When SDMA is use for periphal data transfer, dma request is trigger by
FIFO level. If the SDMA is started after the start of periphal, SDMA
would miss the first pulse and be not able to trigger itself.
In this patch, add interface to trigger a dma request manully.
Chen Liangjun [Wed, 31 Oct 2012 03:11:26 +0000 (11:11 +0800)]
ENGR00231773-5 ASRC: use poll mode to receive last period of ASRC data
ASRC driver use DMA to transfer data from ASRC output FIFO to memory.
However, DMA way require the data number in ASRC output FIFO being larger
than watermark level. Thus a dma request can trigger a DMA burst. For
the last period of output data, its number is possiblely less than output
FIFO watermark level. In this case, the output DMA would pending for the
last period of output data until timeout.
In this patch:
1 divide expected output data length into 2 parts: DMA part
and poll part. Using DMA to get the DMA part data and poll mode to
get the poll part.
2 to prevent user from processing these 2 parts above, kernel
buffers would be untouchable. User application only need send its data
buffer address to driver instead of query the kernel buffer.
Chen Liangjun [Fri, 26 Oct 2012 09:43:30 +0000 (17:43 +0800)]
ENGR00231773-4 ASRC: use scatter list and stall bit for asrc convert
In the origin code, ASRC driver use cyclic way to process DMA task
transfering data to/from ASRC input/output FIFO. In this case, it is
necessary that user application should promise that the input buffer
flow is continuous. If not, there would be 0 data be inserted into data
flow. The output data would be noisy.
In this patch,
1 use scatter list instead of cyclic SDMA: with scatter list,
SDMA would stop when the applied scatter list nents are finished.
2 set stall bit for ASRC "memory->ASRC->memory" convert to stop
ASRC convert when input data is not send into ASRC input FIFO in time.
Chen Liangjun [Fri, 26 Oct 2012 07:10:09 +0000 (15:10 +0800)]
ENGR00231773-3 ASRC: use kzalloc to allocate buffer to support scatterlist
For ASRC's "memory -> ASRC -> memory" using, new driver would support
model below: user input one buffer into ASRC and an corresponding output
buffer would be poped out. There is no timing requirement between this
input buffer and next input buffer. Thus driver would not use the cyclic
way to config SDMA and scatterlist is used. buffer allocated by
dma_alloc_coherent() can't support scatterlist well.
In this patch, use kzalloc to allocate buffer to support scatterlist.
Chen Liangjun [Fri, 26 Oct 2012 05:42:35 +0000 (13:42 +0800)]
ENGR00231773-2 ASRC: add work struct for asrc output data receive
SDMA driver can't promise receive all output data generated. Cause when
the data in output FIFO is less than ASRC output FIFO watermark, there
would be no DMA request generated and thus no SDMA transfer would
happens.
In this patch, add work struct to support ASRC driver receive last part
of data in OUTPUT FIFO in polling way.
Chen Liangjun [Fri, 26 Oct 2012 08:07:41 +0000 (16:07 +0800)]
ENGR00231773-1 ASRC: remove queue operation in ASRC driver
According to ASRC memory->ASRC->memory requirement, driver should
satisfy the feature below: user application would passed into one buffer
and waiting until the output buffer is generated. In this case, only one
buffer is on processing and it is no necessary to use the queue to do
the convert. What is worse, queue operation would make the ASRC driver
hard to understand and maintain.
In this patch, remove the queue operation in ASRC driver.
IPG_PERCLK is the parent of I2C. I2C needs a minimum of
12.8MHz as its input clock to achieve 400KHz speed. Hence
change the IPG_PERCLK speed accordingly.
MX6DQ/MX6DL - Set IPG_PERCLK at 22MHz (sourced from IPG_CLK)
MX6SL - Set IPG_PERCLK to 24MHz(Sourced from 24MHz XTAL).
Israel Perez [Wed, 31 Oct 2012 22:13:36 +0000 (17:13 -0500)]
ENGR00225875-2 i.MX6Q/Solo Sabreauto Bluetooth H4 fix uart rx timeouts.
Bluetooth H4 protocol is very susceptible to data reception timeouts.
DMA transfer only happen when ICD or AGTIM interrutps are trigger.
ICD only happen when a N idle frames are present on rx line
and rx FIFO is empty.
Meanwhile Aging timer is trigger when data in FIFO was been sitting for a
period of 8 frames.
This patch enable both in order to dispatch data as fast is possible only
when the select uart have enable DMA flag.
Signed-off-by: Israel Perez <B37753@freescale.com>
Israel Perez [Wed, 31 Oct 2012 22:06:44 +0000 (17:06 -0500)]
ENGR00225875-1 i.MX6Q/Solo SabreAuto Infineon Bluetooth uart3 config
Configure MUX settings for bluetooth operation over UART3.
Enable RTS,CTS and DMA only for uart3.
Affected files :
arch/arm/mach-mx6/board-mx6q_sabreauto.c
arch/arm/mach-mx6/board-mx6q_sabreauto.h
arch/arm/mach-mx6/board-mx6solo_sabreauto.h
arch/arm/plat-mxc/include/mach/iomux-mx6q.h
On behalf of Francisco Munoz <francisco.munoz@freescale.com>.
Some modification are needed also on hciattach tool.
