- 10 Mar, 2020 3 commits
-
-
Yong Zhao authored
ALLOC_MEM_FLAGS_* used are the same as the KFD_IOC_ALLOC_MEM_FLAGS_*, but they are interweavedly used in kernel driver, resulting in bad readability. For example, KFD_IOC_ALLOC_MEM_FLAGS_COHERENT is not referenced in kernel, and it functions implicitly in kernel through ALLOC_MEM_FLAGS_COHERENT, causing unnecessary confusion. Replace all occurrences of ALLOC_MEM_FLAGS_* with KFD_IOC_ALLOC_MEM_FLAGS_* to solve the problem. Signed-off-by:
Yong Zhao <Yong.Zhao@amd.com> Reviewed-by:
Felix Kuehling <Felix.Kuehling@amd.com> Signed-off-by:
Alex Deucher <alexander.deucher@amd.com>
-
Yong Zhao authored
People are inclined to think of the previous pr_warn message as an error, so use pre_debug instead. Signed-off-by:
-
Nirmoy Das authored
If there are no high priority compute queues available then set normal priority sched array to compute_prio_sched[AMDGPU_GFX_PIPE_PRIO_HIGH] Signed-off-by:
Nirmoy Das <nirmoy.das@amd.com> Reviewed-by:
Christian König <christian.koenig@amd.com> Signed-off-by:
-
- 09 Mar, 2020 21 commits
-
-
Nirmoy Das authored
AMDGPU statically sets priority for compute queues at initialization so remove all the functions responsible for changing compute queue priority dynamically. Signed-off-by:
-
Nirmoy Das authored
Switch to appropriate sched list for an entity on priority override. Signed-off-by:
-
Nirmoy Das authored
Implement drm_sched_entity_modify_sched() which modifies existing sched_list with a different one. This is going to be helpful when userspace changes priority of a ctx/entity then the driver can switch to the corresponding HW scheduler list for that priority. Signed-off-by:
-
Nirmoy Das authored
We were changing compute ring priority while rings were being used before every job submission which is not recommended. This patch sets compute queue priority at mqd initialization for gfx8, gfx9 and gfx10. Policy: make queue 0 of each pipe as high priority compute queue High/normal priority compute sched lists are generated from set of high/normal priority compute queues. At context creation, entity of compute queue get a sched list from high or normal priority depending on ctx->priority Signed-off-by:
-
Xiaojie Yuan authored
Fix version for navi12. Signed-off-by:
Xiaojie Yuan <xiaojie.yuan@amd.com> Reviewed-by:
Hawking Zhang <Hawking.Zhang@amd.com> Signed-off-by:
-
Aly-Tawfik authored
Problem Description: Currently we are checking internal fused rev id with pci rev id. However, fused internal rev id is the same on all raven2 parts (in which Dali and Pollock were based on too), thus Pollock detection fails Fix: use the pci rev to preform the detection for bandwidth calculations. Reviewed-by:
Feifei Xu <Feifei.Xu@amd.com> Signed-off-by:
Aly-Tawfik <altawfik@amd.com> Signed-off-by:
-
Aly-Tawfik authored
Use the pci revision id rather than the asic silicon revision id. Reviewed-by:
-
Andrey Grodzovsky authored
CRTC in DPMS state off calls for low power state entry. Support both atomic mode setting and pre-atomic mode setting. v2: move comment Acked-by:
Andrey Grodzovsky <andrey.grodzovsky@amd.com> Signed-off-by:
-
Melissa Wen authored
The dpp2_get_optimal_number_of_taps function is never used. Removing just for code cleaning up. Signed-off-by:
Melissa Wen <melissa.srw@gmail.com> Reviewed-by:
Rodrigo Siqueira <Rodrigo.Siqueira@amd.com> Signed-off-by:
-
Aric Cyr authored
Signed-off-by:
Aric Cyr <aric.cyr@amd.com> Acked-by:
-
Wenjing Liu authored
[why] FEC capability query should not be affected by debugging decision on whether to disable FEC. We should not determine if display supports FEC by checking debug option. Signed-off-by:
Wenjing Liu <Wenjing.Liu@amd.com> Reviewed-by:
Ashley Thomas <Ashley.Thomas2@amd.com> Acked-by:
-
Martin Leung authored
