drm/amd/display: Add ABM command structs to DMCUB

[Why]
Moving ABM from DMCU to DMCUB.

[How]
Add ABM command structs.

Signed-off-by: Wyatt Wood <wyatt.wood@amd.com>
Reviewed-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: 3.2.75

Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Clear link settings on MST disable connector

[Why]
If we have a single MST display and we disconnect it, we dont disable that
link. This causes the old link settings to still exist

Now on a replug for MST we think its a link loss and will try to reallocate
mst payload which will fail, throwing warning below.

[  129.374192] [drm] Failed to updateMST allocation table forpipe idx:0
[  129.374206] ------------[ cut here ]------------
[  129.374284] WARNING: CPU: 14 PID: 1710 at
drivers/gpu/drm/amd/amdgpu/../dal-dev/dc/core/dc_link.c:3153
dc_link_allocate_mst_payload+0x1f7/0x220 [amdgpu]

[  129.374285] Modules linked in: amdgpu(OE) amd_iommu_v2 gpu_sched ttm
drm_kms_helper drm fb_sys_fops syscopyarea sysfillrect sysimgblt
binfmt_misc nls_iso8859_1 edac_mce_amd snd_hda_codec_realtek
snd_hda_codec_generic ledtrig_audio kvm snd_hda_codec_hdmi snd_hda_intel
snd_intel_nhlt snd_hda_codec irqbypass snd_hda_core snd_hwdep snd_pcm
snd_seq_midi snd_seq_midi_event snd_rawmidi crct10dif_pclmul snd_seq
crc32_pclmul ghash_clmulni_intel snd_seq_device snd_timer snd aesni_intel
eeepc_wmi crypto_simd asus_wmi joydev cryptd sparse_keymap input_leds
soundcore video glue_helper wmi_bmof mxm_wmi k10temp ccp mac_hid
sch_fq_codel parport_pc ppdev lp parport ip_tables x_tables autofs4
hid_generic usbhid hid igb i2c_algo_bit ahci dca i2c_piix4 libahci
gpio_amdpt wmi gpio_generic

[  129.374318] CPU: 14 PID: 1710 Comm: kworker/14:2 Tainted: G        W  OE     5.4.0-rc7bhawan+ #480
[  129.374318] Hardware name: System manufacturer System Product Name/PRIME X370-PRO, BIOS 0515 03/30/2017
[  129.374397] Workqueue: events dm_irq_work_func [amdgpu]
[  129.374468] RIP: 0010:dc_link_allocate_mst_payload+0x1f7/0x220 [amdgpu]
[  129.374470] Code: 52 20 e8 1c 63 ad f4 48 8b 5d d0 65 48 33 1c 25 28 00
00 00 b8 01 00 00 00 75 16 48 8d 65 d8 5b 41 5c 41 5d 41 5e 41 5f 5d c3
<0f> 0b e9 fa fe ff ff e8 ed 5b d6 f3 41 0f b6 b6 c4 02 00 00 48 c7
[  129.374471] RSP: 0018:ffff9f9141e7fcc0 EFLAGS: 00010246
[  129.374472] RAX: 0000000000000000 RBX: ffff91ef0762f800 RCX: 0000000000000000
[  129.374473] RDX: 0000000000000005 RSI: ffffffffc0c4a988 RDI: 0000000000000004
[  129.374474] RBP: ffff9f9141e7fd10 R08: 0000000000000005 R09: 0000000000000000
[  129.374475] R10: 0000000000000002 R11: 0000000000000001 R12: ffff91eebd510c00
[  129.374475] R13: ffff91eebd510e58 R14: ffff91ef052c01b8 R15: 0000000000000006
[  129.374476] FS:  0000000000000000(0000) GS:ffff91ef0ef80000(0000) knlGS:0000000000000000
[  129.374477] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  129.374478] CR2: 000055623ea01d50 CR3: 0000000408a8c000 CR4: 00000000003406e0
[  129.374479] Call Trace:
[  129.374550]  dc_link_reallocate_mst_payload+0x12e/0x150 [amdgpu]
[  129.374617]  dc_link_handle_hpd_rx_irq+0x6d4/0x6e0 [amdgpu]
[  129.374693]  handle_hpd_rx_irq+0x77/0x310 [amdgpu]
[  129.374768]  dm_irq_work_func+0x53/0x70 [amdgpu]
[  129.374774]  process_one_work+0x1fd/0x3f0
[  129.374776]  worker_thread+0x255/0x410
[  129.374778]  kthread+0x121/0x140
[  129.374780]  ? process_one_work+0x3f0/0x3f0
[  129.374781]  ? kthread_park+0x90/0x90
[  129.374785]  ret_from_fork+0x22/0x40

[How]
when we disable MST we should clear the cur link settings (lane_count=0 is
good enough). This will cause us to not reallocate payloads earlier than
expected and not throw the warning

Signed-off-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Reviewed-by: Hersen Wu <hersenxs.wu@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add 'disable FEC for specific monitor' infrastructure to DC

Disabling FEC for specific monitors is sometimes required for debugging
while in the monitor bringup phase

Signed-off-by: Nikola Cornij <nikola.cornij@amd.com>
Reviewed-by: Alvin Lee <Alvin.Lee2@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Set clock optimization required after update clocks

[WHY]
We see an issue that caused clk_optimized_required to be set true in
certain cases, causing passive flips to fail. This is because of a typo
where wm_optimized_required was set twice.

[HOW]
Set clk_optimized_required to false after updating clocks.

Signed-off-by: Sung Lee <sung.lee@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: change number of cursor policy for dml calculation.

[Why]
vstartup calculation is incorrect due to use 2 number of cursors and
result in an underflow when playing video in full screen mode and
combines graphic plane and video plane.

[How]
Apply new policy for dml calculation.
1 cursor for graphic plane, 0 cursor for video plane.

Signed-off-by: Yongqiang Sun <yongqiang.sun@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix dcc swath size calculations on dcn1

[Why]
Swath sizes are being calculated incorrectly. The horizontal swath size
should be the product of block height, viewport width, and bytes per
element, but the calculation uses viewport height instead of width. The
vertical swath size is similarly incorrectly calculated. The effect of
this is that we report the wrong DCC caps.

[How]
Use viewport width in the horizontal swath size calculation and viewport
height in the vertical swath size calculation.

