Commit ee144e64 added
dsi_pll_calc_ddrfreq() which calculates PLL dividers based on given DSI
bus clock speed. The function works ok, but it can be improved for the
DISPC clock calc.

The current version calculates the clock going from the PLL to the DISPC
simply by setting the clock as close to DISPC maximum as possible, and
the pixel clock is calculated based on that.

This patch changes the function to calculate DISPC clock more
dynamically, iterating through different DISPC clocks and pixel clock
values, and thus we'll get more suitable pixel clocks.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

The framebuffers are cleared with the function omapfb_clear_fb(), which
internally calls cfb_fillrect(). The boot logo is copied on to the
framebuffer when the fb device is registered with fb framework.

omapfb_clear_fb() is called after the framebuffer is registered, leading to
the boot logo getting cleared. Clear the framebuffers using omapfb_clear_fb()
before registering the framebuffer devices.
Signed-off-by: NArchit Taneja <archit@ti.com>
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

When we removed fifomerge support, we also changed dss_ovl_disable so
that it doesn't wait for the hardware to be finished with the overlay.
This may cause a problem when changing the overlay's manager, as
changing the manager is an immediate change. Thus if the overlay is
still being used by the HW when the manager is changed, there may be
glitches on the screen.

This patch adds a wait into dss_ovl_unset_manager, which ensures the
overlays are disabled in the HW.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

set_timings function of DSS's output drivers are not consistent. Some of
them disable the output, set the timings, and re-enable the output. Some
set the timings on the fly, while the output is enabled. And some just
store the given timings, so that they will be taken into use next time
the output is enabled.

We require the DISPC output to be disabled when changing the timings,
and so we can change all the output drivers' set_timings to just store
the given timings.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

OMAP4's GFX overlay has smaller fifo than the rest of the overlays
(including writeback "overlay"). This seems to be the reason for
underflows in some more demanding scenarios.

We can avoid the problems by using the WB fifo for GFX overlay, and vice
versa. WB usage is not supported yet, but when it will, it should
perform just fine with smaller fifo as there are no hard realtime
constraints with WB.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

OMAP4+ allows assigning the overlay FIFOs freely, but that is not
supported by omapdss yet. This patch takes a step forward by improving
the fifo management to be more flexible.

dispc.c is changed to keep track of the sizes of each fifo, and also the
overlay using each fifo.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

This reverts commit fb011974.

Adding fifo merge feature as an omapdss internal configuration was a
mistake. We cannot hide from the users of omapdss the complexities of
fifo merge.

The previous commit removed fifo merge itself, and this removes the
remaining fifo merge support functions.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

This reverts commit 1d71f42b.

Adding fifo merge feature as an omapdss internal configuration was a
mistake. We cannot hide from the users of omapdss the complexities of
fifo merge.

This commit removes the fifo merge support, which luckily is easily done
as it was handled totally inside apply.c. Note that this is not a 1:1
revert, but some resolving was needed for the dss_ovl_setup_fifo.

The plan is to try fifo merge again later when it is more clear how the
hardware acts in various situations, and how the omapdrm wants to use
fifo merge.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

dss_mgr_set_timings() can only be called when the output is not active.
This means that most of the code in the function is extra, as there's no
need to write the values to registers, etc, because that will be handled
when the output will be enabled.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

dss_mgr_set_lcd_config() can only be called when the output is not
active. This means that most of the code in the function is extra, as
there's no need to write the values to registers, etc, because that will
be handled when the output will be enabled.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Separate sysfs code for managers to a separate file. This is a bit
cleaner, and will allow us later to easily switch off the sysfs support
via Kconfig option.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Separate sysfs code for overlays to a separate file. This is a bit
cleaner, and will allow us later to easily switch off the sysfs support
via Kconfig option.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Use devm_ version instead of the regular gpio_request_one to simplify
the error handling.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Reorganize taal driver to make it easier to integrate device tree code.
Instead of storing the panel's platform data, we'll "parse" the platform
data and store the required information in driver's own data. This way
adding device tree data parsing is simple.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Currently vram.c clears the allocated memory automatically using OMAP
system DMA. In an effort to reduce OMAP dependencies, we'll do the
memory clear with CPU from now on.

The previous patch implemented memory clear in the omapfb driver, and
this patch removes the now obsolete clear functionality from vram.c.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>
Cc: Russell King - ARM Linux <linux@arm.linux.org.uk>

Currently vram.c clears the allocated memory automatically using OMAP
system DMA. In an effort to reduce OMAP dependencies, we'll do the
memory clear with CPU from now on.

