From mboxrd@z Thu Jan 1 00:00:00 1970 From: kdorfman@codeaurora.org Subject: Re: [PATCH v8] mmc: support BKOPS feature for eMMC Date: Fri, 11 May 2012 03:06:25 -0700 (PDT) Message-ID: References: <4FABC495.8060708@samsung.com> <4FACA5D2.3050706@samsung.com> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7BIT Return-path: Received: from wolverine02.qualcomm.com ([199.106.114.251]:64444 "EHLO wolverine02.qualcomm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932416Ab2EKKG0 (ORCPT ); Fri, 11 May 2012 06:06:26 -0400 In-Reply-To: <4FACA5D2.3050706@samsung.com> Sender: linux-mmc-owner@vger.kernel.org List-Id: linux-mmc@vger.kernel.org Cc: kdorfman@codeaurora.org, Jaehoon Chung , linux-mmc , Chris Ball , Kyungmin Park , Hanumath Prasad , Per FORLIN , Sebastian Rasmussen , "Dong, Chuanxiao" , "svenkatr@ti.com" , Saugata Das , Adrian Hunter > Hi Konstantin. > > On 05/10/2012 11:46 PM, kdorfman@codeaurora.org wrote: > >> Hi Jaehoon, >> I see 2 issues with new patch: >> 1. New code starts BKOPS always with waiting for BUSY. (see comment >> below >> in the code) >> 2. Runtime suspend (mmc_suspend_host) do not checks that BKOPS is in >> progress, >> this means, that CMD5 (slee) will fail. > > You're right. it needs to check in progress. > I will add them. OK, I will send you patch for runtime suspend flow (it is good to do righ before calling ops->suspend callback in mmc_suspend_host() function. > >> >> >> In your previous patch (v7) was code: >> >> - cmd.flags = MMC_RSP_SPI_R1B | MMC_RSP_R1B | MMC_CMD_AC; >> + cmd.flags = MMC_CMD_AC; >> + if (index == EXT_CSD_BKOPS_START && >> + card->ext_csd.raw_bkops_status < EXT_CSD_BKOPS_LEVEL_2) >> + cmd.flags |= MMC_RSP_SPI_R1 | MMC_RSP_R1; >> + else >> + cmd.flags |= MMC_RSP_SPI_R1B | MMC_RSP_R1B; >> >> When level of raw_bkops_status is not critical, the host controller >> flags >> used >> without waiting for BUSY line release. This means, that mmc queue thread >> will continue >> to fetch requests while BKOPS still running on the card. >> Without this code, BKOPS will always be blocking till BKOPS done and >> using >> HPI for interrupt BKOPS on >> new request is not need. >> Why you removed this code? > > You're right. it's my mistake. Actually i modified to use with > __mmc_switch(). > As I remember, maybe you mentioned this point. right? > So i tried to use based on your review. it's missing in this patch. > yes > Thank you for your comment. any other comments? > If you have others, let me know, plz. I will check the patch and update you about it. Thanks, Kostya