[Why]
For later use by the DC VM implementation
Signed-off-by: NEryk Brol <eryk.brol@amd.com>
Reviewed-by: NJun Lei <Jun.Lei@amd.com>
Acked-by: NLeo Li <sunpeng.li@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

[Why]
Seamless boot code not working on all ASICs because of
some underflow issues caused by some uninitialized HW
state.

[How]
Keep some logical and power gating init code in hw_init.
Move some per pipe init code to enable accelerated mode
Signed-off-by: NAnthony Koo <Anthony.Koo@amd.com>
Reviewed-by: NAric Cyr <Aric.Cyr@amd.com>
Acked-by: NLeo Li <sunpeng.li@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

[Why]
DM has impelemented new pp_smu interface. 'Else' is not longer needed.
Signed-off-by: NFatemeh Darbehani <fatemeh.darbehani@amd.com>
Reviewed-by: NEric Yang <eric.yang2@amd.com>
Acked-by: NLeo Li <sunpeng.li@amd.com>
Acked-by: NYongqiang Sun <yongqiang.sun@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

This enum definition is shared, so move it to a shared location.
Signed-off-by: NEric Bernstein <eric.bernstein@amd.com>
Reviewed-by: NTony Cheng <Tony.Cheng@amd.com>
Acked-by: NLeo Li <sunpeng.li@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Signed-off-by: Nmmcgarri <mark.mcgarrity@amd.com>
Reviewed-by: NTony Cheng <Tony.Cheng@amd.com>
Acked-by: NLeo Li <sunpeng.li@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

[Why]
Make dml and integration with pplib clearer.

[How]
Change the way the dml formula is initialized to make its values more
clear. Restructure DC interface with pplib into rv_funcs.
Cap clocks received from pplib.
Signed-off-by: NJun Lei <Jun.Lei@amd.com>
Signed-off-by: NEryk Brol <eryk.brol@amd.com>
Reviewed-by: NDmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Acked-by: NLeo Li <sunpeng.li@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Added a "disable_tri_buf" DC debug option. When set to 1  feature will
be off.
Signed-off-by: NCharlene Liu <charlene.liu@amd.com>
Reviewed-by: NDmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Acked-by: NLeo Li <sunpeng.li@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

In VRR mode, keep track of the vblank count of the last
completed pageflip in amdgpu_crtc->last_flip_vblank, as
recorded in the pageflip completion handler after each
completed flip.

Use that count to prevent mmio programming a new pageflip
within the same vblank in which the last pageflip completed,
iow. to throttle pageflips to at most one flip per video
frame, while at the same time allowing to request a flip
not only before start of vblank, but also anywhere within
vblank.

The old logic did the same, and made sense for regular fixed
refresh rate flipping, but in vrr mode it prevents requesting
a flip anywhere inside the possibly huge vblank, thereby
reducing framerate in vrr mode instead of improving it, by
delaying a slightly delayed flip requests up to a maximum
vblank duration + 1 scanout duration. This would limit VRR
usefulness to only help applications with a very high GPU
demand, which can submit the flip request before start of
vblank, but then have to wait long for fences to complete.

With this method a flip can be both requested and - after
fences have completed - executed, ie. it doesn't matter if
the request (amdgpu_dm_do_flip()) gets delayed until deep
into the extended vblank due to cpu execution delays. This
also allows clients which want to regulate framerate within
the vrr range a much more fine-grained control of flip timing,
a feature that might be useful for video playback, and is
very useful for neuroscience/vision research applications.

In regular non-VRR mode, retain the old flip submission
behavior. This to keep flip scheduling for fullscreen X11/GLX
OpenGL clients intact, if they use the GLX_OML_sync_control
extensions glXSwapBufferMscOML(, ..., target_msc,...) function
with a specific target_msc target vblank count.

glXSwapBuffersMscOML() or DRI3/Present PresentPixmap() will
not flip at the proper target_msc for a non-zero target_msc
if VRR mode is active with this patch. They'd often flip one
frame too early. However, this limitation should not matter
much in VRR mode, as scheduling based on vblank counts is
pretty futile/unusable under variable refresh duration
anyway, so no real extra harm is done.