Signed-off-by: Israel Perez <B37753@freescale.com>
Robby Cai [Fri, 2 Nov 2012 07:58:40 +0000 (15:58 +0800)]
ENGR00232000: Fix "dmaengine: failed to get dma1chan0: (-22)" when boot
The log from [MX6DL/S_SD]:
...
mxc_sdc_fb mxc_sdc_fb.1: register mxc display driver ldb
dmaengine: failed to get dma1chan0: (-22)
dmaengine: failed to get dma1chan1: (-22)
dmaengine: failed to get dma1chan2: (-22)
dmaengine: failed to get dma1chan3: (-22)
dmaengine: failed to get dma1chan4: (-22)
dmaengine: failed to get dma1chan5: (-22)
dmaengine: failed to get dma1chan6: (-22)
dmaengine: failed to get dma1chan7: (-22)
dmaengine: failed to get dma1chan8: (-22)
dmaengine: failed to get dma1chan9: (-22)
dmaengine: failed to get dma1chan10: (-22)
dmaengine: failed to get dma1chan11: (-22)
dmaengine: failed to get dma1chan12: (-22)
dmaengine: failed to get dma1chan13: (-22)
dmaengine: failed to get dma1chan14: (-22)
dmaengine: failed to get dma1chan15: (-22)
...
It happens when there are many DMA-engine drivers in the system and
dmaengine_get() is called. dmaengine_get() will call dma_chan_get(), which will
call device_alloc_chan_resources() literally on channels of available dma
drivers unless reach -ENODEV. device_alloc_chan_resources() is implemented
in the individual dma drivers, which could return -EINVAL rather than -ENODEV,
then the above messages print out (doesn't hurt, however).
Indeed, the dmaengine_get() and dmaengine_put() is not needed and thus removed.
Lionel Xu [Wed, 31 Oct 2012 08:40:37 +0000 (16:40 +0800)]
ENGR00231778 MX6 ESAI: Adjust channel support capability of cpu/codec dai
To adjust the channel support capability, codec dai does not support mono
playback and record, while esai dai does, thus making the whole audio
codec only support stereo and above channel playback/record.
When our DDR size is small or reserved memory are large and
the lowmem can cover all the available pages for kernel,
the highmem pages will not be setup. That means the page_pool
for bounce queue can not be create in init_emergency_pool().
And page_pool will stay NULL without initialized.
In the mmc/card/queue.c the blk_queue_bounce_limit()
function will be called in mmc_init_queue() to
initialize the request_queue and it's bounce_gfp.
If we do not define the DMA mask for our platform,
then the BLK_BOUNCE_HIGH (lowmem pfn) will be set
as limit to queue bounce, which means the blk_queue_bounce
will use page_pool to iterate over the bio segment.
Under the circumstances that highmem is not setup,
the page_pool is null, and causes kernel crash.
After set the DMA mask for esdhci device, the page_pool
will not be used to iterate over the bio segment.
Hongzhang Yang [Thu, 1 Nov 2012 11:26:44 +0000 (19:26 +0800)]
ENGR00232087-1 MX6: Enable PU LDO gating.
1. Revert ENGR00231910 Do not disable PU regulator,revert the PU
regulator patch;
2. VPU reset register address is different on MX6 and MX5. It can
fix ENGR00230203 [Android_MX6DL_SD] Gallery: System hang after resume
from suspend during video playback. 20%
Signed-off-by: Hongzhang Yang <Hongzhang.Yang@freescale.com>
for a big nand chip Micron MT29F64G08AFAAAWP(8GB), we got the following wrong
result:
.............................................
"mtd: partition size too small (0)"
.............................................
We can not get any partition.
The "4g(test)" partition triggers a overflow of the "size". The memparse()
returns 4g to the "size", but the size is "unsigned long" type, so a overflow
occurs, the "size" becomes zero in the end.
This patch changes the "size"/"offset" to "unsigned long long" type,
and replaces the UINT_MAX with ULLONG_MAX for macros SIZE_REMAINING and
OFFSET_CONTINUOUS.
Tony LIU [Thu, 1 Nov 2012 02:15:18 +0000 (10:15 +0800)]
ENGR00231965 MX6 USB CV 3.0 test fail
- For USB CV 3.0 test, the gap between the ACK of set_address and
the subsequent setup packet may be very little, say 500us, and
if the latency we handle the ep completion is greater than this
gap, there is no response to the subsequent packet. It will
cause CV test fail
- There is another way to set the address, it should set the bit 24
to 1 with the right address, and then IC controller will set the
address when the IN req complete instead of SW do it. It is more
fast so it can fix the CV 3.0 test fail issue
Signed-off-by: Tony LIU <junjie.liu@freescale.com>
Matthieu CASTET [Mon, 19 Mar 2012 14:35:25 +0000 (15:35 +0100)]
mtd: support ONFI multi lun NAND
With onfi a flash is organized into one or more logical units (LUNs).
A" logical unit (LUN) is the minimum unit that can independently execute
commands and report status.
Mtd does not exploit LUN, so make it see a big single flash where size is
lun_size * number_of_lun.
Without this patch MT29F8G08ADBDAH4 size is 512MiB instead of 1GiB.
Robin Gong [Wed, 31 Oct 2012 13:41:16 +0000 (21:41 +0800)]
ENGR00231910 PU regulator: do not disable PU regulator
If system enter suspend/resume during VPU encoding on Rigel, there will be
"VPU blocking: timeout." error . But there is ok if enter suspend/resume
during VPU decoding and enter suspend/resume during encoding/decoding on
Arik, until now we didn't know the root cause, so revert it firstly.