This patch implements clearing of the framebuffer in omapfb, using
cfb_fillrect() to do the actual clear.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Remove unnecessary includes from omapdss.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

The old omapfb driver contains a table for DMA element types, which is
not used. Remove it.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Remove unnecessary includes from the old omapfb driver.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Use devm_ functions in panel-taal.c's probe when possible. Also reorder
the initialization sequence so that devm_ allocations are done before
things that require explicit freeing. This simplifies the probe and
remove functions.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Recent commit dca2b152 (OMAPDSS: DSI:
Maintain copy of operation mode in driver data) broke DSI for video mode
displays. The commit changed the way dssdev->caps are initialized, and
the result was that every DSI display is initialized with manual-update
and tear-elim caps.

The code that sets dssdev->caps is not very good, even when fixed.
omapdss driver shouldn't be writing dssdev->caps at all.

This patch fixes the problem with video mode displays by moving the
initialization of dssdev->caps to the panel driver. The same change is
done for RFBI.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Currently the way to configure clocks related to DSI (both DSI and DISPC
clocks) happens via omapdss platform data. The reason for this is that
configuring the DSS clocks is a very complex problem, and it's
impossible for the SW to know requirements about things like
interference.

However, for general cases it should be fine to calculate the dividers
for clocks in the SW. The calculated clocks are probably not perfect,
but should work.

This patch adds support to calculate the dividers when using DSI command
mode panels. The panel gives the required DDR clock rate and LP clock
rate, and the DSI driver configures itself and DISPC accordingly.

This patch is somewhat ugly, though. The code does its job by modifying
the platform data where the clock dividers would be if the board file
gave them. This is not how it's going to be in the future, but allows us
to have quite simple patch and keep the backward compatibility.

It also allows the developer to still give the exact dividers from the
board file when there's need for that, as long as the panel driver does
not override them.

There are also other areas for improvement. For example, it would be
better if the panel driver could ask for a DSI clock in a certain range,
as, at least command mode panels, the panel can work fine with many
different clock speeds.

While the patch is not perfect, it allows us to remove the hardcoded
clock dividers from the board file, making it easier to bring up a new
panel and to use device tree from omapdss.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Add max value of DSI functional clock to dss_features, so that DSI
driver can use it.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

The HDMI driver requires vdda_hdmi_dac power for operation, but does not
enable it. This has worked because the regulator has been always
enabled.

But this may not always be the case, as I encountered when implementing
HDMI device tree support.

This patch changes the HDMI driver to use the vdda_hdmi_dac.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

TPD12S015A spec says to wait 300us after setting CT_CP_HPD gpio for the
5V power output to reach 90% of the voltage. This patch adds the delay
to the driver.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

We currently manage HDMI GPIOs in the board files via
platform_enable/disable calls. This won't work with device tree, and in
any case the correct place to manage the GPIOs is in the HDMI driver.

This patch moves the handling of the GPIOs to the HDMI driver. The GPIO
handling is moved to the common hdmi.c file, and this probably needs to
be revisited when adding OMAP5 HDMI support to see if the GPIO handling
needs to be moved to IP specific files.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>
Acked-by: NTony Lindgren <tony@atomide.com>

omapfb does not currently set pseudo palette correctly for color depths
above 16bpp, making red text invisible, command like
  echo -e '\e[0;31mRED' > /dev/tty1
will display nothing on framebuffer console in 24bpp mode.
This is because temporary variable is declared incorrectly, fix it.
Signed-off-by: NGrazvydas Ignotas <notasas@gmail.com>
Cc: stable@vger.kernel.org # v3.x
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>
Signed-off-by: NFlorian Tobias Schandinat <FlorianSchandinat@gmx.de>

Commit f476ae9d (OMAPDSS: APPLY: Remove
DISPC writes to manager's lcd parameters in interface) broke the SDI
output, as it causes the SDI PLL locking to fail.

LCLK and PCLK divisors are located in shadow registers, and we normally
write them to DISPC registers when enabling the output.  However, SDI
uses pck-free as source clock for its PLL, and pck-free is affected by
the divisors. And as we need the PLL before enabling the output, we need
to write the divisors early.

It seems just writing to the DISPC register is enough, and we don't need
to care about the shadow register mechanism for pck-free. The exact
reason for this is unknown.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>
Reported-by: NAaro Koskinen <aaro.koskinen@iki.fi>
Signed-off-by: NFlorian Tobias Schandinat <FlorianSchandinat@gmx.de>

Do a sanity check on these before using them as divisors.
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Acked-by: NAnatolij Gustschin <agust@denx.de>
Signed-off-by: NFlorian Tobias Schandinat <FlorianSchandinat@gmx.de>

Using kfree to free data allocated with devm_kzalloc causes double frees.