[why] on some displays that prefer swapped polarity we were seeing L/R images swapped because OTG_STEREO_SYNC_OUTPUT_POLARITY would always be mapped to 0 [how] fix initial dal3 implementation to properly update the polarity field according to the crtc_stereo_flags (same as OTG_STEREO_EYE_FLAG_POLARITY) Signed-off-by:
Martin Leung <martin.leung@amd.com> Reviewed-by:
-
Rodrigo Siqueira authored
Raven provides retimer feature support that requires i2c interaction in order to make it work well, all settings required for this configuration are loaded from the Atom bios which include the i2c address. If the retimer feature is not available, we should abort the attempt to set this feature, otherwise, it makes the following line return I2C_CHANNEL_OPERATION_NO_RESPONSE: i2c_success = i2c_write(pipe_ctx, slave_address, buffer, sizeof(buffer)); ... if (!i2c_success) ASSERT(i2c_success); This ends up causing problems with hotplugging HDMI displays on Raven, and causes retimer settings to warn like so: WARNING: CPU: 1 PID: 429 at drivers/gpu/drm/amd/amdgpu/../dal/dc/core/dc_link.c:1998 write_i2c_retimer_setting+0xc2/0x3c0 [amdgpu] Modules linked in: edac_mce_amd ccp kvm irqbypass binfmt_misc crct10dif_pclmul crc32_pclmul ghash_clmulni_intel snd_hda_codec_realtek snd_hda_codec_generic ledtrig_audio snd_hda_codec_hdmi snd_hda_intel amdgpu(+) snd_hda_codec snd_hda_core snd_hwdep snd_pcm snd_seq_midi snd_seq_midi_event snd_rawmidi aesni_intel snd_seq amd_iommu_v2 gpu_sched aes_x86_64 crypto_simd cryptd glue_helper snd_seq_device ttm drm_kms_helper snd_timer eeepc_wmi wmi_bmof asus_wmi sparse_keymap drm mxm_wmi snd k10temp fb_sys_fops syscopyarea sysfillrect sysimgblt soundcore joydev input_leds mac_hid sch_fq_codel parport_pc ppdev lp parport ip_tables x_tables autofs4 igb i2c_algo_bit hid_generic usbhid i2c_piix4 dca ahci hid libahci video wmi gpio_amdpt gpio_generic CPU: 1 PID: 429 Comm: systemd-udevd Tainted: G W 5.2.0-rc1sept162019+ #1 Hardware name: System manufacturer System Product Name/ROG STRIX B450-F GAMING, BIOS 2605 08/06/2019 RIP: 0010:write_i2c_retimer_setting+0xc2/0x3c0 [amdgpu] Code: ff 0f b6 4d ce 44 0f b6 45 cf 44 0f b6 c8 45 89 cf 44 89 e2 48 c7 c6 f0 34 bc c0 bf 04 00 00 00 e8 63 b0 90 ff 45 84 ff 75 02 <0f> 0b 42 0f b6 04 73 8d 50 f6 80 fa 02 77 8c 3c 0a 0f 85 c8 00 00 RSP: 0018:ffffa99d02726fd0 EFLAGS: 00010246 RAX: 0000000000000000 RBX: ffffa99d02727035 RCX: 0000000000000006 RDX: 0000000000000000 RSI: 0000000000000002 RDI: ffff976acc857440 RBP: ffffa99d02727018 R08: 0000000000000002 R09: 000000000002a600 R10: ffffe90610193680 R11: 00000000000005e3 R12: 000000000000005d R13: ffff976ac4b201b8 R14: 0000000000000001 R15: 0000000000000000 FS: 00007f14f99e1680(0000) GS:ffff976acc840000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007fdf212843b8 CR3: 0000000408906000 CR4: 00000000003406e0 Call Trace: core_link_enable_stream+0x626/0x680 [amdgpu] dce110_apply_ctx_to_hw+0x414/0x4e0 [amdgpu] dc_commit_state+0x331/0x5e0 [amdgpu] ? drm_calc_timestamping_constants+0xf9/0x150 [drm] amdgpu_dm_atomic_commit_tail+0x395/0x1e00 [amdgpu] ? dm_plane_helper_prepare_fb+0x20c/0x280 [amdgpu] commit_tail+0x42/0x70 [drm_kms_helper] drm_atomic_helper_commit+0x10c/0x120 [drm_kms_helper] amdgpu_dm_atomic_commit+0x95/0xa0 [amdgpu] drm_atomic_commit+0x4a/0x50 [drm] restore_fbdev_mode_atomic+0x1c0/0x1e0 [drm_kms_helper] restore_fbdev_mode+0x4c/0x160 [drm_kms_helper] ? _cond_resched+0x19/0x40 drm_fb_helper_restore_fbdev_mode_unlocked+0x4e/0xa0 [drm_kms_helper] drm_fb_helper_set_par+0x2d/0x50 [drm_kms_helper] fbcon_init+0x471/0x630 visual_init+0xd5/0x130 do_bind_con_driver+0x20a/0x430 do_take_over_console+0x7d/0x1b0 do_fbcon_takeover+0x5c/0xb0 fbcon_event_notify+0x6cd/0x8a0 notifier_call_chain+0x4c/0x70 blocking_notifier_call_chain+0x43/0x60 fb_notifier_call_chain+0x1b/0x20 register_framebuffer+0x254/0x360 __drm_fb_helper_initial_config_and_unlock+0x2c5/0x510 [drm_kms_helper] drm_fb_helper_initial_config+0x35/0x40 [drm_kms_helper] amdgpu_fbdev_init+0xcd/0x100 [amdgpu] amdgpu_device_init+0x1156/0x1930 [amdgpu] amdgpu_driver_load_kms+0x8d/0x2e0 [amdgpu] drm_dev_register+0x12b/0x1c0 [drm] amdgpu_pci_probe+0xd3/0x160 [amdgpu] local_pci_probe+0x47/0xa0 pci_device_probe+0x142/0x1b0 really_probe+0xf5/0x3d0 driver_probe_device+0x11b/0x130 device_driver_attach+0x58/0x60 __driver_attach+0xa3/0x140 ? device_driver_attach+0x60/0x60 ? device_driver_attach+0x60/0x60 bus_for_each_dev+0x74/0xb0 ? kmem_cache_alloc_trace+0x1a3/0x1c0 driver_attach+0x1e/0x20 bus_add_driver+0x147/0x220 ? 