Signed-off-by: Josip Pavic <Josip.Pavic@amd.com>
Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Disable freesync borderless on Renoir

[WHY]
Freesync borderless is not meant to be enabled on any APUs

[HOW]
Add is_apu cap to dcn21_resource_construct for correct recognition

Signed-off-by: Michael Strauss <michael.strauss@amd.com>
Reviewed-by: Yongqiang Sun <yongqiang.sun@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Remove DISPCLK Limit Floor for Certain SMU Versions

[WHY]
SMU FW previously had an issue with lowering display clock to below 100
MHz, and a workaround was put in to limit it.  Newest SMU FW does not
have this issue, and no longer needs the 100MHz cap.

[HOW]
Remove the 100MHz cap based on the SMU FW version.

Signed-off-by: Sung Lee <sung.lee@amd.com>
Reviewed-by: Yongqiang Sun <yongqiang.sun@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Workaround to do HDCP authentication twice on certain displays

[Why]
When transitioning from SST to MST, the HDCP repeater in some MST
displays will enter a bad state. The HDCP repeater is recovered after
failing and performing authentication again.

[How]
Add monitor patch to trigger HDCP authentication failure after
encryption is enabled and re-authenticate.

Signed-off-by: George Shen <george.shen@amd.com>
Reviewed-by: Wenjing Liu <Wenjing.Liu@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix image corruption with ODM 2:1 DSC 2 slice

[why]
When combining two or more pipes in DSC mode, there will always be more
than 1 slice per line.  In this case, as per DSC rules, the sink device
is expecting that the ICH is reset at the end of each slice line (i.e.
ICH_RESET_AT_END_OF_LINE must be configured based on the number of
slices at the output of ODM).  It is recommended that software set
ICH_RESET_AT_END_OF_LINE = 0xF for each DSC in the ODM combine.  However
the current code only set ICH_RESET_AT_END_OF_LINE = 0xF when number of
slice per DSC engine is greater than 1 instead of number of slice per
output after ODM combine.

[how]
Add is_odm in dsc config. Set ICH_RESET_AT_END_OF_LINE = 0xF if either
is_odm or number of slice per DSC engine is greater than 1.

Signed-off-by: Wenjing Liu <Wenjing.Liu@amd.com>
Reviewed-by: Nikola Cornij <Nikola.Cornij@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: program DPG_OFFSET_SEGMENT for odm_pipe

[why]
When test pattern is enabled with ODM combine, test pattern is generated
by piecing multiple DPGs image together.  The current code will program
all DPGs with horizontal offset of 0. This will cause all DPGs to output
the beginning of the pattern. Instead each DPG should program a
horizontal offset of its x position to form a continous pattern when
pieced together.

Signed-off-by: Wenjing Liu <Wenjing.Liu@amd.com>
Reviewed-by: Nikola Cornij <Nikola.Cornij@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add vsc update support for test pattern request

[how]
Allow vsc info packet if vsc is supported.  Update vsc based on test
pattern request.  Remove dpg_is_blanked polling, apply hardware global
lock instead to ensure double buffered dpg is updated with vsc in one
frame

Signed-off-by: Wenjing Liu <Wenjing.Liu@amd.com>
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix typo "to found" -> "to find"

[Why]
Typo in amdgpu_dm error message:
"Failed to found connector for link!"
[How]
1. Replace with:
"Failed to find connector for link!"
2. Fix indentation checkpatch warnings.

Signed-off-by: Roman Li <roman.li@amd.com>
Reviewed-by: Zhan Liu <Zhan.Liu@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Make clock table struct more accessible

[WHY & HOW]
In order to correctly intepret clock table, num_states is also needed.
This field did not get moved with clock_table but should next to it for
easier access/viewing.

Signed-off-by: Sung Lee <sung.lee@amd.com>
Reviewed-by: Eric Yang <eric.yang2@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add worst case dcc meta pitch to fake plane

[why]
When we have single channel memory, we can not light up 2 4k displays
with a 1080p edp, because we don't have enough bw by a small margin.
this small margin comes from dcc meta being too large. We however don't
have this dcc meta when we create fake planes so, before the flip we
will not filter out the mode for 2 4k displays with a 1080p edp

[how]
Change get_default_swizzle_mode to something more general so we don't
end up with a separate function for every missing field in the fake
plane. Add a reasonable dcc meta to the fake plane when it is filled in,
so we filter out modes that don't have enough bandwidth. To do this, we
take the screen width and align it to 1024(8k 60)

Signed-off-by: Joseph Gravenor <joseph.gravenor@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: only include FEC overhead if both asic and display support FEC

[why]
Some asics don't support FEC but FEC overhead is added into link
bandwidth calculation by mistake. This causes certain timing cannot be
validated.

[how]
Only include FEC overhead if both asic and display support FEC.

Signed-off-by: Wenjing Liu <Wenjing.Liu@amd.com>
Reviewed-by: Ashley Thomas <Ashley.Thomas2@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add driver support for enabling PSR on DMCUB

[Why]
We want to be able to enable PSR on DMCUB, and fallback to
DMCU when necessary.

[How]
Move psr_on_dmub flag from dc_debug_options to dc_config.

Signed-off-by: Wyatt Wood <wyatt.wood@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Move mod_hdcp_displays to mod_hdcp struct

[Why]
Reset connection is called before remove display in mod_hdcp. When
remove display is called, the display structure has been zeroed from
reset connection. Since no displays can be found, remove display does
not properly reset the hardware. This causes validation errors when
another display is plugged into a different port afterwards.

[How]
Moved displays structure out of connections structure. Displays
structure is now directly within mod_hdcp. Displays is no longer zeroed
when reset connection is called.

Signed-off-by: Isabel Zhang <isabel.zhang@amd.com>
Reviewed-by: Wenjing Liu <Wenjing.Liu@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Acked-by: Tony Cheng <Tony.Cheng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Link training TPS1 workaround add back in dpcd

[Why]
Previously implemented early_cr_pattern we mistook dp_hw_link_settings
for a redundant call of dpcd_set_link_settings

[How]
revert the changes to dpcd_set_link_settings calls for this workaround.
Do not need to revert the entire change since it only affects patched
case

Signed-off-by: Martin Leung <martin.leung@amd.com>
Reviewed-by: David Galiffi <David.Galiffi@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Rearm IRQ in Navi10 SR-IOV if IRQ lost

Ported from Vega10. SDMA stress tests sometimes see IRQ lost.

