Age | Commit message (Collapse) | Author |
|
Firmwares are currently released at the time we close the channel
device node. This causes unexpected side effects in cases where
the finalize_poweron is called only once to get the device booted
during the first time the device is accessed.
This patch fixes the issue by not releasing the firmware after it
has once loaded.
Bug 200033244
Change-Id: I8fc6748846e56ac0cc347feb170b239533a47252
Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
Reviewed-on: http://git-master/r/495220
(cherry-picked from commit 69bce0a17ad1e5457cfce0d66f6e1b4add7e9820)
Reviewed-on: http://git-master/r/670671
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Matthew Pedro <mapedro@nvidia.com>
|
|
This far the subdomain registration has been allowed for drivers that
use nvhost_acm framework. However, gk20a handles its own power domain bu
it needs to register it as a subdomain of host1x. This patch exports
necessary functionality in public header.
Bug 200003112
Change-Id: Ie3457345c101329cadac4ef8aa71e448629a419b
Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
Reviewed-on: http://git-master/r/407542
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
Add below two new ioctls for client managed syncpts:
1) NVHOST_IOCTL_CHANNEL_GET_CLIENT_MANAGED_SYNCPOINT
To provide a client managed syncpt to user space
2) NVHOST_IOCTL_CHANNEL_FREE_CLIENT_MANAGED_SYNCPOINT
To free a client managed syncpt from user space
Bug 1482249
Bug 1305024
Change-Id: Ie9376c761024f2ad6e8a35633017153ca5cc8fe7
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: http://git-master/r/394158
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Tested-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
Implement interface to notify user space when errors.
Bug 1469448
Change-Id: Id59e312ed69dcd11672ca5d1c2b71fde21972b94
Signed-off-by: Shridhar Rasal <srasal@nvidia.com>
Reviewed-on: http://git-master/r/395164
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Tested-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
Extend VIC driver into a common Falcon driver and rename it to flcn.
Change-Id: Ic4d8c39e2ac1e7ac571746d7cf1ff8d8d34621d4
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: http://git-master/r/288780
Reviewed-by: Ken Adams <kadams@nvidia.com>
|
|
- Add support to have multiple channels per device.
- set max number of channels that can be assigned to
device in platform data.
- On channel open request assign new channel to device
till max number of channel count.
- suspend all assigned channels on device disable clk.
- un-map all channels on device release callback.
Bug 1259844
Bug 1436477
Change-Id: Ibc1de8b036cde238b292e1ffa660dc2a796f65e8
Signed-off-by: Shridhar Rasal <srasal@nvidia.com>
Reviewed-on: http://git-master/r/387096
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Tested-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
Add ioctl NVHOST_IOCTL_CTRL_SYNC_FENCE_SET_NAME to set
name for a sync fence
Bug 1473041
Change-Id: Iabc1c472f6fddc87dac75d3a37b75a83a834dc1e
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: http://git-master/r/391271
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Tested-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
This reverts commit 3cc76ad71851437c8f5d4d5b27a8df392e95a53b
also fix issues seen in kernel_submit :
- Re-use put/getchannel only for channel ref-counting
- Make put/getchannel calls atomic
- Move device init/deinit to channel map/unmap
- Move keepalive checks to map/unmap
- Uses channel bitmap in circular way to avoid race
- Remove legacy alloc/free channel callbacks
- Remove static reference to hw channel
- Free channels and list on nvhost remove
Bug 1259844
Change-Id: I2be58540370cb7b8a42e72c64ae9cb4588faafd8
Signed-off-by: Shridhar Rasal <srasal@nvidia.com>
Reviewed-on: http://git-master/r/381310
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
Char device region needs to be released when we unloaded module,
otherwise it cause allocation failure and kernel oops.
Bug 1490541
Change-Id: Ib95aef54f90d4f743c208f497db1ed69873cf13d
Signed-off-by: Bryan Wu <pengw@nvidia.com>
Reviewed-on: http://git-master/r/389972
Reviewed-by: Automatic_Commit_Validation_User
GVS: Gerrit_Virtual_Submit
Reviewed-by: Shridhar Rasal <srasal@nvidia.com>
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
Remove support for T114 and its units 2D and 3D. Remove user settable
context switching.
Change-Id: If3e194fe1907ccde1ca0537a549da35a366637b0
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: http://git-master/r/390120
|
|
Make T114 3D driver perform the bootup reset by itself and remove the
generic powerup reset.