Because the previous patch about PU regulator is composed with four commits
and hard to revert, now we adopt simplest way that do not disable PU regulator
in low level. The negative impact is there will several mA increasment in
suspend, we will fix it ASAP. Signed-off-by: Robin Gong <b38343@freescale.com>
Alejandro Sierra [Wed, 24 Oct 2012 20:32:37 +0000 (15:32 -0500)]
ENGR00231063 Sabreauto: fix share pins SD Card and NFC
SD card card detection and NFC controller CS2 share
the same pin on ARD platform. However CS2 is not
connected to the socket. This signal was removed
from the sabreauto board file.
Signed-off-by: Alejandro Sierra <b18039@freescale.com>
Steve Cornelius [Fri, 19 Oct 2012 21:43:41 +0000 (14:43 -0700)]
ENGR00230538-5: CAAM: Add SM register defs
Add SM register defs, and expanded driver-private storage.
These add changes to the driver private areas for the CAAM
controller and CAAM Secure Memory subsystems, and expand register
definitions to include the Secure Memory subsystems as reflected
in multiple areas (controller, rings, secure memory itself).
Signed-off-by: Steve Cornelius <steve.cornelius@freescale.com> Signed-off-by: Terry Lv <r65388@freescale.com>
Steve Cornelius [Fri, 19 Oct 2012 20:27:24 +0000 (13:27 -0700)]
ENGR00230538-2: CAAM: Add Secure Memory and SNVS properties
Add Secure Memory and SNVS properties to MX6 configuration.
Previous configurations of MX6 platform device definition lacked
specific propeties for CAAM Secure Memory and SNVS. Added these
properties to define register ranges for both entities.
Also corrected the name for the offset of the address range for
CAAM Secure Memory to more accurately reflect it's purpose.
Signed-off-by: Steve Cornelius <steve.cornelius@freescale.com> Signed-off-by: Terry Lv <r65388@freescale.com>
Steve Cornelius [Fri, 19 Oct 2012 20:18:37 +0000 (13:18 -0700)]
ENGR00230538-1: CAAM: Correct shifting offset for CAAM IPG clock selection
3 pairs of clock enable bits are required for CAAM clocking:
(1) wrapper IPG clock
(2) wrapper ACLK
(3) secure memory clock
IPG enable happened to be using an incorrect shift selection, which
had the net effect of leaving secure memory unclocked. Added the correct
shift selection in so that all 3 clock enable pairs are turned on.
Signed-off-by: Steve Cornelius <steve.cornelius@freescale.com> Signed-off-by: Terry Lv <r65388@freescale.com>
Robin Gong [Fri, 26 Oct 2012 11:19:43 +0000 (19:19 +0800)]
ENGR00230981-3 pfuze: add suspend voltage set interface
Implement set_suspend_voltage for buck switch of PF100, and set_suspend_enable
/set_suspend_disable interface for LDO(VGENx). Signed-off-by: Robin Gong <b38343@freescale.com>
Robin Gong [Wed, 24 Oct 2012 10:28:59 +0000 (18:28 +0800)]
ENGR00230981-2 pfuze:standby voltage increase for PFM
There is 6% tolerance for PFM momde in standby so we need set 0.975V(>0.9V+%6)
for VDDSOC and VDDARM which maybe impact system resume ability.
Another two change is:
1.set VDDARM and VDDSOC standby voltage by setting PFUZE register directly,it
is not very friendly.So use more common "state_mem" in constrain of regulator
to set standby voltage.
2.align sabreauto code with sabresd Signed-off-by: Robin Gong <b38343@freescale.com>
Huang Shijie [Fri, 26 Oct 2012 08:04:03 +0000 (16:04 +0800)]
ENGR00231331 mtd: gpmi: add kernel command line to enable gpmi in arm2 board
In mx6q arm2 board, the gpmi conflicts with SD module.
But the defconfig has enabled the gpmi by default.
So we have to add a kernel cmdline to enable the gpmi by hand in arm2 board.
make shi [Thu, 18 Oct 2012 07:59:18 +0000 (15:59 +0800)]
ENGR00230167 MX6 regulator: enable and raise the voltage of USB 3p0 LDO
The USB FS eye test will fail in MX6 board if the 3V USB phy LDO is not enabled.
Setting enable bit (bit-0) of LDO 3p0 will make 3p0 LDO to use bandgap output as
reference voltage, LDO output will be accurate. And HW team suggest that it is
better to raise the voltage of USB 3p0 phy LDO 3.2V to pass the USB compliance
testing.
- Implement vdd3p0 regulator enable and disable function to support
enable and disable the LDO 3p0 regulator.
- Use regulator API to enable the USB 3p0 phy LDO and raise the LDO
to 3.2V during system boot up. And disable the LDO before system
enter suspend and enable the LDO again after system resume.
Wayne Zou [Wed, 24 Oct 2012 01:36:00 +0000 (09:36 +0800)]
ENGR00230910 IPU: wrong display to downsize large resolution frame on split mode
Fix bug: IPU IC resize ratio overflow when downsizing large resolution frame
using split mode, for example downsize 4080x2720 frame into 1920x1080 frame.
