cachepc-linux

Fork of AMDESE/linux with modifications for CachePC side-channel attack
git clone https://git.sinitax.com/sinitax/cachepc-linux
Log | Files | Refs | README | LICENSE | sfeed.txt

nbio_v4_3.c (13054B)


      1/*
      2 * Copyright 2021 Advanced Micro Devices, Inc.
      3 *
      4 * Permission is hereby granted, free of charge, to any person obtaining a
      5 * copy of this software and associated documentation files (the "Software"),
      6 * to deal in the Software without restriction, including without limitation
      7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
      8 * and/or sell copies of the Software, and to permit persons to whom the
      9 * Software is furnished to do so, subject to the following conditions:
     10 *
     11 * The above copyright notice and this permission notice shall be included in
     12 * all copies or substantial portions of the Software.
     13 *
     14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
     15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
     16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
     17 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
     18 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
     19 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
     20 * OTHER DEALINGS IN THE SOFTWARE.
     21 *
     22 */
     23#include "amdgpu.h"
     24#include "amdgpu_atombios.h"
     25#include "nbio_v4_3.h"
     26
     27#include "nbio/nbio_4_3_0_offset.h"
     28#include "nbio/nbio_4_3_0_sh_mask.h"
     29#include <uapi/linux/kfd_ioctl.h>
     30
     31static void nbio_v4_3_remap_hdp_registers(struct amdgpu_device *adev)
     32{
     33	WREG32_SOC15(NBIO, 0, regBIF_BX0_REMAP_HDP_MEM_FLUSH_CNTL,
     34		adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL);
     35	WREG32_SOC15(NBIO, 0, regBIF_BX0_REMAP_HDP_REG_FLUSH_CNTL,
     36		adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_REG_FLUSH_CNTL);
     37}
     38
     39static u32 nbio_v4_3_get_rev_id(struct amdgpu_device *adev)
     40{
     41	u32 tmp = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_DEV0_EPF0_STRAP0);
     42
     43	tmp &= RCC_STRAP0_RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0_MASK;
     44	tmp >>= RCC_STRAP0_RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0__SHIFT;
     45
     46	return tmp;
     47}
     48
     49static void nbio_v4_3_mc_access_enable(struct amdgpu_device *adev, bool enable)
     50{
     51	if (enable)
     52		WREG32_SOC15(NBIO, 0, regBIF_BX0_BIF_FB_EN,
     53			     BIF_BX0_BIF_FB_EN__FB_READ_EN_MASK |
     54			     BIF_BX0_BIF_FB_EN__FB_WRITE_EN_MASK);
     55	else
     56		WREG32_SOC15(NBIO, 0, regBIF_BX0_BIF_FB_EN, 0);
     57}
     58
     59static u32 nbio_v4_3_get_memsize(struct amdgpu_device *adev)
     60{
     61	return RREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF0_RCC_CONFIG_MEMSIZE);
     62}
     63
     64static void nbio_v4_3_sdma_doorbell_range(struct amdgpu_device *adev, int instance,
     65					  bool use_doorbell, int doorbell_index,
     66					  int doorbell_size)
     67{
     68	if (instance == 0) {
     69		u32 doorbell_range = RREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_2_CTRL);
     70
     71		if (use_doorbell) {
     72			doorbell_range = REG_SET_FIELD(doorbell_range,
     73						       S2A_DOORBELL_ENTRY_2_CTRL,
     74						       S2A_DOORBELL_PORT2_ENABLE,
     75						       0x1);
     76			doorbell_range = REG_SET_FIELD(doorbell_range,
     77						       S2A_DOORBELL_ENTRY_2_CTRL,
     78						       S2A_DOORBELL_PORT2_AWID,
     79						       0xe);
     80			doorbell_range = REG_SET_FIELD(doorbell_range,
     81						       S2A_DOORBELL_ENTRY_2_CTRL,
     82						       S2A_DOORBELL_PORT2_RANGE_OFFSET,
     83						       doorbell_index);
     84			doorbell_range = REG_SET_FIELD(doorbell_range,
     85						       S2A_DOORBELL_ENTRY_2_CTRL,
     86						       S2A_DOORBELL_PORT2_RANGE_SIZE,
     87						       doorbell_size);
     88			doorbell_range = REG_SET_FIELD(doorbell_range,
