Disable PCIe L1.2 on CP crash for stable operation.

Bug: 195651167
Test: Simulated over 3000 crashes and CP could resume.

Signed-off-by: Kisang Lee <kisang80.lee@samsung.com>
Change-Id: I8d8e5b1b6b632547fd6badbca20c15e4286b17fa

Bug: 196583508
Signed-off-by: George Lee <geolee@google.com>
Change-Id: I77da00481d59c1adc1f6153bd79eec5b490cd50d

Update settings from test data.

Bug: 196583508
Signed-off-by: George Lee <geolee@google.com>
Change-Id: Ifa8e5d631942478a6c84bb6b7e5cace6bf58ccde

Bug: 196889889
Test: 1. Phone sanity test
      2. Emergency call with multiple PDNs connected
Signed-off-by: Namkyu Kim <namkyu78.kim@samsung.com>
Change-Id: I84d6859327b8d61598acb8ecb554c604749e6d6e
Signed-off-by: Mars Lin <marslin@google.com>

Bug: 196889889
Test: 1. Phone sanity test
      2. Emergency call with multiple PDNs connected
Signed-off-by: Namkyu Kim <namkyu78.kim@samsung.com>
Change-Id: I8560badf35c90603fd8d20487d2664f59383e6d6
Signed-off-by: Mars Lin <marslin@google.com>

Bug: 196889889
Test: 1. Phone sanity test
      2. Emergency call with multiple PDNs connected
Signed-off-by: Namkyu Kim <namkyu78.kim@samsung.com>
Change-Id: I10d32fbe330b069075bfd01394568cec5e9ee73d
Signed-off-by: Mars Lin <marslin@google.com>

Bug: 196889889
Test: 1. Phone sanity test
      2. Emergency call with multiple PDNs connected
Signed-off-by: Namkyu Kim <namkyu78.kim@samsung.com>
Change-Id: I728cedb1dfc2a2e9c7d3581beee27cf5b01fb8a8
Signed-off-by: Mars Lin <marslin@google.com>

Bug:196889889
Test: 1. Phone sanity test
      2. Emergency call with multiple PDNs connected
Signed-off-by: Namkyu Kim <namkyu78.kim@samsung.com>
Change-Id: Icddf3534bed4d166f51a1687e10d3fadb06b9ff6
Signed-off-by: Mars Lin <marslin@google.com>

dit_init would get a skb alloc failure in dit_fill_rx_dst_data_buffer.
dit_init will check the below condition on the next call and there
will be no error.

dit_fill_rx_dst_data_buffer:
[...]
  if (initial && desc_info->dst_skb_buf[ring_num])
    return 0;
[...]

Some entries of both dst_desc and dst_skb_buf can be empty. Add a flag
to check if dst_skb_buf is filled.

Bug: 192712042
Test: Simulate skb alloc failure, then use perftest to send UDP data.

Signed-off-by: Dongseok Yi <dseok.yi@samsung.com>
Change-Id: I800971926c13fafd6a0c805844a5c443bb69c0d9
Signed-off-by: Mars Lin <marslin@google.com>

Bug: 196583508
Signed-off-by: George Lee <geolee@google.com>
Change-Id: I6834860a26e77e1fb810b6da819c89be65cc0b92

Update settings from test data.

Bug: 196583508
Signed-off-by: George Lee <geolee@google.com>
Change-Id: Idfd75344bc70900f0cbaf5936d410ad9ae893bbf

This reverts commit 601f3f03.

Reason for revert: Large increase in GPU crashes
Bug: 196300136
Bug: 196269975

Change-Id: Ice98b68e3f2706713f99f4a6a035ee251d52bae7

Bug: 196108695
Test: build
Signed-off-by: Wei Wang <wvw@google.com>
Change-Id: Ie8449e5bfa4f4f2bec3bebfcc5f3df6876e79f12

Move smp_rmb() to ensure tpu_data must have value.

