Sandor Yu [Fri, 11 Jan 2013 07:28:51 +0000 (15:28 +0800)]
ENGR00239734 Mx6 HDMI PHY: Add 2 variable to pass board specific config
The PHY register 0x9 and 0xe should setting
to different value in different board to pass HCT.
Add variable phy_reg_vlev and phy_reg_cksymtx to pass
phy config data.
Huang Shijie [Thu, 17 Jan 2013 08:15:45 +0000 (16:15 +0800)]
ENGR00240571 mtd: gpmi: fix the compiler warning
The current code may print out the following warning:
....................................................................
drivers/mtd/nand/gpmi-nand/gpmi-lib.c: In function gpmi_begin:
drivers/mtd/nand/gpmi-nand/gpmi-lib.c:1163: warning: hw.use_half_periods
may be used uninitialized in this function
drivers/mtd/nand/gpmi-nand/gpmi-lib.c:1163: warning: hw.sample_delay_factor
may be used uninitialized in this function
....................................................................
When disabling IPU channels, it needs less than 200us to wait for stop
Using msleep, it often sleep longer(above 10ms). So the extra delay
decrease the performance.
For 720p video playback on 1080p display(60Hz), the performance is about 40fps
With this patch, it can achieve around 60fps.
Richard Zhu [Mon, 7 Jan 2013 05:26:03 +0000 (13:26 +0800)]
ENGR00239905 PCIe Enable PCIe switch support
PCIe switch access mechanism:
- CfgRd0/CfgWr0 is used to access the CFG space of the EP device
or the upstream port of PCIe switch that is connected to RC directly.
- CfgRd1/CfgWr1 is used to access the CFG space of the downstream port
of PCIe switch and so on cases.
UR and kernel crash problem:
i.MX6 PCIe maps UR(Unsupported Request)err to AXI SLVERR err, which would
cause the arm data abort exception.
There is one "Received Master Abort" in iMX6 Root complex Secondary
status register when a requester receives a Completion
with Unsupported Request Completion Status.
In this case, the Linux kernel would be crashed.
Workaround: correct this imprecise external abort.
make shi [Thu, 10 Jan 2013 05:16:24 +0000 (13:16 +0800)]
ENGR00232755 USB: disable clock and abnormal wakeup when remove gadget driver
- In current bsp, the usb clock mismatch when rmmod gadget class driver. The
clock should be turn off when gadget class driver unregister.
- There is an abnormal usb wakeup interrupt happen if phy is no power without
VBUS. If we unplug the usb cable after unregister usb gadget driver, it is
difficult to handle the unexpected usb wakeup interrupt. SO we must call
dr_discharge_line() to make sure no abnormal usb wakeup interrupt happen in
usb unregister gadget class driver.
Sheng Nan [Tue, 8 Jan 2013 09:32:25 +0000 (17:32 +0800)]
ENGR00237682-3: mxc_v4l2_capture: ov5640: support scaling modes
The method for change between scaling and subsampling mode is different
from ov5640_mipi.
(image bigger than 1280*960 is scaling mode, smaller is subsampling).
According to OV5640 Auto Focus Camera Module Application Notes
(with DVP Interface) R2.14.pdf,
change back from QSXGA to VGA, don't need to do exposure calculation.
According to the test result, if we do exposure calculation when change
back from scaling to subsampling mode, the image would be dark.
So the method is:
Change to scaling mode, go through exposure calcuation.
Change to or back to subsampling mode, change mode directly.
Supported mode:
- QSXGA@7.5fps
- 1080P@7.5fps
Can't make 1080P works at 15fps. Here is a reply from ov fae:
because of scaling down, max frame for 1080P is the same as 5M, both are 15fps.
so if 5M can runs up to 7.5fps on your demo, then 1080P is the same 7.5fps max.
Sheng Nan [Wed, 19 Dec 2012 10:39:46 +0000 (18:39 +0800)]
ENGR00237682-2: mxc_v4l2_capture: ov5640: support all subsampling modes
Supported the following modes, verified image quality and frame rate
- VGA 30/15fps
- QVGA 30/15fps
- NTSC 30/15fps
- PAL 30/15fps
- 720P 30/15fps
Note: use the same setting as app note of ov5640 dvp
- QCIF 30/15fps
- XGA 22.5/15fps
Note: cannot make XGA work on 30fps. Just a reference of ov5640 datasheet:
1280*960 YUV422 maximum at 22.5fps.
