Vishal Sagar [Thu, 15 Nov 2018 07:15:18 +0000 (12:45 +0530)]
drm: xlnx: crtc: Fix max width and height overflow
The maximum resolution supported is shown as -1x-1 in Xorg.log when
trying to run Xorg. This is occurring due to width and height variables
overflowing as they are assigned UINT_MAX instead of INT_MAX.
This patch fixes this by correcting the data types of width and height
variables and correctly assigning INT_MAX as maximum width and height.
Vishal Sagar [Thu, 15 Nov 2018 07:15:17 +0000 (12:45 +0530)]
drm: xlnx: dsi: Fix kernel panic when "halt" is run
When "halt" is run, the kernel panics and dumps a kernel call stack
which showing it originating in xlnx_dsi_unbind(). On closer
examination, it is found that the xlnx_dsi_host_detach() incorrectly
sets the "panel_mode" device node member pointer to NULL instead of
setting the drm_panel member pointer "panel" to NULL.
clk: Fix compilation errors of silicon labs si5324 driver on zynq platform
This patch fixes the following linker errors on zynq(32-bit cpu) platform
by replacing the div and mod operations with math lib api.
drivers/clk/si5324drv.o: In function `si5324_rate_approx':
si5324drv.c:(.text+0xa8): undefined reference to `__aeabi_uldivmod'
si5324drv.c:(.text+0xdc): undefined reference to `__aeabi_uldivmod'
si5324drv.c:(.text+0x13c): undefined reference to `__aeabi_uldivmod'
ASoC: xlnx: Disable clock wizard usage in audio driver
Current solution which uses clocking wizard, couldn't generate
accurate clock due to inability of clock wizard driver to program
fractional value of clock divider value. This is resulting in audio
loss in a simulataneous capture and playback audio usecase.
This patch disables its usage.
Mauro DeLuca [Thu, 1 Nov 2018 17:43:25 +0000 (17:43 +0000)]
misc: xilinx-sdfec: Add kernel-doc style comments
This patch updates the existing comments for enums, structs, macros and
ioctls to the kernel-doc style. Descriptions are also provided for
parameters passed through these data types.
staging: xroeframer: add support for multi-port Ethernet
Support for up to 4 Ethernet ports. This wraps all the "store" & "show"
functions of sysfs entries in a function which calculates the current
Ethernet port (when applicable) and adds the correct offset to the
address' register, which is ADDRESS + (0x100 * port_num). It defaults
to 0 for the non-applicable cases so no other entries are affected.
Signed-off-by: Vasileios Bimpikas <vasileios.bimpikas@xilinx.com> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
ASoC: xlnx: disable strict check of AES parameters
AES audio stream parameters is observed to be not always available
to compare against user provided parameters during capture. So, disable
comparing and just provide AES stream parameters as information to user.
Vishal Sagar [Tue, 23 Oct 2018 10:19:26 +0000 (15:49 +0530)]
xilinx: v4l: dma: Stop subdevs in reverse order of starting
Patch fixes the incorrect sequence of stopping subdevs in a media
pipeline. The order of stopping should be reverse the order of starting
a video pipe. The video pipe should be started from sink to source and
be disabled from source to sink.
Michael Gill [Fri, 19 Oct 2018 00:54:10 +0000 (17:54 -0700)]
staging: apf: add preliminary support for accelerator interrupts
This patch introduces support for dynamic allocation of
interrupts to c-callable IP. This is not full support for the
SDSoC adapter IP interrupts, or general HLS interrupt support.
Signed-off-by: Michael Gill <michael.gill@xilinx.com> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Suresh Gupta [Fri, 12 Oct 2018 09:51:44 +0000 (15:21 +0530)]
v4l: xilinx-multi-scaler: Select taps as per scaling ratio
The following are some recommendations for how many taps to use:
- Upscale 6 taps
- Down scale to 1.5 6 taps
- Down scale > 1.5 <= 2.5 8 taps
- Down scale > 2.5 <= 3.5 10 taps
- Down scale > 3.5 12 taps
This patch implements the same.
