We don't support interlace modes properly on OMAP5+ HDMI, so we need to
reject interlace timings.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Only OMAP5+ has REFSEL field, but at the moment it's set also on OMAP4.

Fix this by adding a "has_refsel" field, and setting the REFSEL based on
that.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Now that we have the common DSS PLL support, change DSI to use it. This
results in quite a lot of changes, but almost all of them are trivial
name changes.

The functions to calculate and program the PLL settings can be removed
from dsi.c, as the common PLL API contains the same functionality.

We also need to create struct dss_pll_hw entries for PLL hardware
features for different OMAP platforms, instead of using the
dss_features.c as the old code does.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

OMAP DSS currently contains two different PLLs: DSI PLL (Type A PLL) and
HDMI PLL (Type B PLL). When DRA7 support is added, we will also support
Video PLLs (Type A).

The driver currently handles all PLLs totally separately. This patch
adds common DSS PLL code, which

a) lets us have common code for the PLLs
b) lets the users of the PLLs use a common API, instead of DSI API or
   HDMI API.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

When DPI uses the DSI PLL for pixel clock, the DPI code will call
dsi_runtime_get/put to keep the DSI block enabled. A much simpler way to
handle this is to do dsi_runtime_get/put in DSI's dsi_pll_init() and
dsi_pll_uninit(), thus removing the need for DSI to call the runtime PM
functions.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

The HSDIV outputs of DSI PLL (and also other PLLs) all have the same
bit width for the divider value.

Simplify the code by merging HSDIV divider widths into one width.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

We are creating a common DSS PLL code, so having fixed DSI specific
hsdiv fields in the clock information is not ok.

This patch changes the hsdiv fields to arrays, so that we can use all
the 4 possible hsdiv outputs (DSI only usees 2), and we have generic way
to access the hsdivs.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

We are creating a common DSS PLL code, so rename 'clkin4ddr' field,
which is DSI specific name, to 'clkdco' which is a generic name.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Now that dsi_clock_info only contains information about the PLL, we can
just copy the whole struct when storing the clock information, instead
of copying individual fields.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

We have pll_locked field in struct dsi_data, but it doesn't have any
meaningful use anymore, and can be removed.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

struct dsi_clock_info contains clkin field, which is the rate of the
PLL's input clock. This field is not needed, as it can be easily
retrieved by using the clk_get_rate().

This patch removes the clkin field.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

struct dsi_clock_info represents the clocks handled by the DSI, mostly
PLL related clocks. In an effort to create common PLL code, we need to
remove all the non-PLL items from dsi_clock_info.

This patch removes LP clock related fields from dsi_clock_info, and
creates a new struct dsi_lp_clock_info for holding clock info for the LP
clock.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

The DSS PLL has support to power on the PLL's highspeed clock output
and HSDIV output separately. In practice both need to powered on, as in
most OMAP's that's the only working configuration. We already do that in
dsi_pll_init(), by overriding the passed arguments so that both are
always powered.

Simplify the code by removing the support for choosing which outputs to
power on.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

With the previous patch "OMAPDSS: DSI: wait for hsdiv clocks when
enabling PLL",  dsi_wait_pll_hsdiv_dispc_active and
dsi_wait_pll_hsdiv_dsi_active are no longer needed, so they and the
callers can be removed.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

At the moment we have two functions to wait for the HSDIV clocks to get
active, dsi_wait_pll_hsdiv_dispc_active and
dsi_wait_pll_hsdiv_dsi_active. Instead of such inconvenient functions,
let's just make sure that the hsdiv clocks are active after the pll has
been enabled.

This patch adds code to dsi_pll_set_clock_div() to wait until HSDIV
clocks are active.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Add a 'port' parameter in dpi_select_source. The param tells the port
number of the DPI instance that we want to configure. We use this number
to select the overlay manager for that DPI instance.
Signed-off-by: NArchit Taneja <archit@ti.com>
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Register DPI outputs, and assign the port_num to them as specified by the
'reg' property in the DPI ports in DT.

To support multiple DPI instances, dpi_get_channel needs to take the DPI
instance's port number to get the corresponding channel. Make it take this
argument. We just pass 0 in the non-DT path, since we don't support multiple
instances in the non-DT case.
Signed-off-by: NArchit Taneja <archit@ti.com>
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

In omapdss_of_find_source_for_first_ep, we retrieve a source endpoint's DT node,
and then see what omapdss output has the matching device_node pointer in
omap_dss_find_output_by_node.

