From: Arnd Bergmann <arnd@arndb.de>
To: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Cc: linux-arm-kernel@lists.infradead.org, linux-arm-msm@vger.kernel.org,
        linux-kernel@vger.kernel.org, devicetree@vger.kernel.org,
        dmaengine@vger.kernel.org, Vinod Koul <vinod.koul@intel.com>,
        Mark Rutland <mark.rutland@arm.com>,
        Archit Taneja <architt@codeaurora.org>,
        Pawel Moll <pawel.moll@arm.com>,
        Ian Campbell <ijc+devicetree@hellion.org.uk>,
        Rob Herring <robh+dt@kernel.org>, Andy Gross <agross@codeaurora.org>
Subject: Re: [PATCH 2/4] dmaengine: qcom_bam_dma: clear BAM interrupt only if it is rised
Date: Wed, 02 Dec 2015 14:05:28 +0100
Message-ID: <11168224.Usic9kvUpA@wuerfel>
User-Agent: KMail/4.11.5 (Linux/3.16.0-10-generic; KDE/4.11.5; x86_64; ; )
In-Reply-To: <565EEA99.3070207@linaro.org>
References: <1448961299-15161-1-git-send-email-stanimir.varbanov@linaro.org> <6461444.oollnBsbrb@wuerfel> <565EEA99.3070207@linaro.org>
MIME-Version: 1.0
Content-Transfer-Encoding: 7Bit
Content-Type: text/plain; charset="us-ascii"
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1912
Lines: 47

On Wednesday 02 December 2015 14:56:57 Stanimir Varbanov wrote:
> On 12/01/2015 12:29 PM, Arnd Bergmann wrote:
> > On Tuesday 01 December 2015 11:14:57 Stanimir Varbanov wrote:
> >> +       if (srcs & BAM_IRQ) {
> >>                 clr_mask = readl_relaxed(bam_addr(bdev, 0, BAM_IRQ_STTS));
> >>  
> >> -       /* don't allow reorder of the various accesses to the BAM registers */
> >> -       mb();
> >> +               /*
> >> +                * don't allow reorder of the various accesses to the BAM
> >> +                * registers
> >> +                */
> >> +               mb();
> >>  
> >> -       writel_relaxed(clr_mask, bam_addr(bdev, 0, BAM_IRQ_CLR));
> >> +               writel_relaxed(clr_mask, bam_addr(bdev, 0, BAM_IRQ_CLR));
> >> +       }
> >>
> > 
> > I think the comment here should be moved: change the writel_relaxed()
> > to writel(), which already includes the appropriate barriers, and
> 
> If we agree with such a change it should be subject to another patch.

Correct.

> > add a comment at the readl_relaxed() to explain why it doesn't need
> > a barrier.
> 
> Infact I'm not sure that readl_relaxed(BAM_IRQ_STTS) does not need
> barrier. If I read the code above correctly the mb() should guarantee
> that all load and store operations before it are happened before the
> write to BAM_IRQ_CLR register, and on the other hand if we replace
> writel_relaxed with writel, the writel has wmb() which guarantees only
> store operations. Did I miss something?

You are right, we only guarantee that stores to memory are complete
before we writel() an MMIO register.

What do you gain from synchronizing reads before an MMIO write?

	Arnd
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/