There are two type of temp WARs in the nvidia repo:
1) Some functions are defined in driver files that depend on
CONFIG_TEGRA_GRHOST_NVCSI. So these function call should
have dependency protection in the header file.
2) The powergate header file is adding an unnecessary define
check for tegra194-powergate.h and this broke two places
that were referring one of the defines in it.
So this change fix the WARs above by:
a) Removing those temp WARs by adding dummy NOP functions for
those platofrms that don't have CONFIG_TEGRA_GRHOST_NVCSI.
b) Removing the define check in tegra194-powergate.h and the
version check when referring TEGRA194_POWER_DOMAIN_DISP.
Bug 2284925
Change-Id: Iad86bfe20f2981b3d31bccf5f51184799fe7e289
Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1803439
GVS: Gerrit_Virtual_Submit
Reviewed-by: David Bang <dbang@nvidia.com>
Reviewed-by: Sachin Nikam <snikam@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
drivers inside platform tegra and nvdisp
point to display repo.
changing that to nvidia/
Bug 200295104
Change-Id: I91e0356555c251df3eae98218a69f4bc7e3cd208
Signed-off-by: Ishan Mittal <imittal@nvidia.com>
To prevent too many bufferred message from flushing out to serial when
we turn on/off display and blocking other high priority task due to
disabled IRQ, we change some prints to debug level to workaround that
problem as a short-term solution. The long-term will be implemented
later.
bug 200205349
Change-Id: I6609844d5c2ee58b2df6a51e9537bb7dd1448995
Signed-off-by: Vince Hsu <vinceh@nvidia.com>
Reviewed-on: http://git-master/r/1326683
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Chun Xu <chunx@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Juha Tukkinen <jtukkinen@nvidia.com>
use generic APIs tegra_unpowergate_partition() and
tegra_powergate_partition() to unpowergate DISP power
domain needed for CEC without enabling disp clocks so
that display pll clocks are disabled in idle state even
though DISP partition is awake.
Bug 1810989
Change-Id: I37428b9012729a4458724314732f80a571057fd6
Signed-off-by: Santosh Reddy Galma <galmar@nvidia.com>
Reviewed-on: http://git-master/r/1317241
(cherry picked from commit dffb3d3b5fb67cb6920fdbfc72cd3d6c37f0061c)
Provide error recovery from IOCTL to user space.
Bug 1866338
Change-Id: I705ada6c8d4cb13f1c882993468f467da2908fdf
Signed-off-by: Chun XU <chunx@nvidia.com>
Reviewed-on: http://git-master/r/1300499
(cherry picked from commit a646f4903c8794641432fa838a27ee5584944eb5)
Turn on the TV as soon as possible for better user experience
by sending <Text View On> and <ACTIVE SOURCE> CEC commands if
the previous reboot is not recovery. If the previous reboot
reason is recovery, don't send <Text View On> to avoid turning
on TV when there is an auto OTA on-going.
CEC driver uses RESERVED2 logical address as the source address.
Later, the upper layer will send OneTouchPlay command by using
its valid logical address.
Bug 200253563
Change-Id: Iecd759bb50c1f21549f55906e68e3faab905f86b
Signed-off-by: Chun XU <chunx@nvidia.com>
Reviewed-on: http://git-master/r/1277092
(cherry picked from commit 2beab2dcfc4e40712004500df99633d59ae0e70c)
Powergate identifiers are different on Tegra210 and Tegra186,
so add per-SoC match data to use the correct value on each
SoC.
bug 200257351
Change-Id: I81b938386ab9cbaec735baac44ee74fccf39731c
Signed-off-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Laxman Dewangan <ldewangan@nvidia.com>
Reviewed-on: http://git-master/r/1297136
(cherry picked from commit 5438b348e036c31dcf9db0d1364fa6c6ac053428)
Create a symlink for tegra_cec if it is not under platform bus or
it has been created with different name.
Bug 200214224
Bug 200185815
Bug 1764073
Change-Id: Ib2f89db016982ca916894c37e47bbdf177fb617e
Signed-off-by: Chun XU <chunx@nvidia.com>
Reviewed-on: http://git-master/r/1175137
(cherry picked from commit 506e6370d234c4770afb22372ef1955faa7822cf)
Use tegra_nvdisp_unpowergate_partition/tegra_nvdisp_powergate_partition
in Tegra CEC driver to control DISA power domain.
Bug 200214236
Bug 200214232
Bug 200185815
Change-Id: Ib860570f901278e9211b5262fddf8c63a700c5f9
Signed-off-by: Chun XU <chunx@nvidia.com>
Reviewed-on: http://git-master/r/1176912
(cherry picked from commit 8e9f75ac1dadf04a8f89a74dac3e67cd1a06f886)
Unpowergate DISA power domain if it is powergated during CEC driver initialization.