Signed-off-by: Samir Dhume <samir.dhume@amd.com>
Reviewed-by: Monk Liu <monk.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: stop using sratch_reg in IB test

scratch_reg0 is used by RLCG for register access usage
in SRIOV case.

both CP firmware and driver can invoke RLCG to do
certain register access (through scratch_reg0/1/2/3)
but rlcg now dosen't have race concern so if two
clients are in parallel doing the RLCG reg access
then we are colliding,

GFX IB test is a runtime work, so it is forbidden
to use scrach_reg0/1/2/3 during IB test period

note:
Although we can only have this change for SRIOV, but
looks it doesn't worth the effort to differentiate
bare-metal with SRIOV on the GFX ib test

Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix IB test MCBP bug

1)for gfx IB test we shouldn't insert DE meta data

2)we should make sure IB test finished before we
send event 3 to hypervisor otherwise the IDLE from
event 3 will preempt IB test, which is not designed
as a compatible structure for MCBP

Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: disable 3D pipe 1 on Navi1x

[why]
CP firmware decide to skip setting the state for 3D pipe 1 for Navi1x as there
is no use case.

[how]
Disable 3D pipe 1 on Navi1x.

Reviewed-by: Feifei Xu <Feifei.Xu@amd.com>
Reviewed-by: Monk Liu <monk.liu@amd.com>
Signed-off-by: Tianci.Yin <tianci.yin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add debugfs interface to set arbitrary sclk for navi14 (v2)

add debugfs interface amdgpu_force_sclk
to set arbitrary sclk for navi14

v2: Add lock

Signed-off-by: Chengming Gui <Jack.Gui@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add lock option for smu_set_soft_freq_range()

Add lock_needed param for smu_set_soft_freq_range()

Signed-off-by: Chengming Gui <Jack.Gui@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: dc_link: code clean up on detect_dp function

Removes codestyle issues on detect_dp function as suggested by
checkpatch.pl.

CHECK: Lines should not end with a '('
WARNING: Missing a blank line after declarations
WARNING: line over 80 characters
CHECK: Alignment should match open parenthesis

Signed-off-by: Melissa Wen <melissa.srw@gmail.com>
Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Reviewed-by: Zhan Liu <zhan.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: dc_link: code clean up on enable_link_dp function

Coding style clean up on enable_link_dp function as suggested by
checkpatch.pl:

CHECK: Lines should not end with a '('
WARNING: line over 80 characters
WARNING: suspect code indent for conditional statements (8, 24)
CHECK: braces {} should be used on all arms of this statement
ERROR: else should follow close brace '}'
CHECK: Comparison to NULL could be written
       "link->preferred_training_settings.fec_enable"

Signed-off-by: Melissa Wen <melissa.srw@gmail.com>
Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix dmub_psr_destroy()

This is freeing the wrong variable so it will crash.  It should be
freeing "*dmub" instead of "dmub".

Fixes: 4c1a1335dfe0 ("drm/amd/display: Driverside changes to support PSR in DMCUB")
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: fix indentation issue on a hunk of code

There are multiple statements that are indented incorrectly. Add
in the missing tabs.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: fix indentation issue

There is a statement that is indented with spaces instead of a tab.
Replace spaces with a tab.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Write blocked CP registers using RLC on VF

This change programs CP_ME_CNTL and RLC_CSIB_* through RLC

Signed-off-by: Rohit Khaire <Rohit.Khaire@amd.com>
Reviewed-by: Monk Liu <monk.liu@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: clean wptr on wb when gpu recovery

The TDR will be randomly failed due to compute ring
test failure. If the compute ring wptr & 0x7ff(ring_buf_mask)
is 0x100 then after map mqd the compute ring rptr will be
synced with 0x100. And the ring test packet size is also 0x100.
Then after invocation of amdgpu_ring_commit, the cp will not
really handle the packet on the ring buffer because rptr is equal to wptr.

Signed-off-by: Yintian Tao <yttao@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Monk Liu <Monk.Liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/smu: Add message sending lock

This adds a message lock to the smu_send_smc_msg* implementations to
protect against concurrent access to the mmu registers used to
communicate with the SMU

v2: Implement for smu_v12_0 as well

v3: Add mutex_init for message_lock

Signed-off-by: Matt Coffin <mcoffin13@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/powerplay: Remove deprecated smc_read_arg

The new interface reads the argument in the call to send the message, so
this is no longer needed, and shouldn't be used for concurrency safety
reasons.

Signed-off-by: Matt Coffin <mcoffin13@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/powerplay: Refactor SMU message handling for safety

Move the responsibility for reading argument registers into the
smu_send_smc_msg* implementations, so that adding a message-sending lock
to protect the SMU registers will result in the lock still being held
when the argument is read.

v2: transition smu_v12_0, it's asics, and vega20

Signed-off-by: Matt Coffin <mcoffin13@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/powerplay: nv1x, renior copy dcn clock settings of watermark to smu during boot up

dc to pplib interface is changed for navi1x, renoir.
display_config_changed is not called by dc anymore.
smu_write_watermarks_table is not executed for navi1x, renoir
during boot up.

solution: call smu_write_watermarks_table just after dc pass
watermark clock settings to pplib

Signed-off-by: Hersen Wu <hersenxs.wu@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: release drm_device after amdgpu_driver_unload_kms