     89						       S2A_DOORBELL_ENTRY_2_CTRL,
     90						       S2A_DOORBELL_PORT2_AWADDR_31_28_VALUE,
     91						       0x3);
     92		} else
     93			doorbell_range = REG_SET_FIELD(doorbell_range,
     94						       S2A_DOORBELL_ENTRY_2_CTRL,
     95						       S2A_DOORBELL_PORT2_RANGE_SIZE,
     96						       0);
     97
     98		WREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_2_CTRL, doorbell_range);
     99	}
    100}
    101
    102static void nbio_v4_3_vcn_doorbell_range(struct amdgpu_device *adev, bool use_doorbell,
    103					 int doorbell_index, int instance)
    104{
    105	u32 doorbell_range;
    106
    107	if (instance)
    108		doorbell_range = RREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_5_CTRL);
    109	else
    110		doorbell_range = RREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_4_CTRL);
    111
    112	if (use_doorbell) {
    113		doorbell_range = REG_SET_FIELD(doorbell_range,
    114					       S2A_DOORBELL_ENTRY_4_CTRL,
    115					       S2A_DOORBELL_PORT4_ENABLE,
    116					       0x1);
    117		doorbell_range = REG_SET_FIELD(doorbell_range,
    118					       S2A_DOORBELL_ENTRY_4_CTRL,
    119					       S2A_DOORBELL_PORT4_AWID,
    120					       instance ? 0x7 : 0x4);
    121		doorbell_range = REG_SET_FIELD(doorbell_range,
    122					       S2A_DOORBELL_ENTRY_4_CTRL,
    123					       S2A_DOORBELL_PORT4_RANGE_OFFSET,
    124					       doorbell_index);
    125		doorbell_range = REG_SET_FIELD(doorbell_range,
    126					       S2A_DOORBELL_ENTRY_4_CTRL,
    127					       S2A_DOORBELL_PORT4_RANGE_SIZE,
    128					       8);
    129		doorbell_range = REG_SET_FIELD(doorbell_range,
    130					       S2A_DOORBELL_ENTRY_4_CTRL,
    131					       S2A_DOORBELL_PORT4_AWADDR_31_28_VALUE,
    132					       instance ? 0x7 : 0x4);
    133	} else
    134		doorbell_range = REG_SET_FIELD(doorbell_range,
    135					       S2A_DOORBELL_ENTRY_4_CTRL,
    136					       S2A_DOORBELL_PORT4_RANGE_SIZE,
    137					       0);
    138
    139	if (instance)
    140		WREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_5_CTRL, doorbell_range);
    141	else
    142		WREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_4_CTRL, doorbell_range);
    143}
    144
    145static void nbio_v4_3_gc_doorbell_init(struct amdgpu_device *adev)
    146{
    147	WREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_0_CTRL, 0x30000007);
    148	WREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_3_CTRL, 0x3000000d);
    149}
    150
    151static void nbio_v4_3_enable_doorbell_aperture(struct amdgpu_device *adev,
    152					       bool enable)
    153{
    154	WREG32_FIELD15_PREREG(NBIO, 0, RCC_DEV0_EPF0_RCC_DOORBELL_APER_EN,
    155			BIF_DOORBELL_APER_EN, enable ? 1 : 0);
    156}
    157
    158static void nbio_v4_3_enable_doorbell_selfring_aperture(struct amdgpu_device *adev,
    159							bool enable)
    160{
    161	u32 tmp = 0;
    162
    163	if (enable) {
    164		tmp = REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL,
    165				    DOORBELL_SELFRING_GPA_APER_EN, 1) |
    166		      REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL,
    167				    DOORBELL_SELFRING_GPA_APER_MODE, 1) |
    168		      REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL,
    169				    DOORBELL_SELFRING_GPA_APER_SIZE, 0);
    170
    171		WREG32_SOC15(NBIO, 0, regBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_BASE_LOW,
    172			     lower_32_bits(adev->doorbell.base));
    173		WREG32_SOC15(NBIO, 0, regBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_BASE_HIGH,
    174			     upper_32_bits(adev->doorbell.base));
    175	}
    176
    177	WREG32_SOC15(NBIO, 0, regBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL,
    178		     tmp);
    179}
    180
    181static void nbio_v4_3_ih_doorbell_range(struct amdgpu_device *adev,
    182					bool use_doorbell, int doorbell_index)
    183{
    184	u32 ih_doorbell_range = RREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_1_CTRL);
    185
    186	if (use_doorbell) {