Change-Id: Ife95d2cdef8370daa9a5ddff02fd9b5c9f115c11
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: http://git-master/r/375978
Reviewed-by: Juha Tukkinen <jtukkinen@nvidia.com>
Tested-by: Juha Tukkinen <jtukkinen@nvidia.com>
|
|
Change-Id: Ic176da5d95657f8b6f36922824449b8e189221e4
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: http://git-master/r/375977
Reviewed-by: Juha Tukkinen <jtukkinen@nvidia.com>
Tested-by: Juha Tukkinen <jtukkinen@nvidia.com>
|
|
This reverts commit 1bf4f9a5e3334756dc789dfc9408e5c64b121f59.
Commit "1bf4f9a" was hard coding some syncpts and was needed
to support T114 on dev-kernel
Since now T114 support is removed from dev-kernel/main branch,
this change can be reverted
Bug 1305024
Change-Id: I5ff5f1190b700fb168b59b6fcff0ec5d66844861
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: http://git-master/r/384265
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Tested-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
This patch modifies the code so that private nvhost headers are never
included inside the gk20a driver. The public header is still used and
hence this patch leaves some dependencies against nvhost driver.
Change-Id: I599930b379af7679a0cdc06a0b5632024503f8e3
Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
Reviewed-on: http://git-master/r/383710
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Tested-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
Currently, vpr prototypes are defined inside nvhost.h but the correct
place for the headers is inside gk20a.h. This patch makes the necessary
modifications to the headers.
Change-Id: I779e242d36402102a5bb39440dd6d6d455937fc3
Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
Reviewed-on: http://git-master/r/381426
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
Currently, nvhost sync objects are considered private and they are
defined inside nvhost private headers. However, we may need to access
the internal data of the nvhost sync objects (hw syncpoint id and
threshold) as the hardware allows using this data directly from also
other devices.
Change-Id: I18dafd9d805ee85e9aa67cad91feb8a7f20950da
Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
Reviewed-on: http://git-master/r/378344
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
This patch extends syncpoint API to support some commonly used functions.
In addition, this patch modifies the current API so that it does not
require giving the host1x client device but it allows giving also host1x
device pointer directly.
Change-Id: Ie3ea53b0faaaa8209639a41f7c375b025b3a277e
Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
Reviewed-on: http://git-master/r/378343
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
This far the nvhost driver has published interface for waiting
syncpoint values synchronously, however, we have need to do similar
waiting asynchronously (i.e. trigger event when certain syncpoint
value has been reached). This patch adds public function
nvhost_intr_register_notifier() that allows registering callbacks.
Change-Id: I3e8a12fad229b264a2ace5bb7f4726e17c90f112
Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
Reviewed-on: http://git-master/r/378342
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
- hard code syncpts for gr3d, gr2d and syncpt no. 17 for T114
- this is to maintain user space dependency on hard-coded syncpts
- syncpt no. 17 is wrongly used by some of the graphics tests which
require this syncpt to be reserved as client_managed syncpt
- keep this hard-coded part for t114 under #ifdef
CONFIG_ARCH_TEGRA_11x_SOC
Bug 1305024
Change-Id: I18dc36018ca5cb95a191dd6f3eaa145d90391608
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: http://git-master/r/381961
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Tested-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
remove static mappings of syncpt ids, syncpt names, and
client managed syncpts
Bug 1305024
Change-Id: I7e297d2bf4c438fc030937f8bdca39068a83d285
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: http://git-master/r/381960
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Tested-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
- implement dynamic syncpoint assignment framework
- provide API nvhost_get_syncpt_client_managed() to get client managed
syncpt which is requested by display, avp, and tegra_camera drivers
- provide API nvhost_get_syncpt_host_managed() for host managed syncpts
- this API is used by host1x clients and gpu
- provide API nvhost_free_syncpt() to free a syncpt
- both client managed and host managed APIs call nvhost_get_syncpt() to
find a free syncpt and reserve it
- remove static syncpt mappings from nvhost_device_data of all clients
- keep VBLANK syncpts hard-coded until we resolve the conflicts with
hard coded syncpt increments from bootloader
Bug 1305024
Change-Id: I02fdfd81d62d215cbb98f4b3f2695bd745fe2402
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: http://git-master/r/381954
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Tested-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
nvhost_syncpt_read_ext is replaced by nvhost_syncpt_read_ext_check
So removing un-used function
Bug 1469913
Change-Id: Ibf2194cacf73f71573024ff737466f56c946b226
Signed-off-by: Shridhar Rasal <srasal@nvidia.com>
Reviewed-on: http://git-master/r/382926
Reviewed-by: Automatic_Commit_Validation_User
GVS: Gerrit_Virtual_Submit
Reviewed-by: Sachin Nikam <snikam@nvidia.com>
|
|
- add new prototype nvhost_syncpt_read_ext_check to
differenciate between error and return value.