If we release drm_device before amdgpu_driver_unload_kms,
then it will raise the error below. Therefore, we need to
place it before amdgpu_driver_unload_kms.
[   43.055736] Memory manager not clean during takedown.
[   43.055777] WARNING: CPU: 1 PID: 2807 at /build/linux-hwe-9KJ07q/linux-hwe-4.18.0/drivers/gpu/drm/drm_mm.c:913 drm_mm_takedown+0x24/0x30 [drm]
[   43.055778] Modules linked in: amdgpu(OE-) amd_sched(OE) amdttm(OE) amdkcl(OE) amd_iommu_v2 drm_kms_helper drm i2c_algo_bit fb_sys_fops syscopyarea sysfillrect sysimgblt snd_hda_codec_generic nfit kvm_intel kvm irqbypass crct10dif_pclmul crc32_pclmul snd_hda_intel snd_hda_codec snd_hda_core snd_hwdep snd_pcm ghash_clmulni_intel snd_seq_midi snd_seq_midi_event pcbc snd_rawmidi snd_seq snd_seq_device aesni_intel snd_timer joydev aes_x86_64 crypto_simd cryptd glue_helper snd soundcore input_leds mac_hid serio_raw qemu_fw_cfg binfmt_misc sch_fq_codel nfsd auth_rpcgss nfs_acl lockd grace sunrpc parport_pc ppdev lp parport ip_tables x_tables autofs4 hid_generic floppy usbhid psmouse hid i2c_piix4 e1000 pata_acpi
[   43.055819] CPU: 1 PID: 2807 Comm: modprobe Tainted: G           OE     4.18.0-15-generic #16~18.04.1-Ubuntu
[   43.055820] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.12.0-1 04/01/2014
[   43.055830] RIP: 0010:drm_mm_takedown+0x24/0x30 [drm]
[   43.055831] Code: 84 00 00 00 00 00 0f 1f 44 00 00 48 8b 47 38 48 83 c7 38 48 39 c7 75 02 f3 c3 55 48 c7 c7 38 33 80 c0 48 89 e5 e8 1c 41 ec d0 <0f> 0b 5d c3 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 55 48 89 e5 41
[   43.055857] RSP: 0018:ffffae33c1393d28 EFLAGS: 00010286
[   43.055859] RAX: 0000000000000000 RBX: ffff9651b4a29800 RCX: 0000000000000006
[   43.055860] RDX: 0000000000000007 RSI: 0000000000000096 RDI: ffff9651bfc964b0
[   43.055861] RBP: ffffae33c1393d28 R08: 00000000000002a6 R09: 0000000000000004
[   43.055861] R10: ffffae33c1393d20 R11: 0000000000000001 R12: ffff9651ba6cb000
[   43.055863] R13: ffff9651b7f40000 R14: ffffffffc0de3a10 R15: ffff9651ba5c6460
[   43.055864] FS:  00007f1d3c08d540(0000) GS:ffff9651bfc80000(0000) knlGS:0000000000000000
[   43.055865] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   43.055866] CR2: 00005630a5831640 CR3: 000000012e274004 CR4: 00000000003606e0
[   43.055870] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[   43.055871] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[   43.055871] Call Trace:
[   43.055885]  drm_vma_offset_manager_destroy+0x1b/0x30 [drm]
[   43.055894]  drm_gem_destroy+0x19/0x40 [drm]
[   43.055903]  drm_dev_fini+0x7f/0x90 [drm]
[   43.055911]  drm_dev_release+0x2b/0x40 [drm]
[   43.055919]  drm_dev_unplug+0x64/0x80 [drm]
[   43.055994]  amdgpu_pci_remove+0x39/0x70 [amdgpu]
[   43.055998]  pci_device_remove+0x3e/0xc0
[   43.056001]  device_release_driver_internal+0x18a/0x260
[   43.056003]  driver_detach+0x3f/0x80
[   43.056004]  bus_remove_driver+0x59/0xd0
[   43.056006]  driver_unregister+0x2c/0x40
[   43.056008]  pci_unregister_driver+0x22/0xa0
[   43.056087]  amdgpu_exit+0x15/0x57c [amdgpu]
[   43.056090]  __x64_sys_delete_module+0x146/0x280
[   43.056094]  do_syscall_64+0x5a/0x120

v2: put drm_dev_put after pci_set_drvdata

Signed-off-by: Yintian Tao <yttao@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: no need to clean debugfs at amdgpu

drm_minor_unregister will invoke drm_debugfs_cleanup
to clean all the child node under primary minor node.
We don't need to invoke amdgpu_debugfs_fini and
amdgpu_debugfs_regs_cleanup to clean agian.
Otherwise, it will raise the NULL pointer like below.
[   45.046029] BUG: unable to handle kernel NULL pointer dereference at 00000000000000a8
[   45.047256] PGD 0 P4D 0
[   45.047713] Oops: 0002 [#1] SMP PTI
[   45.048198] CPU: 0 PID: 2796 Comm: modprobe Tainted: G        W  OE     4.18.0-15-generic #16~18.04.1-Ubuntu
[   45.049538] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.12.0-1 04/01/2014
[   45.050651] RIP: 0010:down_write+0x1f/0x40
[   45.051194] Code: 90 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 55 48 89 e5 53 48 89 fb e8 ce d9 ff ff 48 ba 01 00 00 00 ff ff ff ff 48 89 d8 <f0> 48 0f c1 10 85 d2 74 05 e8 53 1c ff ff 65 48 8b 04 25 00 5c 01
[   45.053702] RSP: 0018:ffffad8f4133fd40 EFLAGS: 00010246
[   45.054384] RAX: 00000000000000a8 RBX: 00000000000000a8 RCX: ffffa011327dd814
[   45.055349] RDX: ffffffff00000001 RSI: 0000000000000001 RDI: 00000000000000a8
[   45.056346] RBP: ffffad8f4133fd48 R08: 0000000000000000 R09: ffffffffc0690a00
[   45.057326] R10: ffffad8f4133fd58 R11: 0000000000000001 R12: ffffa0113cff0300
[   45.058266] R13: ffffa0113c0a0000 R14: ffffffffc0c02a10 R15: ffffa0113e5c7860
[   45.059221] FS:  00007f60d46f9540(0000) GS:ffffa0113fc00000(0000) knlGS:0000000000000000
[   45.060809] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   45.061826] CR2: 00000000000000a8 CR3: 0000000136250004 CR4: 00000000003606f0
[   45.062913] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[   45.064404] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[   45.065897] Call Trace:
[   45.066426]  debugfs_remove+0x36/0xa0
[   45.067131]  amdgpu_debugfs_ring_fini+0x15/0x20 [amdgpu]
[   45.068019]  amdgpu_debugfs_fini+0x2c/0x50 [amdgpu]
[   45.068756]  amdgpu_pci_remove+0x49/0x70 [amdgpu]
[   45.069439]  pci_device_remove+0x3e/0xc0
[   45.070037]  device_release_driver_internal+0x18a/0x260
[   45.070842]  driver_detach+0x3f/0x80
[   45.071325]  bus_remove_driver+0x59/0xd0
[   45.071850]  driver_unregister+0x2c/0x40
[   45.072377]  pci_unregister_driver+0x22/0xa0
[   45.073043]  amdgpu_exit+0x15/0x57c [amdgpu]
[   45.073683]  __x64_sys_delete_module+0x146/0x280
[   45.074369]  do_syscall_64+0x5a/0x120
[   45.074916]  entry_SYSCALL_64_after_hwframe+0x44/0xa9

v2: remove all debugfs cleanup/fini code at amdgpu
v3: squash in unused variable removal

Signed-off-by: Yintian Tao <yttao@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Initialize SPM_VMID with 0xf (v2)

SPM_VMID is a global resource, SPM access the video memory according to
SPM_VMID. The initial valude of SPM_VMID is 0 which is used by kernel.
That means UMD can overwrite the memory of VMID0 by enabling SPM, that
is really dangerous.

