From: Lee Jones Subject: Re: [PATCH 34/39] dmaengine: ste_dma40: Convert data_width from register bit format to value Date: Thu, 16 May 2013 08:35:53 +0100 Message-ID: <20130516073553.GF3269@gmail.com> References: <1368611522-9984-1-git-send-email-lee.jones@linaro.org> <1368611522-9984-35-git-send-email-lee.jones@linaro.org> <20130516063244.GA14863@intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org, balbi@ti.com, linux-usb@vger.kernel.org, linux-crypto@vger.kernel.org, davem@davemloft.net, herbert@gondor.hengli.com.au, arnd@arndb.de, linus.walleij@stericsson.com, srinidhi.kasagar@stericsson.com, Dan Williams , Per Forlin , Rabin Vincent To: Vinod Koul Return-path: Content-Disposition: inline In-Reply-To: <20130516063244.GA14863@intel.com> Sender: linux-kernel-owner@vger.kernel.org List-Id: linux-crypto.vger.kernel.org On Thu, 16 May 2013, Vinod Koul wrote: > On Wed, May 15, 2013 at 10:51:57AM +0100, Lee Jones wrote: > > When a DMA client requests and configures a DMA channel, it request= s > > data_width in Bytes. The DMA40 driver then swiftly converts it over= to > > the necessary register bit value. Unfortunately, for any subsequent > > calculations we have to shift '1' by the bit pattern (1 << data_wid= th) > > times to make any sense of it. > >=20 > > This patch flips the semantics on its head and only converts the va= lue > > to its respective register bit pattern when writing to registers. T= his > > way we can use the true data_width (in Bytes) value. > >=20 > > Cc: Vinod Koul > > Cc: Dan Williams > > Cc: Per Forlin > > Cc: Rabin Vincent > > Signed-off-by: Lee Jones > > --- > =20 > > @@ -2804,14 +2781,24 @@ static int d40_set_runtime_config(struct dm= a_chan *chan, > > src_maxburst =3D dst_maxburst * dst_addr_width / src_addr_width; > > } > > =20 > > + /* Only valid widths are; 1, 2, 4 and 8. */ > > + if (src_addr_width <=3D DMA_SLAVE_BUSWIDTH_UNDEFINED || > > + src_addr_width > DMA_SLAVE_BUSWIDTH_8_BYTES || > > + dst_addr_width <=3D DMA_SLAVE_BUSWIDTH_UNDEFINED || > > + dst_addr_width > DMA_SLAVE_BUSWIDTH_8_BYTES || > > + ((src_addr_width > 1) && (src_addr_width & 1)) || > > + ((dst_addr_width > 1) && (dst_addr_width & 1))) > > + return -EINVAL; > how about a simple macro to check above.. I thought about it, but as this code appears only once, I considered it to be unnecessary abstraction. > > + > > + cfg->src_info.data_width =3D src_addr_width; > > + cfg->dst_info.data_width =3D dst_addr_width; > > + > > ret =3D dma40_config_to_halfchannel(d40c, &cfg->src_info, > > - src_addr_width, > > src_maxburst); > > if (ret) > > return ret; > > =20 > > ret =3D dma40_config_to_halfchannel(d40c, &cfg->dst_info, > > - dst_addr_width, > > dst_maxburst); > > if (ret) > > return ret; > > diff --git a/drivers/dma/ste_dma40_ll.c b/drivers/dma/ste_dma40_ll.= c > > index 5ddd724..a035dfe 100644 > > --- a/drivers/dma/ste_dma40_ll.c > > +++ b/drivers/dma/ste_dma40_ll.c > > @@ -10,6 +10,18 @@ > > =20 > > #include "ste_dma40_ll.h" > > =20 > > +u8 d40_width_to_bits(enum dma_slave_buswidth width) > > +{ > > + if (width =3D=3D DMA_SLAVE_BUSWIDTH_1_BYTE) > > + return STEDMA40_ESIZE_8_BIT; > > + else if (width =3D=3D DMA_SLAVE_BUSWIDTH_2_BYTES) > > + return STEDMA40_ESIZE_16_BIT; > > + else if (width =3D=3D DMA_SLAVE_BUSWIDTH_8_BYTES) > > + return STEDMA40_ESIZE_64_BIT; > > + else > > + return STEDMA40_ESIZE_32_BIT; > > +} > > + > Switch looks better for this and how about > return fls(width); >=20 > as your defines are 0...3 and dmaengine define 1,2,..8 for same thing > then you can also get rid of STEDMA40_XXX_WIDTH macros! I like it. Will you let me do it as a follow-up patch? > > @@ -70,13 +70,6 @@ enum stedma40_flow_ctrl { > > STEDMA40_FLOW_CTRL, > > }; > > =20 > > -enum stedma40_periph_data_width { > > - STEDMA40_BYTE_WIDTH =3D STEDMA40_ESIZE_8_BIT, > > - STEDMA40_HALFWORD_WIDTH =3D STEDMA40_ESIZE_16_BIT, > > - STEDMA40_WORD_WIDTH =3D STEDMA40_ESIZE_32_BIT, > > - STEDMA40_DOUBLEWORD_WIDTH =3D STEDMA40_ESIZE_64_BIT > > -}; > nice >=20 --=20 Lee Jones Linaro ST-Ericsson Landing Team Lead Linaro.org =E2=94=82 Open source software for ARM SoCs =46ollow Linaro: Facebook | Twitter | Blog