The semantic patch that fixes this problem is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@@
expression x;
@@

x = devm_kzalloc(...)
...
?-kfree(x);
// </smpl>
Signed-off-by: NJulia Lawall <Julia.Lawall@lip6.fr>
Signed-off-by: NFlorian Tobias Schandinat <FlorianSchandinat@gmx.de>

Switch to kmalloc(,GFP_ATOMIC) in bit_putcs to fix below trace:

[    9.771812] BUG: sleeping function called from invalid context at /usr/src/linux-git/mm/slub.c:943
[    9.771814] in_atomic(): 1, irqs_disabled(): 1, pid: 1063, name: mount
[    9.771818] Pid: 1063, comm: mount Not tainted 3.5.0-jupiter-00003-g8d858b1-dirty #2
[    9.771819] Call Trace:
[    9.771838]  [<c104f79b>] __might_sleep+0xcb/0xe0
[    9.771844]  [<c10c00d4>] __kmalloc+0xb4/0x1c0
[    9.771851]  [<c1041d4a>] ? queue_work+0x1a/0x30
[    9.771854]  [<c1041dcf>] ? queue_delayed_work+0xf/0x30
[    9.771862]  [<c1205832>] ? bit_putcs+0xf2/0x3e0
[    9.771865]  [<c1041e01>] ? schedule_delayed_work+0x11/0x20
[    9.771868]  [<c1205832>] bit_putcs+0xf2/0x3e0
[    9.771875]  [<c12002b8>] ? get_color.clone.14+0x28/0x100
[    9.771878]  [<c1200d2f>] fbcon_putcs+0x11f/0x130
[    9.771882]  [<c1205740>] ? bit_clear+0xe0/0xe0
[    9.771885]  [<c1200f6d>] fbcon_redraw.clone.21+0x11d/0x160
[    9.771889]  [<c120383d>] fbcon_scroll+0x79d/0xe10
[    9.771892]  [<c12002b8>] ? get_color.clone.14+0x28/0x100
[    9.771897]  [<c124c0b4>] scrup+0x64/0xd0
[    9.771900]  [<c124c22b>] lf+0x2b/0x60
[    9.771903]  [<c124cc95>] vt_console_print+0x1d5/0x2f0
[    9.771907]  [<c124cac0>] ? register_vt_notifier+0x20/0x20
[    9.771913]  [<c102b335>] call_console_drivers.clone.5+0xa5/0xc0
[    9.771916]  [<c102c58e>] console_unlock+0x2fe/0x3c0
[    9.771920]  [<c102ca16>] vprintk_emit+0x2e6/0x300
[    9.771924]  [<c13f01ae>] printk+0x38/0x3a
[    9.771931]  [<c112e8fe>] reiserfs_remount+0x2ae/0x3e0
[    9.771934]  [<c112e650>] ? reiserfs_fill_super+0xb00/0xb00
[    9.771939]  [<c10ca0ab>] do_remount_sb+0xab/0x150
[    9.771943]  [<c1034476>] ? ns_capable+0x46/0x70
[    9.771948]  [<c10e059c>] do_mount+0x20c/0x6b0
[    9.771955]  [<c10a7044>] ? strndup_user+0x34/0x50
[    9.771958]  [<c10e0acc>] sys_mount+0x6c/0xa0
[    9.771964]  [<c13f2557>] sysenter_do_call+0x12/0x26

According to comment in bit_putcs() that kammloc() call only happens
when fbcon is drawing to a monochrome framebuffer (which is my case with
hid-picolcd).
Signed-off-by: NBruno Prémont <bonbons@linux-vserver.org>
Signed-off-by: NFlorian Tobias Schandinat <FlorianSchandinat@gmx.de>

Signed-off-by: NPaul Cercueil <paul@crapouillou.net>
Signed-off-by: NFlorian Tobias Schandinat <FlorianSchandinat@gmx.de>

The OMAP3 checks have been removed and replaced by a dss feature
FEAT_DPI_USES_VDDS_DSI for cleaner implementation. The patches
"OMAP: DSS2: enable VDDS_DSI when using DPI" (8a2cfea8) by Tomi Valkeinen
<tomi.valkeinen@nokia.com> and "ARM: omap: fix oops in
drivers/video/omap2/dss/dpi.c" (40410715) by Russell King
<rmk+kernel@arm.linux.org.uk> had introduced these checks. As it is evident
from these patches a dependency exists for some DSS pins on VDDS_DSI which is
better shown by dss feature FEAT_DPI_USES_VDDS_DSI.
Signed-off-by: NChandrabhanu Mahapatra <cmahapatra@ti.com>
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