According to some testing already done with this patch by
Nicholas on top of my tests, IGT tests didn't report any
problems. If fixes stuttering and flickering when flipping
at rates below the minimum vrr refresh rate.

Fixes: bb47de73 ("drm/amdgpu: Set FreeSync state using drm VRR
properties")
Signed-off-by: NMario Kleiner <mario.kleiner.de@gmail.com>
Cc: <stable@vger.kernel.org>
Cc: Harry Wentland <harry.wentland@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Michel Dänzer <michel@daenzer.net>
Tested-by: NBruno Filipe <bmilreu@gmail.com>
Reviewed-by: NNicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: NNicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Clear the VM PDs/PTs only after initializing all the structures.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NFelix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Clang warns when an expression that equals zero is used as a null
pointer constant (in lieu of NULL):

drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm.c:4435:3:
warning: expression which evaluates to zero treated as a null pointer
constant of type 'const enum color_transfer_func *'
[-Wnon-literal-null-conversion]
                TRANSFER_FUNC_UNKNOWN,
                ^~~~~~~~~~~~~~~~~~~~~
1 warning generated.

This warning is caused by commit bb47de73 ("drm/amdgpu: Set FreeSync
state using drm VRR properties") and it could be solved by using NULL
instead of TRANSFER_FUNC_UNKNOWN or casting TRANSFER_FUNC_UNKNOWN as a
pointer. However, after looking into it, there doesn't appear to be a
good reason to pass app_tf by reference as it is never mutated along the
way. This is the only code path in which app_tf is used:

mod_freesync_build_vrr_infopacket ->
    build_vrr_infopacket_v2 ->
        build_vrr_infopacket_fs2_data

Neither mod_freesync_build_vrr_infopacket or build_vrr_infopacket_v2
modify app_tf's value and build_vrr_infopacket_fs2_data expects just
the value so we can avoid dereferencing anything by just passing in
app_tf's value to mod_freesync_build_vrr_infopacket and
build_vrr_infopacket_v2.

There is no functional change because build_vrr_infopacket_fs2_data
doesn't do anything if TRANSFER_FUNC_UNKNOWN is passed to it, the same
as not calling build_vrr_infopacket_fs2_data at all like before this
change when NULL was used for app_tf.
Reviewed-by: NHarry Wentland <harry.wentland@amd.com>
Signed-off-by: NNathan Chancellor <natechancellor@gmail.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

This reverts commit 2172b89e.
Signed-off-by: NCandice Li <candice.li@amd.com>
Reviewed-by: NHawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Instead of the hard-coded ones from VBIOS.
Signed-off-by: NEvan Quan <evan.quan@amd.com>
Acked-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Set the soft/hard max settings as max possible to
not violate the OD settings.
Signed-off-by: NEvan Quan <evan.quan@amd.com>
Acked-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Set the default fclk as what we got from VBIOS.
Signed-off-by: NEvan Quan <evan.quan@amd.com>
Acked-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

There will be some needs to retrieve clock information from other
sysplls also except default 0.
Signed-off-by: NEvan Quan <evan.quan@amd.com>
Acked-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

For UCLK_FMAX OD feature, SMU overwrites the highest UCLK DPM level freq.
Therefore it can only take values that are greater than the second highest
DPM level freq.
Signed-off-by: NEvan Quan <evan.quan@amd.com>
Acked-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

This can fix possible screen freeze on high resolution displays.
Signed-off-by: NEvan Quan <evan.quan@amd.com>
Acked-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

As these settings got reset during above phm_apply_clock_adjust_rules.
Signed-off-by: NEvan Quan <evan.quan@amd.com>
Acked-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

As these are already set during apply_clocks_adjust_rules.
Signed-off-by: NEvan Quan <evan.quan@amd.com>
Acked-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

if use the legacy method to allocate object, when mqd_hiq need to run
uninit code, it will be cause WARNING call trace.

