drm/amd/display: Add passive dongle support for HPD Rearch

Add HPD delay timer support to
1. Single/dual link DVI.
2. DP to HDMI passive dongle
3. DP to DVI passive dongle.

Signed-off-by: John Barberiz <jbarberi@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix active dongle hotplug

Clean fake sink flag after detecting link on downstream port.
Fixing display light-up after  "hot-unplug&plug again" downstream
of an active dongle.

Signed-off-by: Roman Li <Roman.Li@amd.com>
Reviewed-by: Harry Wentland <Harry.Wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: move MST branch initialize to before link training

some MST capable scaler doesn't like recieving CLEAR_PAYLOAD_ID_TABLE after
link training.  move branch initialize to before link training

Signed-off-by: Hersen Wu <hersenxs.wu@amd.com>
Signed-off-by: Tony Cheng <tony.cheng@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon/mkregtable: Delete unused list functions and macros

The util mkregtable includes a copy of the kernel API for linked lists,
only a small subset of it is used. Delete the unused functions and macros.

Reviewed-by: Guenter Roeck <groeck@chromium.org>
Signed-off-by: Matthias Kaehlcke <mka@chromium.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: used cached pcie gen info for SI (v2)

Rather than querying it every time we need it.
Also fixes a crash in VM pass through if there is no
root bridge because the cached value fetch already checks
this properly.

v2: fix includes

Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=105244
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Rex Zhu<rezhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

drm/amd/amdgpu: Mask rptr as well in ring debugfs

The read/write pointers on sdma4 devices increment
beyond the ring size and should be masked.  Tested
on my Ryzen 2400G.

Signed-off-by: Tom St Denis <tom.stdenis@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: try again kiq access if not in IRQ(v4)

sometimes GPU is switched to other VFs and won't swich
back soon, so the kiq reg access will not signal within
a short period, instead of busy waiting a long time(MAX_KEQ_REG_WAIT)
and returning TMO we can istead sleep 5ms and try again
later (non irq context)

And since the waiting in kiq_r/weg is busy wait, so MAX_KIQ_REG_WAIT
shouldn't set to a long time, set it to 10ms is more appropriate.

if gpu already in reset state, don't retry the KIQ reg access
otherwise it would always hang because KIQ was already die usually.

v2:
replace schedule() with msleep() for the wait

v3:
use while loop for the wait repeating
use macros for the sleep period
more description for it

v4:
drop unused variable

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

drm/amdgpu: cleanups for vram lost handling

1)create a routine "handle_vram_lost" to do the vram
recovery, and put it into amdgpu_device_reset/reset_sriov,
this way no need of the extra paramter to hold the
VRAM LOST information and the related macros can be removed.

3)show vram_recover failure if time out, and set TMO equal to
lockup_timeout if vram_recover is under SRIOV runtime mode.

4)report error if any ip reset failed for SR-IOV

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

drm/amdgpu: stop all rings before doing gpu recover

found recover_vram_from_shadow sometimes get executed
in paralle with SDMA scheduler, should stop all
schedulers before doing gpu reset/recover

Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Tested-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix module parameter descriptions

Some were missing the close parens around options.

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

drm/amdgpu: Map all visible VRAM at startup

When using CPU to update page table, we need to kmap all the PDs/PTs after
they are allocated and that requires a TLB shot down on each CPU, which is
quite heavy.

Instead, we map the whole visible VRAM to a kernel address at once. Pages
can be obtained from the offset.

v2: move the mapping base from gmc to amdgpu_mman structure, and the
    implementation in amdgpu_ttm_* functions

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

drm/amd/display: check for ipp before calling cursor operations

Currently all cursor related functions are made to all
pipes that are attached to a particular stream.
This is not applicable to pipes that do not have cursor plane
initialised like underlay.
Hence this patch allows cursor related operations on a pipe
only if ipp in available on that particular pipe.

The check is added to set_cursor_position & set_cursor_attribute.

Signed-off-by: Shirish S <shirish.s@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

drm/amd/display: make dm_dp_aux_transfer return payload bytes instead of size

The drm layer expects aux->transfer() to return the payload bytes read.
Currently dm_dp_aux_transfer() returns the payload size which does not
gets updated during the read, hence not giving the right data for the
drm layer to pars edid. This leads to the drm layer to conclude as the
edid is BAD and hence some monitors/devices dont get detected properly.

This patch changes the return type of dm_dp_aux_transfer() to actual
bytes read during DP_AUX_NATIVE_READ & DP_AUX_I2C_READ.