Suresh Gupta [Fri, 12 Oct 2018 09:51:41 +0000 (15:21 +0530)]
v4l: xilinx-multi-scaler: Move tap parameter to xm2msc_dev from channel context
HW tap value is constant for all the channels and do not have
individual values in HW. So, this patch move tap parameter
from channel context to device context.
Michal Simek [Wed, 17 Oct 2018 07:20:09 +0000 (09:20 +0200)]
serial: uartlite: Fix compilation issue on !OF platforms
There is missing ulite_of_match on !OF platforms.
Use of_match_ptr() to fix it.
Warning log:
drivers/tty/serial/uartlite.c: In function ‘ulite_get_id’:
drivers/tty/serial/uartlite.c:774:33: error: ‘ulite_of_match’ undeclared
(first use in this function)
ret = of_alias_get_alias_list(ulite_of_match, "serial",
^
drivers/tty/serial/uartlite.c:774:33: note: each undeclared identifier
is reported only once for each function it appears in
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
On 64-bit:
drivers/of/configfs.c: In function ‘cfs_overlay_item_dtbo_read’:
include/linux/kern_levels.h:5:18: warning: format ‘%u’ expects
argument of type ‘unsigned int’, but argument 4 has type
‘size_t {aka long unsigned int}’ [-Wformat=]
Signed-off-by: Nava kishore Manne <nava.manne@xilinx.com> Reviewed-by: Radhey Shyam Pandey <radhey.shyam.pandey@xilinx.com> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
gpio: Fix the Dual Channel gpio driver removal issue
When axi-gpio is configured as a dual-channel with the
existing implementation OF the driver, we are not freeing
the 2nd channel during un-registration of the driver.
Signed-off-by: Nava kishore Manne <nava.manne@xilinx.com> Reviewed-by: Shubhrajyoti Datta <Shubhrajyoti.datta@xilinx.com> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Checkpatch reports warnining because "xlnx,zynqmp-pm"
DT compatible string is not documented in DT binding doc.
Fix this by defining deprecated "xlnx,zynqmp-pm" in
DT binding doc.
Signed-off-by: Rajan Vaja <rajan.vaja@xilinx.com> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Suresh Gupta [Fri, 12 Oct 2018 10:10:22 +0000 (15:40 +0530)]
v4l: xilinx-multi-scaler: Remove volatile
As per Documentation/process/volatile-considered-harmful.rst,
using accessor functions do not require volatile as accessors
are written to prevent unwanted optimization.
And we are using ioread32 and iowrite32, so volatile is not require.
This patch removes the volatile from code and remove
the below warning also
warning: passing argument 1 of 'ioread32' discards 'const volatile'
qualifier from pointer target type [-Wdiscarded-qualifiers]
return ioread32(addr);
^~~~
Vishal Sagar [Sat, 13 Oct 2018 06:32:42 +0000 (12:02 +0530)]
drm: xlnx: pl_disp: Enable early callback
Patch adds support for enabling early callback. This enables the
user space applications like kmssink plugin that queue up only one
buffer at a time to not drop any frames due to timeout receiving
vblank event from the driver.
The side effect of enabling this behavior is that the buffer whose
callback function has been called may actually still be scanning
out.
Vishal Sagar [Sat, 13 Oct 2018 06:32:41 +0000 (12:02 +0530)]
dma: xilinx: Add early callback support to Framebuffer Driver
This patch adds support for the callback per descriptor to be called
after the staging phase itself rather than waiting for the descriptor
to transition to the active stage and done list.
This behavior is enabled by an API which lets dma engine client decide
whether the callback needs to called as usual or earlier after staging
state.
Harini Katakam [Sat, 13 Oct 2018 11:21:38 +0000 (16:51 +0530)]
net: macb: Fix phy path in macb resume
A soft reset on the PHY is required after resume, especially with some
devices like TI that implement their own SW reset bits apart from
generic phy reset. This is cleanly accesible to the driver via the
phy_init_hw function. Hence call the same to fix ethernet link after
resume. Otherwise the link will be unstable and packets will not be
received.
