Expand the GPUVM documentation to better describe the
hardware functionality and use cases it serves.

v2: Fixed a couple of spelling mistakes.
Reviewed-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
Link: https://lore.kernel.org/r/20221201214153.8453-2-alexander.deucher@amd.comSigned-off-by: NLuben Tuikov <luben.tuikov@amd.com>
Reviewed-by: NLuben Tuikov <luben.tuikov@amd.com>

Add definitions to clarify GPU virtual memory.

v2: clarify the terms a bit more
Reviewed-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NLuben Tuikov <luben.tuikov@amd.com>
Suggested-by: NPeter Maucher <bellosilicio@gmail.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
Link: https://lore.kernel.org/r/20221201214153.8453-1-alexander.deucher@amd.com

In the SDMA s0ix save process requires to turn off SDMA ring buffer for
avoiding the SDMA in-flight request, otherwise will suffer from SDMA page
fault which causes by page request from in-flight SDMA ring accessing at
SDMA restore phase.

Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2248
Cc: stable@vger.kernel.org # 6.0,5.15+
Fixes: f8f4e2a5 ("drm/amdgpu: skipping SDMA hw_init and hw_fini for S0ix.")
Signed-off-by: NPrike Liang <Prike.Liang@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>
Tested-by: NMario Limonciello <mario.limonciello@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Trigger Mid-Command Buffer Preemption according to the priority of the software
rings and the hw fence signalling condition.

The muxer saves the locations of the indirect buffer frames from the software
ring together with the fence sequence number in its fifo queue, and pops out
those records when the fences are signalled. The locations are used to resubmit
packages in preemption scenarios by coping the chunks from the software ring.

v2: Update comment style.
v3: Fix conflict caused by previous modifications.
v4: Remove unnecessary prints.
v5: Fix corner cases for resubmission cases.
v6: Refactor functions for resubmission, calling fence_process in irq handler.
v7: Solve conflict for removing amdgpu_sw_ring.c.
v8: Add time threshold to judge if preemption request is needed.
v9: Correct comment spelling. Set fence emit timestamp before rsu assignment.

Cc: Christian Koenig <Christian.Koenig@amd.com>
Cc: Luben Tuikov <Luben.Tuikov@amd.com>
Cc: Andrey Grodzovsky <Andrey.Grodzovsky@amd.com>
Cc: Michel Dänzer <michel@daenzer.net>
Signed-off-by: NJiadong.Zhu <Jiadong.Zhu@amd.com>
Acked-by: NLuben Tuikov <luben.tuikov@amd.com>
Acked-by: NHuang Rui <ray.huang@amd.com>
Acked-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

1. Modify the unmap_queue package on gfx9. Add trailing fence to track the
   preemption done.
2. Modify emit_ce_meta emit_de_meta functions for the resumed ibs.

v2: Restyle code not to use ternary operator.
v3: Modify code format.
v4: Enable Mid-Command Buffer Preemption for gfx9 by default.
v5: Optimize the flag bit set for emit_fence.
v6: Modify log message for preemption timeout.

Cc: Christian Koenig <Christian.Koenig@amd.com>
Cc: Michel Dänzer <michel@daenzer.net>
Cc: Luben Tuikov <Luben.Tuikov@amd.com>
Signed-off-by: NJiadong.Zhu <Jiadong.Zhu@amd.com>
Acked-by: NChristian König <christian.koenig@amd.com>
Acked-by: NHuang Rui <ray.huang@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Set ring functions with software ring callbacks on gfx9.

The software ring could be tested by debugfs_test_ib case.

v2: Set sw_ring 2 to enable software ring by default.
v3: Remove the parameter for software ring enablement.
v4: Use amdgpu_ring_init/fini for software rings.
v5: Update for code format. Fix conflict.
v6: Remove unnecessary checks and enable software ring on gfx9 by default.
v7: Use static array for software ring names and priorities.
v8: Stop creating software rings if no gfx ring existed.