1280*720 YUV422 maximum at 30fps.
Need to confirm later.
Sheng Nan [Wed, 19 Dec 2012 07:58:58 +0000 (15:58 +0800)]
ENGR00237682-1: mxc_v4l2_capture: ov5640: use global initialization
The current code struct of parallel ov5640 set mode directly.
The newest settings need to go through global initialization.
New settings are provided by ov company
So this patch does:
- Make parallel ov5640 mode settings go through global initialization.
- Only VGA (640 * 480) are provided as a validation of the new setting.
- Other modes will be provided in the later patches.
* Fix adv7180 tvin powerdown function
gpio power pin already exported in io-mux setup function
no need to request/free gpio
* Update copyrigth year 2013.
Signed-off-by: Adrian Alonso <aalonso@freescale.com>
Jason Liu [Tue, 8 Jan 2013 08:13:07 +0000 (16:13 +0800)]
fix echo 1 > compact_memory return error issue
when run the folloing command under shell, it will return error
sh/$ echo 1 > /proc/sys/vm/compact_memory
sh/$ sh: write error: Bad address
After strace, I found the following log:
...
write(1, "1\n", 2) = 3
write(1, "", 4294967295) = -1 EFAULT (Bad address)
write(2, "echo: write error: Bad address\n", 31echo: write error: Bad address
) = 31
This tells system return 3(COMPACT_COMPLETE) after write data to compact_memory.
The fix is to make the system just return 0 instead 3(COMPACT_COMPLETE) from
sysctl_compaction_handler after compaction_nodes finished.
Signed-off-by: Jason Liu <r64343@freescale.com> Suggested-by: David Rientjes <rientjes@google.com> Cc: Mel Gorman <mgorman@suse.de> Cc: Rik van Riel <riel@redhat.com> Cc: Minchan Kim <minchan@kernel.org> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Peter Chan [Tue, 8 Jan 2013 08:04:41 +0000 (16:04 +0800)]
ENGR00232879 mx6sl: EPDC VDDH and VPOS power on/off sequence is wrong
VDDH should only be ON after VPOS when power up and should be off
before VPOS when power down. Set the appropriate MAX17135 timing
parameters for the correct power up/down sequence
Robby Cai [Wed, 26 Dec 2012 07:26:39 +0000 (15:26 +0800)]
ENGR00238237-1 mx6sl: csi/v4l: fix camera picture flickering issue
Flickering issue happens when there's no buffer to be processed(e.g., the
pace of QBUF is much slower than DQBUF). The cause is the hardware is using
double buffering, while the driver has no good protection at above case and
thus the CSI will fill the buffer not in the right order. The way to fix is
refining the output of the working_q buffer list, that is, if there's no buffer
to be processed then output to a dummy buffer.
Another important change is to only do DMA reflash operation when SOF is
detected in streamon. Remove this operation is CSI interrupt handler because
it violates to the SPEC (only do DMA reflash before DMA is enabled but NOT at
the time or after DMA's enabled).
Signed-off-by: LiGang <b41990@freescale.com> Signed-off-by: Robby Cai <R63905@freescale.com>
(cherry picked from commit 0c4584763fa44b01a2f48198fa27c9206a116164)
Lionel Xu [Wed, 26 Dec 2012 05:54:49 +0000 (13:54 +0800)]
ENGR00238281 MX6SL_EVK: Add rfkill interface to bluetooth
MX6SL EVK board uses Silex SX-SDMAN board for bluetooth.
Add rfkill interface to control SX-SDMAN reset.
The reset signal is required before using bluetooth.
Gary Zhang [Fri, 4 Jan 2013 02:09:00 +0000 (10:09 +0800)]
ENGR00238809-2 mx6sl: ssi: add IRAM support
locate SSI playback buffer into IRAM in mx6sl.
because left IRAM room is not enough to contain record
buffer, if IRAM allocation for record fails, record
buffer will use external ram
Chen Liangjun [Fri, 28 Dec 2012 08:42:42 +0000 (16:42 +0800)]
ENGR00238439 ASRC: add delay before reading ASRC FIFO status
ASRC driver would read the sample number of ASRC output FIFO to fetch
the data from ASRC output FIFO. However, SDMA's fetching operation may
not finished before ASRC's reading. In this case, ASRC driver may read a
error data from the register.