Bug: 194307914
Test: boot to home and tpu pause/resume worked properly
Change-Id: I607925c56f6cd616f9907806389c2683a5ae165b
Signed-off-by: David Chao <davidchao@google.com>

With the most recent pipeline_input release for Tuscany, an additional
context model was used (v2.1 for auth) but enrollment is still using
v2.0 (intentional). That pushed the pipeline_input size from 46MB to
~53MB, just over the dma_heap famodel limit of ~52MB.

Bug: 196409876
Signed-off-by: Jidong Sun <jidong@google.com>
Change-Id: I9890ab87ec7752c308a599a1fc0077c407f87d61

CMA could be allocated from zram, move the xhci_dma reserved memory from
CMA to DMA coherent region.

Bug: 195433739
Test: build and boot pass
Change-Id: I37d218322e81922d2bdd6f8597b221a66007cdb3
Signed-off-by: Howard Yen <howardyen@google.com>

PDP need to access IPP when cleanup. Add sysmmu_ipp to PDP.

Bug: 191114282
Test: GCA, CTS
Signed-off-by: Holmes Chou <holmeschou@google.com>
Change-Id: I453749c913e3a2dc948ce6adc8e9a549bfbfcf59

Keep a copy of current RPMB write counter in ufs host
structure.

If we see the same RPMB write counter value in two
consecutive RPMB authenticated write requests,
print the stack to identify the driver flow.

Bug: 195243667
Test: suspend/resume test
Signed-off-by: Randall Huang <huangrandall@google.com>
Change-Id: I43e611cdfa002054f2341bb0e83121cd9bf91000

This reverts commit 6d4d98c2.

Reason for revert: b/196108695

Bug: 196108695
Signed-off-by: Wei Wang <wvw@google.com>
Change-Id: I038a7c559c23069f6a9529f21895d80e10cbd86a

Limit the max number of instances in BO kernel
driver.

Bug: 186772223

Signed-off-by: Ruofei Ma <ruofeim@google.com>
Change-Id: I594a6ba79b89d428da5cd4345f89e25d0cd46568

This is for fix PMU kernel panic when power down blkpwr_ipp
due to LH still transmission between PDP and IPP.
Add ipp core block to pdp to make pdp can write ipp stall
registers before PDP ip reset to prevent this issue.

Bug: 191114282
Test: GCA, CTS
Signed-off-by: Holmes Chou <holmeschou@google.com>
Change-Id: I6ae2214286e32073f0228808b6505a75cf4b65c4

When the core pci driver calls exynos_d3_sleep_hook(), the error value
that is passed in needs to be cleared, since there is an error check in
the calling code.

There are 2 instances of the pcie driver that are probed, one for wifi
and the other for modem. The vendor hook needs to be registered just
once or it ends up being called twice. Add a check to ensure that the
registration is just done once in the probe.

Bug: 196109714
Test: Verified using trace_printk that pci_pm_resume_noirq() takes 10ms.
Change-Id: Id08d9bc2e855b96adf9dbb0d22fc98165d043196
Signed-off-by: Sajid Dalvi <sdalvi@google.com>

Bug: 195267101
Change-Id: Id4b21dd904b25a1f74e8e60ef134d39a9de3ff59
Signed-off-by: Sunyoung Kang <sy0816.kang@samsung.com>

We will skip sync wake up if the cpu capacity is lower than task's
uclamp setting.

Bug: 194533502
Test: Build
Signed-off-by: Wei Wang <wvw@google.com>
Change-Id: If6c713dfa26a7f9d9406a70f6bc3a2ab6d24f7c3

Rarely, 0x0 is filled from the middle of dst_skb_buf till the end. Not
sure kvzalloc has an allocation issue. Use devm_kzalloc and remove
uncertainty.

00019670: 0008 1111 88ff ffff 000e 1111 88ff ffff  ................
00019680: 000b 1111 88ff ffff 0018 3811 88ff ffff  ..........8.....
00019690: 0000 0000 0000 0000 0000 0000 0000 0000  ................
000196a0: 0000 0000 0000 0000 0000 0000 0000 0000  ................