Also remove netif_carrier_off/on call in suspend/resume path.
These are already taken care of by the link change function.
Harini Katakam [Sat, 13 Oct 2018 11:21:37 +0000 (16:51 +0530)]
net: macb: Do not call PM sync in mdio helpers
Do not bring up clocks in MDIO helpers when the net device is down.
When suspended, there is no need to perform PHY read/write operations
except if GEM is a wake device.
Martin Lund [Fri, 12 Oct 2018 09:16:42 +0000 (14:46 +0530)]
mtd: nand: Arasan: Fix initialization of controller structure
Make sure that the controller structure is initialized using the
nand_hw_control_init() function. This way the spinlock, waitqueue
elements etc. of the structure will be correctly initialized.
Without this fix the nand driver have been seen stalling, especially
when stress tested.
Signed-off-by: Martin Lund <malu@gomspace.com> Reviewed-by: Naga Sureshkumar Relli <naga.sureshkumar.relli@xilinx.com> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Durga Challa [Thu, 11 Oct 2018 10:52:53 +0000 (16:22 +0530)]
nvmem: zynqmp: Added support to access efuse
This patch provides support to read/write efuse memory map of zynqmp. So
modified the driver accordingly, 0th offset is for Soc version and from
0xc onwards used for efuse access, total size of nvmem file is 0x100
bytes.
Signed-off-by: Durga Challa <vnsl.durga.challa@xilinx.com> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Sam Bobrowicz [Thu, 27 Sep 2018 06:28:08 +0000 (23:28 -0700)]
media: xilinx: csi2rxss: store format size
Currently set_format does not store the dimensions when setting a
non-RAW format. Although they are irrelevant to the operation of the
CSI RX IP, this can cause link_validate to fail and also unexpected
results when calling get_format. Fix this by storing the format size.
Signed-off-by: Sam Bobrowicz <sam@elite-embedded.com> Reviewed-by: Vishal Sagar <vishal.sagar@xilinx.com> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Kalyani Akula [Wed, 10 Oct 2018 09:20:39 +0000 (14:50 +0530)]
crypto: zynqmp-aes: Fix for segfault seen with large sets of data
This patch resolves the issue seen when encrypt/decrypt of large sets of
data.
Issue is seen due to improper handing of blkcipher_walk. When large
amount of data is sent for encrypt/decrypt, af_alg framework copies it
to the TSGL (TX SGL) kernel. When driver does a walk on the scatterlist,
walk returns nbytes to be processed. nbytes can be complete data sent
from user space or few bytes. So, when data is received in multiple
walks, we are accumulating the data and sending complete data at once.
Sam Bobrowicz [Thu, 11 Oct 2018 16:23:30 +0000 (09:23 -0700)]
media: ov5640: Add additional media bus formats
Add support for 1X16 yuv media bus formats (v4l2_mbus_framefmt).
These formats are equivalent to the 2X8 formats that are already
supported, both of which accurately describe the data present on
the CSI2 interface. This change will increase compatibility with
CSI2 RX drivers that only advertise support for the 1X16 formats.
Signed-off-by: Sam Bobrowicz <sam@elite-embedded.com> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Sam Bobrowicz [Thu, 11 Oct 2018 16:23:26 +0000 (09:23 -0700)]
media: ov5640: fix resolution update
set_fmt was not properly triggering a mode change when
a new mode was set that happened to have the same format
as the previous mode (for example, when only changing the
frame dimensions). Fix this.
Signed-off-by: Sam Bobrowicz <sam@elite-embedded.com> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Sam Bobrowicz [Thu, 11 Oct 2018 16:23:25 +0000 (09:23 -0700)]
media: ov5640: set DPHY timing registers
Register 0x4837 is used to calculate the DPHY timing
parameters such as HS_SETTLE. It needs to be set to the
current pclk period. It is possible for some CSI receivers
to tune this on the RX side, but some do not.
Currently this value is not set, so fix this by setting it
whenever setting the rest of the PLL settings.