In this patch, add delay before reading ASRC FIFO status to prevent
noise.
guoyin.chen [Fri, 7 Dec 2012 02:39:57 +0000 (10:39 +0800)]
ENGR00236141 csi:Add stride alignment setting from userspace
In android, IPU fills the I420 buffer. And GPU shows the buffer to display.
mx6's GPU has 32 Y-stride alignment for I420. The stride alignment will
be passed through by bytesperline. This update is only for
csi->smfc->mem channel.
Robin Gong [Wed, 19 Dec 2012 10:48:45 +0000 (18:48 +0800)]
ENGR00237742 busfreq:fix IPG_PERCLK will be decreased to 6M once exit low bus
on Sabresd board, IPG_PERCLK will be fixed on 6Mhz once system enter low bus,
and never restore to 22Mhz which be set in boot. It means some device clock
which sourcing from IPG_PERCLK such as I2C will be slow down. The root cause is
that there is workaround for GPT timer of Arik TO1.0 in mx6_ddr_freq.S. GPT
clock source from IPG_PERCLK on TO1.0 and should be fixed on 6Mhz. But for
TO1.1 and TO1.2 ,the workaround should be removed. Signed-off-by: Robin Gong <B38343@freescale.com>
Liu Ying [Wed, 19 Dec 2012 06:46:06 +0000 (14:46 +0800)]
ENGR00237678 IPUv3:Clean up sync and error interrupt
1) Split sync and error interrupt handler into 2 different
handlers, so that we may save several CPU cycles to
handler sync interrupt which is triggered in most IPU
usecases.
2) So far, every IPUv3 version in different SoCs has sync
and error interrupts, so we return -ENODEV if they cannot
be gotten from platform resource. Then, we may request
them directly but do not check more.
3) Clean up free irq code by removing unnecessary check.
Fugang Duan [Fri, 7 Dec 2012 10:48:34 +0000 (18:48 +0800)]
ENGR00236240 i2c: i2c performance optimization
It is unnecessary to calculate and update i2c divider during
every transaction. Only do it if current i2c clock is different
with the previous clock frequency.
Hongzhang Yang [Thu, 13 Dec 2012 08:44:16 +0000 (16:44 +0800)]
ENGR00236879 Enhance VPU driver to handle API call sequence abnormal abort
Some application may exit without calling neccessay API to wrap up VPU
after it receives error message.
This could lead to system hang because driver will power off VPU
(vpu_release) while VPU may still be busy.
We require application to strictly follow the API call sequence even in
error handling case. Meanwhile, we enhance VPU driver to protect against
such abnormal abort, to prevent system hang at least.
If the last instance is closed, VPU will gate off or power off only if
VPU is idle.
Signed-off-by: Hongzhang Yang <Hongzhang.Yang@freescale.com>
LiGang [Wed, 14 Nov 2012 02:12:53 +0000 (10:12 +0800)]
ENGR00231808: Add epdc pmic shut down feature
For some cases, system maybe restart with epdc pmic on.
If epdc pmic on for sometime(2~3s), the current epdc pmic consume will increase
obviously, then the total current of board maybe exceed to the current limit,
which will pull down the input voltage to lead to the system pmic reset,
an un-expected POR reset occurs.
So it is necessary to close epdc pmic before system restart.
This patch is added by Robby Cai<r63905@freescale.com>
Robby Cai [Mon, 10 Dec 2012 09:13:45 +0000 (17:13 +0800)]
ENGR00236722 mx6sl: csi: Ensure dma reflash operation done when dma is disabled
If do dma reflash operation when dma is enabled, the system will hang and we
can not connect to the core through jtag. The reason is the reflash signal
(DMA_REFLASH_RFF) will initialize the AHB bus signals and it indeed seems to
modify the AHB address on the clock as soon as the programmable register value
is changed, the bus may not respond.
This patch revised it according to the RM:
"Reflash DMA Controller for RxFIFO. This bit reflash the embedded DMA controller
for RxFIFO. It should be reflashed before the embedded DMA controller starts
to work."
Robby Cai [Thu, 29 Nov 2012 09:58:43 +0000 (17:58 +0800)]
ENGR00229855 mx6sl: csi: can not support two camera instances
run two unit test instances as follows fails (sometimes kernel dump).