    187		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
    188						  S2A_DOORBELL_ENTRY_1_CTRL,
    189						  S2A_DOORBELL_PORT1_ENABLE,
    190						  0x1);
    191		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
    192						  S2A_DOORBELL_ENTRY_1_CTRL,
    193						  S2A_DOORBELL_PORT1_AWID,
    194						  0x0);
    195		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
    196						  S2A_DOORBELL_ENTRY_1_CTRL,
    197						  S2A_DOORBELL_PORT1_RANGE_OFFSET,
    198						  doorbell_index);
    199		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
    200						  S2A_DOORBELL_ENTRY_1_CTRL,
    201						  S2A_DOORBELL_PORT1_RANGE_SIZE,
    202						  2);
    203		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
    204						  S2A_DOORBELL_ENTRY_1_CTRL,
    205						  S2A_DOORBELL_PORT1_AWADDR_31_28_VALUE,
    206						  0x0);
    207	} else
    208		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
    209						  S2A_DOORBELL_ENTRY_1_CTRL,
    210						  S2A_DOORBELL_PORT1_RANGE_SIZE,
    211						  0);
    212
    213	WREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_1_CTRL, ih_doorbell_range);
    214}
    215
    216static void nbio_v4_3_ih_control(struct amdgpu_device *adev)
    217{
    218	u32 interrupt_cntl;
    219
    220	/* setup interrupt control */
    221	WREG32_SOC15(NBIO, 0, regBIF_BX0_INTERRUPT_CNTL2, adev->dummy_page_addr >> 8);
    222
    223	interrupt_cntl = RREG32_SOC15(NBIO, 0, regBIF_BX0_INTERRUPT_CNTL);
    224	/*
    225	 * BIF_BX0_INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=0 - dummy read disabled with msi, enabled without msi
    226	 * BIF_BX0_INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=1 - dummy read controlled by IH_DUMMY_RD_EN
    227	 */
    228	interrupt_cntl = REG_SET_FIELD(interrupt_cntl, BIF_BX0_INTERRUPT_CNTL,
    229				       IH_DUMMY_RD_OVERRIDE, 0);
    230
    231	/* BIF_BX0_INTERRUPT_CNTL__IH_REQ_NONSNOOP_EN_MASK=1 if ring is in non-cacheable memory, e.g., vram */
    232	interrupt_cntl = REG_SET_FIELD(interrupt_cntl, BIF_BX0_INTERRUPT_CNTL,
    233				       IH_REQ_NONSNOOP_EN, 0);
    234
    235	WREG32_SOC15(NBIO, 0, regBIF_BX0_INTERRUPT_CNTL, interrupt_cntl);
    236}
    237
    238static void nbio_v4_3_update_medium_grain_clock_gating(struct amdgpu_device *adev,
    239						       bool enable)
    240{
    241	uint32_t def, data;
    242
    243	def = data = RREG32_SOC15(NBIO, 0, regCPM_CONTROL);
    244	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_BIF_MGCG)) {
    245		data |= (CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK |
    246			 CPM_CONTROL__TXCLK_DYN_GATE_ENABLE_MASK |
    247			 CPM_CONTROL__TXCLK_LCNT_GATE_ENABLE_MASK |
    248			 CPM_CONTROL__TXCLK_REGS_GATE_ENABLE_MASK |
    249			 CPM_CONTROL__TXCLK_PRBS_GATE_ENABLE_MASK |
    250			 CPM_CONTROL__REFCLK_REGS_GATE_ENABLE_MASK);
    251	} else {
    252		data &= ~(CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK |
    253			  CPM_CONTROL__TXCLK_DYN_GATE_ENABLE_MASK |
    254			  CPM_CONTROL__TXCLK_LCNT_GATE_ENABLE_MASK |
    255			  CPM_CONTROL__TXCLK_REGS_GATE_ENABLE_MASK |
    256			  CPM_CONTROL__TXCLK_PRBS_GATE_ENABLE_MASK |
    257			  CPM_CONTROL__REFCLK_REGS_GATE_ENABLE_MASK);
    258	}
    259
    260	if (def != data)
    261		WREG32_SOC15(NBIO, 0, regCPM_CONTROL, data);
    262}
    263
    264static void nbio_v4_3_update_medium_grain_light_sleep(struct amdgpu_device *adev,
    265						      bool enable)
    266{
    267	uint32_t def, data;
    268
    269	/* TODO: need update in future */
    270	def = data = RREG32_SOC15(NBIO, 0, regPCIE_CNTL2);
    271	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_BIF_LS)) {
    272		data |= PCIE_CNTL2__SLV_MEM_LS_EN_MASK;
    273	} else {
    274		data &= ~PCIE_CNTL2__SLV_MEM_LS_EN_MASK;
    275	}
    276
    277	if (def != data)
    278		WREG32_SOC15(NBIO, 0, regPCIE_CNTL2, data);
    279}
    280
    281static void nbio_v4_3_get_clockgating_state(struct amdgpu_device *adev,
    282					    u64 *flags)
    283{
    284	int data;
    285
    286	/* AMD_CG_SUPPORT_BIF_MGCG */
