID: <1536172759.3024.44.camel@codethink.co.uk> Subject: Re: [PATCH 4.4 026/124] iwlwifi: pcie: fix race in Rx buffer allocator From: Ben Hutchings To: Shaul Triebitz , Luca Coelho Cc: stable@vger.kernel.org, Sasha Levin , Greg Kroah-Hartman , LKML Date: Wed, 05 Sep 2018 19:39:19 +0100 In-Reply-To: <20180804082703.421702408@linuxfoundation.org> References: <20180804082702.434482435@linuxfoundation.org> <20180804082703.421702408@linuxfoundation.org> Content-Type: text/plain; charset="UTF-8" Mime-Version: 1.0 Content-Transfer-Encoding: 8bit Sender: stable-owner@vger.kernel.org List-ID: On Sat, 2018-08-04 at 11:00 +0200, Greg Kroah-Hartman wrote: > 4.4-stable review patch.  If anyone has any objections, please let me know. > > ------------------ > > From: Shaul Triebitz > > [ Upstream commit 0f22e40053bd5378ad1e3250e65c574fd61c0cd6 ] > > Make sure the rx_allocator worker is canceled before running the > rx_init routine.  rx_init frees and re-allocates all rxb's pages.  The > rx_allocator worker also allocates pages for the used rxb's.  Running > rx_init and rx_allocator simultaniously causes a kernel panic.  Fix > that by canceling the work in rx_init. > > Signed-off-by: Shaul Triebitz > Signed-off-by: Luca Coelho > Signed-off-by: Sasha Levin > Signed-off-by: Greg Kroah-Hartman > --- >  drivers/net/wireless/iwlwifi/pcie/rx.c |    2 ++ >  1 file changed, 2 insertions(+) > > --- a/drivers/net/wireless/iwlwifi/pcie/rx.c > +++ b/drivers/net/wireless/iwlwifi/pcie/rx.c > @@ -713,6 +713,8 @@ int iwl_pcie_rx_init(struct iwl_trans *t >   WQ_HIGHPRI | WQ_UNBOUND, 1); >   INIT_WORK(&rba->rx_alloc, iwl_pcie_rx_allocator_work); >   > + cancel_work_sync(&rba->rx_alloc); > + >   spin_lock(&rba->lock); >   atomic_set(&rba->req_pending, 0); >   atomic_set(&rba->req_ready, 0); This seems to be an incomplete fix. INIT_WORK() will overwrite the list pointers in the work item, so there is still a race condition. I think that the RX initialisation and reinitialisation/reset cases need to be explicitly distinguished. In the initialisation case the driver should do only INIT_WORK() and in the reset case it should do only cancel_work_sync() (although it would be safe to do INIT_WORK() *after* that). Ben. -- Ben Hutchings, Software Developer   Codethink Ltd https://www.codethink.co.uk/ Dale House, 35 Dale Street Manchester, M1 2HF, United Kingdom