1. 25 Sep, 2012 4 commits
    • Chandrabhanu Mahapatra's avatar
      OMAPDSS: DISPC: Add predecimation limit for TILER based rotations · d557a9cf
      Chandrabhanu Mahapatra authored
      
      
      In OMAP4 and OMAP5 when TILER 2D burst mode is used, a maximum of one line can
      be skipped as per the respective TRMs. The MBlockStride OCP signal, which is
      sum of ROWINC and image width in memory, is only 17 bits wide. In 2D mode TILER
      supports 8192, 16384, 32768 and 65536 values of MBlockStride. In case when 2 or
      more lines are skipped the ROWINC value exceeds 65536 resulting in OCP errors.
      So, maximum vertical predecimation achievable is 2.
      
      Signed-off-by: default avatarChandrabhanu Mahapatra <cmahapatra@ti.com>
      Signed-off-by: default avatarTomi Valkeinen <tomi.valkeinen@ti.com>
      d557a9cf
    • Tomi Valkeinen's avatar
      Merge OMAP5 DSS changes to omapdss · 406f7b8b
      Tomi Valkeinen authored
      This series adds basic OMAP5 DSS functionality, mainly related to DSS core, DPI
      and DSI.
      
      * omap5-dss:
        OMAPDSS: DSI: make OMAP2_DSS_DSI depend on ARCH_OMAP5
        OMAPDSS: DSI: Add code to disable PHY DCC
        OMAPDSS: DSI: Add new linebuffer size for OMAP5
        OMAPDSS: DSI: Add FEAT_DSI_PLL_REFSEL
        OMAPDSS: DSI: Add FEAT_DSI_PLL_SELFREQDCO
        OMAPDSS: Add support for DPI source selection
        OMAPDSS: move dss feats to the end of dss.c
        OMAPDSS: Add basic omap5 features to dss and dispc
        OMAPDSS: DSI: improve DSI clock calcs for DISPC
      406f7b8b
    • Tomi Valkeinen's avatar
      Merge omapdss single-dssdev series · c0ca7c38
      Tomi Valkeinen authored
      This series contains patches that change how omapdss's panel devices
      (omap_dss_device) are initialized and registered. There are two patches that
      change behaviour, the rest are just cleanups:
      
      The patch "omap_dss_register_device() doesn't take display index" affects the
      number for the "displayX" sysfs files. This hopefully doesn't affect the
      userspace, as the number has never been a clear indication of what the
      particular display is.
      
      The patch "register only one display device per output" affects how panel
      devices are created. Currently we support multiple panels per output, i.e. you
      could have DVI and an LCD displays using the same DPI output, as long as the
      DVI and LCD are not used at the same time.
      
      This patch changes the omapdss driver to only register one display device per
      output. If there are multiple displays for the output, either the first one is
      picked or, if def_display has been defined in kernel parameters and the
      def_display is one of the displays for this output, the def_display is picked.
      See the patch for more information.
      
        OMAPDSS: alloc dssdevs dynamically
        OMAPDSS: cleanup dss_recheck_connections further
        OMAPDSS: cleanup dss_recheck_connections
        OMAPDSS: handle errors in dss_init_device
        OMAPDSS: explicitely initialize dssdev->channel for new displays
        OMAPDSS: register only one display device per output
        OMAPDSS: Add dss_get_default_display_name()
        OMAPDSS: omap_dss_register_device() doesn't take display index
      c0ca7c38
    • Raphaël Assénat's avatar
      OMAPDSS: Do not require a VDDS_DSI regulator on AM35xx · 524d9f48
      Raphaël Assénat authored
      
      
      On our AM3505 based board, dpi.c complains that there is no VDDS_DSI
      regulator and the framebuffer cannot be enabled. However, this check
      does not seem to apply to AM3505/17 chips.
      
      This patch adds new features list for AM35xxx, which is the same as for
      OMAP3 except the VDDS_DSI is removed.
      
      Signed-off-by: default avatarRaphael Assenat <raph@8d.com>
      Signed-off-by: default avatarTomi Valkeinen <tomi.valkeinen@ti.com>
      524d9f48
  2. 24 Sep, 2012 9 commits
  3. 18 Sep, 2012 8 commits
    • Tomi Valkeinen's avatar
      OMAPDSS: alloc dssdevs dynamically · 5274484b
      Tomi Valkeinen authored
      
      
      We currently create omap_dss_devices statically in board files, and use
      those devices directly in the omapdss driver. This model prevents us
      from having the platform data (which the dssdevs in board files
      practically are) as read-only, and it's also different than what we will
      use with device tree.
      
      This patch changes the model to be in line with DT model: we allocate
      the dssdevs dynamically, and initialize them according to the data in
      the board file's dssdev (basically we memcopy the dssdev fields).
      
      The allocation and registration is done in the following steps in the
      output drivers:
      
      - Use dss_alloc_and_init_device to allocate and initialize the device.
        The function uses kalloc and device_initialize to accomplish this.
      - Call dss_copy_device_pdata to copy the data from the board file's
        dssdev
      - Use dss_add_device to register the device.
      
      Signed-off-by: default avatarTomi Valkeinen <tomi.valkeinen@ti.com>
      5274484b
    • Tomi Valkeinen's avatar
      OMAPDSS: cleanup dss_recheck_connections further · 5eeb55f8
      Tomi Valkeinen authored
      
      
      Cleanup dss_recheck_connections, move and rename it to a static
      dss_init_connections function inside display.c. Improve the function to
      return errors, and implement a matching dss_uninit_connections that can
      be used to free the mgr->dssdev link.
      