Signed-off-by: Shirish S <shirish.s@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: disable GFX ring and disable PQ wptr in hw_fini

otherwise there will be DMAR reading error comes out from CP since
GFX is still alive and CPC's WPTR_POLL is still enabled, which would
lead to DMAR read error.

fix:
we can hault CPG after hw_fini, but cannot halt CPC becaues KIQ
stil need to be alive to let RLCV invoke, but its WPTR_POLL could
be disabled.

Signed-off-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: cleanup SA inti and fini(v2)

should use bo_create_kernel instead of split to two
function that create and pin the SA bo

issue:
before this patch, there are DMAR read error in host
side when running SRIOV test, the DMAR address dropped
in the range of SA bo.

fix:
after this cleanups of SA init and fini, above DMAR
eror gone.

v2:
keep sa_bo's fini instead of suspend, to keep
reporting error

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

drm/amdgpu: Correct sdma_v4 get_wptr(v2)

the original method will change the wptr value in wb.
v2:
furthur cleanup

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

drm/amdgpu: adjust timeout for ib_ring_tests(v2)

issue:
sometime GFX/MM ib test hit timeout under SRIOV env, root cause
is that engine doesn't come back soon enough so the current
IB test considered as timed out.

fix:
for SRIOV GFX IB test wait time need to be expanded a lot during
SRIOV runtimei mode since it couldn't really begin before GFX engine
come back.

for SRIOV MM IB test it always need more time since MM scheduling
is not go together with GFX engine, it is controled by h/w MM
scheduler so no matter runtime or exclusive mode MM IB test
always need more time.

v2:
use ring type instead of idx to judge

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

drm/amdgpu: don't use MM idle_work for SRIOV(v2)

SRIOV doesn't give VF cg/pg feature so the MM's idle_work
is skipped for SR-IOV

v2:
remove superfluous changes
since idle_work is not scheduled for SR-IOV so the condition
check for SR-IOV inside idle_work also can be dropped

v3:
drop the SRIOV check in amdgpu_vce/uvd_suspend

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

drm/amd/pp: Print more smu failed info on Vega10

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

drm/amd/powerplay: fix power over limit on Fiji

power containment disabled only on Fiji and compute
power profile. It violates PCIe spec and may cause power
supply failed. Enabling it will fix the issue, even the
fix will drop performance of some compute tests.

Signed-off-by: Eric Huang <JinHuiEric.Huang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

drm/amdgpu:Fixed wrong emit frame size for enc

Emit frame size should match with corresponding function,
uvd_v6_0_enc_ring_emit_vm_flush has 5 amdgpu_ring_write

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

drm/amdgpu: increase gart size to 512MB

256MB is too small consider PTE/PDE shadow and TTM
eviction activity

Signed-off-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: move WB_FREE to correct place

WB_FREE should be put after all engines's hw_fini
done, otherwise the invalid wptr/rptr_addr would still
be used by engines which trigger abnormal bugs.

This fixes couple DMAR reading error in host side for SRIOV
after guest kmd is unloaded.

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

drm/amdgpu: change gfx9 ib test to use WB

two reasons to switch SCRATCH reg method to WB method:

1)Because when doing IB test we don't want to involve KIQ health
status affect, and since SCRATCH register access is go through
KIQ that way GFX IB test would failed due to KIQ fail.

2)acccessing SCRATCH register cost much more time than WB method
because SCRATCH register access runs through KIQ which at least could
begin after GPU world switch back to current Guest VF

Signed-off-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: cond_exec only for schedule with a job

issue:
under SR-IOV sometimes the iB test will fail on
gfx ring

fix:
with cond_exec inserted in RB the gfx engine would
skip part packets if RLCV issue PREEMPT on gfx engine
if gfx engine is prior to COND_EXEC packet, this is
okay for regular command from UMD, but for the ib test
since the whole dma format doesn't support PREEMPT
so must remove the COND_EXEC from it.

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

drm/amdgpu: only flush hotplug work without DC

since hotplug_work is initialized under the case of
no dc support

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

drm/amdgpu: skip ECC for SRIOV in gmc late_init

Signed-off-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: fix&cleanups for wb_clear

fix:
should do right shift on wb before clearing

cleanups:
1,should memset all wb buffer
2,set max wb number to 128 (total 4KB) is big enough

Signed-off-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/powerplay/smu7: use proper dep table for mclk

For mclk od, use the vdd dependency on mclk table.  Looks
like a cut and paste typo.