Otherwise, the downsized frame is wrong.
Chen Liangjun [Wed, 24 Oct 2012 04:31:59 +0000 (12:31 +0800)]
ENGR00230920-1 HDMI AUDIO: add support for HDMI audio module build
HDMI driver can be divided into 3 parts: machine driver, platform
driver, codec driver.To support HDMI AUDIO loadable module build, HDMI
machine driver should be built as loadable.
In this patch, adjust HDMI audio driver's struct: move HDMI audio platform
driver to snd-soc-imx-objs(snd-soc-imx-objs would always be
build-in).In this case, user need only build HDMI AUDIO machine driver
as loadable.
Robby Cai [Mon, 15 Oct 2012 13:36:44 +0000 (21:36 +0800)]
ENGR00229665 pxp: correct crop setting
The settings in the PXP_PS_BUF, PXP_OUT_PS_ULC, and PXP_OUT_PS_LRC will
determine the subset of the PS buffer, or clipped PS source buffer, that
will be used in the output buffer.
HW_PXP_OUT_PS_LRC should set the scaled output size rather than the origin
size when scaling.
Please refer to the "Clipping source images" section in RM for how it works.
Michael Minnick [Tue, 16 Oct 2012 23:07:20 +0000 (18:07 -0500)]
ENGR00227965 EPDC: Init sequence leaves EDPC clocks on
A small logic bug prevents the init sequence from properly turning
off the clocks. This leads to the clocks being always on
if the first update does not complete due to the screen being blanked.
Signed-off-by: Michael Minnick <michael.minnick@freescale.com>
Robby Cai [Thu, 11 Oct 2012 09:18:15 +0000 (17:18 +0800)]
ENGR00227890 ts: fix elan touch screen gets no response upon suspend/resume
To reproduce:
1. let system enter suspend mode
2. touch the screen
3. after the system resumes, touch screen does not respond again.
The cause:
The touch screen interrupt is triggered by falling edge. During suspend stage,
once the screen has ever been touched, then the interrupt line will be always
pulled low. Since elan ts chip is always powered on and the interrupt gets no
chance to be handled during suspend stage, the interrupt line can not recover
to high to detect a new one.
Workaround:
Read out the pending data to make the touch screen come back alive.
Signed-off-by: LiGang <b41990@freescale.com> Signed-off-by: Robby Cai <R63905@freescale.com>
Alejandro Sierra [Mon, 15 Oct 2012 23:13:47 +0000 (18:13 -0500)]
ENGR00229725 Sabreauto: Support NAND SPINOR NOR SD on same config
Configuration file modified to support NAND flash, SPI-NOR,
WEIM NOR and SD card on the same image.
Bootloader arguments will be used to choose between them.
Arguments on uboot are:
spi-nor
weim-nor
By default NAND is configured if neither spi-nor or weim-nor are used
Signed-off-by: Alejandro Sierra <b18039@freescale.com>
Nicolin Chen [Tue, 16 Oct 2012 08:20:03 +0000 (16:20 +0800)]
ENGR00225520 SDMA:fix kernel dump occasionally during I2C stress test
Stress test with I2C devices occasionally caused kernel dump and panic:
==========================dump=start==========================
v4l_capture_testapp 0 TINFO :
Color space conversion YUV420->RGB565X success!
v4l_capture_testapp 0 TINFO :
Color space conversion YUV420->RGB565X success!
clean up environment...VPU interrupt received.
Unable to handle kernel paging request at virtual address ffdf401a
pgd = ba2a4000
[ffdf401a] *pgd=4fe1a811, *pte=00000000, *ppte=00000000
Internal error: Oops: 7 [#1] PREEMPT SMP
Modules linked in: mxc_v4l2_capture ipu_still ipu_bg_overlay_sdc
ipu_prp_enc ipu_fg_overlay_sdc ipu_csi_enc ov5642_camera
camera_sensor_clock [last unloaded: ipu_csi_enc]
CPU: 0 Not tainted (3.0.35-2039-g267e004 #1)
PC is at sdma_int_handler+0x144/0x1a4
LR is at sdma_int_handler+0x70/0x1a4
pc : [<802663f4>] lr : [<80266320>] psr: 60000193
sp : ba3e7ca8 ip : bfee2100 fp : 00000001
r10: 80a67200 r9 : 80acbcf0 r8 : 00000003
r7 : 00000001 r6 : 00000001 r5 : 00000002 r4 : bfee20e0
r3 : ffdf4000 r2 : 00010104 r1 : ffdf4018 r0 : bfee2104
Flags: nZCv IRQs off FIQs on Mode SVC_32 ISA ARM Segment user
Control: 10c53c7d Table: 4a2a404a DAC: 00000015
Process mxc_vpu_test.ou (pid: 3277, stack limit = 0xba3e62f0)
Stack: (0xba3e7ca8 to 0xba3e8000)
7ca0: 80038f40bfee2000002977e3bf9cda80 80a6724c00000000
7cc0: 000000000000002280acbcf080a6720000000001800a5cb8 0000f08f00000000
[<802663f4>] (sdma_int_handler+0x144/0x1a4)
from [<800a5cb8>] (handle_irq_event_percpu+0x50/0x180)
[<800a5cb8>] (handle_irq_event_percpu+0x50/0x180)
from [<800a5e24>] (handle_irq_event+0x3c/0x5c)
[<800a5e24>] (handle_irq_event+0x3c/0x5c)
from [<800a81a8>] (handle_fasteoi_irq+0xbc/0x154)
[<800a81a8>] (handle_fasteoi_irq+0xbc/0x154)
from [<800a5620>] (generic_handle_irq+0x28/0x3c)
[<800a5620>] (generic_handle_irq+0x28/0x3c)