Cc: Christian Koenig <Christian.Koenig@amd.com>
Cc: Luben Tuikov <Luben.Tuikov@amd.com>
Cc: Andrey Grodzovsky <Andrey.Grodzovsky@amd.com>
Cc: Michel Dänzer <michel@daenzer.net>
Cc: Likun Gao <Likun.Gao@amd.com>
Signed-off-by: NJiadong.Zhu <Jiadong.Zhu@amd.com>
Acked-by: NLuben Tuikov <luben.tuikov@amd.com>
Acked-by: NHuang Rui <ray.huang@amd.com>
Acked-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

The software ring is created to support priority context while there is only
one hardware queue for gfx.

Every software ring has its fence driver and could be used as an ordinary ring
for the GPU scheduler.
Multiple software rings are bound to a real ring with the ring muxer. The
packages committed on the software ring are copied to the real ring.

v2: Use array to store software ring entry.
v3: Remove unnecessary prints.
v4: Remove amdgpu_ring_sw_init/fini functions,
using gtt for sw ring buffer for later dma copy
optimization.
v5: Allocate ring entry dynamically in the muxer.
v6: Update comments for the ring muxer.
v7: Modify for function naming.
v8: Combine software ring functions into amdgpu_ring_mux.c
v9: Use kernel-doc comment on the get_rptr function.

Cc: Christian Koenig <Christian.Koenig@amd.com>
Cc: Luben Tuikov <Luben.Tuikov@amd.com>
Cc: Andrey Grodzovsky  <Andrey.Grodzovsky@amd.com>
Cc: Michel Dänzer <michel@daenzer.net>
Signed-off-by: NJiadong.Zhu <Jiadong.Zhu@amd.com>
Acked-by: NHuang Rui <ray.huang@amd.com>
Acked-by: NLuben Tuikov <luben.tuikov@amd.com>
Acked-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Currently, userspace doesn't have a way to communicate selective updates
to displays. So, enable support for FB_DAMAGE_CLIPS for DCN ASICs newer
than DCN301, convert DRM damage clips to dc dirty rectangles and fill
them into dirty_rects in fill_dc_dirty_rects().
Reviewed-by: NLeo Li <sunpeng.li@amd.com>
Signed-off-by: NHamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

The amdgpu kernel module has supported RDNA for a while,
mention that in the module description.

v2: Add CDNA as well (Alex)
Signed-off-by: NPeter Maucher <bellosilicio@gmail.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Document difference between amdgpu.gartsize and amdgpu.gttsize
module parameters, as initially explained by Alex Deucher here:
https://lists.freedesktop.org/archives/dri-devel/2022-October/375358.html

v2: minor cleanups (Alex)
Signed-off-by: NPeter Maucher <bellosilicio@gmail.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Replace the open-code with sysfs_emit() to simplify the code.
Reviewed-by: NLuben Tuikov <luben.tuikov@amd.com>
Signed-off-by: Nye xingchen <ye.xingchen@zte.com.cn>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

This fixes DMCU initialization in APU GPU passthrough.  The
DMCU needs the GPU physical address, not the CPU physical
address.  This ends up working out on bare metal because
we always use the physical address, but doesn't work in
passthrough because the addresses are different.
Reviewed-by: NHuang Rui <ray.huang@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Under virtualization guest needs to receive notification from host to
perform reset in some cases. Add nv mailbox irq in soc21.
Signed-off-by: NYuBiao Wang <YuBiao.Wang@amd.com>
Reviewed-by: NHawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

So that uses PSP to initialize HW.

Fixes: 0c2c02b6 ("drm/amdgpu/vcn: add firmware support for dimgrey_cavefish")
Signed-off-by: NLeo Liu <leo.liu@amd.com>
Reviewed-by: NJames Zhu <James.Zhu@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Configure related registers.
Signed-off-by: NTao Zhou <tao.zhou1@amd.com>
Reviewed-by: NHawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Enable reg active poll in mes11.
Signed-off-by: NJack Xiao <Jack.Xiao@amd.com>
Reviewed-by: NHawking Zhang <Hawking.Zhang@amd.com>
Tested-and-acked-by: NEvan Quan <evan.quan@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Update the api def of mes11.
Signed-off-by: NJack Xiao <Jack.Xiao@amd.com>
Reviewed-by: NHawking Zhang <Hawking.Zhang@amd.com>
Tested-and-acked-by: NEvan Quan <evan.quan@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

This commit fixes logic error in function 'amdgpu_hw_ip_info':
       - value 'uvd' might be 'vcn'.
Signed-off-by: NKonstantin Meskhidze <konstantin.meskhidze@huawei.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

This patch fixes potential memory leakage and seg fault
in  _gpuvm_import_dmabuf() function

Fixes: d4ec4bdc ("drm/amdkfd: Allow access for mmapping KFD BOs")
Signed-off-by: NKonstantin Meskhidze <konstantin.meskhidze@huawei.com>
Signed-off-by: NFelix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: NFelix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Fix documentation build errors for amdgpu: correct the filename.