Initialize SPM_VMID with 0xf, it messes up other user mode process at
most.

v2: squash in indentation fix

Signed-off-by: Jacob He <jacob.he@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/sriov: Use kiq to copy the gpu clock

For vega10 sriov, the register is blocked, use
copy data command to fix the issue.

v2: Rename amdgpu_kiq_read_clock to gfx_v9_0_kiq_read_clock.

Signed-off-by: Emily Deng <Emily.Deng@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: change SDMA MQD memory type

SDMA MQD memory type is NC that causes MQD data overwritten
accidentally by an old stable cache line. Changing it to UC
default for GART will fix the issue.

The mqd_gfx9 parameter is meant for control stacks that are
allocated together with user mode queue MQDs. Setting
mqd_gfx9 to true maps the control stack pages as NC.
Here it was accidentally applied to SDMA MQDs,
which are allocated together with the HIQ MQD. Setting
the mqd_gfx9 to false avoids that.

Signed-off-by: Eric Huang <jinhuieric.huang@amd.com>
Acked-by: Yong Zhao <Yong.Zhao@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Make get_tile_config() generic

Given we can query all the asic specific information from amdgpu_gfx_config,
we can make get_tile_config() generic.

Signed-off-by: Yong Zhao <Yong.Zhao@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add num_banks and num_ranks to gfx config structure

The two members will be used by KFD later.

Signed-off-by: Yong Zhao <Yong.Zhao@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Merge tag 'drm-misc-next-2020-02-27' of git://anongit.freedesktop.org/drm/drm-misc into drm-next

drm-misc-next for 5.7

UAPI Changes:

Cross-subsystem Changes:

Core Changes:
  - bridge: huge rework to get rid of omap_dss custom display drivers

Driver Changes:
  - hisilicon: some fixes related to modes it can deal with / default to
  - virtio: shmem and gpu context fixes and enhancements
  - sun4i: Support for LVDS on the A33

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Maxime Ripard <maxime@cerno.tech>
Link: https://patchwork.freedesktop.org/patch/msgid/20200227113222.cdwzy4cvcqjtbmou@gilmour.lan

Merge tag 'amd-drm-next-5.7-2020-02-26' of git://people.freedesktop.org/~agd5f/linux into drm-next

amd-drm-next-5.7-2020-02-26:

amdgpu:
- Rework VM update handling in preparation for HMM support
- HDCP srm support
- PSR fixes
- DC watermark fixes
- OLED panel support
- SR-IOV fixes
- BACO fixes
- Optimize debugging vram access
- RAS fixes
- Use BACO for runtime pm
- HDCP fixes
- XGMI fixes
- DDC fixes
- DC clock programming optimizations and fixes
- PSP fw loading sequence updates
- Drop DRIVER_USE_AGP
- Remove legacy drm load and unload callbacks

amdkfd:
- Add runtime pm support

radeon:
- Drop DRIVER_USE_AGP

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200227043142.4075-1-alexander.deucher@amd.com

drm/virtio: add virtio_gpu_is_shmem helper

The plan is use have both shmem and virtual "vram" running
side-by-side in virtio-gpu. It looks like we'll eventually use
struct drm_gem_object as a base class, and we'll need to convert
to shmem and vram objects on the fly. As a first step, add a
virtio_gpu_is_shmem helper. Thanks to kraxel for suggesting this
approach on Gitlab.

Suggested-by: Gerd Hoffman <kraxel@redhat.com>
Signed-off-by: Gurchetan Singh <gurchetansingh@chromium.org>
Link: http://patchwork.freedesktop.org/patch/msgid/20200227002601.745-3-gurchetansingh@chromium.org
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

drm/virtio: make mmap callback consistent with callbacks

This is a very, very minor cleanup.

Signed-off-by: Gurchetan Singh <gurchetansingh@chromium.org>
Link: http://patchwork.freedesktop.org/patch/msgid/20200227002601.745-2-gurchetansingh@chromium.org
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

drm/hisilicon: Fixed pcie resource conflict between drm and firmware

use the drm_fb_helper_remove_conflicting_pci_framebuffer to remove
the framebuffer initialized by fireware/bootloader to avoid resource
conflict.

Signed-off-by: Tian Tao <tiantao6@hisilicon.com>
Acked-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Xinliang Liu <xinliang.liu@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/1582524112-5628-1-git-send-email-tiantao6@hisilicon.com

drm/hisilicon: Set preferred mode resolution and maximum resolution

set the preferred mode resolution to 1024 * 768 and maximum
resolution to 1920 * 1200.

Signed-off-by: Tian Tao <tiantao6@hisilicon.com>
Signed-off-by: Gong junjie <gongjunjie2@huawei.com>
Signed-off-by: Xinliang Liu <xinliang.liu@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/1582080707-18825-1-git-send-email-tiantao6@hisilicon.com

drm/hisilicon: Add the mode_valid function

add mode_valid function, we can make sure the resolution is valid.

Signed-off-by: Tian Tao <tiantao6@hisilicon.com>
Signed-off-by: Gong junjie <gongjunjie2@huawei.com>
Reviewed-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Xinliang Liu <xinliang.liu@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/1578476501-45807-1-git-send-email-tiantao6@hisilicon.com

drm/hisilicon: fixed the wrong resolution configurations

The maximum resolution supported by hibmc is 1920 * 1200 instead of
1920 * 1440, this patch fixed this problem

Signed-off-by: Tian Tao <tiantao6@hisilicon.com>
Signed-off-by: Gong junjie <gongjunjie2@huawei.com>
Signed-off-by: Xinliang Liu <xinliang.liu@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/1578023756-46567-1-git-send-email-tiantao6@hisilicon.com

drm/hisilicon: Enable the shadowfb for hibmc

set the prefer_shadow as 1,so we use the shadowfb to acceleration.