- add new func to read syncpoint val and for error check
Bug 1469913
Change-Id: Icd13a741992154177eedcd71e6e65f80025ab03f
Signed-off-by: Shridhar Rasal <srasal@nvidia.com>
Reviewed-on: http://git-master/r/375431
Reviewed-by: Jon Mayo <jmayo@nvidia.com>
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
This patch adds debug interface implementation inside gk20a. As
part of the process, this patch separates debug dump facilities
from the nvhost driver.
Bug 1468086
Change-Id: I8bb6eb90117e8f2fcda86d3936aafce9bc86b271
Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
Reviewed-on: http://git-master/r/376650
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
With dynamic syncpoint allocation ap_systemsw fails
intermittently for TN8. so reverting series of dynamic
syncpoint patches.
Bug 1305024
Bug 1467163
Revert "video: tegra: host: add pbdma syncpoint headers"
commit 62efc18329c5da0e6dc90f088c145e1099213f20.
Revert "video: tegra: host: print syncpt id in debug spew"
commit c918cc4494008fe764e746142060d896436a1d3c.
Revert "video: tegra: host: remove static from get_syncpt_name()"
commit c098ce933738d892bbeba003ab35505c4b7d7951.
Revert "video: tegra: host: remove unused variable"
commit a4dad0592c95165e8366ebe939367a1ece2cb9b9.
Revert "video: tegra: host: Hardcode AVP sync point"
commit 8da84e56de66f20806cb706ee005f263ca5e213f.
Revert "video: tegra: host: hard-code syncpts for t114"
commit 410536f668a62d94603bc3287efaef3dcffb1226.
Revert "video: tegra: host: remove static syncpt mappings"
commit 4cee1d17b01019cc122acf21604d4f01ac6f7474.
Revert "media: tegra_camera: use dynamic syncpts"
commit eae55c8e4cbb65e3ed1a8c8ef8150d8a6ba938bf.
Revert "video: tegra: dc: use dynamic syncpt"
commit 03d7c261ceb62bf1eb93cae48aaa84881d6de5b9.
Revert "video: tegra: host: use dynamic syncpt for hwctx"
commit 03136f549f54a26d77987f3751bf12d202614372.
Revert "video: tegra: host: implement dynamic synpt"
commit 6c2d1b43e51f2e0424b793c9da1c25adf63ea569.
Change-Id: I19a63e812680b7521bed3c1188858b2165a88c89
Signed-off-by: Shridhar Rasal <srasal@nvidia.com>
Reviewed-on: http://git-master/r/374169
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Tested-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
To make dynamic channel mapping easier, moving address space
node to nvhost_device_data.
Bug 1259844
Change-Id: I36f30e97441d26216c5536bb2bc160a9d9afc3d7
Signed-off-by: Shridhar Rasal <srasal@nvidia.com>
Reviewed-on: http://git-master/r/367105
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Tested-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
This far we have delivered minimum frequency, maximum frequency and
the last event type in private_data field for the governor. All these
fields are already available in some other format so it makes no
sense to duplicate this information.
This patch makes necessary modifications to remove dependency to
ext_stat structure.
Bug 1454499
Change-Id: I076aaeb0b9204b810f345ab03b648e238cdab3ff
Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
Reviewed-on: http://git-master/r/365422
Reviewed-by: Hiroshi Doyu <hdoyu@nvidia.com>
Tested-by: Hiroshi Doyu <hdoyu@nvidia.com>
|
|
nvhost never creates these nodes, the gk20a does internally.
Bug 1434573
Change-Id: I8dc48287b2d1cd150ae78b30aaf225fcdf9fdb31
Signed-off-by: Lauri Peltonen <lpeltonen@nvidia.com>
Reviewed-on: http://git-master/r/353300
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Arto Merilainen <amerilainen@nvidia.com>
Reviewed-by: Hiroshi Doyu <hdoyu@nvidia.com>
Tested-by: Hiroshi Doyu <hdoyu@nvidia.com>
|
|
Variable 'syncpt_base' (per device syncpt base) in
struct nvhost_device_data is now unused since we moved
to dynamic syncpt assignment.
Hence remove it.
Bug 1305024
Change-Id: I57e77da5867b14db3e5f768215400513bf12b452
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: http://git-master/r/368273
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Sachin Nikam <snikam@nvidia.com>
|
|
Return AVP sync point as hard coded.