/unit_tests/csi_v4l2_capture.out &
/unit_tests/csi_v4l2_capture.out
Fix by improving the resource lock.
- 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.
- Use queue_int_lock spinlock to protect cam->ping_pong_csi, since it can be
modified either in irq handler or in queue event.
- linked list should be protected by the lock:
-- 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
different threads at same time.
-- Protect ready_q with queue_int_lock spinlock in streamon(), in case,
VIDIOC_STREAMON and VIDIOC_QBUF ioctrls are called on different threads
at same time.
ENGR00236837 MX6SL-Fix random crash caused by incorrect setting of IPG clk rate.
Need to ensure that bus frequency setpoint is changed only if
the system is not already at the requested setpoint.
Changing the bus freq to high setpoint when its already at
high setpoint causes the AHB/IPG dividers to be set incorrectly.
Then when the system enters WAIT mode, the 12:5 ratio of
ARM_CLK:IPG_CLK is no longer maintained.
This causes random crashes.
Fix is to return immediately if the bus is already at the
requested setpoint
make shi [Mon, 10 Dec 2012 08:40:05 +0000 (16:40 +0800)]
ENGR00236169 MX6 USB :kfree udc_controller when remove udc driver
Kree and reset udc_controller should be done when remove udc driver to avoid
kernel dump during modprobe gadget driver after modprobe and rmmod udc driver.
Gary Zhang [Tue, 11 Dec 2012 04:42:42 +0000 (12:42 +0800)]
ENGR00236020-1 ALSA: add calling of trigger in machine driver
soc_pcm_trigger() calls trigger functions of cpu_dai, codec_dai
and platform, but the trigger function of machine is not called.
add calling of trigger in machine driver in soc_pcm_trigger()
Zhang Jiejing [Tue, 11 Dec 2012 07:34:25 +0000 (15:34 +0800)]
ENGR00235817 mx6: use SNVS LPGPR register to store boot mode value.
After using POR reset, the content in SRC will be reset.
See RM: 63.5.1.2.3 IPP_RESET_B(POR)
Because POR reset will reset most of register in IC, so use
SNVS_LP General Purpose Register (LPGPR) to store the boot mode value.
Below copy from SNVS_BlockGuide.pdf:
The SNVS_LP General Purpose Register provides a 32 bit read write
register, which can be used by any application for retaining 32 bit
data during a power-down mode
Sheng Nan [Wed, 5 Dec 2012 02:22:35 +0000 (10:22 +0800)]
ENGR00235665: mxc_v4l2_capture: add YV12 format support in camera driver
Android CTS verifier have a must requirement for YV12 format. Since IPUv3
common driver has supported IPU_PIX_FMT_YVU420P pixel format, add the
support of YV12 format in mxc_v4l2_capture.
make shi [Thu, 6 Dec 2012 09:19:30 +0000 (17:19 +0800)]
ENGR00236031 MX6 USB :Change default USB H1 and OTG driver load order
In current linux BSP USB H1 driver default load before otg driver load,
which cause USBx not match the ehci controller number. like bellow:
root@freescale /sys/devices/platform/fsl-ehci.0$ ls
driver modalias pools power subsystem uevent usb2
root@freescale /sys/devices/platform/fsl-ehci.1$ ls
driver modalias pools power subsystem uevent usb1
Hongzhang Yang [Tue, 6 Nov 2012 06:49:38 +0000 (14:49 +0800)]
ENGR00232530 Refine VPU suspend/resume according to open_count
1. Refine VPU suspend/resume according to open_count to completely
fix bug: ENGR00230203 [Android_MX6DL_SD] Gallery: System hang
after resume from suspend during video playback. 20%
open_count == 0 case can be simplified because VPU is released
(all instances are freed), so
- clock is already off
- context is no longer needed
- power is already off on MX6
VPU reset is removed from resume because power is ensured to be off
before entering resume on MX6 by calling regulator API.
2. Fix bug: VPU always busy after suspend/resume
Error log (VPU refused to suspend due to VPU busy):
pm_op(): platform_pm_suspend+0x0/0x54 returns -11
PM: Device mxc_vpu failed to suspend: error -11
PM: Some devices failed to suspend
Root cause:
- Suspend happened during vpu_Init(), somewhere after VPU lib started
to download FW (when PC == 0), but before run FW. (BIT_BUSY_FLAG=1,
BIT_CODE_RUN=1).