0xffffffffc0cb9000 driver_register+0x60/0x100 ? 0xffffffffc0cb9000 __pci_register_driver+0x5a/0x60 amdgpu_init+0x74/0x83 [amdgpu] do_one_initcall+0x4a/0x1fa ? _cond_resched+0x19/0x40 ? kmem_cache_alloc_trace+0x3f/0x1c0 ? __vunmap+0x1cc/0x200 do_init_module+0x5f/0x227 load_module+0x2330/0x2b40 __do_sys_finit_module+0xfc/0x120 ? __do_sys_finit_module+0xfc/0x120 __x64_sys_finit_module+0x1a/0x20 do_syscall_64+0x5a/0x130 entry_SYSCALL_64_after_hwframe+0x44/0xa9 RIP: 0033:0x7f14f9500839 Code: 00 f3 c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 1f f6 2c 00 f7 d8 64 89 01 48 RSP: 002b:00007fff9bc4f5a8 EFLAGS: 00000246 ORIG_RAX: 0000000000000139 RAX: ffffffffffffffda RBX: 000055afb5abce30 RCX: 00007f14f9500839 RDX: 0000000000000000 RSI: 000055afb5ace0f0 RDI: 0000000000000017 RBP: 000055afb5ace0f0 R08: 0000000000000000 R09: 000000000000000a R10: 0000000000000017 R11: 0000000000000246 R12: 0000000000000000 R13: 000055afb5aad800 R14: 0000000000020000 R15: 0000000000000000 ---[ end trace c286e96563966f08 ]--- This commit reworks the way that we handle i2c write for retimer in the way that we abort this configuration if the feature is not available in the device. For debug sake, we kept a simple log message in case the retimer is not available. Signed-off-by:
Hersen Wu <hersenxs.wu@amd.com> Acked-by:
-
Wenjing Liu authored
[why] In HDCP Uninitialzed State, a CPIRQ event would cause log output internal policy error because the CPIRQ event is not recognized as unexpected event. [how] CPIRQ is issued in HDCP uninitialized state is unexpected. We should set unexpected event flag in event ctx. Signed-off-by:
-
Wenjing Liu authored
[why] Some RX doesn't like us to read rx id list upto max rx id list size. As discussed, we decided to read rx id list based on device count. [how] According to HDCP specs the actual size of rx id list is calculated as rx id list size = 2+3+16+5*device_count. We will read 16 bytes at a time until it reached or exceeded rx id list size. Signed-off-by:
-
Nikola Cornij authored
[why] Link or DIG BE can't be exposed to a higher stream bandwidth than they can handle. When DSC is required to fit the stream into the link bandwidth, DSC has to be programmed during timing programming to ensure this. Without it, intermittent issues such as black screen after S3 or a hot-plug can be seen. [how] Move DSC programming from enabling stream on link to timing setup. Signed-off-by:
Nikola Cornij <nikola.cornij@amd.com> Reviewed-by:
Tony Cheng <Tony.Cheng@amd.com> Acked-by:
-
Yongqiang Sun authored
[Why] System isn't able to enter S0i3 due to not send display count 0 to smu. When dpms off, clk changed flag is cleared alreay, and it is checked when doing optimized bandwidth, and update clocks is bypassed due to the flag is unset. [How] Remove check flag incide the function since watermark values and clocks values are checked during update to determine whether to perform it, no need to check it again outside the function. Signed-off-by:
Yongqiang Sun <yongqiang.sun@amd.com> Reviewed-by:
-
Braden Bakker authored
[What] Need debug options to control lightl/deep sleep [How] Add registry for memory power control Signed-off-by:
Braden Bakker <Braden.Bakker@amd.com> Reviewed-by:
Charlene Liu <Charlene.Liu@amd.com> Acked-by:
-
Wenjing Liu authored
[why] It is possible even if sink signal is MST but driver enables SST stream. We should not determine if we should do MST authentication based on sink's capability. Instead we should determine whether to do MST authentication based on what we have enabled in stream. Signed-off-by:
-
Isabel Zhang authored
[Why] So the values of stay count and bstatus can be easily viewed during debugging. [How] Add stay count and bstatus values to be outputted in HDCP log Signed-off-by:
Isabel Zhang <isabel.zhang@amd.com> Reviewed-by:
-
Martin Leung authored
[why] nv14 previously inherited soc bb from generic dcn 2, did not match watermark values according to memory team [how] add nv14 specific soc bb: copy nv2 generic that it was using from before, but changed num channels to 8 Signed-off-by:
Jun Lei <Jun.Lei@amd.com> Acked-by:
-
- 06 Mar, 2020 16 commits
-
-
Felix Kuehling authored
Otherwise BOs may wait for the fence indefinitely and never be destroyed. v2: Signal the fence right after destroying queues to avoid unnecessary delaye-delete in kfd_process_wq_release Signed-off-by:xinhui pan <xinhui.pan@amd.com> Acked-by:
-
Monk Liu authored
and disable MC resum in VCN2.0 as well those are not concerned by VF driver Singed-off-by:
darlington Opara <darlington.opara@amd.com> Signed-off-by:
Jinage Zhao <jiange.zhao@amd.com> Signed-off-by:
Monk Liu <Monk.Liu@amd.com> Acked-by:
Leo Liu <leo.liu@amd.com> Signed-off-by:
-
Monk Liu authored
support IB test on dec/enc ring disable ring test on dec/enc ring (MMSCH limitation) v2: squash in unused variable warning fix Singed-off-by:
-
Monk Liu authored
something need to do for VCN2.0 enablement on SRIOV: 1)use one dec ring and one enc ring 2)allocate MM table for MMSCH usage 3)implement SRIOV version vcn_start which orgnize vcn programing with patcket format and implement start mmsch for to run those packet 4)doorbell is changed for SRIOV Singed-off-by:
Emily Deng <Emily.Deng@amd.com> Signed-off-by:
-
Monk Liu authored
MMSCH doesn't support jpeg ring on SRIOV Signed-off-by:
-
Monk Liu authored
Signed-off-by:
-
Yong Zhao authored
Because too many things are involved in this workaround, we need more comments to avoid pitfalls. Signed-off-by:
Philip Yang <philip.yang@amd.com> Signed-off-by:
-
Yong Zhao authored
Add "CP" to AMDGPU_GEM_CREATE_MQD_GFX9 to indicate it is only for CP MQD buffer. Signed-off-by:
-
Andrey Grodzovsky authored
Problem: During GU reset PSP's sysfs was being wrongly reinitilized during call to amdgpu_device_ip_late_init which was failing with duplicate error. Fix: Move psp_sysfs_init to psp_sw_init to avoid this. Add guards in sysfs file's read and write hook agains premature call if PSP is not finished initialization. Signed-off-by:
-
Jacob He authored
SPM access the video memory according to SPM_VMID. It should be updated with the job's vmid right before the job is scheduled. SPM_VMID is a global resource Signed-off-by:
Jacob He <jacob.he@amd.com> Reviewed-by:
-
John Clements authored
check UMC status and exit prior to making and erroneus register access this resolved unexpected behaviour with UMC indexing mode broadcasting writes Reviewed-by:
John Clements <john.clements@amd.com> Signed-off-by:
-
Guchun Chen authored
On arcturus, DF-Cstate needs to be toggled off/on before and after accessing UMC error counter and error address registers, otherwise, clearing such registers may fail. Signed-off-by:
Guchun Chen <guchun.chen@amd.com> Reviewed-by:
Tao Zhou <tao.zhou1@amd.com> Signed-off-by:
-
John Clements authored
mitigates race condition on BACO reset between GPU bootcode and driver reload Reviewed-by:
-
Hawking Zhang authored
add arcturus xgmi/wafl pcs err status group to support PCS error detection and report on arcturus Signed-off-by:
-
Hawking Zhang authored
Now driver will report XGMI/WAFL PCS error through sysfs xgmi_wafl_err_count node on Vega20 Signed-off-by:
-
Hawking Zhang authored
Since from vega20, hardware supports run-time detect and report XGMI/WAFL PCS ras error. Add helper functions to walkthrough every type of ras error and report it if any. Signed-off-by:
-