Bug 1305024
Change-Id: I727c896ac97d7a6e7a527dc18f1de86a87f169e3
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: http://git-master/r/368075
Reviewed-by: Deepak Nibade <dnibade@nvidia.com>
Tested-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Bharat Nihalani <bnihalani@nvidia.com>
|
|
- hard code syncpts for gr3d, gr2d and syncpt no. 17 for T114
- this is to maintain user space dependency on hard-coded syncpts
- syncpt no. 17 is wrongly used by some of the graphics tests which
require this syncpt to be reserved as client_managed syncpt
- keep this hard-coded part for t114 under #ifdef
CONFIG_ARCH_TEGRA_11x_SOC
Bug 1305024
Change-Id: I08eeb0acb6c85569f354c0233cc127362b4c0d8d
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: http://git-master/r/365977
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Shridhar Rasal <srasal@nvidia.com>
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
remove static mappings of syncpt ids, syncpt names, and
client managed syncpts
Bug 1305024
Change-Id: Ie457462e621e1fb39b484872fcf7556824fedc67
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: http://git-master/r/360966
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Shridhar Rasal <srasal@nvidia.com>
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
- implement dynamic syncpoint assignment framework
- provide API nvhost_get_syncpt_client_managed() to get client managed
syncpt which is requested by display, avp, and tegra_camera drivers
- provide API nvhost_get_syncpt_host_managed() for host managed syncpts
- this API is used by host1x clients and gpu
- provide API nvhost_free_syncpt() to free a syncpt
- both client managed and host managed APIs call nvhost_get_syncpt() to
find a free syncpt and reserve it
- remove static syncpt mappings from nvhost_device_data of all clients
- keep VBLANK syncpts hard-coded until we resolve the conflicts with
hard coded syncpt increments from bootloader
Bug 1305024
Change-Id: I403aa3f246de190707764fcac6ae295ff5f738f0
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: http://git-master/r/360963
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Shridhar Rasal <srasal@nvidia.com>
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
Do not crash if pm_runtime_get_sync() fails.
Change-Id: Ifb1e271fc7ccc22f3679d9643eab77506c0f9fbc
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: http://git-master/r/363789
|
|
For some devices we use QoS to define the minimum required frequency.
As this mechanism bypasses totally devfreq and the usual device profile,
we need to trigger postscale callback also in these cases to ensure that
i.e. EMC is scaled correctly at the same time.
Bug 1441874
Change-Id: I33545101157b015db240bfe9bb8a5c404469803c
Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
Reviewed-on: http://git-master/r/362487
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Tested-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
For dynamic channel mapping, when ref_count is zero,
it is unclear to which device channel maps to.
So moving channel device node to nvhost_device_data from
nvhost_channel.
Bug 1259844
Change-Id: I402d0b6ac4dcf72260f94e5ed387174fcfe0d2ab
Signed-off-by: Shridhar Rasal <srasal@nvidia.com>
Reviewed-on: http://git-master/r/360061
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
Introduce a new function to let gk20a fetch VRP info.
Bug 1409861
Change-Id: I6151d3439d22c0460fedfa7b5083beb1bf684580
Signed-off-by: Hiroshi Doyu <hdoyu@nvidia.com>
Signed-off-by: Vandana Salve <vsalve@nvidia.com>
Reviewed-on: http://git-master/r/353277
|
|
Introduce one set of functions to read/write host1x and host1x client
registers, and remove redundant ones.
Change-Id: I42ac706ec6c6f030175d0ee1d6a343ccaa7b7143
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: http://git-master/r/345613
|
|
Implement handling for gk20a semaphore_timeout, illegal_notify.
Report correct gk20a error codes to userspace drivers using
error notification interface.
Remove incorrectly done engine reset from non-mmu error handlers.
These errors already trigger mmu fault and engine reset is part of
mmu fault recovery.
Don't skip reset if gk20a_gr_isr() handles several interrupt
flags.
bug 1402282
bug 1355640
Signed-off-by: Kirill Artamonov <kartamonov@nvidia.com>
Change-Id: I622a26ccd1fb5efd1344083e3c1ffc0a79d7f8e4
Reviewed-on: http://git-master/r/336286
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
Add support for sync_fence interface for kernel modules.
Bug 1356564.
Change-Id: I851ac2e8abcc29d512ce4caf52ece502832f8742
Signed-off-by: Rakesh Iyer <riyer@nvidia.com>
Reviewed-on: http://git-master/r/336298
Reviewed-by: Jon Mayo <jmayo@nvidia.com>
Tested-by: Jon Mayo <jmayo@nvidia.com>
|
|
Create as devices only for units that support as. At the moment that
is only gk20a.