Signed-off-by: Tian Tao <tiantao6@hisilicon.com>
Signed-off-by: Gong junjie <gongjunjie2@huawei.com>
Signed-off-by: Xinliang Liu <xinliang.liu@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/1578712146-16327-1-git-send-email-tiantao6@hisilicon.com

drm/hisilicon: Add new clock/resolution configurations

Add the three new pll config for corresponding resolution 1440x900 and
1600x900, 640x480 for hibmc

Signed-off-by: Tian Tao <tiantao6@hisilicon.com>
Signed-off-by: Gong junjie <gongjunjie2@huawei.com>
Signed-off-by: Xinliang Liu <xinliang.liu@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/1577774571-60493-1-git-send-email-tiantao6@hisilicon.com

Merge tag 'drm-intel-next-2020-02-25' of git://anongit.freedesktop.org/drm/drm-intel into drm-next

- A backmerge of drm-next solving conflicts on i915/gt/intel_lrc.c
- Clean up shadow batch after I915_EXEC_SECURE
- Drop assertion that active->fence is unchanged

Here goes drm-intel-next-2020-02-25:
- A backmerge of drm-next solving conflicts on i915/gt/intel_lrc.c
- Clean up shadow batch after I915_EXEC_SECURE
- Drop assertion that active->fence is unchanged
drm-intel-next-2020-02-24-1:
- RC6 fixes - Chris
- Add extra slice common debug register - Lionel
- Align virtual engines uabi_class/instance with i915_drm.h - Tvrtko
- Avoid potential division by zero in computing CS timestamp - Chris
- Avoid using various globals - Michal Winiarski, Matt Auld
- Break up long lists of GEM object reclaim - Chris
- Check that the vma hasn't been closed before we insert it - Chris
- Consolidate SDVO HDMI force_dvi handling - Ville
- Conversion to new logging and warn macros and functions - Pankaj, Wambul, Chris
- DC3CO fixes - Jose
- Disable use of hwsp_cacheline for kernel_context - Chris
- Display IRQ pre/post uninstall refactor - Jani
- Display port sync refactor for robustness and fixes - Ville, Manasi
- Do not attempt to reprogram IA/ring frequencies for dgfx - Chris
- Drop alpha_support for good in favor of force_probe - Jani
- DSI ACPI related fixes and refactors - Vivek, Jani, Rajat
- Encoder refactor for flexibility to add more information, especiallly DSI related - Jani, Vandita
- Engine workarounds refactor for robustness around resue - Daniele
- FBC simplification and tracepoints
- Various fixes for build - Jani, Kees Cook, Chris, Zhang Xiaoxu
- Fix cmdparser - Chris
- Fix DRM_I915_GEM_MMAP_OFFFSET - Chris
- Fix i915_request flags - Chris
- Fix inconsistency between pfit enable and scaler freeing - Stanislav
- Fix inverted warn_on on display code - Chris
- Fix modeset locks in sanitize_watermarks - Ville
- Fix OA context id overlap with idle context id - Umesh
- Fix pipe and vblank enable for MST - Jani
- Fix VBT handling for timing parameters - Vandita
- Fixes o kernel doc - Chris, Ville
- Force full modeset whenever DSC is enabled at probe - Jani
- Various GEM locking simplification and fixes - Jani , Chris, Jose
  - Including some changes in preparation for making GEM execbuf parallel - Chris
- Gen11 pcode error codes - Matt Roper
- Gen8+ interrupt handler refactor - Chris
- Many fixes and improvements around GuC code - Daniele, Michal Wajdeczko
- i915 parameters improvements sfor flexible input and better debugability - Chris, Jani
- Ice Lake and Elkhart Lake Fixes and workarounds - Matt Roper, Jose, Vivek, Matt Atwood
- Improvements on execlists, requests and other areas, fixing hangs and also
  improving hang detection, recover and debugability - Chris
  - Also introducing offline GT error capture - Chris
- Introduce encoder->compute_config_late() to help MST - Ville
- Make dbuf configuration const - Jani
- Few misc clean ups - Ville, Chris
- Never allow userptr into the new mapping types - Janusz
- Poison rings after use and GTT scratch pages - Chris
- Protect signaler walk with RCU - Chris
- PSR fixes - Jose
- Pull sseu context updates under gt - Chris
- Read rawclk_freq earlier - Chris
- Refactor around VBT handling to allow geting information through the encoder - Jani
- Refactor l3cc/mocs availability - Chris
- Refactor to use intel_connector over drm_connector - Ville
- Remove i915_energy_uJ from debugfs - Tvrtko
- Remove lite restore defines - Mika Kuoppala
- Remove prefault_disable modparam - Chris
- Many selftests fixes and improvements - Chris
- Set intel_dp_set_m_n() for MST slaves - Jose
- Simplify hot plug pin handling and other fixes around pin and polled modes - Ville
- Skip CPU synchronization on dma-buf attachments - chris
- Skip global serialization of clear_range for bxt vtd - Chris
- Skip rmw for marked register - Chris
- Some other GEM Fixes - Chris
- Some small changes for satisfying static code analysis - Colin, Chris
- Suppress warnings for unused debugging locals
- Tiger Lake enabling, including re-enable -f RPS, workarounds and other display fixes and changes - Chris, Matt Roper, Mika Kuoppala, Anshuman, Jose, Radhakrishna, Rafael.
- Track hw reported context runtime - Tvrtko
- Update bug filling URL - Jani
- Use async bind for PIN_USER into bsw/bxt ggtt - Chris
- Use the kernel_context to measuer the breadcrumb size - Chris
- Userptr fixes and robustness for big pages - Matt Auld
- Various Display refactors and clean-ups, specially around logs and use of drm_i915_private - Jani, Ville
- Various display refactors and fixes, especially around cdclk, modeset, and encoder - Chris, Jani
- Various eDP/DP fixes around DPCD - Lyude
- Various fixes and refactors for better Display watermark handling - Ville, Stanislav
- Various other display refactors - Ville
- Various refactor for better handling of display plane states - Ville
- Wean off drm_pci_alloc/drm_pci_free - Chris
- Correctly terminate connector iteration- Ville
- Downgrade gen7 (ivb, byt, hsw) back to aliasing-ppgtt - Chris