- In such case, VPU resume downloaded FW and run VPU to idle because
suspend was triggered after VPU was opened (active).
- vpu_Init run VPU again with BIT_BUSY_FLAG=1. So VPU was trapped in
idle loop but BIT_BUSY_FLAG was never cleared. VPU lib regarded VPU
as always busy.
Solution (in VPU resume):
- run VPU FW only if VPU was opened and PC before suspend is not 0
- restore host register is required
- download FW is required, because program memory is lost after power
off.
Signed-off-by: Hongzhang Yang <Hongzhang.Yang@freescale.com>
Fugang Duan [Thu, 29 Nov 2012 08:55:12 +0000 (08:55 +0000)]
ENGR00235626 FEC: Enable phy pause frame feature
Since some ethernet MAC flow control is relied on phy pause
status, both link partners exchange information via auto neg to
determine if both parties are capable of flow control.
Advertise phy pause frame to avoid pause frame is not responsed
by the other net node.
make shi [Wed, 5 Dec 2012 06:31:41 +0000 (14:31 +0800)]
ENGR00235630 MX6 USB :fix USB does not work when plug in device during suspend
USB does not work when plug in a usb device during system suspend. Under this
case, USB driver will be in low power mode, but WIE bit not be set if usb wake
up is not enabled.So there are only ID change interrupt no USB wakeup interrupt
after system resume.In current bsp, after system resume ID change status not be
clear,and ID change interrupt will continue happen, which cause the system busy.
No checking WIR bit if ID change interrupt happen when USB in low power mode to
fix this issue.
make shi [Thu, 29 Nov 2012 08:28:25 +0000 (16:28 +0800)]
ENGR00234722 USB: fix Kernel dump issue after USB driver loadable
- It is better to disable otgsc and wake up interrupt to avoid an
abnormal interrupt happen during USB driver being removed.
- If the USB host is already at low power mode, only need turn on
the clock, no need turn off the clock.
- Need discharge dp and dm during USB driver being removed ,in order
to avoid a wakeup interrupt happen. And if the USB otg is in host
mode, we should clear discharge dp and dm in fsl_otg_set_host()
during system boot up.
I2C driver call the function "static void __init i2c_imx_set_clk()"
in runtime, the function is linked to init.text section, and don't
be used after kernel bootup. Remove the "__init" statement to fix
the issue.
Alejandro Sierra [Wed, 21 Nov 2012 16:01:26 +0000 (10:01 -0600)]
ENGR00234466 UART: Fix disablement of CTS signal
On Uart driver, CTS signal were never disabled
on the imx_set_mctrl function since the register was
written inside of the conditional.
if (mctrl & TIOCM_RTS) {
temp |= UCR2_CTS;
writel(temp, sport->port.membase + UCR2);
}
Signed-off-by: Alejandro Sierra <b18039@freescale.com>
Fugang Duan [Wed, 28 Nov 2012 02:40:52 +0000 (10:40 +0800)]
ENGR00235086 I2C: update i2c clock divider for each transaction
Currently on Arik/Rigel, the I2C clk is from IPG_PERCLK which is
sourced from IPG_CLK. Under normal operation, ipg_perclk is at 22MHz
so that we can get 400KHz i2c speed. In low bus freq mode, IPG_CLK is
at 12MHz and IPG_PERCLK is down to 4MHz.
So the I2C driver must update the divider register for each transaction.
Fugang Duan [Tue, 27 Nov 2012 09:44:19 +0000 (17:44 +0800)]
ENGR00235090 FEC: Workaround for FEC RX hang with stress test
When do Ethernet UDP stress overnight test with abundance of
data transmission, RX path may hang-on.
Dump the RX BD, found all BD "Empty" bit is cleared, which means
CPU read BD status is not right and waiting here.
Change BD memroy attribute from Normal to strongly ordered:
changes the memory attribute of C=0, B=0 instead of C=0, B=1.
Apply the change, the issue cannot be reproduced.
Ryan QIAN [Tue, 27 Nov 2012 23:54:32 +0000 (07:54 +0800)]
ENGR00234933 mmc: sdhci: invalid cd_gpio for always_present host controller
Issue:
By default, cd_gpio is 0 for always presented host controller, which is a
valid gpio. Then it will result to free_irq for 0 in esdhc_pltfm_exit for
these always_present host controllers.