Error: Cannot open file ../drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c
Error: Cannot open file ../drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c
Error: Cannot open file ../drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c

WARNING: kernel-doc '../scripts/kernel-doc -rst -enable-lineno -sphinx-version 5.3.0 -function MMU Notifier ../drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c' failed with return code 1
WARNING: kernel-doc '../scripts/kernel-doc -rst -enable-lineno -sphinx-version 5.3.0 -internal ../drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c' failed with return code 2

Fixes: d9483ecd ("drm/amdgpu: rename the files for HMM handling")
Signed-off-by: NRandy Dunlap <rdunlap@infradead.org>
Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Felix Kuehling <Felix.Kuehling@amd.com>
Cc: David Airlie <airlied@gmail.com>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: amd-gfx@lists.freedesktop.org
Cc: dri-devel@lists.freedesktop.org
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

As comment of pci_get_class() says, it returns a pci_device with its
refcount increased and decreased the refcount for the input parameter
@from if it is not NULL.

If we break the loop in amdgpu_atrm_get_bios() with 'pdev' not NULL, we
need to call pci_dev_put() to decrease the refcount. Add the missing
pci_dev_put() to avoid refcount leak.

Fixes: d38ceaf9 ("drm/amdgpu: add core driver (v4)")
Signed-off-by: NXiongfeng Wang <wangxiongfeng2@huawei.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Runtime PM can happen pretty frequently, as these printings
may be annoyed, switch to dev_dbg.
Suggested-by: NLijo Lazar <lijo.lazar@amd.com>
Signed-off-by: NGuchun Chen <guchun.chen@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

[Description]
- When transitioning FRL / DP2 is not required, we will always request
  DTBCLK = 0Mhz, but PMFW returns the min freq
- This causes us to make DTBCLK requests every time we call optimize
  after transitioning from FRL to non-FRL
- If DTBCLK is not required, request the min instead (then we only need
  to make 1 extra request at boot time)
- Also when programming PIPE_DTO_SRC_SEL, don't programming for DP
  first, just programming once for the required selection (programming
  DP on an HDMI connection then switching back causes corruption)
Reviewed-by: NDillon Varone <Dillon.Varone@amd.com>
Acked-by: NJasdeep Dhillon <jdhillon@amd.com>
Signed-off-by: NAlvin Lee <Alvin.Lee2@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

[Why]
Tests need to tell if display is connected via USB4 DPIA link.
Currently this is only possible via analyzing dmesg logs.

[How]
Create a per-connector debugfs entry to report if the link is
tunneled via USB4 DPIA.
Reviewed-by: NWayne Lin <Wayne.Lin@amd.com>
Acked-by: NJasdeep Dhillon <jdhillon@amd.com>
Signed-off-by: NStylon Wang <stylon.wang@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

[Description]
Output transfer function must be programmed per pipe as part of a front end
update when the plane changes, or output transfer function changes for a
given plane.
Reviewed-by: NAlvin Lee <Alvin.Lee2@amd.com>
Acked-by: NJasdeep Dhillon <jdhillon@amd.com>
Signed-off-by: NDillon Varone <Dillon.Varone@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

[Description]
Calculations for determining DCC meta size should be pitch*height*bpp/256.
Reviewed-by: NAlvin Lee <Alvin.Lee2@amd.com>
Acked-by: NJasdeep Dhillon <jdhillon@amd.com>
Signed-off-by: NDillon Varone <Dillon.Varone@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

