From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jack Mitchell Subject: Re: AM335x BeagleBone SPI Issues Date: Mon, 10 Dec 2012 16:18:47 +0000 Message-ID: <50C60B67.20709@communistcode.co.uk> References: <50C5E23D.5040605@communistcode.co.uk> <20121210135337.GA10219@arwen.pp.htv.fi> <50C5F6A8.5060701@communistcode.co.uk> <20121210145947.GK11038@arwen.pp.htv.fi> <50C5FD73.1000408@communistcode.co.uk> Reply-To: ml@communistcode.co.uk Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: Received: from eumx.net ([91.82.101.43]:42978 "EHLO eumx.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752316Ab2LJQQc (ORCPT ); Mon, 10 Dec 2012 11:16:32 -0500 In-Reply-To: Sender: linux-omap-owner@vger.kernel.org List-Id: linux-omap@vger.kernel.org To: Shubhrajyoti Datta Cc: balbi@ti.com, linux-omap@vger.kernel.org On 10/12/12 15:26, Shubhrajyoti Datta wrote: > On Mon, Dec 10, 2012 at 8:49 PM, Jack Mitchell wrote: >> On 10/12/12 14:59, Felipe Balbi wrote: >>> Hi, >>> >>> On Mon, Dec 10, 2012 at 02:50:16PM +0000, Jack Mitchell wrote: >>>>> On Mon, Dec 10, 2012 at 01:23:09PM +0000, Jack Mitchell wrote: >>>>>> Hi, >>>>>> >>>>>> I am currently having issues with the SPI driver on the beaglebone >>>>>> using the 3.7-rc8 kernel[1]. I have probed the SPI pins and I have >>>>>> found that writing works however reading doesn't. When using DMA the >>>>>> program seems to lock hard and no data is sent on the bus. I am >>>>>> testing the bus using spidev and the spidev_test[2] application, >>>>>> however I first came across spi issues with a custom spi driver which >>>>>> stopped working when I transitioned from 3.2-psp to 3.7-rc8. >>>>>> >>>>>> The current output I am seeing from the spidev_test program is just a >>>>>> series of 0x00 data, which looks to me like no data is getting in at >>>>>> all. The spidev_test program is not using DMA as the buffer size is >>>>>> too low, so I forced the dma on when buffer size is > 1 and the >>>>>> program hangs hard with the system still responding to other >>>>>> commands.I have briged the pins 18 and 21 on the BeagleBone P9 >>>>>> header. >>>>>> >>>>>> Has anyone seen issues like this, or if not if someone could please >>>>>> test the latest 3.7-rc8 from [1] and let me know if it works for them >>>>>> and the issue is at my end. >>>>>> >>>>>> To get spidev working with devicetree I applied the patch from [3] >>>>>> and changed the dtb as in the patch pasted below. >>>>>> >>>>>> [1] https://github.com/beagleboard/kernel/tree/3.7 >>>>>> [2] http://lxr.linux.no/#linux+v3.6.9/Documentation/spi/spidev_test.c >>>>>> [3] >>>>>> http://www.mail-archive.com/spi-devel-general@lists.sourceforge.net/msg09958.html >>>>> do you have any debugging output from that driver ? It would be cool to >>>>> see if DMA is at least being kicked properly for small transfers. >>>> When I run the spidev program with dma for transfers > 1, the program >>>> hangs and the only output in dmesg is: >>>> >>>> [ 12.613952] libphy: 4a101000.mdio:00 - Link is Up - 100/Full <---- >>>> Last line from initial log in [2] >>>> [ 47.669202] spidev spi1.0: setup: speed 24000000, sample leading >>>> edge, clk normal >>>> [ 47.669246] spidev spi1.0: setup mode 0, 8 bits/w, 24000000 Hz max --> >>>> 0 >>>> [ 47.669260] spidev spi1.0: spi mode 00 >>>> [ 47.669283] spidev spi1.0: setup: speed 24000000, sample leading >>>> edge, clk normal >>>> [ 47.669300] spidev spi1.0: setup mode 0, 16 bits/w, 24000000 Hz max >>>> --> 0 >>>> [ 47.669312] spidev spi1.0: 16 bits per word >>>> [ 47.669330] spidev spi1.0: setup: speed 24000000, sample leading >>>> edge, clk normal >>>> [ 47.669347] spidev spi1.0: setup mode 0, 16 bits/w, 24000000 Hz max >>>> --> 0 >>>> [ 47.669358] spidev spi1.0: 24000000 Hz (max) >>>> [ 47.673811] spidev spi1.0: setup: speed 24000000, sample leading >>>> edge, clk normal >>>> >>>> The initial dmesg statup log is at [2]. >>> can you apply the debugging patch below to spi driver and show me the >>> output again ? >>> >>> Note that I'm allowing DMA for as small as 1-byte transfers (as you >>> needed) and I'm also disabling DMA Request line before calling >>> complete() as I think the current way could race, but likely wouldn't >>> cause issues. Anyway, please show me the output. >>> >>> diff --git