Fix:
Invalid cd_gpio if the controller is indicated to be always present.
Acked-by: Dong Aisheng <b29396@freescale.com> Signed-off-by: Ryan QIAN <b32804@freescale.com>
Michal Hocko [Tue, 10 Jan 2012 23:08:02 +0000 (15:08 -0800)]
mm: fix off-by-two in __zone_watermark_ok()
Commit 88f5acf88ae6 ("mm: page allocator: adjust the per-cpu counter
threshold when memory is low") changed the form how free_pages is
calculated but it forgot that we used to do free_pages - ((1 << order) -
1) so we ended up with off-by-two when calculating free_pages.
Reported-by: Wang Sheng-Hui <shhuiw@gmail.com> Signed-off-by: Michal Hocko <mhocko@suse.cz> Acked-by: Mel Gorman <mgorman@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Robin Gong [Mon, 26 Nov 2012 03:25:55 +0000 (11:25 +0800)]
ENGR00234685-2 mx6q_sabreauto: change Sabreauto board to LDO-ENABLED mode
Per hardware design, we can't set LDO bypass mode on Sabreauto board,otherwise,
system will can't reset,if cpu freq run in 400Mhz. Signed-off-by: Robin Gong <b38343@freescale.com>
Robin Gong [Mon, 26 Nov 2012 03:23:28 +0000 (11:23 +0800)]
ENGR00234685-1 cpufreq:fix one bug in cpufreq driver if I2C transfer error
Currently, if we used LDO bypass, will set pfuze register by I2C bus to modify
voltage according to different cpu frequency, if I2C transfer error, we should
restore to old cpu frequency, not only in cpufreq driver but also cpufreq core.
Ge Lei [Fri, 23 Nov 2012 06:36:04 +0000 (14:36 +0800)]
ENGR00233569 SDMA: Add support for SDMA M2M copy
Our SDMA code did not support SDMA M2M copy function before, we add
SDMA M2M copy function in this patch, you can use 'sg' to use this
function, you can refer to 'linux-test/module_test/mxc_sdma_memcopy_test.c'
for how to use this function.
Ryan QIAN [Thu, 22 Nov 2012 07:38:46 +0000 (15:38 +0800)]
ENGR00234519 mmc: support eMMC v4.5 memory cards
Bypass eMMC version checking, so that eMMC v4.5 can work on current kernel as
eMMC v4.4 cards, no specific v4.5 feature supported. Only basic read/write
operations are supported, also ddr mode is supported.
Acked-by: Dong Aisheng <b29396@freescale.com> Signed-off-by: Ryan QIAN <b32804@freescale.com>
Sheng Nan [Fri, 23 Nov 2012 02:59:22 +0000 (10:59 +0800)]
ENGR00234362 Camera: ov5640_mipi: wait for sensor stable before streamon
ov5642 add some delay to wait for sensor stable after S_PARM.
And ov5640_mipi should keep the same behavior.
So the upper layer can trust the first frame comes out of ov5640_mipi.
- delay added according to the recommended time from ov company
Steve Cornelius [Tue, 20 Nov 2012 23:21:26 +0000 (16:21 -0700)]
ENGR00234401: CAAM: Fix incorrect invalidate call for output ring
The job ring driver exhibited a hang condition in the top of
caam_jr_dequeue() where a BUG_ON statement looks for a condition
where the output ring is said to have valid entries by the ring logic,
but the ring entries apparently have NULL descriptor pointers.
In the initial ARM port of this driver, the cache flush call
of the output ring content occured before the output ring read index
register read occurred, exposing a condition where the driver sensed valid
output entries, yet the entries written by the ring hardware were not
invalidated, and therefore were not visible to the processor, appearing
as false NULL entries.
This patch relocates the invalidate call to immediately follow the
check of the output read index, where it is required.
Signed-off-by: Vicki Milhoan <vicki.milhoan@freescale.com> Signed-off-by: Steve Cornelius <steve.cornelius@freescale.com> Signed-off-by: Terry Lv <r65388@freescale.com>
Robin Gong [Thu, 22 Nov 2012 05:31:05 +0000 (13:31 +0800)]
ENGR00234411-1 Sabreauto: fix error print COULD NOT SET GP VOLTAGE.