[Description]
Need to subtract unused section of the viewport when calculating required space
in MALL for subvp instead of adding, to prevent over allocation.
Reviewed-by: NAlvin Lee <Alvin.Lee2@amd.com>
Acked-by: NJasdeep Dhillon <jdhillon@amd.com>
Signed-off-by: NDillon Varone <Dillon.Varone@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

[Description]
- When entering into a SubVP config that requires a minimal
  transition we need to retain phantom pipes and also restore
  the mall config
- This is because the min transition will remove phantom pipes
  from the context (shallow copy) and not restore it's original
  state
- This is just a workaround, and needs a proper fix
Reviewed-by: NJun Lei <Jun.Lei@amd.com>
Acked-by: NJasdeep Dhillon <jdhillon@amd.com>
Signed-off-by: NAlvin Lee <Alvin.Lee2@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

[Description]
- This is a workaround to avoid concurrency issues -- a fast update
  creates a shallow copy of the dc current_state, and removes all
  subvp/phantom related flags.
- We want to prevent the fast update thread from removing those
  flags in case there's another thread running that requires
  the info for proper programming
Reviewed-by: NJun Lei <Jun.Lei@amd.com>
Acked-by: NJasdeep Dhillon <jdhillon@amd.com>
Signed-off-by: NAlvin Lee <Alvin.Lee2@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

The 'commit 52e4fdf09ebc ("drm/amd/display: use low clocks for no plane
configs")' introduced a change that set low clock values for DCN31 and
DCN32. As a result of these changes, DC started to spam the log with the
following warning:

------------[ cut here ]------------
WARNING: CPU: 8 PID: 1486 at
drivers/gpu/drm/amd/amdgpu/../display/dc/dcn20/dcn20_dccg.c:58
dccg2_update_dpp_dto+0x3f/0xf0 [amdgpu]
[..]
CPU: 8 PID: 1486 Comm: kms_atomic Tainted: G W 5.18.0+ #1
RIP: 0010:dccg2_update_dpp_dto+0x3f/0xf0 [amdgpu]
RSP: 0018:ffffbbd8025334d0 EFLAGS: 00010206
RAX: 00000000000001ee RBX: ffffa02c87dd3de0 RCX: 00000000000a7f80
RDX: 000000000007dec3 RSI: 0000000000000000 RDI: ffffa02c87dd3de0
RBP: ffffbbd8025334e8 R08: 0000000000000001 R09: 0000000000000005
R10: 00000000000331a0 R11: ffffffffc0b03d80 R12: ffffa02ca576d000
R13: ffffa02cd02c0000 R14: 00000000001453bc R15: ffffa02cdc280000
[..]
dcn20_update_clocks_update_dpp_dto+0x4e/0xa0 [amdgpu]
dcn32_update_clocks+0x5d9/0x650 [amdgpu]
dcn20_prepare_bandwidth+0x49/0x100 [amdgpu]
dcn30_prepare_bandwidth+0x63/0x80 [amdgpu]
dc_commit_state_no_check+0x39d/0x13e0 [amdgpu]
dc_commit_streams+0x1f9/0x3b0 [amdgpu]
dc_commit_state+0x37/0x120 [amdgpu]
amdgpu_dm_atomic_commit_tail+0x5e5/0x2520 [amdgpu]
? _raw_spin_unlock_irqrestore+0x1f/0x40
? down_trylock+0x2c/0x40
? vprintk_emit+0x186/0x2c0
? vprintk_default+0x1d/0x20
? vprintk+0x4e/0x60

We can easily trigger this issue by using a 4k@120 or a 2k@165 and
running some of the kms_atomic tests. This warning is triggered because
the per-pipe clock update is not happening; this commit fixes this issue
by ensuring that DPPCLK is updated when calculating the watermark and
dlg is invoked.

Fixes: 2641c7b7 ("drm/amd/display: use low clocks for no plane configs")
Reported-by: NMark Broadworth <mark.broadworth@amd.com>
Reviewed-by: NRodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: NDmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

[Why]
This fix was intended for improving on coding style but in the process
uncovers a race condition, which explains why we are getting incorrect
length in DPIA AUX replies. Due to the call path of DPIA AUX going from
DC back to DM layer then again into DC and the added complexities on top
of current DC AUX implementation, a proper fix to rely on current dc_lock
to address the race condition is difficult without a major overhual
on how DPIA AUX is implemented.