from [<80040830>] (handle_IRQ+0x4c/0xac)
[<80040830>] (handle_IRQ+0x4c/0xac)
from [<8003f9cc>] (__irq_svc+0x4c/0xe8)
[<8003f9cc>] (__irq_svc+0x4c/0xe8)
from [<800764f4>] (__do_softirq+0x4c/0x140)
[<800764f4>] (__do_softirq+0x4c/0x140)
from [<80076a90>] (irq_exit+0x94/0x9c)
[<80076a90>] (irq_exit+0x94/0x9c)
from [<8003a1b4>] (do_local_timer+0x70/0x90)
[<8003a1b4>] (do_local_timer+0x70/0x90)
from [<8003f9cc>] (__irq_svc+0x4c/0xe8)
Exception stack(0xba3e7de8 to 0xba3e7e30)
[<8003f9cc>] (__irq_svc+0x4c/0xe8)
from [<80071a88>] (vprintk+0x328/0x4a8)
[<80071a88>] (vprintk+0x328/0x4a8)
from [<804ddb28>] (printk+0x1c/0x2c)
[<804ddb28>] (printk+0x1c/0x2c)
from [<80390de0>] (vpu_ioctl+0x2cc/0x864)
[<80390de0>] (vpu_ioctl+0x2cc/0x864)
from [<800fc314>] (do_vfs_ioctl+0x80/0x54c)
[<800fc314>] (do_vfs_ioctl+0x80/0x54c)
from [<800fc818>] (sys_ioctl+0x38/0x5c)
[<800fc818>] (sys_ioctl+0x38/0x5c)
from [<8003ff80>] (ret_fast_syscall+0x0/0x30)
Code: e594101ce5943038e0811081e0831101 (e5d13002)
---[ end trace 82daf36a5a07d470 ]---
Kernel panic - not syncing: Fatal exception in interrupt
Rebooting in 60 seconds..
==========================dump=end==========================
This kernel dump only happened after one period of stress-test's done.
From the dump info above, we just located the issue happened in SDMA driver.
Regularly, it'd not be any problem when sdma_int_handler()'s called. But after
tracing, we found that in those occasional times, the last one irq of a channel
hadn't been responded while sdma_free_chan_resources() was already done.
sdma_free_chan_resources() should be called in the end of the procedure. Any
irq wouldn't occur after its resources're freed.
But considering about stress test, the test scripts uses "kill" cmd to close
aplay, which means pcm_free() might be called before last buffer's transmission
was finished. Plus, many modules're working in the same time during the test.
So CPU0, the only core can handle irq, would be busy with irq-handlings, while
the other CPU cores(i.e. CPU1~3) might be idle and deal with free() much faster
than CPU0's irq-handling. Then kernel panic.
Since we know, in some extreme circumstances, the irq would not be handled in
time, we can manually handle the irq ONLY IF we could still detect one irq to
the channel in the beginning of free(), right before its resources's gonna be
freed.
This Patch added checking code in the beginning of sdma_free_chan_resources()
to detect when the channel's gonna be freed if there's still one irq pended.
If so, just handle the irq manually before we free it.
Again, considering about sdma_int_handler() might be running at the same time,
and if it already cleared the value of reg but hadn't handled the irq yet, also
added code to pend free() until irq to the channel was handled.
Sheng Nan [Wed, 17 Oct 2012 06:55:10 +0000 (14:55 +0800)]
ENGR00229962 Capture: ov5642/ov5640: update sensor params even if s_parm failed
ioctl_s_parm for ov5642 and ov5640, it didn't check if sensor changed mode
successfully.
So it updates the sensor parameters with new framerate and new mode even
if the sensor failed to change mode.
The original framerate and mode is useful for the exposure calculation.
It should keep consistent with sensor actual work mode.
- This patch checks the return value of function which changes sensor mode
If it succeed, update sensor parameters.
ENGR00229695 MX6x-Set RBC counters correctly in STOP mode.
The REG_BYPASS_COUNTER(RBC) holds off interrupts when the PGC
block is sending signals to power gate the core. This is apart
from the RBC counter's basic functionality to act as counter to
power down the analog portions of the chip.
But the counter needs to be set/cleared only when no interrupts
are pending. And also for correct hold off the interrupts, enable the
counter as close to WFI as possible.
The RBC counts CKIL cycles (32KHz)
So follow the following steps to set the counter
in suspend/resume in mx6_suspend.S:
1. Mask all the GPC interrupts.
2. Write the counter value to the RBC
3. Enable the RBC
4. Unmask all the interrupts.
5. Busy wait for a few usecs to wait for RBC to start counting
in case an interrupt is pending.
4. Execute WFI
Reset the counter after resume in pm.c:
1. Mask all the GPC interrupts.
2. Disable the counter.
3. Set the RBC counter to 0.
4. Wait for 80usec for the write to get accepted.
5. Unmask all the interrupts.
With the above steps, we can minimize the PDNSCR and PUPSCR counters
in the GPC. The basic condition for the RBC counter:
RBC count >= 25 * IPG_CLK + PDNSCR_SW2ISO.