Also, a change to the init register blob that helps ensure
the following DPHY spec requirement is met:
MIN HS_ZERO + MIN HS_PREPARE > 145 + t_UI * 10
Signed-off-by: Sam Bobrowicz <sam@elite-embedded.com> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Sam Bobrowicz [Thu, 11 Oct 2018 16:23:24 +0000 (09:23 -0700)]
media: ov5640: calculate PLL settings for modes
Remove the PLL settings from the register blobs and
calculate them based on required clocks. This allows
more mode and input clock (xclk) configurations.
Signed-off-by: Sam Bobrowicz <sam@elite-embedded.com> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Devarsh Thakkar [Tue, 9 Oct 2018 23:30:23 +0000 (16:30 -0700)]
xilinx: v4l: m2m: Set default mode for framebuffer IP
- Set the framebuffer IP to default mode instead of auto-restart
mode for mem2mem usecases due to reason explained below:
- By framebuffer IP is programmed by driver to use a special mode
called auto_restart mode or free-running mode in which core processes
data periodically irrespective of input rate.
- So in this scenario, if input is stalled for sometime,
then framebuffer IP will still process with previous data
and transmit/receive the DMA descriptors.
- This behaviour of auto_restart or free-running mode is fine for
streaming usecases where display device needs to be provided
with data at a consistent rate but is not acceptable for mem2mem usecases
where output needs to have exact 1-to-1 relationship with input
without any duplication of frames.
- Furthermore this gives a performance benefit for mem2mem usecases
as the default mode doesn't require to stage a descriptor and
instead the active descriptor can be marked as complete after
interrupt is received to process next descriptor.
Devarsh Thakkar [Tue, 9 Oct 2018 23:30:22 +0000 (16:30 -0700)]
dma: xilinx: Add API to set framebuffer IP mode
This adds API so that client driver can set operation mode
for framebuffer IP core to either auto_restart or default mode
of operation :
- Auto_restart or free-running mode is more suitable for streaming
usecases which have a strict requirement to maintain uniform
throughput to display devices. In this mode framebuffer IP
core is restarted automatically after completing first transaction
even if input is delayed or blocked, where core will use same descriptor
again thereby maintaining consistent throughput by duplicating frames.
- Default mode of operation is suitable for non-streaming usecases
like mem2mem usecases where there has to be a 1-to-1 relationship
for frame sequence and frame numbers with respect to input and
so if input is delayed core should wait until it gets new descriptor.
- Unless explicitly set by client driver framebuffer IP core should
use auto_restart mode as previously, so set mode as auto_restart during
channel probe.
- Add enum in framebuffer channel structure to select operation mode.
- Use operation mode enum to decide mode during framebuffer start and halt.
- Add a function "frmbuf_find_chan" to find corresponding framebuffer
channel structure from dma channel structure.
- Export an API "xilinx_xdma_set_mode" to set operation mode of framebuffer
IP core to either auto-restart or default.
- Unlike Auto-restart mode which follows a 3-stage model
where it marks the descriptor as staging on submit,
as active when next interrupt is received, and eventually
as complete after next interrupt is received, there is no
such requirement in default mode, so for default mode
we directly mark the current descriptor submitted to IP
as active and then mark it as complete when next
interrupt is received.
- Fix some typos in xilinx_xdma_drm_config and xilinx_xdma_v4l2_config
description.
Devarsh Thakkar [Tue, 9 Oct 2018 23:30:21 +0000 (16:30 -0700)]
dma: xilinx: Mark previous active decriptor as complete in irq handler
Moving the block to mark previous active descriptor as complete
to irq handler as it's more appropriate to do it in interrupt
context which confirms that previous transaction was completed.
There is no semantic change in the flow,
previously descriptor was marked as completed in
xilinx_frmbuf_start_transfer() assuming it's is called by
the done interrupt handler which didn't seem required,
so we now mark it as completed in the interrupt handler itelf.
Vishal Sagar [Thu, 11 Oct 2018 06:04:31 +0000 (11:34 +0530)]
drm: xlnx: Select VIDEOMODE_HELPERS when DRM_XLNX_PL_DISP is selected
Select the VIDEOMODE_HELPERS when DRM_XLNX_PL_DISP is selected in
menuconfig. This fixes the compilation error of undefined reference to
drm_display_mode_to_videomode in the Xilinx PL Display driver.