a/drivers/spi/spi-omap2-mcspi.c b/drivers/spi/spi-omap2-mcspi.c >>> index 3542fdc..d2b1af2 100644 >>> --- a/drivers/spi/spi-omap2-mcspi.c >>> +++ b/drivers/spi/spi-omap2-mcspi.c >>> @@ -108,7 +108,7 @@ struct omap2_mcspi_dma { >>> /* use PIO for small transfers, avoiding DMA setup/teardown overhead and >>> * cache operations; better heuristics consider wordsize and bitrate. >>> */ >>> -#define DMA_MIN_BYTES 160 >>> +#define DMA_MIN_BYTES 1 >>> /* >>> @@ -298,10 +298,11 @@ static void omap2_mcspi_rx_callback(void *data) >>> struct omap2_mcspi *mcspi = spi_master_get_devdata(spi->master); >>> struct omap2_mcspi_dma *mcspi_dma = >>> &mcspi->dma_channels[spi->chip_select]; >>> - complete(&mcspi_dma->dma_rx_completion); >>> - >>> /* We must disable the DMA RX request */ >>> + dev_info(&spi->dev, "Disabling RX Request Line\n"); >>> omap2_mcspi_set_dma_req(spi, 1, 0); >>> + >>> + complete(&mcspi_dma->dma_rx_completion); >>> } >>> static void omap2_mcspi_tx_callback(void *data) >>> @@ -310,10 +311,11 @@ static void omap2_mcspi_tx_callback(void *data) >>> struct omap2_mcspi *mcspi = spi_master_get_devdata(spi->master); >>> struct omap2_mcspi_dma *mcspi_dma = >>> &mcspi->dma_channels[spi->chip_select]; >>> - complete(&mcspi_dma->dma_tx_completion); >>> - >>> /* We must disable the DMA TX request */ >>> + dev_info(&spi->dev, "Disabling TX Request Line\n"); >>> omap2_mcspi_set_dma_req(spi, 0, 0); >>> + >>> + complete(&mcspi_dma->dma_tx_completion); >>> } >>> static void omap2_mcspi_tx_dma(struct spi_device *spi, >>> @@ -328,6 +330,7 @@ static void omap2_mcspi_tx_dma(struct spi_device *spi, >>> void __iomem *chstat_reg; >>> struct omap2_mcspi_cs *cs = spi->controller_state; >>> + dev_info(&spi->dev, "kicking TX DMA\n"); >>> mcspi = spi_master_get_devdata(spi->master); >>> mcspi_dma = &mcspi->dma_channels[spi->chip_select]; >>> count = xfer->len; >>> @@ -359,7 +362,9 @@ static void omap2_mcspi_tx_dma(struct spi_device *spi, >>> dma_async_issue_pending(mcspi_dma->dma_tx); >>> omap2_mcspi_set_dma_req(spi, 0, 1); >>> + dev_info(&spi->dev, "Waiting for TX Completion\n"); >>> wait_for_completion(&mcspi_dma->dma_tx_completion); >>> + dev_info(&spi->dev, "TX Completed\n"); >>> dma_unmap_single(mcspi->dev, xfer->tx_dma, count, >>> DMA_TO_DEVICE); >>> @@ -392,6 +397,7 @@ omap2_mcspi_rx_dma(struct spi_device *spi, struct >>> spi_transfer *xfer, >>> word_len = cs->word_len; >>> l = mcspi_cached_chconf0(spi); >>> + dev_info(&spi->dev, "kicking RX DMA\n"); >>> if (word_len <= 8) >>> element_count = count; >>> else if (word_len <= 16) >>> @@ -428,7 +434,9 @@ omap2_mcspi_rx_dma(struct spi_device *spi, struct >>> spi_transfer *xfer, >>> dma_async_issue_pending(mcspi_dma->dma_rx); >>> omap2_mcspi_set_dma_req(spi, 1, 1); >>> + dev_info(&spi->dev, "Waiting for RX Completion\n"); >>> wait_for_completion(&mcspi_dma->dma_rx_completion); >>> + dev_info(&spi->dev, "RX Completed\n"); >>> dma_unmap_single(mcspi->dev, xfer->rx_dma, count, >>> DMA_FROM_DEVICE); >>> omap2_mcspi_set_enable(spi, 0); >>> >>>>> It would also be nice to have a clear picture of what "custom spi >>>>> driver" you're talking about. >>>> The custom SPI driver is for connecting and reading registers from an >>>> in house FPGA design and can be found at [1]. It's fairly rudimentary >>>> and also in the development stages, I'm very new to Linux kernel >>>> programming so please take that into account :) >>> no problem ;-) >>> >>>> However it did work flawlessly with 3.2-psp. >>> yeah, it could be some regression introduced somewhere, or just a bugfix >>> which was done on 3.2-psp and was missed upstream... annoying when those >>> happen :-p >>> > Can you try the following > > git://git.kernel.org/pub/scm/linux/kernel/git/broonie/misc.git spi-next > > Alternatively you can use the linux-next. > > And let me know if you still see the issue. > -- > To unsubscribe from this list: send the line "unsubscribe linux-omap" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html Neither of those trees seem to have SPI enabled via device tree, how can I make the same use case as I also cannot see a board file to enable the spi? Sorry, I'm quite new at this so bear with me if I'm being a bit naive! Cheers, -- Jack Mitchell (jack@embed.me.uk) Embedded Systems Engineer http://www.embed.me.uk --