From: marex@denx.de (Marek Vasut) Date: Tue, 2 Apr 2013 01:30:51 +0200 Subject: [PATCH 2/5] spi/mxs: Fix chip select control bits in DMA mode In-Reply-To: References: <1364570381-17605-1-git-send-email-tpiepho@gmail.com> <201304020113.42124.marex@denx.de> Message-ID: <201304020130.51951.marex@denx.de> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org Dear Trent Piepho, > On Mon, Apr 1, 2013 at 4:13 PM, Marek Vasut wrote: > >> Change the code to set the CS bits in the message transfer function > >> once. Now the DMA and PIO txrx functions don't need to care about CS > >> at all. > > > > Ok, lemme ask this one more time -- will the DMA work with long transfers > > where the CTRL0 will be overwritten on each turn? Did you actually test > > this? > > I don't have SPI flash, so I can't test SPI flash. I can test other > spi devices and by capturing the pins with a logic analyzer. It does > work after the patch, and does not work before the patch. The error > should be obvious from looking at the code. > > ctrl0 is not overwritten, from scratch, each on transfer. The ctrl0 > value in the PIO part of the DMA is based on the current value of > ctrl0 with additional bits ORed in. The flaw here is that bits that > should not be set are not masked out. > > >> -static uint32_t mxs_spi_cs_to_reg(unsigned cs) > >> +static u32 mxs_spi_cs_to_reg(unsigned cs) > >> > >> { > >> > >> - uint32_t select = 0; > >> + u32 select = 0; > > I'll make it a separate patch. This is completely irrelevant change, please just submit the relevant patches. Best regards, Marek Vasut