PDNSCR_SW2ISO = PDNSCR_ISO = 1 (counts in IPG_CLK)
PUPSCR_SW2ISO = PUPSCR_ISO = 2 (counts in 32K)
Sheng Nan [Mon, 15 Oct 2012 09:49:27 +0000 (17:49 +0800)]
ENGR00224964-1 Capture: ov5642: 5M mode works at low frame rate
current setting of 5M (QSXGA) mode, sensor works at 2.5fps.
the expected frame rate is 7.5fps.
- use new ov5642 QSXGA firmware get from ov
change sensor PLL settings 0x3010/0x3012
QSXGA frame rate changes from 2.5 -- 7.5fps
- change mode between QSXGA@15fps and VGA@15fps go through quick change path.
modify QSXGA_VGA quick change firmware due to the QSXGA PLL setting changes.
keep value of 0x3010/0x3012 the same as VGA@15fps original value.
Anson Huang [Mon, 15 Oct 2012 23:01:17 +0000 (19:01 -0400)]
ENGR00229630 vpu: need to manage pu regulator in suspend/resume
If VPU is working before suspend, we need to disable its regulator
to make sure regulator can be off before suspend, then enable
its regulator before resume to work, we check vpu's open_count
to determine whether to disable/enable its regulator.
Michael Minnick [Fri, 12 Oct 2012 18:52:36 +0000 (13:52 -0500)]
ENGR00229291 EPDC: MX6: Treat fully-collided VOID update as a collision
The EPDC set the UPD_VOID (i.e. cancelled) bit in two cases:
1. No pixels needed updating
2. All pixels collided (COL bit also set)
The driver was miss-handling case 2. This fix causes case 2
to be treated as a collision and the update to be resubmitted.
Signed-off-by: Michael Minnick <michael.minnick@freescale.com>
ENGR00229441 MX6SL-Reset MMDC read FIFO in low power IDLE
MMDC can clock in bad data due to the glitches caused by
changing the setting of various DDR IO pads in low power
IDLE to save power. Solution is to reset the MMDC read FIFO
before the DDR exits self-refresh.
Jack Lee [Wed, 3 Oct 2012 05:31:47 +0000 (13:31 +0800)]
ENGR00223348 EPDC: Unable to enable DISPLAY regulator
In the maxim 17135 driver, the power good is confirmed by the
power good GPIO polarity change when comparing the status at
the beginning of driver probe and display regulator enabled.
However, it is not reliable since the initial value of the GPIO
is not constant. Normally, it is 1 but it can be 0 after system reset
unexpectedly. Now, it is changed to POK bit checking in FAULT register.
ENGR00229470-2 MX6SL-Add support for debug UART to be sourced from 24MHz.
If "debug_uart" is specified in the command line, uart will
be sourced from 24MHz XTAL. This is required for getting the
correct power measurements on MX6SL.
Certain analog power optimizations are done only if ALL PLLs
are bypassed on MX6SL. To verify this path, we need to ensure
that UART is not sourced from PLL3.
ENGR00229470-1 MX6SL-Add support for debug UART to be sourced from 24MHz.
If "debug_uart" is specified in the command line, uart will
be sourced from 24MHz XTAL. This is required for getting the
correct power measurements on MX6SL.
Certain analog power optimizations are done only if ALL PLLs
are bypassed on MX6SL. To verify this path, we need to ensure
that UART is not sourced from PLL3.
Sheng Nan [Sat, 13 Oct 2012 06:33:13 +0000 (14:33 +0800)]
ENGR00211376 Capture: ov5640_mipi: the QVGA is brighter
change ov5640_init_mode sequence according to ov's suggestion
ov5640 support two method of size switching, scaling and subsampling
exposure calculation when change size between scaling and subsampling
- scaling: image size bigger than 1280*960
- subsampling: image size smaller than 1280*960
This patch changes the sequence of ov5640_init_mode()
1. setting mipi csi2 (no change).
2. check mode
- if it is in INIT_MODE, go throught initial procedure
- if sensor changes between scaling and subsampling,
go through exposure calcualtion
- otherwise, configure mode directly.
3. other procedures keep the same.
ENGR00229464 MX6SL-Update the SOC voltages based on datasheet
Update the VDDARM and VDDSOC voltages based on IMX6SLCEC_Rev0
datasheet.
As the voltages for ARM @ 198MHz and ARM @ 396MHz are the same
remove the 198MHz working point.
Michael Minnick [Tue, 9 Oct 2012 23:06:31 +0000 (18:06 -0500)]
ENGR00229290 EPDC: MX6: Adjust number of LUTs for 5-bit waveform
When a 5-bit waveform is loaded, the maximum number of available
LUTs is 16. The LUT allocator must account for this.
Note that 5-bit waveform loading is currently not supported in the
driver. However, this fix makes sure the LUT allocator is correct
when 5-bit support is added.
Signed-off-by: Michael Minnick <michael.minnick@freescale.com>
Ryan QIAN [Mon, 8 Oct 2012 23:44:34 +0000 (07:44 +0800)]
ENGR00227420 mmc: sdhci: shorten the delay on disabling clk
- change the delay from 10 * HZ to 1 * HZ, though (1 * HZ) might not be the
best, just as a base.