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200225185853.GA3282832@intel.com

Merge tag 'drm-misc-next-2020-02-21' of git://anongit.freedesktop.org/drm/drm-misc into drm-next

drm-misc-next for 5.7:

UAPI Changes:

Cross-subsystem Changes:

Core Changes:
  - crtc: Drop get_crtc callback
  - dp: Add support for DP1.4 EDID corruption test
  - edid: Improve CEA detailed timings support
  - format-helper: Move to a GPL2/MIT dual license
  - mode: Drop drm_display_mode.private
  - vblank: Convert drm-driver vblank related code to CRTC equivalents and
    cleanup the core.
  - drm_global_mutex reworks

Driver Changes:
  - hibmc: Add gamma_set support and improve DPMS support
  - pl111: Support Integrator IM-PD1
  - sun4i: LVDS support for the A20, improvements to panel handling in DSI
  - virtio: job batching improvements, mem handling code rework
  - panel: Support for Rocktech RK101II01D-CT
  - bridge: Support for ADV7535, tc358768, improvements to ti-sn65dsi86

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Maxime Ripard <maxime@cerno.tech>
Link: https://patchwork.freedesktop.org/patch/msgid/20200221100928.anipic2v7q5wswkb@gilmour.lan

drm/amdgpu: drop legacy drm load and unload callbacks

We've moved the debugfs handling into a centralized place
so we can remove the legacy load an unload callbacks.

Tested-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/display: don't call drm_dp_mst_connector_late_register (v2)

Nothing else calls it.  Not sure it's necessary.

v2: remove unused port variable.

Reviewed-by: Jerry (Fangzhi) Zuo <Jerry.Zuo@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/display: split dp connector registration (v4)

Split into init and register functions to avoid a segfault
in some configs when the load/unload callbacks are removed.

v2:
- add back accidently dropped has_aux setting
- set dev in late_register

v3:
- fix dp cec ordering

v4:
- squash in kdev reference fix

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Tested-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/display: add a late register connector callback

To handle debugfs setup on non DP MST connectors.

Tested-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: move dpcd debugfs members setup

Into the function that creates the debugfs files rather
than setting them explicitly in the callers.

Tested-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/display: move debugfs init into core amdgpu debugfs (v2)

In order to remove the load and unload drm callbacks,
we need to reorder the init sequence to move all the drm
debugfs file handling.  Do this for display.

v2: add config guard for DC

Tested-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Harry Wentland <harry.wentland@amd.com> (v1)
Acked-by: Christian König <christian.koenig@amd.com> (v1)
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: don't call drm_connector_register for non-MST ports

The core does this for us now.

Tested-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/ring: move debugfs init into core amdgpu debugfs

In order to remove the load and unload drm callbacks,
we need to reorder the init sequence to move all the drm
debugfs file handling.  Do this for rings.

Tested-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/firmware: move debugfs init into core amdgpu debugfs

In order to remove the load and unload drm callbacks,
we need to reorder the init sequence to move all the drm
debugfs file handling.  Do this for firmware.

Tested-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/regs: move debugfs init into core amdgpu debugfs

In order to remove the load and unload drm callbacks,
we need to reorder the init sequence to move all the drm
debugfs file handling.  Do this for register access files.

Tested-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/gem: move debugfs init into core amdgpu debugfs

In order to remove the load and unload drm callbacks,
we need to reorder the init sequence to move all the drm
debugfs file handling.  Do this for gem.

Tested-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/fence: move debugfs init into core amdgpu debugfs

In order to remove the load and unload drm callbacks,
we need to reorder the init sequence to move all the drm
debugfs file handling.  Do this for fence handling.

Tested-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/sa: move debugfs init into core amdgpu debugfs

In order to remove the load and unload drm callbacks,
we need to reorder the init sequence to move all the drm
debugfs file handling.  Do this for SA (sub allocator).

Tested-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/pm: move debugfs init into core amdgpu debugfs

In order to remove the load and unload drm callbacks,
we need to reorder the init sequence to move all the drm
debugfs file handling.  Do this for pm.

Tested-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/ttm: move debugfs init into core amdgpu debugfs

In order to remove the load and unload drm callbacks,
we need to reorder the init sequence to move all the drm
debugfs file handling.  Do this for ttm.

Tested-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: rename amdgpu_debugfs_preempt_cleanup

to amdgpu_debugfs_fini.  It will be used for other things in
the future.

Tested-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Increase timout on emulator to tenfold instead of twice

Since emulators are slower, sometime some operations like flushing tlb
through FM need more than twice the regular timout of 100ms, so increase
the timeout to 1s on emulators.

Signed-off-by: Yong Zhao <Yong.Zhao@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: add DFCstate control pptable func for arct

This is the callback function that is going to be invoked
when amdgpu_dpm_set_df_cstate is called to toggle DFCstate

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: update arcturus ppsmc header to 54.15.0

two new smc messages added for arcturus with pmfw 54.15.0

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/display: clean up hdcp workqueue handling

Use the existence of the workqueue itself to determine when to
enable HDCP features rather than sprinkling asic checks all over
the code.  Also add a check for the existence of the hdcp
workqueue in the irq handling on the off chance we get and HPD
RX interrupt with the CP bit set.  This avoids a crash if
the driver doesn't support HDCP for a particular asic.

Fixes: 96a3b32e67236f ("drm/amd/display: only enable HDCP for DCN+")
Bug: https://bugzilla.kernel.org/show_bug.cgi?id=206519
Reviewed-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Delete unnecessary unmap queue package submissions

The previous way of using SDMA queue count to infer whether we should unmap
SDMA engines has bugs. The reason it did not cause issues is because MEC
firmware unmaps all queues (CP + SDMA) when a unmap package for compute
engine is received. Becasue of that, only one unmap queue package
is needed, instead of one unmap queue package for CP and each SDMA engine,
which results in much simpler driver code.

Signed-off-by: Yong Zhao <Yong.Zhao@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Delete excessive printings

Those printings are duplicated or useless.

Signed-off-by: Yong Zhao <Yong.Zhao@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Fix a memory leak in queue creation error handling

When the queue creation failed, some resources were not freed. Fix it.

Signed-off-by: Yong Zhao <Yong.Zhao@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Count active CP queues directly

The previous code of calculating active CP queues is problematic if
some SDMA queues are inactive. Fix that by counting CP queues directly.