Didn't take more care about non-pfuze board, and there is two place in BSP will
call "mx6_cpu_regulator_init". It means regulator_get will be called twice on
every vddcore/vddsoc regulator. Then one value need set twice ,because from
regulator core view, there is two regulators share the same regulator. The non-
validate one will return error and print "COULD NOT SET GP VOLTAGE!!!!." on
Sabreauto board. The same as Sabrelite and ARM2 board.
Meanwhile, Sabreauto need be configured LDO bypass default. Signed-off-by: Robin Gong <b38343@freescale.com>
Michael Minnick [Tue, 13 Nov 2012 18:51:13 +0000 (12:51 -0600)]
ENGR00233494 EPDC: Driver only supports 16 LUTs
This bug was introduced by ENGR00229290 which fixed
the problem of greater than 16 LUTs used when 5-bit
waveform loaded. The bug is that now the driver is also
restricted to using 16 LUTs in 4-bit mode.
The fix is to correct the test of the EPDC_FORMAT
register used to determine if a 5-bit waveform
is loaded.
Also removed the while loop in favor of a bitwise OR
used to determine if a chosen LUT has yet to be
acknowledged by the interrupt handler.
Signed-off-by: Michael Minnick <michael.minnick@freescale.com>
Robin Gong [Tue, 20 Nov 2012 07:08:12 +0000 (15:08 +0800)]
ENGR00234217 cpufreq:fix loops_per_jiffy wrong on new suspend flow of cpufreq
Currently, we use pm_notifier to enter suspend/resume flow. But in the notifier
we only set cpufreq, didn't tell CPUFREQ core what the current cpufreq setting
now. So in the next time if CPUFREQ core find the current cpu frequncy is not
the value that CPUFREQ core want to set before. CPUFREQ core will force to set
the freqs.old with its own rule, which means the freqs.old will be MODIFYED
unexpectedly, and this will cause wrong loops_per_jiffy. We need add cpufreq_
notify_transition in the suspend/resume interface of cpufreq.
Robby Cai [Mon, 12 Nov 2012 08:17:14 +0000 (16:17 +0800)]
ENGR00229852 csi/v4l: lower the priority of "working queue empty" message
When pause the capture test program, the "working queue empty" message
shows up repeatedly. However this message is expected to show up because
there's no QBUF called. Change pr_err to pr_debug to keep it as debug level.
Robby Cai [Thu, 8 Nov 2012 06:45:56 +0000 (14:45 +0800)]
ENGR00232893 mx6sl: pxp/v4l2: add dependency on CONFIG_FB_MXC_ELCDIF_FB
Without this dependency, we have to manually disable
CONFIG_VIDEO_MXC_PXP_V4L2 when disable CONFIG_FB_MXC_ELCDIF_FB.
Otherwise, a build error shows up. This patch fixed it.
Robin Gong [Mon, 19 Nov 2012 02:04:46 +0000 (10:04 +0800)]
ENGR00233366-4 WDOG LDO_BYPASS: fix wdog2 to reset external pmic in ldo bypass
On Sabresd board design, the WDOG_B output to reset external pmic source from
GPIO_2 pad which can be configured as WDOG2_WDOG_B, so if in ldo bypass mode,
we should use WDOG2 reset signal to reset pmic, not WDOG1. Also, configure the
related pins.
Robin Gong [Tue, 13 Nov 2012 07:06:16 +0000 (15:06 +0800)]
ENGR00233366-2 mx6q_sabresd mx6sl_arm2 mx6sl_evk: config in LDO bypass
U-boot will not care about ldo bypass, move these code from u-boot to kernel.
Move the workaround for PFUZE1.0 to kernel too. Signed-off-by: Robin Gong <b38343@freescale.com>
Robin Gong [Mon, 19 Nov 2012 02:21:00 +0000 (10:21 +0800)]
ENGR00233366-1 Anatop PFUZE: LDO bypass can be configed by cmdline
Currently, use CONFIG_MX6_INTER_LDO_BYPASS to enable/disable LDO bypass, and
use the same macro in u-boot too. It's not very friendly ,it will be more
flexible if use dynamic configure by command line input by u-boot.
Two ways to enable LDO bypass:
1. use command line:
You can set "ldo_active=on" or "ldo_active=off" in command line to enable/
disable LDO bypass.
