1. 05 12月, 2017 1 次提交
  2. 28 8月, 2017 1 次提交
  3. 24 5月, 2017 3 次提交
    • L
      dmaengine: pl08x: Add support for Faraday Technology FTDMAC020 · 1e1cfc72
      Linus Walleij 提交于
      After reading the specs for the Faraday Technology FTDMAC020 found
      in the Gemini platform, it becomes pretty evident that this is just
      another PL08x derivative, and should be handled like such by simply
      extending the existing PL08x driver to handle the quirks in this
      hardware.
      
      This patch makes memcpy work and has been tested on the Gemini and
      also regression-tested on the Nomadik NHK15 using dmatest with
      10 threads per channel without a hinch for hours.
      
      I have not implemented slave DMA in those codepaths, because this
      device (Gemini) does not use slave DMA, and it seems like devices
      using FTDMAC020 for device DMA have a slightly different register
      layout so some real hardware is needed to proceed with this. I
      left some FIXME etc in the code for this.
      
      I had to do some refactorings of some helper functions, but I have
      not split those into separate patches because these refactorings
      do not make much sense without the increased complexity of handling
      the FTDMAC020.
      
      The DMA test would hang the platform on me on the Gemini after a
      few thousand iterations, however after turning of the caches the
      problem immediately disappeared and I could run the DMA engine
      with 10 threads pers physical channel for days in a row without
      a crash. I think there is no problem with the DMA driver: instead
      it is something fishy in the FA526 cache handling code that get
      pretty heavily exercised by the DMA engine and we need to go and
      fix that instead.
      Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>
      Signed-off-by: NVinod Koul <vinod.koul@intel.com>
      1e1cfc72
    • L
      dmaengine: pl08x: Make slave engine optional · ebe9b300
      Linus Walleij 提交于
      If the vendor data does not specify any signals, we do not
      have to support slave DMA. Make the registration of the slave
      DMA engine optional, so we can use this for the FTDMAC020
      in the Gemini that only has memcpy support.
      Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>
      Signed-off-by: NVinod Koul <vinod.koul@intel.com>
      ebe9b300
    • L
      ARM/dmaengine: pl08x: pass reasonable memcpy settings · 4166a56a
      Linus Walleij 提交于
      We cannot use bits from configuration registers as API between
      platforms and driver like this, abstract it out to two enums
      and mimic the stuff passed as device tree data.
      
      This is done to make it possible for the driver to generate the
      ccfg word on-the-fly so we can support more PL08x derivatives.
      Acked-by: NOlof Johansson <olof@lixom.net>
      Acked-by: NArnd Bergmann <arnd@arndb.de>
      Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>
      Signed-off-by: NVinod Koul <vinod.koul@intel.com>
      4166a56a
  4. 04 5月, 2017 4 次提交
  5. 17 11月, 2016 1 次提交
  6. 22 6月, 2016 1 次提交
  7. 06 4月, 2016 1 次提交
  8. 19 8月, 2015 1 次提交
  9. 08 5月, 2015 1 次提交
  10. 19 3月, 2015 2 次提交
  11. 17 3月, 2015 1 次提交
  12. 22 12月, 2014 1 次提交
  13. 06 11月, 2014 1 次提交
  14. 04 8月, 2014 2 次提交
  15. 28 1月, 2014 1 次提交
  16. 19 12月, 2013 1 次提交
  17. 04 12月, 2013 1 次提交
  18. 15 11月, 2013 1 次提交
  19. 14 11月, 2013 1 次提交
  20. 31 10月, 2013 1 次提交
  21. 25 10月, 2013 1 次提交
  22. 13 10月, 2013 1 次提交
  23. 02 9月, 2013 8 次提交
  24. 23 8月, 2013 2 次提交
  25. 05 7月, 2013 1 次提交