From mboxrd@z Thu Jan  1 00:00:00 1970
From: tomasz.figa@gmail.com (Tomasz Figa)
Date: Thu, 27 Jun 2013 00:04:23 +0200
Subject: [RFC PATCH v2 03/12] dmaengine: PL08x: Rework LLI handling to be
 less fragile
In-Reply-To: <CACRpkdYm7gFVLzRpQ2J97M+-zQAug_ftJhyi_kniTgCk=GT_bQ@mail.gmail.com>
References: <1371933764-24875-1-git-send-email-tomasz.figa@gmail.com>
 <1371933764-24875-4-git-send-email-tomasz.figa@gmail.com>
 <CACRpkdYm7gFVLzRpQ2J97M+-zQAug_ftJhyi_kniTgCk=GT_bQ@mail.gmail.com>
Message-ID: <2212577.G2ANCYONVo@flatron>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On Tuesday 25 of June 2013 00:23:09 Linus Walleij wrote:
> On Sat, Jun 22, 2013 at 10:42 PM, Tomasz Figa <tomasz.figa@gmail.com> 
wrote:
> > +enum {
> > +       PL080_LLI_SRC,
> > +       PL080_LLI_DST,
> > +       PL080_LLI_LLI,
> > +       PL080_LLI_CCTL,
> > +
> > +       PL080_LLI_WORDS
> > +};
> 
> I usually don't like it when enums are not given names, and are
> implicitly cast to integers.
> 
> I think it'd be better to just use #define for these so we know what
> is going on.

OK. I didn't have any preference for one or the other way and I randomly 
chose to use enum, but since you seem to have one (and hard to disagree 
with), I will change this patch to use #define.

> > @@ -181,7 +177,7 @@ struct pl08x_txd {
> > 
> >         struct virt_dma_desc vd;
> >         struct list_head dsg_list;
> >         dma_addr_t llis_bus;
> > 
> > -       struct pl08x_lli *llis_va;
> > +       u32 *llis_va;
> 
> It's nice that you use the u32 * here for proper indexing into an
> array.
> 
> > -/* Size (bytes) of each LLI buffer allocated for one transfer */
> > -# define PL08X_LLI_TSFR_SIZE   0x2000
> > -
> > -/* Maximum times we call dma_pool_alloc on this pool without freeing
> > */ -#define MAX_NUM_TSFR_LLIS      (PL08X_LLI_TSFR_SIZE/sizeof(struct
> > pl08x_lli)) +/*
> > + * Number of LLIs in each LLI buffer allocated for one transfer
> > + * (maximum times we call dma_pool_alloc on this pool without
> > freeing)
> > + */
> > +#define MAX_NUM_TSFR_LLIS      512
> 
> And I like this nice side effect that we allocate a number of LLIs
> rather than a fixed-size buffer.

As we're at it, is there any reason to choose this particular number of 
LLis? I have calculated it from the buffer size defined originally, but I 
wonder if 512 LLIs for single transfer isn't too much?

> > +static void pl08x_write_lli(struct pl08x_driver_data *pl08x,
> > +               struct pl08x_phy_chan *phychan, const u32 *lli, u32
> > ccfg) +{
> > +       dev_vdbg(&pl08x->adev->dev,
> > +               "WRITE channel %d: csrc=0x%08x, cdst=0x%08x, "
> > +               "clli=0x%08x, cctl=0x%08x, ccfg=0x%08x\n",
> > +               phychan->id, lli[PL080_LLI_SRC], lli[PL080_LLI_DST],
> > +               lli[PL080_LLI_LLI], lli[PL080_LLI_CCTL], ccfg);
> > +
> > +       writel(lli[PL080_LLI_SRC], phychan->base + PL080_CH_SRC_ADDR);
> > +       writel(lli[PL080_LLI_DST], phychan->base + PL080_CH_DST_ADDR);
> > +       writel(lli[PL080_LLI_LLI], phychan->base + PL080_CH_LLI);
> > +       writel(lli[PL080_LLI_CCTL], phychan->base + PL080_CH_CONTROL);
> > +
> > +       writel(ccfg, phychan->reg_config);
> 
> Take this opportunity to replace the first four writel() with
> writel_relaxed(), keep the last one to make sure all hit the hardware.

OK.

> Apart from that this looks nice!

Thanks.

Best regards,
Tomasz