2. set enable_ldo_mode value in board file:
If you didn't set the param in command line, every board
will use its default value, for example, you can find below code in arch/arm/
mach-mx6/mx6q_sabresd_pmic_pfuze100.c:
static int pfuze100_init(struct mc_pfuze *pfuze)
{
....
/*use default mode(ldo bypass) if no param from cmdline*/
if (enable_ldo_mode == LDO_MODE_DEFAULT)
enable_ldo_mode = LDO_MODE_BYPASSED;
....
}
Note:
1.You should know clearly ldo bypass can be only enabled in the board
that mounted with external pmic to supply VDDARM_IN/VDDSOC_IN power rail, and
you should implement related external regulator firstly, such as:
in arch/arm/mach-mx6/board-mx6q_sabresd.c
static struct mxc_dvfs_platform_data sabresd_dvfscore_data = {
.reg_id = "VDDCORE",
.soc_id = "VDDSOC",
....
}
otherwise, you have to use internal ldo which is the default configuration.
2.one special case, if the chip is 1.2Ghz, it can't be set LDO bypass.
Robin Gong [Mon, 19 Nov 2012 02:16:22 +0000 (10:16 +0800)]
ENGR00234040 FUSE 1.2G: add fuse bit for 1.2G
Before, we use "arm_freq" in command line to set 1.2G. Now we will read the
fuse bit and "arm_freq", get the mini value to be used as "arm_max_freq".And:
1. you can easily set CPU max freq on what frequency you want by cmdline.
2. if you didn't set arm_freq in cmdline, kernel will read the fuse bit
(0x021bc440) to set the right arm_max_freq.
At the same time, add 1Ghz setpoint if chip max freq is 1.2Ghz.
Jason Liu [Thu, 15 Nov 2012 06:51:10 +0000 (14:51 +0800)]
ENGR00233770 i.mx6q/vpu: Add fuse check for VPU_DISABLE feature
This patch add the fuse check for VPU_DISABLE feature. If the fuse
bit for VPU_DISABLE is 1, which means VPU is disabled, then we will
not register VPU device to the kernel.
ENGR00233705 [MX6SL] -Fix suspend/resume issue when SD1 is used to boot.
Setting certain IOMUX settings on SD1 prevents the system from
entering suspend. These pins are already configured as GPIO, so
it does not help to reconfigure them during suspend.
Terry Lv [Fri, 16 Nov 2012 07:41:01 +0000 (15:41 +0800)]
ENGR00233929: add kernel command line to enable snvs
In MX6Q/DL, originally GPIO_0 is used as CKO pin function. when SNVS
module is enabled, CKO output stops suddenly.
Both CKO clock config register CCOSR and GPIO_0 IOMUX register value are
not changed. But because ALT7 of GPIO_0 pad is SNVS_VIO_5 function. I
doubt that when SNVS module is enabled, GPIO_0 pad is automatically
changed to SNVS instance by SoC.
Thus we add option for snvs enable/disable.
Terry Lv [Thu, 15 Nov 2012 08:50:41 +0000 (16:50 +0800)]
ENGR00233800: CAAM: running sha_speed in cryptodev crashed
The reason is that when switching from SHA1 to SHA256, cryptodev will
create a new session.
But in this new session, the __ctx in allocated req is not fully initialized.
Thus dma_buf in __ctx will be a random value.
If the value is 0 or some address in DMA memory, that will be ok,
otherwise, it will crashed in dma_unmap_single().
The calling sequence is:
ahash_final_ctx=>try_buf_map_to_sec_sg()=>dma_unmap_single()
When calling dma_unmap_single(), the parameter buf_dma is invalid in
crash case.
The error msg is:
kernel BUG at arch/arm/mm/dma-mapping.c:478!
Chen Liangjun [Mon, 12 Nov 2012 05:41:13 +0000 (13:41 +0800)]
ENGR00233780 ASRC: limit output buffer size to avoid kernel dump
For ASRC memory to memory transfer, user would send driver input buffer
and driver would copy converted output buffer into user's buffer.
However, ASRC can't promise the ratio of output buffer size/input buffer
size being equal to output sample rate/input sample rate.e.g, for
convert from 8k to 48k and 1000 bytes input buffer size, ASRC may pop
out 5999 bytes or 6001 bytes. If driver copy all 6001 bytes into user's
buffer, kernel dump may happens cause of accessing unexisted buffer.
In this patch, if ASRC output buffer size is larger than user's buffer
size, discard exact part.