Bug 200175747
Change-Id: I8b3d879fe0c33eaef74c1b4d3b549559c4977e4d
Signed-off-by: Chun XU <chunx@nvidia.com>
Reviewed-on: http://git-master/r/1142593
Reviewed-on: http://git-master/r/1164149
(cherry picked from commit c341176715597fe8268b20b291bcef07bb429500)
Bug 1591149
Initialization of cec engine is slow.
Short-circuit the init in event of suspend.
Change-Id: Ibfbd6f36883a7bf45fdb5137120b041a52f42086
Signed-off-by: Xia Yang <xiay@nvidia.com>
Reviewed-on: http://git-master/r/714423
(cherry picked from commit d5f4a6602678c87e8caa5032e43f2415c314d800)
Reviewed-on: http://git-master/r/716496
Reviewed-on: http://git-master/r/1164147
(cherry picked from commit 82a640a1ea11e850f769b78b5d0c531731b1ffd8)
Auxdata still have to be provided in of_platform_populate()
during early boot device detection to get the device correctly
named as "tegra_cec". Failing to do so will result in device
probe failure as of_device_make_bus_id() will name the device
as "[phys_base].tegra_cec" and cause clock assignment failure.
Bug 200013561
Change-Id: Ib0f898f89146cba7f02cffb768c6a2eb558400ce
Signed-off-by: Xia Yang <xiay@nvidia.com>
Reviewed-on: http://git-master/r/433871
Reviewed-on: http://git-master/r/1164144
(cherry picked from commit 43402b18ee6def21d7d55420f5acd22403e86f1d)
Change write() to work per frame.
Change write() to return -1 on error and set up errno
write() API:
-Userspace is responsible for re-transmission.
-Read from user-space byte by byte, each byte representing
a block, up to 16 bytes as specified by HDMI standard.
-Return 0 on success transmission, -1 otherwise, with errno
setup as follows:
EIO - TX_REGISTER_UNDERRUN, should not happen, otherwise
driver is have serious timing issue.
ECOMM - BUS arbitration failure or anomaly BUS activity.
Transmission is abandoned.
ECONNRESET - For broadcast message only, someone on the BUS
asserted NAK during transmission.
EHOSTUNREACH - For direct message only, message was not ACK'd.
(Required by logical address allocation)
EMSGSIZE - Message size > 16 bit.
EFAULT - Page fault accessing message buffer.
EINTR - call interrupted by singal.
read() API is unchanged, works per packet with no error report.
Change-Id: Iabdd92b5658dd63c7b500a7ec88d79a64c8c0a43
Signed-off-by: Xia Yang <xiay@nvidia.com>
Reviewed-on: http://git-master/r/304664
(cherry picked from commit 6ab7a446c4a3e8e4970ceec5a3c715453a24e4a5)
Signed-off-by: Xia Yang <xiay@nvidia.com>
Reviewed-on: http://git-master/r/347779
Reviewed-on: http://git-master/r/1164143
(cherry picked from commit 67a6c2d3cfbeeca5cb5ed57aae53f34d07decd4f)
Remove wait_event_interruptible() from interrupt context
Change init_done to be atomic_t and reset as early as possible
Bug 1395893
Change-Id: Ib0cf423a3405293000b0c0d9aa105da5bba22e53
Signed-off-by: Xia Yang <xiay@nvidia.com>
Reviewed-on: http://git-master/r/304631
(cherry picked from commit ef034436a3c8aaf7a9ce5dd9ebaf8dc90dbcce4b)
Signed-off-by: Xia Yang <xiay@nvidia.com>
Reviewed-on: http://git-master/r/346042
Reviewed-on: http://git-master/r/1164141
(cherry picked from commit 76e681f957b27323227c9990679631636084b6ae)
cancel the workqueue in suspend method before turning
off the clock
Bug 1360341
Change-Id: I126da686a6ba0c5eec55b67df1e9f962ce16fc48
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: http://git-master/r/280159
(cherry picked from commit 7cc6da6e41c7889f7ed8da7833d293b3b884bc0e)
Signed-off-by: Xia Yang <xiay@nvidia.com>
Reviewed-on: http://git-master/r/346041
Reviewed-on: http://git-master/r/1164140
(cherry picked from commit e57647e7daacd5a3c22016cbe3f01ddb800d702c)
Add wait_event to make sure that CEC functions
do not execute until CEC init completes
Bug 1283088
Change-Id: I1d26360326338f549a14cbf109a24c2935ebe472
Signed-off-by: Shridhar Rasal <srasal@nvidia.com>
Reviewed-on: http://git-master/r/232566
(cherry picked from commit 6e94e976c1e2121d81d19018b79a97d353b11d70)
Signed-off-by: Xia Yang <xiay@nvidia.com>
Reviewed-on: http://git-master/r/346039
Reviewed-on: http://git-master/r/1164139
(cherry picked from commit 794145e9f8c16c7e36119f3652dd185cf12eb5dc)
In T124, the RX_REGISTER_FULL interrupt is cleared
by reading the RX_REGISTER instead of writing to the
INT_STAT register as before.