This is also added to DRM_XLNX_MIXER as the same API is being used
there.
Sam Bobrowicz [Thu, 11 Oct 2018 00:21:55 +0000 (17:21 -0700)]
xilinx: v4l: dma: Invoke s_power function of subdevice
This patch enables driver to invoke s_power, which is part of subdev
core APIs. This can put subdevice in power saving mode or normal
operation mode.
Signed-off-by: Sam Bobrowicz <sam@elite-embedded.com> Signed-off-by: Satish Kumar Nagireddy <satish.nagireddy.nagireddy@xilinx.com> Signed-off-by: Sai Hari Chandana Kalluri <chandana.kalluri@xilinx.com> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
xilinx: v4l: vipp: Avoid decrementing refcount of a node twice
In the current implementation, driver is calling of_graph_get_next_endpoint()
and of_node_put() functions to get next endpoint and decrement refcount of a
node. As the former function is already decrementing refcount, there is no need
of calling the later function which decrements the refcount again.
This patch avoids calling of_node_put() function in driver, as the same has been
taken care in of_graph_get_next_endpoint() implementation.
arm64: zynqmp: Add reset-on-timeout for all boards and modify default timeout value
This patch adds reset-on-timeout to FPD WDT which will trigger an
interrupt to PMU when watchdog expiry happens and PMU takes the
necessary action. If this property is not enabled, reason will not be
known when watchdog expiry happens.
This patch also modifies the default timeout to 60 seconds. Reason is
that if u-boot enables WDT, it will set the timeout to 10 seconds and
this is not enough to boot till Linux and start the WDT application in
Linux. 60 seconds is the maximum safest value to boot till Linux and
start the WDT application.
Users need to change this timeout value to fit their needs.
Signed-off-by: Mounika Grace Akula <mounika.grace.akula@xilinx.com> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
To reflect a recent change in the hardware design, RADIO_CTRL_BASE has
been changed from 0xA0030000 to 0xA0060000 in xroe_framer.h.
This is to be skipped altogether along with any remaining hard-coded
values when the secondary devices are properly implemented
in the device tree.
Signed-off-by: Vasileios Bimpikas <vasileios.bimpikas@xilinx.com> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Michael Gill [Tue, 9 Oct 2018 19:42:04 +0000 (12:42 -0700)]
staging: apf: Correct kernel-doc structure for xlnk.c
This patch eliminates warnings associated with kernel-doc
formatting in the apf driver. It does this by removing the
partial comment was being detected as being an attempt at a
kernel-doc comment. This patch is only meant to resolve
warnings. Follow up patches will be to comment the driver.
Signed-off-by: Michael Gill <michael.gill@xilinx.com> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Suresh Gupta [Fri, 5 Oct 2018 14:18:08 +0000 (19:48 +0530)]
v4l: xilinx-multi-scaler: Add reserve area between channel 4 and 5
There is reserve area between XM2MSC_DSTIMGBUF0 and XM2MSC_DSTIMGBUF1
of channel 4 due to which the offset for XM2MSC_DSTIMGBUF1 channel 4
is 0x01000 instead of 0x0A00.
Also, this shift the base offset of channels 5 and onward.
So, the effective base offsets for channels are:
Channel 0: 0x100,
Channel 1: 0x300,
Channel 2: 0x500,
Channel 3: 0x700,
Channel 4: 0x900,
Channel 5: 0x1100, <-- Should be 0xb00
Channel 6: 0x1300,
Channel 7: 0x1500;
This patch fixes this problem by increment the offset of channel 5, 6,
and 7 by reserved value. And increments the XM2MSC_DSTIMGBUF1
of channel 4 by reserved value.
This is temporary fix and can be taken care in next version of IP.
Suresh Gupta [Fri, 5 Oct 2018 14:13:06 +0000 (19:43 +0530)]
v4l: xilinx-multi-scaler: Check any channel is still running in streamoff
If an application queued buffer while the device is busy in processing
of previous buffer, the application will not call device_run () again.