      Signed-off-by: default avatarTomi Valkeinen <tomi.valkeinen@ti.com>
      5eeb55f8
    • Tomi Valkeinen's avatar
      OMAPDSS: cleanup dss_recheck_connections · 6b417858
      Tomi Valkeinen authored
      
      
      dss_recheck_connections is quite a mess. With the previous commit that
      initializes the channel field for HDMI and VENC displays, we can greatly
      simplify the dss_recheck_connections.
      
      Signed-off-by: default avatarTomi Valkeinen <tomi.valkeinen@ti.com>
      6b417858
    • Tomi Valkeinen's avatar
      OMAPDSS: handle errors in dss_init_device · 47eb6763
      Tomi Valkeinen authored
      
      
      Add error handling to dss_init_device(), which has, for some reason,
      been missing.
      
      Signed-off-by: default avatarTomi Valkeinen <tomi.valkeinen@ti.com>
      47eb6763
    • Tomi Valkeinen's avatar
      OMAPDSS: explicitely initialize dssdev->channel for new displays · bcb226a9
      Tomi Valkeinen authored
      
      
      HDMI and VENC outputs always use the DIGIT output from DISPC. The dssdev
      struct contains "channel" field which is used to specify the DISPC
      output for the display, but this was not used for HDMI and VENC.
      
      This patch fills the channel field explicitely for HDMI and VENC
      displays so that we can always rely on the channel field.
      
      Signed-off-by: default avatarTomi Valkeinen <tomi.valkeinen@ti.com>
      bcb226a9
    • Tomi Valkeinen's avatar
      OMAPDSS: register only one display device per output · 1521653c
      Tomi Valkeinen authored
      
      
      We have boards with multiple panel devices connected to the same
      physical output, of which only one panel can be enabled at one time.
      Examples of these are Overo, where you can use different daughter boards
      that have different LCDs, and 3430SDP which has an LCD and a DVI output
      and a physical switch to select the active display.
      
      These are supported by omapdss so that we add all the possible display
      devices at probe, but the displays are inactive until somebody enables
      one. At this point the panel driver starts using the DSS, thus reserving
      the physcal resource and excluding the other panels.
      
      This is problematic:
      - Panel drivers can't allocate their resources properly at probe(),
        because the resources can be shared with other panels. Thus they can
        be only reserved at enable time.
      - Managing this in omapdss is confusing. It's not natural to have
        child devices, which may not even exist (for example, a daughterboard
        that is not connected).
      
      Only some boards have multiple displays per output, and of those, only
      very few have possibility of switching the display during runtime.
      Because of the above points:
      - We don't want to make omapdss and all the panel drivers more complex
        just because some boards have complex setups.
      - Only few boards support runtime switching, and afaik even then it's
        not required. So we don't need to support runtime switching.
      
      Thus we'll change to a model where we will have only one display device
      per output and this cannot be (currently) changed at runtime. We'll
      still have the possibility to select the display from multiple options
      during boot with the default display option.
      
      This patch accomplishes the above by changing how the output drivers
      register the display device. Instead of registering all the devices
      given from the board file, we'll only register one. If the default
      display option is set, the output driver selects that display from its
      displays. If the default display is not set, or the default display is
      not one of the output's displays, the output driver selects the first
      display.
      
      Signed-off-by: default avatarTomi Valkeinen <tomi.valkeinen@ti.com>
      1521653c
    • Tomi Valkeinen's avatar
      OMAPDSS: Add dss_get_default_display_name() · 6a03fca9
      Tomi Valkeinen authored
      
      
      Add function dss_get_default_display_name() which returns the name of
      the default display, given from the board file or via module parameters.
      The default display name can be used by output drivers to decide which
      display is the wanted one.
      
      Signed-off-by: default avatarTomi Valkeinen <tomi.valkeinen@ti.com>
      6a03fca9
    • Tomi Valkeinen's avatar
      OMAPDSS: omap_dss_register_device() doesn't take display index · 8768a52f
      Tomi Valkeinen authored
      
      
      We used to have all the displays of the board in one list, and we made a
      "displayX" directory in the sysfs, where X was the index of the display
      in the list.
      
      This doesn't work anymore with device tree, as there's no single list to
      get the number from, and it doesn't work very well even with non-DT as
      we need to do some tricks to get the index nowadays.
      
      This patch changes omap_dss_register_device() so that it doesn't take
      disp_num as a parameter anymore, but uses a private increasing counter
      for the display number.
      
      Signed-off-by: default avatarTomi Valkeinen <tomi.valkeinen@ti.com>
      8768a52f
  4. 11 Sep, 2012 2 commits
  5. 10 Sep, 2012 1 commit
    • Tomi Valkeinen's avatar
      Merge miscellaneous omapdss changes · 6cd05430
      Tomi Valkeinen authored
      Most important changes:
      
      * Remove OMAP4 HDMI gpio handling from board files
      * Add vdda_hdmi_dac supply for HDMI to twl-common.c
      * Calculate DSI clocks dynamically
      * Remove DSS clock dividers from 4430sdp board file
      * vram.c no longer uses OMAP's sDMA to clear the memory
      * Fifomerge has been reverted
      * Swap GFX and WB fifos to avoid underflows
      6cd05430
  6. 07 Sep, 2012 16 commits