OMAP4 checks are removed from VENC to provide it a cleaner interface. These
checks were introduced by patches "HACK: OMAP: DSS2: VENC: disable VENC on OMAP4
to prevent crash" (ba02fa37) by Tomi Valkeinen <tomi.valkeinen@ti.com> and
"OMAPDSS: VENC: fix NULL pointer dereference in DSS2 VENC sysfs debug attr on
OMAP4" (cc1d3e03)  by Danny Kukawka <danny.kukawka@bisect.de> to prevent VENC
from crashing OMAP4 kernel.
Signed-off-by: NChandrabhanu Mahapatra <cmahapatra@ti.com>
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

All the cpu_is checks have been moved to dss_init_features function providing a
much more generic and cleaner interface. The OMAP version and revision specific
initializations in various functions are cleaned and the necessary data are
moved to dss_features structure which is local to dss.c.
Signed-off-by: NChandrabhanu Mahapatra <cmahapatra@ti.com>
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Functions dss_calc_clock_rates() and dss_get_clock_div() are removed as these
functions have become redundant and no longer used.
Signed-off-by: NChandrabhanu Mahapatra <cmahapatra@ti.com>
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

All the cpu_is checks have been moved to dispc_init_features function providing
a much more generic and cleaner interface. The OMAP version and revision
specific functions and data are initialized by dispc_features structure which is
local to dispc.c.
Signed-off-by: NChandrabhanu Mahapatra <cmahapatra@ti.com>
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Commit 7849398f introduced a bug,
causing the following error to be reported:

[  370.827819] cannot lock PLL
[  370.830749] CFG1 0x1e
[  370.833160] CFG2 0x602004
[  370.835876] CFG4 0x40000
[  370.838562] omapdss HDMI: Failed to lock PLL

However, HDMI output is still enabled.

The problem is that we enable the HDMI video output temporarily when
reading EDID or detecting if a HDMI cable is connected (ugh), and the
commit above changes the behavior of the driver so that the video
timings are not yet configured at the point when EDID is read.

This patch fixes the problem by configuring the initial VGA timings at
HDMI probe.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

So we've had a fair few reports of fbcon handover breakage between
efi/vesafb and i915 surface recently, so I dedicated a couple of
days to finding the problem.

Essentially the last thing we saw was the conflicting framebuffer
message and that was all.

So after much tracing with direct netconsole writes (printks
under console_lock not so useful), I think I found the race.

Thread A (driver load)    Thread B (timer thread)
  unbind_con_driver ->              |
  bind_con_driver ->                |
  vc->vc_sw->con_deinit ->          |
  fbcon_deinit ->                   |
  console_lock()                    |
      |                             |
      |                       fbcon_flashcursor timer fires
      |                       console_lock() <- blocked for A
      |
      |
fbcon_del_cursor_timer ->
  del_timer_sync
  (BOOM)

Of course because all of this is under the console lock,
we never see anything, also since we also just unbound the active
console guess what we never see anything.

Hopefully this fixes the problem for anyone seeing vesafb->kms
driver handoff.

v1.1: add comment suggestion from Alan.

Cc: stable@vger.kernel.org
Signed-off-by: NDave Airlie <airlied@redhat.com>

So we've had a fair few reports of fbcon handover breakage between
efi/vesafb and i915 surface recently, so I dedicated a couple of
days to finding the problem.

Essentially the last thing we saw was the conflicting framebuffer
message and that was all.

So after much tracing with direct netconsole writes (printks
under console_lock not so useful), I think I found the race.

  Thread A (driver load)    Thread B (timer thread)
    unbind_con_driver ->              |
    bind_con_driver ->                |
    vc->vc_sw->con_deinit ->          |
    fbcon_deinit ->                   |
    console_lock()                    |
        |                             |
        |                       fbcon_flashcursor timer fires
        |                       console_lock() <- blocked for A
        |
        |
  fbcon_del_cursor_timer ->
    del_timer_sync
    (BOOM)

Of course because all of this is under the console lock,
we never see anything, also since we also just unbound the active
console guess what we never see anything.

Hopefully this fixes the problem for anyone seeing vesafb->kms
driver handoff.
Signed-off-by: NDavid Airlie <airlied@redhat.com>
Acked-by: NAlan Cox <alan@lxorguk.ukuu.org.uk>
Cc: stable@vger.kernel.org
Tested-by: NJosh Boyer <jwboyer@gmail.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>