And at this point if some application running higher channel stops stream,
the device_run () return as there is no buffere for stopping channel.
In this situation the first channel does not get chance to call
device_run() again and will wait for infinite time.
This patch checks if there is/are any continues channels are still running
and if running, start the process again.
Suresh Gupta [Fri, 5 Oct 2018 14:13:04 +0000 (19:43 +0530)]
v4l: xilinx-multi-scaler: Wait for all running channels to get buffer
So, this patch waits until all continuous running channels have buffer.
Driver do not support processing if any in between intermediate
channel stops streaming.
Limitation: If any intermediate channel stops the streaming,
the streaming continues for only channels which fall under a
low number of stop channel. In this case if channel 0,1,2,3,4
are running and channel 3 stops, then the streaming for channels
0,1,2 will continue and driver stop processing 4th channel without
any error indications. So, application running channel 4 may wait
indefinitely if it does not implement timeout.
Suresh Gupta [Fri, 5 Oct 2018 14:13:03 +0000 (19:43 +0530)]
v4l: xilinx-multi-scaler: Do not allow new task when device is busy
Driver supports multiple channel so device_run( ) can be
called for multiple time.
This patch would not allow reprograming and running
of device again when driver is waiting for IP to
complete processing.
Suresh Gupta [Fri, 5 Oct 2018 14:13:02 +0000 (19:43 +0530)]
v4l: xilinx-multi-scaler: Allow stream on for the immediate next channel
This patch enables the use case to add one more stream when driver already
processing streams of some channels.
Start of new channel requires reset and reprogram of HW IP.
Here the condition is that only next immediate channel
will able to start and stop the stream.
Example: Channels 0,1,2,3 are streaming, then the user can stream on
immediate next channel 4.
Limitation: if user adds stream for channel 6 when channels 0,1,2,3
are running, then driver does not start 6th channel and do not return
any error. And if any intermediate channel stops the streaming,
the streaming continues for only channels which fall under a low number
of stop channel. In this case if channel 0,1,2,3,4 are running and
channel 3 stops, then the streaming for channels 0,1,2 will continue
and driver stop processing 4th channel without any error indications.
Suresh Gupta [Fri, 5 Oct 2018 14:12:58 +0000 (19:42 +0530)]
v4l: xilinx-multi-scaler: Allow abort for individual channels
Driver do not support individual stremoff/abort of any channel and
consider abort will come together for all channels.
This patch allows individual channel abort,
by freeing all pending buffers. This will help to
remove/abort the last running channel while all
other channels continue running stream.
Suresh Gupta [Fri, 5 Oct 2018 14:12:52 +0000 (19:42 +0530)]
v4l: xilinx-multi-scaler: Add support for VIDIOC_CREATE_BUFS for Dmabuf io_mode
Gstremer V4L2transform Plugin calls VIDIOC_CREATE_BUFS ioctl
for Dmabuf io_mode followed by VIDIOC_EXPBUF which is used
to export previously created buffer as a DMABUF file.
But VIDIOC_EXPBUF is available only for V4L2_MEMORY_MMAP buffers.
So, enabling VB2_MMAP flags while creating the m2m_buffer queues.
This patch is removing ULITE_NR_PORTS macro which limits number of
ports which can be used. Every instance is registering own struct
uart_driver with minor number which corresponds to alias ID (or 0 now).
and with 1 uart port. The same alias ID is saved to
tty_driver->name_base which is key field for creating ttyULX name.
Because name_base and minor number are setup already there is no need to
setup any port->line number because 0 is the right value.
Signed-off-by: Shubhrajyoti Datta <shubhrajyoti.datta@xilinx.com> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
serial-uartlite: Change logic how console_port is setup
Change logic how console_port is setup by using CON_ENABLED flag
instead of index. There will be unique uart_console
structure that's why code can't use id for console_port
assignment.
Signed-off-by: Shubhrajyoti Datta <shubhrajyoti.datta@xilinx.com> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
projects / zynq / linux.git / log