Bug 1389581
Change-Id: I7e6f9fbdf2ce1fb35cb07cd6c4c46a85247914e8
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: http://git-master/r/304717
Reviewed-by: Juha Tukkinen <jtukkinen@nvidia.com>
Tested-by: Juha Tukkinen <jtukkinen@nvidia.com>
|
|
A slave platform_device will be registered when we initialize the
master device. But if modprobe to init master device again, it will
try to initialize slave platform device again then oops happens.
And if we registered a platform_device in kernel, we normally don't
release it. So introduce a flag of nvhost_device_data to indicate
the status of slave and skip the slave init.
Bug 1377330
Change-Id: Ia4dfe6f8046952e2642343d23e39a14b6eb021b4
Signed-off-by: Bryan Wu <pengw@nvidia.com>
Reviewed-on: http://git-master/r/289326
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
Implement interface to pass gk20a error notification
to userspace.
bug 1297417
bug 1232244
Change-Id: If271a6f9c464e7b558de119d9746a7382d09fd62
Signed-off-by: Kirill Artamonov <kartamonov@nvidia.com>
Reviewed-on: http://git-master/r/299476
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Tested-by: Terje Bergstrom <tbergstrom@nvidia.com>
|
|
Change behavior of keepalive to allow clock gating, but prevent
power/rail gating.
Change-Id: Ia3e832daad77a0049992374a4e3cfe9f87cce663
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: http://git-master/r/288828
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Juha Tukkinen <jtukkinen@nvidia.com>
|
|
Framework for gk20a gpu profiler interface.
Bug 1228597
Change-Id: Id34614d59193e81c9f8ab9ed8ae173b38de71bb7
Signed-off-by: Ken Adams <kadams@nvidia.com>
(cherry picked from commit 9ff625237d1615f2b99c49e0f7b06f9f18e8b615)
Signed-off-by: Ajay Nandakumar <anandakumarm@nvidia.com>
|
|
We used actmon property to indicate whether the current device is
the gpu for edp. However, this is now incorrect as both VIC and
MSENC have actmons and gk20a has no actmon.
This patch adds a separate flag to indicate that the device should
be used as a GPU EDP device.
Change-Id: I0c8c5cfc215788b923c4f9b20b7afed8f2d44129
Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
Reviewed-on: http://git-master/r/280182
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Tested-by: Terje Bergstrom <tbergstrom@nvidia.com>
(cherry picked from commit ce45f8227377693388bb0a8cff1eef9a3c6cba2c)
Signed-off-by: Ajay Nandakumar <anandakumarm@nvidia.com>
|
|
We used to deliver this information by calling directly a function
inside nvhost_pod. This is not suitable for our needs as we want
to be able to use the same governor with several different (graphics)
devices in the same system.
This patch modifies tegra-throughput to use notifier chain for
delivering information about changed throughput hint.
Bug 1364804
Change-Id: Ifce1c84ae68cdcf54b00402efa80aee0b220bc9e
Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
(cherry picked from commit 9f10cf7082055f81c201ab5f5574c2f8fe8c6795)
Signed-off-by: Ajay Nandakumar <anandakumarm@nvidia.com>
|
|
Some units have strict linear relationship between unit clock and
required memory bandwidth, however, currently automatic EMC
relationship computation expects a non-linear relationship. This
patch adds a flag to nvhost device data to indicate if the
EMC should be scaled based using linear relationship.
Bug 1364804
Change-Id: I68fbb06de33fdeeb1c0e3c6b1db1f69c99e0e14e
Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
(cherry picked from commit 9897a5854b24a48d0acace8e4371cdefa44253a4)
Signed-off-by: Ajay Nandakumar <anandakumarm@nvidia.com>
|
|
Framework and implementation of a gk20a
debugger/profiler session interface.
Adds work toward optimized handling of
context patch write sequences. These introduce
cpu map/unmap operations and gpu l2 invaliates.
Unless we take care to coalesce them they occur
*per write*.
Change-Id: I8afc11a6f6782b80996404acbd01bffe9653ebdd
Signed-off-by: Ken Adams <kadams@nvidia.com>
Reviewed-on: http://git-master/r/274416
|
|
Allocate and request IOMEM resources for all clients before calling
nvhost_module_init(). Also moves setting of callbacks to the SoC
specific file.
Bug 1355069
Change-Id: Ic7483dba969be8e5e985d5abbba11393afdc2a2d
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: http://git-master/r/271960
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Juha Tukkinen <jtukkinen@nvidia.com>
|