For all DPI and SDI outputs, the device_node pointer is set as the parent's DSS
device_node pointer. If the source is one of these outputs, the above method
won't work.

To get the correct output for ports within DSS(and in other cases in the future,
where multiple ports might be under one device), we require additional
information which is exclusive to the output port.

We create a new field in omap_dss_device called 'port_num', this provides port
number of the output port corresponding to this device. When searching for the
source endpoint in DT, we extract the 'reg' property from the port corresponding
to the endpoint source. From the list of registered outputs, we pick out that
output which has both dev->of_node and port_num matching with the device_node
pointer and 'reg' of the source endpoint node from DT.

For encoder blocks(the ones which have both an input and output port), we need
to set the port_num as the 'reg' property for the output port as defined in the
DT bindings. We set port_num to 1 in the tfp410 and tpd12s015 encoder drivers.
Signed-off-by: NArchit Taneja <archit@ti.com>
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

The init/uninit port functions are used to set up the DPI and SDI outputs under
the dss platform device. A 'reg' property is used to determine the port number
of the output. This tells us whether the port is DPI or SDI for OMAP34xx DSS
revision. For other DSS revisions, we only have DPI outputs under the dss
platform device.

For multiple DPI output instances(introduced in DRA7xx DSS), we will use the
the port number to specify which DPI output instance is being inited.

The current functions work fine if there is only one DPI output instance in
DSS. For multiple DPI instances, it would get complicated to figure out whether
port number was used to specify whether the output is SDI, or another DPI
instance.

We create a list of port types supported for each DSS rev, with the index of the
port in the list specifying the port number of the output for that DSS revision.
This allows us to have a more generic way to init/uninit ports within DSS, and
also support multiple DPI ports.

We make the uninit_port functions iterative since we will have multiple DPI
ports to uninit in the future.
Signed-off-by: NArchit Taneja <archit@ti.com>
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

DPI and SDI ports are backed by only one parent DSS device. We don't have a
corresponding platform_device for ports under DSS. In order to support multiple
instances of DPI, we need to pass the driver data pointer through the DPI port's
private data ('data' member in device_node struct).

dpi_init_output/dpi_uninit_output are untouched and only used for non-DT case,
these are called when the DPI platform device probed/removed. These funcs will
be removed when non-DT mode is removed.

dpi_init_output_port/dpi_uninit_output_port are created and used for the DT
path, called when DSS inits/uninits it's ports. These new functions retrieve
the dpi_data pointer from 'port->data', and not from the platform device's
data(pdev->dev) like in the non-DT path.

We add some code in dss_uninit_ports() to pass a pointer to the DPI port in
dpi_uninit_port().
Signed-off-by: NArchit Taneja <archit@ti.com>
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Allocate driver data(dpi_data) for each DPI instance. It's allocated in
omap_dpi_probe() when DT isn't used, and in dpi_init_port() when DT is used.
The dpi_data struct instance is no longer global. In the case of DPI ops, it's
retrieved from dpi_get_data_from_dssdev(). 'dssdev' passed by the connected
encoder/panel driver is a pointer to the 'output' member in dpi_data, and thus
can be used to get the DPI instance's driver data. In the case of probe/ini_port
functions, it's set as DPI/DSS device's private data embedded in the
platform_device struct.

Having dpi_data as private data of the platform device will not work for
multiple DPI instances in the DT case. This is because there is no corresponding
platform_device for DPI or SDI, they exist only as ports under the parent DSS
platform_device in the DT case. The DPI port's private data('data' member in
device_node struct) will later be used to store dpi_data.
Signed-off-by: NArchit Taneja <archit@ti.com>
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

DPI related data is currently a static global struct parameter. It is accessed
directly by functions in the driver.

This method won't work if we want the driver to support multiple DPI instances.
Create struct dpi_data, and pass its pointer to functions which need to use it.

