Age | Commit message (Collapse) | Author |
|
This reverts commit 239dac4f58e128248ad971ee518980b3e24e26c6 since it's
not compatible on v6.1.
Bug: 335301050
Change-Id: Iddcc813adce0e9d5f7359c9b6246d2e374f1a61e
Signed-off-by: Will McVicker <willmcvicker@google.com>
|
|
Upstream kernel commit [1] has changed the class_create()
API and dropped the first argument.
Follow this change here.
[1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=1aaba11da9aa
("driver core: class: remove module * from class_create()")
Bug: 283465996
Test: TH
Change-Id: Ic32581c464aaaa010c00c51d00ec0a0c58b77520
Signed-off-by: André Draszik <draszik@google.com>
Signed-off-by: Will McVicker <willmcvicker@google.com>
|
|
This let's you build the module from there.
Change-Id: If77cbe85b216f64c258a4908df62c18c70d82d43
Signed-off-by: Will McVicker <willmcvicker@google.com>
|
|
Change-Id: I9173d528126e5a1e550cdbfe997eb0faf496b316
Signed-off-by: Will McVicker <willmcvicker@google.com>
|
|
This was moved from the soc/gs project as part of tm-qpr2.
Change-Id: I4a9bc08c34a4287805f427765d8a3e5fe390af38
Signed-off-by: Will McVicker <willmcvicker@google.com>
|
|
userspace build will get it from linux/compiler.h
Bug: 262917450
Change-Id: Id1ee69b2bb638634bd94035259a4f8139566c0c8
Signed-off-by: Cyan_Hsieh <cyanhsieh@google.com>
(cherry picked from commit 06ad88e793b2f9cb40b04c5b2716ae4eab02b8ea)
|
|
Driver should always do SMC_PROTECTION_DISABLE regardless of job status.
Bug: 233393972
Bug: 267935059
Change-Id: Id9b78fb1cd7e8d5cdbd68511c8d6fc3a8a04cee2
Signed-off-by: Ruofei Ma <ruofeim@google.com>
(cherry picked from commit be7c095ef86dd9d083af7b0045765fa8406179f4)
|
|
Bug: 233569354
Signed-off-by: Robin Peng <robinpeng@google.com>
Change-Id: I321ed3a3b24a1373ffa2828f88cb39e6a9385719
|
|
By setting instance to idle at open, we a delaying the
power/freq vote until there is activity from this particular
instance. This change can avoid unnecessary power comsumption
on the instances that are opened but have no activity.
Without the change the statement above only works for the
first instance. With this change it works for all instances.
Refer to the bug for details.
Bug: 242975531
Signed-off-by: Ruofei Ma <ruofeim@google.com>
Change-Id: Ia871e9428995fea91a322f9245977210d2da39d9
|
|
Turn off bigocean clocks when it is idle for predefined timout.
This change is to save the power when user holds onto
bigocean instance but are not actually playing any video.
This can happen with video pause or app bug.
Bug: 242975531
Bug: 243650406
Signed-off-by: Ruofei Ma <ruofeim@google.com>
Change-Id: Ie4fa8ac0e3b5dcc9d552b257b8c32649a43d873c
(cherry picked from commit 51480d7d5ad2fbe2b18624ad266c390dded132ce)
|
|
Turn off bigocean clocks when it is idle for predefined timout.
This change is to save the power when user holds onto
bigocean instance but are not actually playing any video.
This can happen with video pause or app bug.
Bug: 242975531
Signed-off-by: Ruofei Ma <ruofeim@google.com>
Change-Id: Ie4fa8ac0e3b5dcc9d552b257b8c32649a43d873c
|
|
Bug: 233569354
Signed-off-by: Robin Peng <robinpeng@google.com>
Change-Id: Ia56497c9fe5dccd12e483208c5031fc84a1ee411
|
|
"inst" should not be dereferenced after it has been freed.
bug: 238462767
Signed-off-by: Vinay Kalia <vinaykalia@google.com>
Change-Id: Ifd5c165ea285d01878563ff21de452e5b523a7df
|
|
Bug: 233569354
Signed-off-by: Robin Peng <robinpeng@google.com>
Change-Id: I596748c902171beccd54fadbee532ac887771c1f
|
|
When all the bigocean instances close, all the BW votes should
be removed.
bug: 233707844
Signed-off-by: Vinay Kalia <vinaykalia@google.com>
Change-Id: I864e105977a965f7376de321df2899541e5f0f36
|
|
Revert the workaround made for b/180443732 as it is no longer
applicable.
bug: 233707844
Signed-off-by: Vinay Kalia <vinaykalia@google.com>
Change-Id: I0e784cced9037f8e0f26097402b3f051a8756e77
|
|
For 10-bit content at max resolutio and fps, the effective
load is more than the max defined load. Fix the MIF floor
for such heavy load usecases to prevent framedrops.
bug: 233707844
Signed-off-by: Vinay Kalia <vinaykalia@google.com>
Change-Id: I8da4eed15df9d76950095c3d1e2c03120363fa46
|
|
Destroying the instance while hardware is active can lead to faults like
SysMMU page faults.
bug: 236515139
Signed-off-by: Vinay Kalia <vinaykalia@google.com>
Change-Id: I20e037204db4a7a09bdcfbb512f8cddc96800768
|
|
Lock is needed when accessing core struct.