    287	data = RREG32_SOC15(NBIO, 0, regCPM_CONTROL);
    288	if (data & CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK)
    289		*flags |= AMD_CG_SUPPORT_BIF_MGCG;
    290
    291	/* AMD_CG_SUPPORT_BIF_LS */
    292	data = RREG32_SOC15(NBIO, 0, regPCIE_CNTL2);
    293	if (data & PCIE_CNTL2__SLV_MEM_LS_EN_MASK)
    294		*flags |= AMD_CG_SUPPORT_BIF_LS;
    295}
    296
    297static u32 nbio_v4_3_get_hdp_flush_req_offset(struct amdgpu_device *adev)
    298{
    299	return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF0_GPU_HDP_FLUSH_REQ);
    300}
    301
    302static u32 nbio_v4_3_get_hdp_flush_done_offset(struct amdgpu_device *adev)
    303{
    304	return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF0_GPU_HDP_FLUSH_DONE);
    305}
    306
    307static u32 nbio_v4_3_get_pcie_index_offset(struct amdgpu_device *adev)
    308{
    309	return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF0_RSMU_INDEX);
    310}
    311
    312static u32 nbio_v4_3_get_pcie_data_offset(struct amdgpu_device *adev)
    313{
    314	return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF0_RSMU_DATA);
    315}
    316
    317const struct nbio_hdp_flush_reg nbio_v4_3_hdp_flush_reg = {
    318	.ref_and_mask_cp0 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP0_MASK,
    319	.ref_and_mask_cp1 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP1_MASK,
    320	.ref_and_mask_cp2 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP2_MASK,
    321	.ref_and_mask_cp3 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP3_MASK,
    322	.ref_and_mask_cp4 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP4_MASK,
    323	.ref_and_mask_cp5 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP5_MASK,
    324	.ref_and_mask_cp6 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP6_MASK,
    325	.ref_and_mask_cp7 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP7_MASK,
    326	.ref_and_mask_cp8 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP8_MASK,
    327	.ref_and_mask_cp9 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP9_MASK,
    328	.ref_and_mask_sdma0 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__SDMA0_MASK,
    329	.ref_and_mask_sdma1 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__SDMA1_MASK,
    330};
    331
    332static void nbio_v4_3_init_registers(struct amdgpu_device *adev)
    333{
    334	return;
    335}
    336
    337static u32 nbio_v4_3_get_rom_offset(struct amdgpu_device *adev)
    338{
    339	u32 data, rom_offset;
    340
    341	data = RREG32_SOC15(NBIO, 0, regREGS_ROM_OFFSET_CTRL);
    342	rom_offset = REG_GET_FIELD(data, REGS_ROM_OFFSET_CTRL, ROM_OFFSET);
    343
    344	return rom_offset;
    345}
    346
    347const struct amdgpu_nbio_funcs nbio_v4_3_funcs = {
    348	.get_hdp_flush_req_offset = nbio_v4_3_get_hdp_flush_req_offset,
    349	.get_hdp_flush_done_offset = nbio_v4_3_get_hdp_flush_done_offset,
    350	.get_pcie_index_offset = nbio_v4_3_get_pcie_index_offset,
    351	.get_pcie_data_offset = nbio_v4_3_get_pcie_data_offset,
    352	.get_rev_id = nbio_v4_3_get_rev_id,
    353	.mc_access_enable = nbio_v4_3_mc_access_enable,
    354	.get_memsize = nbio_v4_3_get_memsize,
    355	.sdma_doorbell_range = nbio_v4_3_sdma_doorbell_range,
    356	.vcn_doorbell_range = nbio_v4_3_vcn_doorbell_range,
    357	.gc_doorbell_init = nbio_v4_3_gc_doorbell_init,
    358	.enable_doorbell_aperture = nbio_v4_3_enable_doorbell_aperture,
    359	.enable_doorbell_selfring_aperture = nbio_v4_3_enable_doorbell_selfring_aperture,
    360	.ih_doorbell_range = nbio_v4_3_ih_doorbell_range,
    361	.update_medium_grain_clock_gating = nbio_v4_3_update_medium_grain_clock_gating,
    362	.update_medium_grain_light_sleep = nbio_v4_3_update_medium_grain_light_sleep,
    363	.get_clockgating_state = nbio_v4_3_get_clockgating_state,
    364	.ih_control = nbio_v4_3_ih_control,
    365	.init_registers = nbio_v4_3_init_registers,
    366	.remap_hdp_registers = nbio_v4_3_remap_hdp_registers,
    367	.get_rom_offset = nbio_v4_3_get_rom_offset,
    368};