[How]
- Add a mutex dpia_aux_lock to protect DPIA AUX transfers
- Remove DMUB_ASYNC_TO_SYNC_ACCESS_* codes and rely solely on
  aux_return_code_type for error reporting and handling
- Separate SET_CONFIG from DPIA AUX transfer because they have quiet
  different processing logic
- Remove unnecessary type casting to and from void * type
Reviewed-by: NNicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: NJasdeep Dhillon <jdhillon@amd.com>
Signed-off-by: NStylon Wang <stylon.wang@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

[Description]
MALL SS allocation calculations should iterate over all pipes to determine the
the allocation size required for HW cursor.
Reviewed-by: NAlvin Lee <Alvin.Lee2@amd.com>
Acked-by: NJasdeep Dhillon <jdhillon@amd.com>
Signed-off-by: NDillon Varone <Dillon.Varone@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

This version brings along following fixes:
-Program output transfer function when required
-Fix arthmetic errror in MALL size caluclations for subvp
-DCC Meta pitch used for MALL allocation
-Debugfs entry to tell if connector is DPIA link
-Use largest vready_offset in pipe group
-Fixes race condition in DPIA Aux transfer
Reviewed-by: NRodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Acked-by: NJasdeep Dhillon <jdhillon@amd.com>
Signed-off-by: NAric Cyr <Aric.Cyr@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

[WHY]
Corruption can occur in LB if vready_offset is not large enough.
DML calculates vready_offset for each pipe, but we currently select the
top pipe's vready_offset, which is not necessarily enough for all pipes
in the group.

[HOW]
Wherever program_global_sync is currently called, iterate through the
entire pipe group and find the highest vready_offset.
Reviewed-by: NDillon Varone <Dillon.Varone@amd.com>
Acked-by: NJasdeep Dhillon <jdhillon@amd.com>
Signed-off-by: NWesley Chalmers <Wesley.Chalmers@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

As comment of pci_get_class() says, it returns a pci_device with its
refcount increased and decreased the refcount for the input parameter
@from if it is not NULL.

If we break the loop in radeon_atrm_get_bios() with 'pdev' not NULL, we
need to call pci_dev_put() to decrease the refcount. Add the missing
pci_dev_put() to avoid refcount leak.

Fixes: d8ade352 ("drm/radeon: handle non-VGA class pci devices with ATRM")
Fixes: c61e2775 ("drm/radeon: split ATRM support out from the ATPX handler (v3)")
Signed-off-by: NXiongfeng Wang <wangxiongfeng2@huawei.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

We re-arranged this code recently so "ret" is always zero at this point.
Signed-off-by: NDan Carpenter <error27@gmail.com>
Signed-off-by: NFelix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: NFelix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

The logical operator '&&' will make
smu->ppt_funcs->set_gfx_power_up_by_imu segment fault when
smu->ppt_funcs is NULL.
Signed-off-by: NYu Songping <yusongping@huawei.com>
Reviewed-by: NEvan Quan <evan.quan@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

release_firmware() checks whether firmware pointer is NULL.
Remove the redundant NULL check in psp_sw_fini().
Signed-off-by: NYushan Zhou <katrinzhou@tencent.com>
Reviewed-by: NLuben Tuikov <luben.tuikov@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Corrected the typo in the 4K resolution parameters.

Fixes: b3a24461 ("amdgpu/nv.c - Added codec query for Beige Goby")
Fixes: 9075096b ("amdgpu/nv.c - Optimize code for video codec support structure")
Fixes: 9ac0edaa ("drm/amdgpu: add vcn_4_0_0 video codec query")
Signed-off-by: NVeerabadhran Gopalakrishnan <veerabadhran.gopalakrishnan@amd.com>
Acked-by: NLuben Tuikov <luben.tuikov@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

In amdgpu_get_xgmi_hive(), we should not call kfree() after
kobject_put() as the PUT will call kfree().

In amdgpu_device_ip_init(), we need to check the returned *hive*
which can be NULL before we dereference it.
Signed-off-by: NLiang He <windhl@126.com>
Reviewed-by: NLuben Tuikov <luben.tuikov@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>