From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail.bootlin.com ([62.4.15.54]:51638 "EHLO mail.bootlin.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1726554AbeJBOHC (ORCPT ); Tue, 2 Oct 2018 10:07:02 -0400 Date: Tue, 2 Oct 2018 09:25:00 +0200 From: Miquel Raynal To: Boris Brezillon Cc: Chris Packham , Daniel Mack , "linux-mtd@lists.infradead.org" , "stable@vger.kernel.org" Subject: Re: [PATCH v2] mtd: rawnand: marvell: check for RDY bits after enabling the IRQ Message-ID: <20181002092500.3dc2ac7a@xps13> In-Reply-To: <20181002084601.2a118f10@xps13> References: <20181001234120.425ae1f5@bbrezillon> <20181002001328.7f64ecbc@bbrezillon> <20181002084601.2a118f10@xps13> MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8BIT Sender: stable-owner@vger.kernel.org List-ID: Hi Daniel, Chris, Miquel Raynal wrote on Tue, 2 Oct 2018 08:46:01 +0200: > Hi Boris, > > Boris Brezillon wrote on Tue, 2 Oct 2018 > 00:13:28 +0200: > > > On Mon, 1 Oct 2018 22:01:27 +0000 > > Chris Packham wrote: > > > > > On 02/10/18 10:41, Boris Brezillon wrote: > > > > On Mon, 1 Oct 2018 22:34:38 +0200 > > > > Boris Brezillon wrote: > > > > > > > >>> > > > >>> I'd previously tried readl() based on the same hunch. No change. > > > >>> > > > >>> I think my snippet above might be misleading. While a delay between > > > >>> readl_relaxed() and the if should not change the outcome, this is also a > > > >>> delay between marvell_nfc_enable_int() and marvell_nfc_disable_int() > > > >>> which is probably more significant. Sure enough if I move the delay to > > > >>> just before the marvell_nfc_disable_int() the error is not seen. > > > >> > > > >> AFAICT, your timeout always happens when waiting for RDREQ, not RDYM. > > > >> So maybe disabling MRDY too early has a side-effect on the RDREQ event. > > > > > > > > Can you try with this patch [1]? It should ensure that NDSR_RDY bits > > > > are cleared before starting an operation. > > > > > > > > [1]http://code.bulix.org/lgs30c-468205 > > > > > > > > > > No luck. I applied that on top of Daniel's and got the same result. > > > > > > One thing that does look promising is the following modification of > > > Daniel's patch[1]. Which moves the RDY check to before where the > > > interrupts are enabled. > > > > Except we still don't know why this is happening, and I'm not sure I > > want to take a fix without understanding why it does fix the problem. > > > > Also, it looks like complete() is not called until the RDDREQ, WRDREQ > > and WRCMDREQ are cleared in the interrupt handler [1], which is weird. > > Miquel, do you happen to remember why you had to do that? > > The RDDREQ, WRDREQ and WRCMDREQ events might potentially happen while > the interrupts are enabled while we only wait for R/B signalling. This > check is to avoid calling complete() on these situations. Actually Boris is right on the fact that while the intention is good, the writing of [1] is not accurate. Daniel, could you please test if the following diff changes something with your setup, without your patch? diff --git a/drivers/mtd/nand/raw/marvell_nand.c b/drivers/mtd/nand/raw/marvell_nand.c index bc2ef5209783..c7573ccdbacd 100644 --- a/drivers/mtd/nand/raw/marvell_nand.c +++ b/drivers/mtd/nand/raw/marvell_nand.c @@ -686,7 +686,7 @@ static irqreturn_t marvell_nfc_isr(int irq, void *dev_id) marvell_nfc_disable_int(nfc, st & NDCR_ALL_INT); - if (!(st & (NDSR_RDDREQ | NDSR_WRDREQ | NDSR_WRCMDREQ))) + if (st & (NDSR_RDY(0) | NDSR_RDY(1))) complete(&nfc->complete); return IRQ_HANDLED; > > > > > [1]https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/mtd/nand/raw/marvell_nand.c?h=v4.19-rc6#n689 > Thanks, Miquèl