Reviewed-by: Rex Zhu<rezhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu: Correct VRAM width for APUs with GMC9

DDR4 has a 64-bit width not 128-bits.  It was reporting
twice the width.  Tested with my Ryzen 2400G.

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/ttm: cleanup ttm_tt_create

Cleanup ttm_tt_create a bit.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Roger He <Hongbo.He@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/ttm: move ttm_tt_create into ttm_tt.c v2

Rename ttm_bo_add_ttm to ttm_tt_create and move it into ttm_tt.c.

v2: separate the cleanup.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Roger He <Hongbo.He@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/ttm: check if free mem space is under the lower limit

the free mem space and the lower limit both include two parts:
system memory and swap space.

For the OOM triggered by TTM, that is the case as below:
first swap space is full of swapped out pages and soon
system memory also is filled up with ttm pages. and then
any memory allocation request will run into OOM.

to cover two cases:
a. if no swap disk at all or free swap space is under swap mem
   limit but available system mem is bigger than sys mem limit,
   allow TTM allocation;

b. if the available system mem is less than sys mem limit but
   free swap space is bigger than swap mem limit, allow TTM
   allocation.

v2: merge two memory limit(swap and system) into one
v3: keep original behavior except ttm_opt_ctx->flags with
    TTM_OPT_FLAG_FORCE_ALLOC
v4: always set force_alloc as tx->flags & TTM_OPT_FLAG_FORCE_ALLOC
v5: add an attribute for lower_mem_limit
v6: set lower_mem_limit as 0 to keep original behavior

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

drm/ttm: drop persistent_swap_storage from ttm_bo_init and co

Never used as parameter, the only driver actually using this is nouveau
and there it is initialized after the BO is initialized.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/ttm: drop ttm->dummy_read_page

Only used by the AGP backend and there it can be easily accessed using
ttm->bdev->glob.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/ttm: drop ttm->glob

The pointer is available as ttm->bdev->glob as well.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/ttm: drop bo->glob

The pointer is available as bo->bdev->glob as well.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

staging: vboxvideo: remove ttm_pool_* wrappers

TTM calls the default implementation now.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/bochs: remove the default ttm_tt_populate callbacks

TTM calls the default implementation now.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/cirrus: remove ttm_pool_* wrappers

TTM calls the default implementation now.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/qxl: remove ttm_pool_* wrappers

TTM calls the default implementation now.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/ast: remove ttm_pool_* wrappers

TTM calls the default implementation now.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/hisilicon: remove ttm_pool_* wrappers

TTM calls the default implementation now.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/mgag200: remove ttm_pool_* wrappers

TTM calls the default implementation now.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/virtio: remove ttm_pool_* wrappers

TTM calls the default implementation now.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/ttm: add default implementations for ttm_tt_(un)populate

Use ttm_pool_populate/ttm_pool_unpopulate if the driver doesn't provide
a function.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pp: Move common dpm check functions to hardwaremanager.c

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

drm/amd/pp: Cleaning up vega10_enable_dpm_tasks function

1. move display num initialize out of dpm enable tasks.
2. do not set/restore smc telemetry if dpm is runing.

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

drm/amd/pp: Refine code in powerplay for Cz/Vega10

Add dpm check functions on CZ/Vega10 to smu backend
function table.

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

drm/amd/pp: Remove dead error checking code on Vega10

when smu failed, print out the error info immediately
for debug. smum_send_msg_to_smu always return true,
so no need to check return value.

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

drm/amd/pp: Add debug info when smu failed on Vega10

When smu msssage failed, print out return value in dmesg.

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

drm/amd/pp: Remove duplicated vega10_is_smc_ram_running calls

Avoid conflicts in reading the same register mmPCIE_INDEX2
with other clients

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

drm/amd: Remove inclusion of non-existing include directories

This patch fix the following build warnings:
  CC [M]  drivers/gpu/drm/amd/amdgpu/amdgpu_drv.o
cc1: warning: ../display/: No such file or directory [-Wmissing-include-dirs]
cc1: warning: ../display/include: No such file or directory [-Wmissing-include-dirs]
  CC [M]  drivers/gpu/drm/amd/amdgpu/amdgpu_device.o
cc1: warning: ../display/: No such file or directory [-Wmissing-include-dirs]
cc1: warning: ../display/include: No such file or directory [-Wmissing-include-dirs]
[...]
This warning is shown for each file in amdgpu directory, so it spams a lot.

Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Corentin Labbe <clabbe@baylibre.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd: remove inclusion of non-existing scheduler directory

The scheduler directory was removed via commit 1b1f42d8fde4 ("drm: move amd_gpu_scheduler into common location")
Remove it from include path.

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Corentin Labbe <clabbe@baylibre.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon: insist on 32-bit DMA for Cedar on PPC64/PPC64LE

In radeon_device_init, set the need_dma32 flag for Cedar chips
(e.g. FirePro 2270).  This fixes, or at least works around, a bug
on PowerPC exposed by last year's commits

8e3f1b1d8255105f31556aacf8aeb6071b00d469 (Russell Currey)

and

253fd51e2f533552ae35a0c661705da6c4842c1b (Alistair Popple)

which enabled the 64-bit DMA iommu bypass.

This caused the device to freeze, in some cases unrecoverably, and is
the subject of several bug reports internal to Red Hat.

Signed-off-by: Ben Crocker <bcrocker@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

drm/amdgpu: Remove duplicate setting of ->need_swiotlb

There's no need to set this before the number of DMA bits has been
properly determined.

Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pp: Add a pp feature mask bit for AutoWattman feature

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

drm/amdgpu: Change default value of module parameter amdgpu_pp_feature_mask

Currently all pp features are enabled by default except
OVERDRIVE

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

drm/amd/powerplay: fix thermal interrupts on vega10

a bug in programming thermal interrupt register masks out
interrupts and driver cannot receive interrupts. Setting
0 to mask bits will fix it.

Signed-off-by: Eric Huang <JinHuiEric.Huang@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add query vram width in CGS query system info

powerplay need vram width to set default mclk optimization
settings(uphyst/downhyst/activity threshold)

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

drm/amdgpu: use the TTM dummy page instead of allocating one

We have a global dummy page in TTM, use that one instead of allocating a
new one.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pp: Fix bug that dpm level was not really locked

Lock the dpm levels when we use SW method to modify
the dpm tables directly to avoid a possible race
with the smu.

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

drm/amd/pp: Fix error handling when smu return failed on Vega10.

Clamp the clock index to a valid range when reading it back

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

drm/ttm: set TTM_OPT_FLAG_FORCE_ALLOC in ttm_bo_force_list_clean

Because ttm_bo_force_list_clean() is only called on two occasions:
1. By ttm_bo_evict_mm() during suspend.
2. By ttm_bo_clean_mm() when the driver unloads.
On both cases we absolutely don't want any memory allocation failure.

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

drm/ttm: add bit flag TTM_OPT_FLAG_FORCE_ALLOC

set TTM_OPT_FLAG_FORCE_ALLOC when we are servicing for page
fault routine.

for ttm_mem_global_reserve if in page fault routine, allow the gtt
pages reservation always. because page fault routing already grabbed
system memory and the allowance of this exception is harmless.
Otherwise, it will trigger OOM killer.

will be used later.

v2: set the FORCE_ALLOC always
v3: minor refine

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

drm/ttm: use bit flag to replace allow_reserved_eviction in ttm_operation_ctx

for saving memory and more bit flag can be used in future

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

radeon: hide pointless #warning when compile testing

In randconfig testing, we sometimes get this warning:

drivers/gpu/drm/radeon/radeon_object.c: In function 'radeon_bo_create':
drivers/gpu/drm/radeon/radeon_object.c:242:2: error: #warning Please enable CONFIG_MTRR and CONFIG_X86_PAT for better performance thanks to write-combining [-Werror=cpp]
 #warning Please enable CONFIG_MTRR and CONFIG_X86_PAT for better performance \

This is rather annoying since almost all other code produces no build-time
output unless we have found a real bug. We already fixed this in the
amdgpu driver in commit 31bb90f1cd08 ("drm/amdgpu: shut up #warning for
compile testing") by adding a CONFIG_COMPILE_TEST check last year and
agreed to do the same here, but both Michel and I then forgot about it
until I came across the issue again now.

For stable kernels, as this is one of very few remaining randconfig
warnings in 4.14.

Cc: stable@vger.kernel.org
Link: https://patchwork.kernel.org/patch/9550009/
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/ttm: set page mapping during allocation

To aid debugging set the page mapping during allocation instead of
during VM faults.

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

Revert "drm/radeon/pm: autoswitch power state when in balanced mode"

This reverts commit 1c331f75aa6ccbf64ebcc5a019183e617c9d818a.

Breaks resume on some systems.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=100759
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

drm/radeon: use drm_gem_private_object_init

We use our own backing store and don't need the shmem file.