We still have a static instance defined for dpi_data, which is accessed by top
level DPI ops. This will be removed when the driver dynamically allocates
dpi_data for each DPI instance.
Signed-off-by: NArchit Taneja <archit@ti.com>
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Without that fix connector-analog-tv driver isn't probed when compiled
as module.
Signed-off-by: NH. Nikolaus Schaller <hns@goldelico.com>
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

PLL_SELFREQDCO bitfield is from bit 3 to 1, but the driver writes bits
from 4 to 1. The bit 4 is 'reserved', so this probably should not cause
any issues, but it's better to fix it.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Preload register is dumped twice for video overlays and mflag register
is not dumped for GFX.

Fix the register dump.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

The register offset for DISPC_OVL_MFLAG_THRESHOLD is wrong, fix it.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

HDMI PLL's REGSD field is only set by the driver if the PLL's output
clock is over 1GHz. This is clearly an error, as REGSD should be set
always.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

The PLL settings are committed by setting GO bit, which is then cleared
by the HW when the settings have been taken into use.

The current PLL code handles this wrong: instead of waiting for the bit
to be cleared, it waits for the bit to be set. Usually, the bit is
always set, as the CPU has just set it before. However, if the CPU takes
enough time between setting the GO bit and checking it, the HW may
already have cleared the bit and this leads to timeout error.

Fix the wait to check the bit properly.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

omapfb disables all the overlays when freeing resources, but it should
also remove those overlays from overlay managers.

Not doing so causes a crash if omapfb is unbound and bound, or omapfb
module is removed and loaded, while keeping omapdss around.

Fix this by calling unset_manager() for all overlays.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

When omapfb is shutting down, it will disable all the overlays. However,
instead of actually disabling all the overlays, it disables only all the
overlays that are currently attached to framebuffers.

On OMAP4+, this leaves the fourth overlay left enabled.

Fix the loop so that we actually go through all the overlays.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

We should wait for any pending updates when an overlay manager is
about to be disabled, because the updates will never be finished if the
manager is disabled too early.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

If we leave __exit annotation, driver can't be unbound
through sysfs.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

omapdss drivers cannot handle devices being unbound while the devices
are part of a connected display pipeline. Module refcounts are used to
prevent unloading the modules, but one can still manually unbind the
devices via sysfs, causing crash.

Set suppress_bind_attrs to disable the bind/unbind support via sysfs.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

without MODULE_ALIAS(), omapfb won't get loaded
automatically.
Signed-off-by: NFelipe Balbi <balbi@ti.com>
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Remove unnecessary mach/cpu.h header to be able to converge to a multiplatform
kernel.
Signed-off-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Acked-by: NNicolas Ferre <nicolas.ferre@atmel.com>
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Resolve shadow warnings that appear in W=2 builds by renaming
the "state" global to "vgastate".
Signed-off-by: NMark Rustad <mark.d.rustad@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Since strsep is used to tokenize the options string, after each option
match the code should use "continue" to get the next token from strsep.
This patch applies this pattern consistently.

Previously, for "scrollback:" and "map:" the parse code would return
(unconditionally: strsep ensures *options != ','), causing any
following option to be ignored, while for "vc:" the parse code would
go on to parse further options within the same token, which could lead
to invalid input being accepted.
Signed-off-by: NMaarten ter Huurne <maarten@treewalker.org>
Acked-by: NPaul Cercueil <paul@crapouillou.net>
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

The drivers should use dmaengine_terminate_all() API instead of
accessing the device_control which will be deprecated soon
Signed-off-by: NVinod Koul <vinod.koul@intel.com>

The kernel used to contain two functions for length-delimited,
case-insensitive string comparison, strnicmp with correct semantics and
a slightly buggy strncasecmp.  The latter is the POSIX name, so strnicmp
was renamed to strncasecmp, and strnicmp made into a wrapper for the new
strncasecmp to avoid breaking existing users.

To allow the compat wrapper strnicmp to be removed at some point in the
future, and to avoid the extra indirection cost, do
s/strnicmp/strncasecmp/g.
Signed-off-by: NRasmus Villemoes <linux@rasmusvillemoes.dk>
Cc: Jean-Christophe Plagniol-Villard <plagnioj@jcrosoft.com>
Cc: Tomi Valkeinen <tomi.valkeinen@ti.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The DEFINE_XENBUS_DRIVER() macro looks a bit weird and causes sparse
errors.

Replace the uses with standard structure definitions instead.  This is
similar to pci and usb device registration.
Signed-off-by: NDavid Vrabel <david.vrabel@citrix.com>