Signed-off-by: Yong Zhao <Yong.Zhao@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Avoid ambiguity by indicating it's cp queue

The queues represented in queue_bitmap are only CP queues.

Signed-off-by: Yong Zhao <Yong.Zhao@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Rename queue_count to active_queue_count

The name is easier to understand the code.

Signed-off-by: Yong Zhao <Yong.Zhao@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd: Extend ROCt to surface UUID for devices that have them

Devices from Arcturus onwards will have their UUID exposed to Thunk.
Adding neccessary functions to the kernel to propagate the uuid.

Signed-off-by: Divya Shikre <DivyaUday.Shikre@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix check for DPM when returning max clock

pp_funcs may not exist, while dpm may be enabled. This change ensures
that KFD topology will report the same as pp_dpm_sclk, as the conditions
for reporting them will be the same.

Otherwise, we may see the issue where KFD reports "100MHz" in topology
as the max speed, while DPM is working correctly.

Signed-off-by: Kent Russell <kent.russell@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Don't write GCVM_L2_CNTL* regs on navi12 VF

This change disables programming of GCVM_L2_CNTL* regs on VF.

Signed-off-by: Rohit Khaire <Rohit.Khaire@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon: Inline drm_get_pci_dev

It's the last user, and more importantly, it's the last non-legacy
user of anything in drm_pci.c.

The only tricky bit is the agp initialization. But a close look shows
that radeon does not use the drm_agp midlayer (the main use of that is
drm_bufs for legacy drivers), and instead could use the agp subsystem
directly (like nouveau does already). Hence we can just pull this in
too.

A further step would be to entirely drop the use of drm_device->agp,
but feels like too much churn just for this patch.

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: "Christian König" <christian.koenig@amd.com>
Cc: "David (ChunMing) Zhou" <David1.Zhou@amd.com>
Cc: amd-gfx@lists.freedesktop.org
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Drop DRIVER_USE_AGP

This doesn't do anything except auto-init drm_agp support when you
call drm_get_pci_dev(). Which amdgpu stopped doing with

commit b58c11314a1706bf094c489ef5cb28f76478c704
Author: Alex Deucher <alexander.deucher@amd.com>
Date:   Fri Jun 2 17:16:31 2017 -0400

    drm/amdgpu: drop deprecated drm_get_pci_dev and drm_put_dev

No idea whether this was intentional or accidental breakage, but I
guess anyone who manages to boot a this modern gpu behind an agp
bridge deserves a price. A price I never expect anyone to ever collect
:-)

Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: "Christian König" <christian.koenig@amd.com>
Cc: Hawking Zhang <Hawking.Zhang@amd.com>
Cc: Xiaojie Yuan <xiaojie.yuan@amd.com>
Cc: Evan Quan <evan.quan@amd.com>
Cc: "Tianci.Yin" <tianci.yin@amd.com>
Cc: "Marek Olšák" <marek.olsak@amd.com>
Cc: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: Use bitwise instead of arithmetic operator for flags

This silences the following coccinelle warning:

"WARNING: sum of probable bitmasks, consider |"

Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Chen Zhou <chenzhou10@huawei.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: remove set but not used variable 'mc_vm_apt_default'

drivers/gpu/drm/amd/amdgpu/../display/dc/dcn21/dcn21_hubp.c:
 In function hubp21_set_vm_system_aperture_settings:
drivers/gpu/drm/amd/amdgpu/../display/dc/dcn21/dcn21_hubp.c:343:23:
 warning: variable mc_vm_apt_default set but not used [-Wunused-but-set-variable]

It is never used, so remove it.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu: Add gfxoff debugfs entry

Write a 32-bit value of zero to disable GFXOFF and write a 32-bit
value of non-zero to enable GFXOFF.

Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: use amdgpu_ring_test_helper when possible

amdgpu_ring_test_helper already handles ring->sched.ready correctly

Signed-off-by: Nirmoy Das <nirmoy.das@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add VM update fences back to the root PD v2

Add update fences to the root PD while mapping BOs.

Otherwise PDs freed during the mapping won't wait for
updates to finish and can cause corruptions.

v2: rebased on drm-misc-next

Signed-off-by: Christian König <christian.koenig@amd.com>
Fixes: 90b69cdc5f159 drm/amdgpu: stop adding VM updates fences to the resv obj
Reviewed-by: xinhui pan <xinhui.pan@amd.com>
Tested-by: Luben Tuikov <luben.tuikov@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: cleanup amdgpu_ring_fini

cleanup amdgpu_ring_fini to check the prerequisites before changing ring->sched.ready

Signed-off-by: Nirmoy Das <nirmoy.das@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add Arcturus D342 page retire support

Check Arcturus SKU type to select I2C address of page retirement EEPROM

Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: John Clements <john.clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: toggle DF-Cstate to protect DF reg access

driver needs to take DF out Cstate before any DF register
access. otherwise, the DF register may not be accessible.

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Acked-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: move get_xgmi_relative_phy_addr to amdgpu_xgmi.c

centralize all the xgmi related function to amdgpu_xgmi.c

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Acked-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add dpm helper function for DF Cstate control

The helper function hides software smu and legacy powerplay
implementation for DF Cstate control.

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: update psp firmwares loading sequence V2

For those ASICs with DF Cstate management centralized to PMFW,
TMR setup should be performed between pmfw loading and other
non-psp firmwares loading.

V2: skip possible SMU firmware reloading

Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Remove kfd eviction fence before release bo (v2)

No need to trigger eviction as the memory mapping will not be used
anymore.

All pt/pd bos share same resv, hence the same shared eviction fence.
Everytime page table is freed, the fence will be signled and that cuases
kfd unexcepted evictions.

v2: squash in 32 bit fix

CC: Christian König <christian.koenig@amd.com>
CC: Felix Kuehling <felix.kuehling@amd.com>
CC: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: xinhui pan <xinhui.pan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/omap: dss: Remove unused omap_dss_device operations

The omap_dss_device .pre_enable(), .post_disable() and .set_timings()
are not used anymore. Remove them.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ti.com>
Tested-by: Sebastian Reichel <sebastian.reichel@collabora.com>
Reviewed-by: Sebastian Reichel <sebastian.reichel@collabora.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ti.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200226112514.12455-55-laurent.pinchart@ideasonboard.com