Signed-off-by: Christian König <christian.koenig@amd.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: use drm_gem_private_object_init

We use our own backing store and don't need the shmem file.

Signed-off-by: Christian König <christian.koenig@amd.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: mitigate workaround for i915

To be able to use DRI_PRIME with amdgpu and i915 we add all our fences
only as exclusive ones.

Disable that behavior when sharing between amdgpu itself cause it
hinders concurrent execution.

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

drm/amdgpu: implement amdgpu_gem_map_(attach/detach)

Instead of the pin/unpin callback implement the attach/detach ones.

Functional identical, but allows us access to the attachment.

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

drm/amdgpu/powerplay/smu7: drop refresh rate checks for mclk switching

The logic has moved to cgs.  mclk switching with DC at higher refresh
rates should work.

Reviewed-by: Eric Huang <JinhuiEric.Huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: Harry Wentland <harry.wentland@amd.com>

drm/amdgpu/cgs: add refresh rate checking to non-DC display code

Clamp the vblank period to 0 if the refresh rate is larger than
120 hz for non-DC.  This allows us to remove the refresh rate
checks from powerplay for mclk switching.

Reviewed-by: Eric Huang <JinhuiEric.Huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay/smu7: allow mclk switching with no displays

If there are no displays attached, there is no reason to disable
mclk switching.

Fixes mclks getting set to high when there are no displays attached.

Reviewed-by: Eric Huang <JinhuiEric.Huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

drm/amd/powerplay/vega10: allow mclk switching with no displays

If there are no displays attached, there is no reason to disable
mclk switching.

Fixes mclks getting set to high when there are no displays attached.

Reviewed-by: Eric Huang <JinhuiEric.Huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

drm/amd/powerplay: use PP_CAP macro for disable_mclk_switching_for_frame_lock

Rather than open coding it.

Reviewed-by: Eric Huang <JinhuiEric.Huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: remove unused headers

All thoses headers are not used by any source files.
Lets just remove them.

Signed-off-by: Corentin Labbe <clabbe@baylibre.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu_gem: fix error handling path in amdgpu_gem_va_update_vm

Currently, if amdgpu_vm_bo_update() fails, the returned error
is being ignored.

Fix this by properly checking _r_ after calling amdgpu_vm_bo_update.
Also, remove redundant code just before label _error_.

Addresses-Coverity-ID: 1464280 ("Unused value")
Fixes: 0abc6878fc2d ("drm/amdgpu: update VM PDs after the PTs")
Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: update the PASID mapping only on demand

Updating the PASID is rather heavyweight and shouldn't be done all the
time.

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

drm/amdgpu: separate PASID mapping from VM flush v2

Stuffing the PASID mapping into the VM flush isn't flexible enough since
the PASID mapping changes not as often as we need a VM flush.

v2: add missing use of gmc_v7_0_emit_pasid_mapping

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

drm/amd/display: Fix increment when sampling OTF in DCE

Previously, the number of software segmets per region was reduced to 16.
This needs to be reflected in the sampling distance (increment) used when
translating to the hardware format.

Signed-off-by: Leo (Sunpeng) Li <sunpeng.li@amd.com>
Reviewed-by: Krunoslav Kovac <Krunoslav.Kovac@amd.com>
Acked-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: De PQ implementation

Some refactoring and optimizations in color module.
Added de gamma 2.2 & 2.4, also re gamma 2.2.
Added interface for diagnostic for de gamma & de pq.

Signed-off-by: Vitaly Prosyak <vitaly.prosyak@amd.com>
Reviewed-by: Krunoslav Kovac <Krunoslav.Kovac@amd.com>
Acked-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Remove unused dm_pp_ interfaces

Signed-off-by: Harry Wentland <harry.wentland@amd.com>
Reviewed-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Acked-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add logging for aux DPCD access

Add basic logging for DPCD access. Does not print
by default.

Currently only prints first byte of the data accessed.

Technical debt: Need to make it so that the entire
data block accessed is printed. Also need to log
address space that's not DPCD.

Signed-off-by: Eric Yang <Eric.Yang2@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Set vsc pack revision when DPCD revision is >= 1.2

Brightness couldn't change when booting up in DC mode.
It was because "psr_enabled" flag was not set to true before
setting vsc packet revision, causing packet rev setup was skipped.
Now instead of checking the psr flag, it checks if the DPCD_REV >= 1.2
and set the vsc packet revision.