1. Since the request has been moved out of interrupt context, there will be no
more calling enable_clk in interrupt context. So it's not necessary to keep
such a long delay on disabling clock in order to save power.
2. Still keeping the 1*HZ of delay is to avoid frequently enabling/disabling
clock.
make shi [Fri, 12 Oct 2012 09:01:33 +0000 (17:01 +0800)]
ENGR00229354 Mx6 USB device: fix wrong handle for invalid USB_FEATURE requests
There is a bug udc driver handle invalid USB_FEATURE requests in current bsp.
The invalid USB_FEATURE request will be handled as a valid USB_FEATURE request.
We should set protocol stall on ep0 to handle invalid USB_FEATURE requests.
Terry Lv [Fri, 12 Oct 2012 06:39:34 +0000 (14:39 +0800)]
ENGR00229327: otp: driver causes warning when enable CONFIG_LOCKDEP
When CONFIG_LOCKDEP is enabled, it will cause warings:
------------[ cut here ]------------
WARNING: at kernel/lockdep.c:2885 sysfs_add_file_mode+0x54/0xc0()
Modules linked in:
[<80046364>] (unwind_backtrace+0x0/0xfc) from [<800758c0>]
(warn_slowpath_common+0x4c/0x64)
[<800758c0>] (warn_slowpath_common+0x4c/0x64) from [<800758f4>]
(warn_slowpath_null+0x1c/0x24)
[<800758f4>] (warn_slowpath_null+0x1c/0x24) from [<801536c4>]
(sysfs_add_file_mode+0x54/0xc0)
[<801536c4>] (sysfs_add_file_mode+0x54/0xc0) from [<8015616c>]
(internal_create_group+0xdc/0x1d8)
[<8015616c>] (internal_create_group+0xdc/0x1d8) from [<80524110>]
(fsl_otp_probe+0x168/0x1d4)
[<80524110>] (fsl_otp_probe+0x168/0x1d4) from [<802b42e8>]
(platform_drv_probe+0x18/0x1c)
[<802b42e8>] (platform_drv_probe+0x18/0x1c) from [<802b2fe4>]
(driver_probe_device+0x98/0x1a4)
[<802b2fe4>] (driver_probe_device+0x98/0x1a4) from [<802b3184>]
(__driver_attach+0x94/0x98)
[<802b3184>] (__driver_attach+0x94/0x98) from [<802b280c>]
(bus_for_each_dev+0x60/0x8c)
[<802b280c>] (bus_for_each_dev+0x60/0x8c) from [<802b2180>]
(bus_add_driver+0x190/0x268)
[<802b2180>] (bus_add_driver+0x190/0x268) from [<802b3788>]
(driver_register+0x78/0x13c)
[<802b3788>] (driver_register+0x78/0x13c) from [<800394ac>]
(do_one_initcall+0x30/0x170)
[<800394ac>] (do_one_initcall+0x30/0x170) from [<800083cc>]
(kernel_init+0x98/0x144)
[<800083cc>] (kernel_init+0x98/0x144) from [<8004003c>]
(kernel_thread_exit+0x0/0x8)
---[ end trace 877415a10b5d9cb1 ]---
also, on imx6sl, it will cause below issue:
BUG: key bffea2e4 not in .data!
BUG: key bffea300 not in .data!
BUG: key bffea31c not in .data!
BUG: key bffea338 not in .data!
BUG: key bffea354 not in .data!
BUG: key bffea370 not in .data!
BUG: key bffea38c not in .data!
BUG: key bffea3a8 not in .data!
BUG: key bffea3c4 not in .data!
BUG: key bffea3e0 not in .data!
BUG: key bffea3fc not in .data!
BUG: key bffea418 not in .data!
BUG: key bffea434 not in .data!
BUG: key bffea450 not in .data!
BUG: key bffea46c not in .data!
BUG: key bffea488 not in .data!
BUG: key bffea4a4 not in .data!
BUG: key bffea4c0 not in .data!
BUG: key bffea4dc not in .data!
We need to call sysfs_attr_init to initlize sysfs attr.
