- 02 Dec, 2022 8 commits
-
-
Alex Deucher authored
Expand the GPUVM documentation to better describe the hardware functionality and use cases it serves. v2: Fixed a couple of spelling mistakes. Reviewed-by:
Christian König <christian.koenig@amd.com> Signed-off-by:
Alex Deucher <alexander.deucher@amd.com> Link: https://lore.kernel.org/r/20221201214153.8453-2-alexander.deucher@amd.comSigned-off-by:
Luben Tuikov <luben.tuikov@amd.com> Reviewed-by:
-
Alex Deucher authored
Add definitions to clarify GPU virtual memory. v2: clarify the terms a bit more Reviewed-by:
Peter Maucher <bellosilicio@gmail.com> Signed-off-by:
-
Prike Liang authored
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:
Prike Liang <Prike.Liang@amd.com> Reviewed-by:
Mario Limonciello <mario.limonciello@amd.com> Signed-off-by:
-
Jiadong.Zhu authored
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:
Jiadong.Zhu <Jiadong.Zhu@amd.com> Acked-by:
Huang Rui <ray.huang@amd.com> Acked-by:
-
Jiadong.Zhu authored
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:
-
Jiadong.Zhu authored
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:
-
Jiadong.Zhu authored
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:
-
Hamza Mahfooz authored
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:
Leo Li <sunpeng.li@amd.com> Signed-off-by:
Hamza Mahfooz <hamza.mahfooz@amd.com> Signed-off-by:
-
- 01 Dec, 2022 9 commits
-
-
Peter Maucher authored
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:
-
Peter Maucher authored
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:
-
ye xingchen authored
Replace the open-code with sysfs_emit() to simplify the code. Reviewed-by:
ye xingchen <ye.xingchen@zte.com.cn> Signed-off-by:
-
Alex Deucher authored
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:
-
YuBiao Wang authored
Under virtualization guest needs to receive notification from host to perform reset in some cases. Add nv mailbox irq in soc21. Signed-off-by:
YuBiao Wang <YuBiao.Wang@amd.com> Reviewed-by:
Hawking Zhang <Hawking.Zhang@amd.com> Signed-off-by:
-
Leo Liu authored
So that uses PSP to initialize HW. Fixes: 0c2c02b6 ("drm/amdgpu/vcn: add firmware support for dimgrey_cavefish") Signed-off-by:
Leo Liu <leo.liu@amd.com> Reviewed-by:
James Zhu <James.Zhu@amd.com> Signed-off-by:
-
Tao Zhou authored
Configure related registers. Signed-off-by:
Tao Zhou <tao.zhou1@amd.com> Reviewed-by:
-
Jack Xiao authored
Enable reg active poll in mes11. Signed-off-by:
Jack Xiao <Jack.Xiao@amd.com> Reviewed-by:
Evan Quan <evan.quan@amd.com> Signed-off-by:
-
Jack Xiao authored
Update the api def of mes11. Signed-off-by:
-
- 29 Nov, 2022 23 commits
-
-
Konstantin Meskhidze authored
This commit fixes logic error in function 'amdgpu_hw_ip_info': - value 'uvd' might be 'vcn'. Signed-off-by:Konstantin Meskhidze <konstantin.meskhidze@huawei.com> Signed-off-by:
-
Konstantin Meskhidze authored
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:
Felix Kuehling <Felix.Kuehling@amd.com> Reviewed-by:
-
Randy Dunlap authored
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:
Randy 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:
-
Xiongfeng Wang authored
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:
Xiongfeng Wang <wangxiongfeng2@huawei.com> Signed-off-by:
-
Guchun Chen authored
Runtime PM can happen pretty frequently, as these printings may be annoyed, switch to dev_dbg. Suggested-by:
Lijo Lazar <lijo.lazar@amd.com> Signed-off-by:
Guchun Chen <guchun.chen@amd.com> Reviewed-by:
-
Alvin Lee authored
[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:
Dillon Varone <Dillon.Varone@amd.com> Acked-by:
Jasdeep Dhillon <jdhillon@amd.com> Signed-off-by:
Alvin Lee <Alvin.Lee2@amd.com> Signed-off-by:
-
Stylon Wang authored
[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:
Wayne Lin <Wayne.Lin@amd.com> Acked-by:
Stylon Wang <stylon.wang@amd.com> Signed-off-by:
-
Dillon Varone authored
[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:
-
Dillon Varone authored
[Description] Calculations for determining DCC meta size should be pitch*height*bpp/256. Reviewed-by:
-
Dillon Varone authored
[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:
-
Alvin Lee authored
[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:
Jun Lei <Jun.Lei@amd.com> Acked-by:
-
Alvin Lee authored
[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:
-
Dmytro Laktyushkin authored
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:Mark Broadworth <mark.broadworth@amd.com> Reviewed-by:
Rodrigo Siqueira <Rodrigo.Siqueira@amd.com> Signed-off-by:
Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com> Signed-off-by:
-
Stylon Wang authored
[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:
Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com> Acked-by:
-
Dillon Varone authored
[Description] MALL SS allocation calculations should iterate over all pipes to determine the the allocation size required for HW cursor. Reviewed-by:
-
Aric Cyr authored
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:
Aric Cyr <Aric.Cyr@amd.com> Signed-off-by:
-
Wesley Chalmers authored
[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:
Wesley Chalmers <Wesley.Chalmers@amd.com> Signed-off-by:
-
Xiongfeng Wang authored
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:
-
Dan Carpenter authored
We re-arranged this code recently so "ret" is always zero at this point. Signed-off-by:
Dan Carpenter <error27@gmail.com> Signed-off-by:
-
Yu Songping authored
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:
Yu Songping <yusongping@huawei.com> Reviewed-by:
-
Yushan Zhou authored
release_firmware() checks whether firmware pointer is NULL. Remove the redundant NULL check in psp_sw_fini(). Signed-off-by:
Yushan Zhou <katrinzhou@tencent.com> Reviewed-by:
-
Veerabadhran Gopalakrishnan authored
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:
Veerabadhran Gopalakrishnan <veerabadhran.gopalakrishnan@amd.com> Acked-by:
-
Liang He authored
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:
Liang He <windhl@126.com> Reviewed-by:
-