eg: (s3 suspend test)
[   34.918944] Call Trace:
[   34.918948]  [<ffffffff92961dc1>] dump_stack+0x19/0x1b
[   34.918950]  [<ffffffff92297648>] __warn+0xd8/0x100
[   34.918951]  [<ffffffff9229778d>] warn_slowpath_null+0x1d/0x20
[   34.918991]  [<ffffffffc03ce1fe>] uninit_mqd_hiq_sdma+0x4e/0x50 [amdgpu]
[   34.919028]  [<ffffffffc03d0ef7>] uninitialize+0x37/0xe0 [amdgpu]
[   34.919064]  [<ffffffffc03d15a6>] kernel_queue_uninit+0x16/0x30 [amdgpu]
[   34.919086]  [<ffffffffc03d26c2>] pm_uninit+0x12/0x20 [amdgpu]
[   34.919107]  [<ffffffffc03d4915>] stop_nocpsch+0x15/0x20 [amdgpu]
[   34.919129]  [<ffffffffc03c1dce>] kgd2kfd_suspend.part.4+0x2e/0x50 [amdgpu]
[   34.919150]  [<ffffffffc03c2667>] kgd2kfd_suspend+0x17/0x20 [amdgpu]
[   34.919171]  [<ffffffffc03c103a>] amdgpu_amdkfd_suspend+0x1a/0x20 [amdgpu]
[   34.919187]  [<ffffffffc02ec428>] amdgpu_device_suspend+0x88/0x3a0 [amdgpu]
[   34.919189]  [<ffffffff922e22cf>] ? enqueue_entity+0x2ef/0xbe0
[   34.919205]  [<ffffffffc02e8220>] amdgpu_pmops_suspend+0x20/0x30 [amdgpu]
[   34.919207]  [<ffffffff925c56ff>] pci_pm_suspend+0x6f/0x150
[   34.919208]  [<ffffffff925c5690>] ? pci_pm_freeze+0xf0/0xf0
[   34.919210]  [<ffffffff926b45c6>] dpm_run_callback+0x46/0x90
[   34.919212]  [<ffffffff926b49db>] __device_suspend+0xfb/0x2a0
[   34.919213]  [<ffffffff926b4b9f>] async_suspend+0x1f/0xa0
[   34.919214]  [<ffffffff922c918f>] async_run_entry_fn+0x3f/0x130
[   34.919216]  [<ffffffff922b9d4f>] process_one_work+0x17f/0x440
[   34.919217]  [<ffffffff922bade6>] worker_thread+0x126/0x3c0
[   34.919218]  [<ffffffff922bacc0>] ? manage_workers.isra.25+0x2a0/0x2a0
[   34.919220]  [<ffffffff922c1c31>] kthread+0xd1/0xe0
[   34.919221]  [<ffffffff922c1b60>] ? insert_kthread_work+0x40/0x40
[   34.919222]  [<ffffffff92974c1d>] ret_from_fork_nospec_begin+0x7/0x21
[   34.919224]  [<ffffffff922c1b60>] ? insert_kthread_work+0x40/0x40
[   34.919224] ---[ end trace 38cd9f65c963adad ]---
Signed-off-by: NKevin Wang <kevin1.wang@amd.com>
Reviewed-by: NOak Zeng <Oak.Zeng@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

This patch uses REG32_PCIE wrapper instead of writting pci_index2 and reading
pci_data2 for psp. This sequence should be protected by pcie_idx_lock.
Signed-off-by: NHuang Rui <ray.huang@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

This patch uses REG32_PCIE wrapper instead of writting pci_index2 and reading
pci_data2 for powerplay. This sequence should be protected by pcie_idx_lock.
Signed-off-by: NHuang Rui <ray.huang@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

[Why]
For MST, link not disabled until all streams disabled

[How]
Add check for stream_count before setting link_active = false for MST
Signed-off-by: NAnthony Koo <Anthony.Koo@amd.com>
Reviewed-by: NWenjing Liu <Wenjing.Liu@amd.com>
Acked-by: NLeo Li <sunpeng.li@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Reference counting in amdgpu_dm_connector for amdgpu_dm_connector::dc_sink
and amdgpu_dm_connector::dc_em_sink as well as in dc_link::local_sink seems
to be out of shape. Thus make reference counting consistent for these
members and just plain increment the reference count when the variable
gets assigned and decrement when the pointer is set to zero or replaced.
Also simplify reference counting in selected function sopes to be sure the
reference is released in any case. In some cases add NULL pointer check
before dereferencing.
At a hand full of places a comment is placed to stat that the reference
increment happened already somewhere else.