Bug: 192712042
Test: Embedded & tethering iperf3 DL/bidirectional test

Signed-off-by: Dongseok Yi <dseok.yi@samsung.com>
Change-Id: I452b7f7ca312ee83914719ca0a77454c8161da85
Signed-off-by: Mars Lin <marslin@google.com>

Removed the redundant smp_rmb() in pause function.

Bug: 194307914
Test: boot to home and tpu pause/resume worked properly
Change-Id: I8f0f8e15bb3c3ae1b32364830d8c5784357f030b
Signed-off-by: David Chao <davidchao@google.com>

Bug: 195663403
Change-Id: Ifa61be579a7439b6544d8a2bf3103652bc7bb9e5
Signed-off-by: Vincent Palomares <paillon@google.com>

Bug: 195663403
Change-Id: If252765f7f325bfc052d71d83f0b401b1d9bc75e
Signed-off-by: Vincent Palomares <paillon@google.com>

The default d3hot_delay() in pci.c uses msleep, which ends up making the
delay 20ms, instead of the PCIe spec value of 10ms.

To optimize resume time, this vendor hook changes the implementation to
usleep_range() to get closer to 10ms.

Bug: 194231641
Change-Id: I9d406845b64cced7e08902041627f2d9f7593a2c
Signed-off-by: Sajid Dalvi <sdalvi@google.com>

Add tpu pause/resume settings

Bug: 194307914
Test: boot to home and tpu pause/resume worked properly
Change-Id: I6150511e700fb36cd16dc1fd9587c137f0484b6d
Signed-off-by: David Chao <davidchao@google.com>

Added gs101-tmu header for defining TMU type in dt.

Bug: 194307914
Test: Build and boot
Change-Id: Ieb79d346718ca491336307d0687560388c38b177
Signed-off-by: David Chao <davidchao@google.com>

Add edgetpu suspend/resume atrace for debugging.

Bug: 194307914
Test: atrace worked properly
Change-Id: Ib9a4dea9ca459e8e09be0c1d9c0eb205c033f4ce
Signed-off-by: David Chao <davidchao@google.com>

Add tpu pause/resume feature

Bug: 194307914
Test: boot to home and tpu pause/resume worked properly
Change-Id: I2178ff75237404a19aa02ce0ed2fe912bf9b683e
Signed-off-by: David Chao <davidchao@google.com>

Related change in platform spcific xhci driver for moving
xhci_vendor_ops to xhci_hcd.

Bug: 194461020
Test: build and boot pass
Change-Id: Id0435af2788a7df181c86656b212122e5830a1a0
Signed-off-by: Howard Yen <howardyen@google.com>

This reverts commit cdc24d42.

Reason for revert: LDFW loads unsigned MFC fw directly

Bug:195214116
Change-Id: Ie90ec98ae712d590d154404bd806fec4604aeb03
Signed-off-by: Jidong Sun <jidong@google.com>

This reverts commit e77a5673.

Reason for revert: Extra memory for header is not needed.

Bug:195214116
Change-Id: I542b7045eacd2d780b0f57ab4108a80791dbb115
Signed-off-by: Jidong Sun <jidong@google.com>

Signed-off-by: Petri Gynther <pgynther@google.com>
Change-Id: I0e7ae38c2ad00091a24f433d21741435313bcf77

Bug: 194533502
Test: Build
Signed-off-by: Wei Wang <wvw@google.com>
Change-Id: I5f7b8e5577f4125d59992bf0b74382b91eaf855d

Not allowing to run disconnect flow when system started SYS_HALT.

Bug: 193406034
Test: USB host and peripheral modes work normally
Signed-off-by: Albert Wang <albertccwang@google.com>
Change-Id: I38b4e03fe589bf7648417791f95c7423054d363c