Signed-off-by: Tao <xtao@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: provide an interface to query firmware version

Signed-off-by: Anthony Koo <Anthony.Koo@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add missing Vega defines to dal_asic_id

Signed-off-by: Harry Wentland <harry.wentland@amd.com>
Reviewed-by: Hersen Wu <hersenxs.wu@amd.com>
Acked-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Treat bad EDID as no EDID

We have an mst dock firmware that will emulate an EDID with bad
checksum.

v2: Tread -> Treat

Signed-off-by: Eric Yang <Eric.Yang2@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Remove duplicate entries from BIOS function table

Identical to the two entries above them.

Signed-off-by: Ken Chalmers <ken.chalmers@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Make FBC work without fbdev emulation

Previously, FBC dynamic allocation relied on connector modes
populated during dm init. This is only the case if
DRM_FBDEV_EMULATION config flag is enabled.
Moving fbc allocation from dm_late_init() to
amdgpu_dm_connector_get_modes() where actual modes init happens.

Signed-off-by: Roman Li <Roman.Li@amd.com>
Reviewed-by: Shirish Shankarappa <Shirish.S@amd.com>
Acked-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Remove delay on disconnect patch

HDMI HPD's generic solution makes the monitor
patch code unnecessary so anything related has
been removed.

Signed-off-by: John Barberiz <jbarberi@amd.com>
Reviewed-by: John Barberiz <jbarberi@amd.com>
Acked-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Skip 2 frames when first reading CRC

Skipping the first frame will prevent uncoooked values most of the time.
However, in some unlikely cases, the second frame will be uncooked as
well.

Signed-off-by: Leo (Sunpeng) Li <sunpeng.li@amd.com>
Reviewed-by: Harry Wentland <Harry.Wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Expose DCE110 CRC functions for DCE8

Implement CRC for DCE8. Registers remain the same, so call DCE110 code
directly.

Signed-off-by: Leo (Sunpeng) Li <sunpeng.li@amd.com>
Reviewed-by: Harry Wentland <Harry.Wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Rework DCE transform bit depth reduction programming.

Clear up the logic, and enable programming truncation as a bit reduction
mode.

Signed-off-by: Leo (Sunpeng) Li <sunpeng.li@amd.com>
Reviewed-by: Harry Wentland <Harry.Wentland@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: When enabling CRC, disable dither & enable truncation

When user-mode is using 8bpc, the hardware represents it internally
using a higher bit depth. This causes problems when comparing CRCs for
color managment tests.

We need to disable dithering as well, since it makes CRC values
non-deterministic.

It's easy to see why dithering needs to be disabled, The reason why
truncation also needs to be enabled is better described with an example.
Consider the folowing which tests the color transform matrix (CTM):

Expected CRC = FB_A -> Degamma (Bypassed) -> CTM (Bypassed)
                                                  |
                                                  v
                            Obtain CRC  <- Regamma(Bypassed)

Actual CRC = FB_B -> Degamma (Bypassed) -> CTM (0.5*Identity)
                                                  |
                                                  v
                            Obtain CRC  <- Regamma(Bypassed)

FB_A contains a solid red color at half intensity (127 @ 8bpc)
FB_B contains a solid red color at full intensity (255 @ 8bpc)

We expect that Expected CRC = Actual CRC, but that's not the case. When
the CTM is applied, the output is at half intensity, but also at a
higher bit depth within hardware. i.e. 255/2 = 127.5: not representable
in 8bpc, but can be at 10bpc. This causes the two CRC's to be different.

The solution is to truncate the output bit depth to the same as input
when enabling CRC capture. Since Linux only supports 8bpc, hard code
that for now.

Signed-off-by: Leo (Sunpeng) Li <sunpeng.li@amd.com>
Reviewed-by: Harry Wentland <Harry.Wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Expose dither setting functionality to Linux

We will need this to disable dither for CRC capture.

Signed-off-by: Leo (Sunpeng) Li <sunpeng.li@amd.com>
Reviewed-by: Harry Wentland <Harry.Wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Refactor max color lut entries into a macro.

Signed-off-by: Leo (Sunpeng) Li <sunpeng.li@amd.com>
Reviewed-by: Harry Wentland <Harry.Wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Hookup color management functions

Hookup new color management functions into amdgpu_dm:

- Notify DRM that we support CRTC color management during CRTC init
- Call color management functions within atomic check to update dc
  states in preparation for a commit

Signed-off-by: Leo (Sunpeng) Li <sunpeng.li@amd.com>
Reviewed-by: Harry Wentland <Harry.Wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>