Huang Shijie [Thu, 11 Oct 2012 01:50:13 +0000 (09:50 +0800)]
ENGR00227835 imx6q: gpmi: fix the warning when no NAND chip exits
If there is no nand chip in the board, the kernel will prints out the
following warning message:
------------[ cut here ]------------
WARNING: at arch/arm/plat-mxc/clock.c:63 clk_disable+0x48/0x90()
clock enable/disable mismatch! clk apbh_dma_clk
Modules linked in:
[<80044f48>] (unwind_backtrace+0x0/0xfc) from
[<80070ac0>] (warn_slowpath_common+0x4c/0x64)
[<80070ac0>] (warn_slowpath_common+0x4c/0x64) from
[<80070b6c>] (warn_slowpath_fmt+0x30/0x40)
[<80070b6c>] (warn_slowpath_fmt+0x30/0x40) from
[<8005ee60>] (clk_disable+0x48/0x90)
[<8005ee60>] (clk_disable+0x48/0x90) from
[<80255e48>] (dma_chan_put+0x4c/0x50)
[<80255e48>] (dma_chan_put+0x4c/0x50) from
[<80255f18>] (dma_release_channel+0x24/0x94)
[<80255f18>] (dma_release_channel+0x24/0x94) from
[<802ad8ec>] (release_resources+0x58/0x6c)
[<802ad8ec>] (release_resources+0x58/0x6c) from
[<80445964>] (gpmi_nand_probe+0x44c/0x4ec)
[<80445964>] (gpmi_nand_probe+0x44c/0x4ec) from
[<80281868>] (platform_drv_probe+0x18/0x1c)
[<80281868>] (platform_drv_probe+0x18/0x1c) from
[<80280590>] (driver_probe_device+0x98/0x1a4)
[<80280590>] (driver_probe_device+0x98/0x1a4) from
[<80280728>] (__driver_attach+0x8c/0x90)
[<80280728>] (__driver_attach+0x8c/0x90) from
[<8027fdd0>] (bus_for_each_dev+0x60/0x8c)
[<8027fdd0>] (bus_for_each_dev+0x60/0x8c) from
[<8027f75c>] (bus_add_driver+0x184/0x25c)
[<8027f75c>] (bus_add_driver+0x184/0x25c) from
[<80280d1c>] (driver_register+0x78/0x13c)
[<80280d1c>] (driver_register+0x78/0x13c) from
[<80022d80>] (gpmi_nand_init+0xc/0x3c)
[<80022d80>] (gpmi_nand_init+0xc/0x3c) from
[<80039478>] (do_one_initcall+0x30/0x16c)
[<80039478>] (do_one_initcall+0x30/0x16c) from
[<80008410>] (kernel_init+0x98/0x144)
[<80008410>] (kernel_init+0x98/0x144) from
[<8003ffb4>] (kernel_thread_exit+0x0/0x8)
---[ end trace c28d32057fe33a29 ]---
This mxs_dma_clk's usecount is not correctly changed which causes the kernel
shows this warning.
This patch adds proper clk_disable_unprepare/clk_prepare_enable in
the mxs-dma driver to balance the mxs_dma_clk's usecount.
Also put the mxs_dma_clk when the gpmi exits.
Liu Ying [Wed, 10 Oct 2012 07:13:30 +0000 (15:13 +0800)]
ENGR00227681 IPUv3:Use spinlock to protect buf ready reg
There are several channels' buffer ready bits controlled
by a single 32bit register-IPU_CHA_BUFx_RDY. These buffer
ready can be write-one-to-set or write-one-to-clear, which
is controlled by IPU_GPR register. v4l2 capture driver will
touch the buffer ready registers in interrupt context, so,
currently, ipu->mutex_lock is bypassed with the context.
Then, a race condition is that v4l2 capture driver interrupt
context tries to set a channel buffer ready, while, another
context tries to disable another channel(clear buffer ready
bit), which may cause v4l2 capture driver fails to set buffer
ready(SMFC0_FRM_LOST error may happen). This patch uses ipu->
rdy_reg_spin_lock to protect buffer ready registers to fix
the race condition issue and rename ipu->spin_lock to ipu->
int_reg_spin_lock.
1) We get busy_lock semaphore before we get a dqueue event, so, when user
is blocked at DQBUF ioctrl, the user will also be blocked at QBUF ioctrl,
then the video performance will drop. This patch changes to get busy_lock
semaphore to protect DQBUF ioctrl until we successfully get a dqueue event.
2) Use queue_int_lock and dqueue_int_lock spinlocks to protect working_q/
ready_q/done_q in the end of frame interrupt handler camera_callback(), in
case, the handler and VIDIOC_QBUF/VIDIOC_DQBUF ioctrls are called on diff-
erent cores at the same time.
3) Protect ready_q with queue_int_lock spinlock in mxc_streamon(), in case,
VIDIOC_STREAMON and VIDIOC_QBUF ioctrls are called on different cores at
the same time.
1) Change to enable/disable mclk only in open, release,
suspend and resume functions, since we may simply think that
sensor or mclk will be used soon after cam->open_count is non-zero.
2) Fix a bug when calling ipu_csi_enable_mclk_if() with wrong
parameter(cam->csi should be cam->mclk_source) in mxc_v4l2_close()
and in mxc_v4l2_s_ctrl() with V4L2_CID_MXC_SWITCH_CAM control id.
Robby Cai [Tue, 9 Oct 2012 11:59:25 +0000 (19:59 +0800)]
ENGR00227568 elcdif: fix fb wait for vsync timeout when suspend and resume
When suspend, the lcdif and panel will be stopped. When resume, fb_set_par()
will be called, in which the lcdif and the panel will be re-initialized.
However, fb_set_par() also checks the parameters via mxc_elcdif_fb_par_equal(),
which will probably make fb_set_par() just return with them un-initialized.
And thus, the interrupt will not come. This patch added a varible to check
whether they're running along with mxc_elcdif_fb_par_equal() checking to
fix the issue. If not running, re-initialization will be forcely done.
Robby Cai [Tue, 9 Oct 2012 08:36:41 +0000 (16:36 +0800)]
ENGR00227502-1 csi/v4l2: Implement probe and remove function for csi v4l2 driver
Because csi_v4l2 driver will only be loaded when needed(by assign 'csi' in
kernel cmdline), we use standard driver framework to easily bind the device and
driver. Otherwise, we will meet the problem like the crash as below when do
suspend/resume due to the resource of csi not assigned at all if 'csi' not
passed in cmdline.
projects / karo-tx-linux.git / log