This actually fixes the following kernel bug on my system when enabling
display core in amdgpu. There are some more similar bug reports around,
so it probably helps at more places.

   kernel BUG at mm/slub.c:294!
   invalid opcode: 0000 [#1] SMP PTI
   CPU: 9 PID: 1180 Comm: Xorg Not tainted 5.0.0-rc1+ #2
   Hardware name: Supermicro X10DAi/X10DAI, BIOS 3.0a 02/05/2018
   RIP: 0010:__slab_free+0x1e2/0x3d0
   Code: 8b 54 24 30 48 89 4c 24 28 e8 da fb ff ff 4c 8b 54 24 28 85 c0 0f 85 67 fe ff ff 48 8d 65 d8 5b 41 5c 41 5d 41 5e 41 5f 5d c3 <0f> 0b 49 3b 5c 24 28 75 ab 48 8b 44 24 30 49 89 4c 24 28 49 89 44
   RSP: 0018:ffffb0978589fa90 EFLAGS: 00010246
   RAX: ffff92f12806c400 RBX: 0000000080200019 RCX: ffff92f12806c400
   RDX: ffff92f12806c400 RSI: ffffdd6421a01a00 RDI: ffff92ed2f406e80
   RBP: ffffb0978589fb40 R08: 0000000000000001 R09: ffffffffc0ee4748
   R10: ffff92f12806c400 R11: 0000000000000001 R12: ffffdd6421a01a00
   R13: ffff92f12806c400 R14: ffff92ed2f406e80 R15: ffffdd6421a01a20
   FS:  00007f4170be0ac0(0000) GS:ffff92ed2fb40000(0000) knlGS:0000000000000000
   CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
   CR2: 0000562818aaa000 CR3: 000000045745a002 CR4: 00000000003606e0
   DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
   DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
   Call Trace:
    ? drm_dbg+0x87/0x90 [drm]
    dc_stream_release+0x28/0x50 [amdgpu]
    amdgpu_dm_connector_mode_valid+0xb4/0x1f0 [amdgpu]
    drm_helper_probe_single_connector_modes+0x492/0x6b0 [drm_kms_helper]
    drm_mode_getconnector+0x457/0x490 [drm]
    ? drm_connector_property_set_ioctl+0x60/0x60 [drm]
    drm_ioctl_kernel+0xa9/0xf0 [drm]
    drm_ioctl+0x201/0x3a0 [drm]
    ? drm_connector_property_set_ioctl+0x60/0x60 [drm]
    amdgpu_drm_ioctl+0x49/0x80 [amdgpu]
    do_vfs_ioctl+0xa4/0x630
    ? __sys_recvmsg+0x83/0xa0
    ksys_ioctl+0x60/0x90
    __x64_sys_ioctl+0x16/0x20
    do_syscall_64+0x5b/0x160
    entry_SYSCALL_64_after_hwframe+0x44/0xa9
   RIP: 0033:0x7f417110809b
   Code: 0f 1e fa 48 8b 05 ed bd 0c 00 64 c7 00 26 00 00 00 48 c7 c0 ff ff ff ff c3 66 0f 1f 44 00 00 f3 0f 1e fa b8 10 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d bd bd 0c 00 f7 d8 64 89 01 48
   RSP: 002b:00007ffdd8d1c268 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
   RAX: ffffffffffffffda RBX: 0000562818a8ebc0 RCX: 00007f417110809b
   RDX: 00007ffdd8d1c2a0 RSI: 00000000c05064a7 RDI: 0000000000000012
   RBP: 00007ffdd8d1c2a0 R08: 0000562819012280 R09: 0000000000000007
   R10: 0000000000000000 R11: 0000000000000246 R12: 00000000c05064a7
   R13: 0000000000000012 R14: 0000000000000012 R15: 00007ffdd8d1c2a0
   Modules linked in: nfsv4 dns_resolver nfs lockd grace fscache fuse vfat fat amdgpu intel_rapl sb_edac x86_pkg_temp_thermal intel_powerclamp coretemp kvm_intel kvm irqbypass crct10dif_pclmul chash gpu_sched crc32_pclmul snd_hda_codec_realtek ghash_clmulni_intel amd_iommu_v2 iTCO_wdt iTCO_vendor_support ttm snd_hda_codec_generic snd_hda_codec_hdmi ledtrig_audio snd_hda_intel drm_kms_helper snd_hda_codec intel_cstate snd_hda_core drm snd_hwdep snd_seq snd_seq_device intel_uncore snd_pcm intel_rapl_perf snd_timer snd soundcore ioatdma pcspkr intel_wmi_thunderbolt mxm_wmi i2c_i801 lpc_ich pcc_cpufreq auth_rpcgss sunrpc igb crc32c_intel i2c_algo_bit dca wmi hid_cherry analog gameport joydev

