From mboxrd@z Thu Jan 1 00:00:00 1970 From: Geert Uytterhoeven Subject: Re: [PATCH] mmc: tmio: Fix hang during suspend Date: Tue, 19 Aug 2014 11:18:33 +0200 Message-ID: References: <1408029833-6970-1-git-send-email-geert+renesas@glider.be> <20140815134917.da77205b68f18341140c6593@codethink.co.uk> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Return-path: Received: from mail-la0-f54.google.com ([209.85.215.54]:57152 "EHLO mail-la0-f54.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750767AbaHSJSj (ORCPT ); Tue, 19 Aug 2014 05:18:39 -0400 In-Reply-To: Sender: linux-mmc-owner@vger.kernel.org List-Id: linux-mmc@vger.kernel.org To: Ulf Hansson Cc: Ian Molton , Geert Uytterhoeven , Chris Ball , Linux MMC List , Linux-sh list , "linux-kernel@vger.kernel.org" Hi Ulf, On Mon, Aug 18, 2014 at 3:10 PM, Ulf Hansson wrote: > On 18 August 2014 14:36, Geert Uytterhoeven wrote: >> On Fri, Aug 15, 2014 at 2:49 PM, Ian Molton wrote: >>> On Thu, 14 Aug 2014 17:23:53 +0200 >>> Geert Uytterhoeven wrote: >>>> On R-Car Gen 2, the SDHI registers cannot be accessed while the SDHI >>>> module clock is disabled. Doing so will cause a lock-up. >>>> >>>> When suspending, enable the module clock before disabling the SDHI >>>> interrupts, and disable the clock again afterwards. >>> >>> This feels wrong. Why can't interrupts be disabled prior to turning the clock off? >> >> The clock is handled by runtime PM. So if SDHI becomes idle, the clock >> is turned off. >> However, the card insertion interrupt is still enabled. >> If all interrupts would be disabled when the clock is turned off, I believe card >> insertion can no longer be detected. >> >> BTW, I'm still wondering why tmio_mmc_host_resume() works without >> enabling the clock, as it calls tmio_mmc_enable_dma(), which also writes to >> an SDHI register (CTL_DMA_ENABLE). > > A while ago, I tried to rework the runtime PM handling of tmio. > > http://www.spinics.net/lists/linux-mmc/msg23177.html > > The hole set never got merged, due to that we couldn't get them tested > on hardware and there were a lack in review. > > We might want to pick them up to see if those may solve our problem. :-) Thanks, I applied patches 4-8 (1-3 did get merged), and tried them on SDHI1 of r8a7791/koelsch. It still seems to work as before. However, I still need my patch to fix suspend. Your "[PATCH 8/8] mmc: tmio: Handle clock gating from runtime PM functions" manages the SDHI interface clock, while my patch manages the platform-specific SDHI module clock. >>>> This fixes suspend and resume on r8a7791/koelsch. >>>> >>>> Signed-off-by: Geert Uytterhoeven >>>> --- >>>> drivers/mmc/host/tmio_mmc_pio.c | 10 ++++++++++ >>>> 1 file changed, 10 insertions(+) >>>> >>>> diff --git a/drivers/mmc/host/tmio_mmc_pio.c b/drivers/mmc/host/tmio_mmc_pio.c >>>> index faf0924e71cb..83192420a7e3 100644 >>>> --- a/drivers/mmc/host/tmio_mmc_pio.c >>>> +++ b/drivers/mmc/host/tmio_mmc_pio.c >>>> @@ -1147,8 +1147,18 @@ int tmio_mmc_host_suspend(struct device *dev) >>>> { >>>> struct mmc_host *mmc = dev_get_drvdata(dev); >>>> struct tmio_mmc_host *host = mmc_priv(mmc); >>>> + struct tmio_mmc_data *pdata = host->pdata; >>>> + >>>> + if (pdata->clk_enable) { >>>> + unsigned int f; >>>> + pdata->clk_enable(host->pdev, &f); >>>> + } >>>> >>>> tmio_mmc_disable_mmc_irqs(host, TMIO_MASK_ALL); >>>> + >>>> + if (pdata->clk_disable) >>>> + pdata->clk_disable(host->pdev); >>>> + >>>> return 0; >>>> } >>>> EXPORT_SYMBOL(tmio_mmc_host_suspend); Gr{oetje,eeting}s, Geert -- Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that. -- Linus Torvalds