Bug 1364229
Change-Id: Ib080a48910304553b0752a3ede55ab0d7653dd77
Signed-off-by: Ankita Garg <ankitag@nvidia.com>
Reviewed-on: http://git-master/r/275982
Reviewed-on: http://git-master/r/1164137
(cherry picked from commit b890e7f50d3d032fc3ede4c4935fdf2c4501f063)
tegra_cec.c: fix warnings as compilation fails after enabling warning as
errors flag
Bug 1258710
Change-Id: Iecd3051d482f5a7488c9f95f6124ad39371f0bd9
Signed-off-by: Benjamin Lu <benjaminl@nvidia.com>
Reviewed-on: http://git-master/r/1164136
(cherry picked from commit 8112e3cd46d11eca51e95cc2fed66d2f7946553a)
READ API:
read API ignores count and will always return 16 bit data.
read API expects user to supply it with min of 16 bits data
it returns CEC packet in following format
bit 0-7: data
bit 8: EOM
bit 9: ACK
WRITE API:
write API ignores count and will always accept 32 bit data.
write API expects user to supply it with min of 32 bits data
it accepts CEC packet supllied in following format
bit 0-7: data
bit 8: EOM
bit 12: Address mode, 0 = Direct, 1 = Broadcast
bit 16: Generate Start bit, 0 = Disable, 1 = Enable
bit 17: Retry frame, 0 = Disable, 1 = Enable
Logical address is set to 4, as of now there is no mechanism to change this
address from userspace.
Driver is disabled by default in Kconfig
Bug 200198493
Change-Id: Ia3835cec0bb717e63dabca5c5fcb1236847bf492
Reviewed-on: http://git-master/r/1164135
(cherry picked from commit 955ec819872e66c4732b38cd74c7ff3a302d95f2)
Signed-off-by: Chun XU <chunx@nvidia.com>
Configure BAR and MSI for Thor.
Add checks for chip id for chip specific code
between Orin and Thor.
Separate Stream ID's are assigned to upstream and downstream
struct device in Rootport client driver.
While accessing Rootport local memory upstream Stream ID should be used.
While accessing memories over BAR downstream Stream ID should be used.
Jira NVIPC-2877
Jira NVIPC-2484
Change-Id: I67df4b78e57b6de36f9bfaf966978f7ee875d596
Signed-off-by: Deepak Badgaiyan <dbadgaiyan@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3226748
Reviewed-by: svcacv <svcacv@nvidia.com>
Reviewed-by: Vipin Kumar <vipink@nvidia.com>
Reviewed-by: Janardhan Reddy AnnapuReddy <jreddya@nvidia.com>
Reviewed-by: Sivagamy Govindasamy <sivagamyg@nvidia.com>
GVS: buildbot_gerritrpt <buildbot_gerritrpt@nvidia.com>
In Linux v6.13, commit cdd30ebb1b9f ("module: Convert symbol namespace
to string literal") updated the MODULE_IMPORT_NS macro to take a string
literal as an argument in Linux v6.13. Update the mods_dmabuf driver as
necessary to fix this.
Bug 4991705
Change-Id: I9326e360fe8daae4c43941812301818d8b35c825
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3263799
Reviewed-by: Chris Dragan <kdragan@nvidia.com>
GVS: buildbot_gerritrpt <buildbot_gerritrpt@nvidia.com>
This reverts commit ee72e995da1ee31177a06053b4c89f3ba6a1060a.
Reason for revert: This breaks building the OOT modules standalone without the GPU sources which are not public. Futhermore, this header
appears to be internal and so should not be included in sources that
are distributed publicly.
Jira: NVIPC-2817
Change-Id: I9a98056b004b7b6454473957f488020718eabfd4
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3263605
Reviewed-by: Brad Griffis <bgriffis@nvidia.com>
GVS: buildbot_gerritrpt <buildbot_gerritrpt@nvidia.com>
pci_client abstraction supports notification to application.
Each endpoint has to register with pci_client for notification which
return link_event_id after registration.
link_event_id is index in lookup table with total size as MAX_ENDPOINT.
This link_event_id is allocated sequential and does not match with ep_id
which is endpoint index with max as MAX_ENDPOINT.
stream-extension is not aware about link_event_id and uses ep_id for
notification which leads to mismatch in look up table finally dropping
notification.
Registering for link event in sequential manner does not help much as
registration against ep_id will make lookup table and endpoint database
one-to-one mapping.
Update endpoint registration for link event based on ep_id and use ep_id
in lookup table for notification purpose.
Bug 4913236
Change-Id: I75bfdbf0e8e5b7b11b0cca7dc266f01f492362f6
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nv-oot/+/3233789
GVS: buildbot_gerritrpt <buildbot_gerritrpt@nvidia.com>
Reviewed-by: Sumeet Gupta <sumeetg@nvidia.com>
Tested-by: Janardhan Reddy AnnapuReddy <jreddya@nvidia.com>