This patch is based on agd5f/drm-next-5.1-wip. This patch does not require
all of that, but agd5f/drm-next-5.1-wip contains at least one more dc_sink
counting fix that I could spot.
Signed-off-by: NMathias Fröhlich <Mathias.Froehlich@web.de>
Reviewed-by: NLeo Li <sunpeng.li@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

This was noticed by Gustavo and his -Wimplicit-fallthrough
patches.  However, in this case, I believe we should have breaks
rather than falling though, that said, in practice we should
never fall through in the first place so there should be no
change in behavior.
Reviewed-by: NEvan Quan <evan.quan@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Simplify the ppfeature mask calculations.
Signed-off-by: NEvan Quan <evan.quan@amd.com>
Reviewed-by: NKenneth Feng <kenneth.feng@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

When DPM for the specific clock is disabled, driver should still print out
current clock info for rocm-smi support on vega20
Signed-off-by: Nshaoyunl <shaoyun.liu@amd.com>
Reviewed-by: NEric Huang <JinhuiEric.Huang@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Fixes for 5.1:
amdgpu:
- Fix missing fw declaration after dropping old CI DPM code
- Fix debugfs access to registers beyond the MMIO bar size
- Fix context priority handling
- Add missing license on some new files
- Various cleanups and bug fixes

radeon:
- Fix missing break in CS parser for evergreen
- Various cleanups and bug fixes

sched:
- Fix entities with 0 run queues
Signed-off-by: NDave Airlie <airlied@redhat.com>
From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190221214134.3308-1-alexander.deucher@amd.com

Signed-off-by: NBas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

s/BOCO/BACO/g
Reviewed-by: NHawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Use a proper return code rather than -1.
Reviewed-by: NHawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Trivial.
Reviewed-by: NHawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

On the display side, cleanups and fixes to enabled modifiers
(QCOM_COMPRESSED).  And otherwise mostly misc fixes all around.
Signed-off-by: NDave Airlie <airlied@redhat.com>
From: Rob Clark <robdclark@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/CAF6AEGuZ5uBKpf=fHvKpTiD10nychuEY8rnE+HeRz0QMvtY5_A@mail.gmail.com

Various fixes/cleanups, along with initial support for SVM features
utilising HMM address-space mirroring and device memory migration.
There's a lot more work to do in these areas, both in terms of
features and efficiency, but these can slowly trickle in later down
the track.
Signed-off-by: NDave Airlie <airlied@redhat.com>
From: Ben Skeggs <skeggsb@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/CACAvsv5bsB4rRY1Gqa_Bp_KAd-v_q1rGZ4nYmOAQhceL0Nr-Xg@mail.gmail.com

Removes the need for temporary VMM mappings.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Removes the need for temporary VMM mappings.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

This add an ioctl to migrate a range of process address space to the
device memory. On platform without cache coherent bus (x86, ARM, ...)
this means that CPU can not access that range directly, instead CPU
will fault which will migrate the memory back to system memory.

This is behind a staging flag so that we can evolve the API.
Signed-off-by: NJérôme Glisse <jglisse@redhat.com>

Device memory can be use in SVM, in which case we do not have any of
the existing buffer object. This commit add infrastructure to allow
use of device memory without nouveau_bo. Again this is a temporary
solution until a rework of GPU memory management.
Signed-off-by: NJérôme Glisse <jglisse@redhat.com>

This uses HMM to mirror a process' CPU page tables into a channel's page
tables, and keep them synchronised so that both the CPU and GPU are able
to access the same memory at the same virtual address.

While this code also supports Volta/Turing, it's only enabled for Pascal
GPUs currently due to channel recovery being unreliable right now on the
later GPUs.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>