Bug: 231912091
Change-Id: Ie63352ca64df3469fb86121d621fda9dd37d66e6
Signed-off-by: Ruofei Ma <ruofeim@google.com>
|
|
|
|
When bigo_worker_thread does kref_put on an instance, if the refcount
changes from 1 to 0, bigo_close will be invoked. If this is the last
instance, on_last_inst_close will be invoked and it will wait for
bigo_worker_thread to exit. It is a dead loop that bigo_worker_thread
is waiting for itself to exit. The fix is to make sure the worker thread
is stopped before closing the last instance.
Bug: 229580907
Signed-off-by: Ruofei Ma <ruofeim@google.com>
Change-Id: I0950a07e072c6d10ab76389bd10ff697c3d1fc79
|
|
stat_with_irq will have stale value if interrupt timeout.
Read status from device register and update stat_with_irq.
Bug: 229580907
Signed-off-by: Ruofei Ma <ruofeim@google.com>
Change-Id: I17635113091fc006737c859415c0ed0bfc5ab3fb
|
|
Limit the max number of instances in BO kernel
driver to 16.
Bug: 226540966
Change-Id: I3f343e3a1f1d9345ad4ba470fc863dd94a214fb7
Signed-off-by: Ruofei Ma <ruofeim@google.com>
|
|
Unregister iommu fault handler if pt_client_register() returns -EPROBE_DEFER.
Bug: 210150963
Change-Id: I5556f86538532fc251a2d6077775a0f9f5f779f0
|
|
Add path to handle error from pt_client_register()
Bug: 210150963
Signed-off-by: Ruofei Ma <ruofeim@google.com>
Change-Id: Ia02ab93ae57d264732a08813b5dc8630c11e5b1e
|
|
Change bigocean kernel driver to async mode by adding
a kernel worker thread. This change will allow kernel
to serve requests based on instance priority.
bug: 197665952
Signed-off-by: Ruofei Ma <ruofeim@google.com>
Change-Id: I801fd1e6fff0f9e795c01dee4d4cfa3802e3d47a
|
|
Commit 6b47a64 ("Makefile: Consolidate build targets") introduced a
regression where it changed the default make target from "modules" to
"clean". As a result, bigocean.ko was not built at all! Change the
default target back to "modules".
Fixes: 6b47a64 ("Makefile: Consolidate build targets")
Signed-off-by: Daniel Mentz <danielmentz@google.com>
Change-Id: I7ca51f81b573eb773ada30dff6c2123d223bd0e9
|
|
The recipes for the build targets clean, modules, modules_install and
headers_install are basically identical. Let's consolidate them.
Signed-off-by: Daniel Mentz <danielmentz@google.com>
Change-Id: Idad42663bb497468bf31a2c9c3ec43fcd849f097
|
|
Previously, this makefile wasn't invoked with the headers_install
target. To work around this, we made modules_install depend on
headers_install. Now, that we have the commit "build.config: Install
UAPI headers of external modules", we no longer need this dependency.
Signed-off-by: Daniel Mentz <danielmentz@google.com>
Change-Id: Ic64a04cdb109d38aa3fc67e3d5d9e106bcd7744e
|
|
With the change aosp/1825055 ("build.sh: Pass INSTALL_HDR_PATH to
makefiles of external modules"), we no longer need to guess the correct
value of INSTALL_HDR_PATH because build.sh sets and exports it.
Signed-off-by: Daniel Mentz <danielmentz@google.com>
Change-Id: I285bb5d3c18d5038b86bea7f74d3a614d5730459
|
|
Test: builds
Bug: 192617687
Change-Id: Iaf40996951b3520bb139f3565994b19f1e087ae9
|
|
Move the BigOcean uapi bigo.h from gs-google
to google-modules and install it.
Bug: 195458188
Signed-off-by: Ruofei Ma <ruofeim@google.com>
Change-Id: Iebb2313e6b5c49d4bc71c6d4bac16434de517e1c
|
|
Bug: 195458188
Signed-off-by: Ruofei Ma <ruofeim@google.com>
Change-Id: I5975a32a5a1ce5593a84f455f54c887191fb00fe
|
|
|