drm/amdgpu: create vega20 ih blocks
authorHawking Zhang <Hawking.Zhang@amd.com>
Fri, 18 Dec 2020 13:48:59 +0000 (21:48 +0800)
committerAlex Deucher <alexander.deucher@amd.com>
Wed, 23 Dec 2020 20:05:35 +0000 (15:05 -0500)
vega20 ih blocks will be used for vega20/arcturus

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Dennis Li <Dennis.Li@amd.com>
Reviewed-by: Feifei Xu <Feifei.Xu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drivers/gpu/drm/amd/amdgpu/Makefile
drivers/gpu/drm/amd/amdgpu/vega20_ih.c [new file with mode: 0644]
drivers/gpu/drm/amd/amdgpu/vega20_ih.h [new file with mode: 0644]

index 6bf6cfa..c626268 100644 (file)
@@ -97,6 +97,7 @@ amdgpu-y += \
        tonga_ih.o \
        cz_ih.o \
        vega10_ih.o \
+       vega20_ih.o \
        navi10_ih.o
 
 # add PSP block
diff --git a/drivers/gpu/drm/amd/amdgpu/vega20_ih.c b/drivers/gpu/drm/amd/amdgpu/vega20_ih.c
new file mode 100644 (file)
index 0000000..c8f92d2
--- /dev/null
@@ -0,0 +1,661 @@
+/*
+ * Copyright 2020 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/pci.h>
+
+#include "amdgpu.h"
+#include "amdgpu_ih.h"
+#include "soc15.h"
+
+#include "oss/osssys_4_2_0_offset.h"
+#include "oss/osssys_4_2_0_sh_mask.h"
+
+#include "soc15_common.h"
+#include "vega20_ih.h"
+
+#define MAX_REARM_RETRY 10
+
+static void vega20_ih_set_interrupt_funcs(struct amdgpu_device *adev);
+
+/**
+ * vega20_ih_init_register_offset - Initialize register offset for ih rings
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Initialize register offset ih rings (VEGA20).
+ */
+static void vega20_ih_init_register_offset(struct amdgpu_device *adev)
+{
+       struct amdgpu_ih_regs *ih_regs;
+
+       if (adev->irq.ih.ring_size) {
+               ih_regs = &adev->irq.ih.ih_regs;
+               ih_regs->ih_rb_base = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_BASE);
+               ih_regs->ih_rb_base_hi = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_BASE_HI);
+               ih_regs->ih_rb_cntl = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL);
+               ih_regs->ih_rb_wptr = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_WPTR);
+               ih_regs->ih_rb_rptr = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_RPTR);
+               ih_regs->ih_doorbell_rptr = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_DOORBELL_RPTR);
+               ih_regs->ih_rb_wptr_addr_lo = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_WPTR_ADDR_LO);
+               ih_regs->ih_rb_wptr_addr_hi = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_WPTR_ADDR_HI);
+               ih_regs->psp_reg_id = PSP_REG_IH_RB_CNTL;
+       }
+
+       if (adev->irq.ih1.ring_size) {
+               ih_regs = &adev->irq.ih1.ih_regs;
+               ih_regs->ih_rb_base = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_BASE_RING1);
+               ih_regs->ih_rb_base_hi = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_BASE_HI_RING1);
+               ih_regs->ih_rb_cntl = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL_RING1);
+               ih_regs->ih_rb_wptr = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_WPTR_RING1);
+               ih_regs->ih_rb_rptr = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_RPTR_RING1);
+               ih_regs->ih_doorbell_rptr = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_DOORBELL_RPTR_RING1);
+               ih_regs->psp_reg_id = PSP_REG_IH_RB_CNTL_RING1;
+       }
+
+       if (adev->irq.ih2.ring_size) {
+               ih_regs = &adev->irq.ih2.ih_regs;
+               ih_regs->ih_rb_base = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_BASE_RING2);
+               ih_regs->ih_rb_base_hi = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_BASE_HI_RING2);
+               ih_regs->ih_rb_cntl = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL_RING2);
+               ih_regs->ih_rb_wptr = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_WPTR_RING2);
+               ih_regs->ih_rb_rptr = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_RPTR_RING2);
+               ih_regs->ih_doorbell_rptr = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_DOORBELL_RPTR_RING2);
+               ih_regs->psp_reg_id = PSP_REG_IH_RB_CNTL_RING2;
+       }
+}
+
+/**
+ * vega20_ih_toggle_ring_interrupts - toggle the interrupt ring buffer
+ *
+ * @adev: amdgpu_device pointer
+ * @ih: amdgpu_ih_ring pointet
+ * @enable: true - enable the interrupts, false - disable the interrupts
+ *
+ * Toggle the interrupt ring buffer (VEGA20)
+ */
+static int vega20_ih_toggle_ring_interrupts(struct amdgpu_device *adev,
+                                           struct amdgpu_ih_ring *ih,
+                                           bool enable)
+{
+       struct amdgpu_ih_regs *ih_regs;
+       uint32_t tmp;
+
+       ih_regs = &ih->ih_regs;
+
+       tmp = RREG32(ih_regs->ih_rb_cntl);
+       tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_ENABLE, (enable ? 1 : 0));
+       /* enable_intr field is only valid in ring0 */
+       if (ih == &adev->irq.ih)
+               tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, ENABLE_INTR, (enable ? 1 : 0));
+       if (amdgpu_sriov_vf(adev)) {
+               if (psp_reg_program(&adev->psp, ih_regs->psp_reg_id, tmp)) {
+                       dev_err(adev->dev, "PSP program IH_RB_CNTL failed!\n");
+                       return -ETIMEDOUT;
+               }
+       } else {
+               WREG32(ih_regs->ih_rb_cntl, tmp);
+       }
+
+       if (enable) {
+               ih->enabled = true;
+       } else {
+               /* set rptr, wptr to 0 */
+               WREG32(ih_regs->ih_rb_rptr, 0);
+               WREG32(ih_regs->ih_rb_wptr, 0);
+               ih->enabled = false;
+               ih->rptr = 0;
+       }
+
+       return 0;
+}
+
+/**
+ * vega20_ih_toggle_interrupts - Toggle all the available interrupt ring buffers
+ *
+ * @adev: amdgpu_device pointer
+ * @enable: enable or disable interrupt ring buffers
+ *
+ * Toggle all the available interrupt ring buffers (VEGA20).
+ */
+static int vega20_ih_toggle_interrupts(struct amdgpu_device *adev, bool enable)
+{
+       struct amdgpu_ih_ring *ih[] = {&adev->irq.ih, &adev->irq.ih1, &adev->irq.ih2};
+       int i;
+       int r;
+
+       for (i = 0; i < ARRAY_SIZE(ih); i++) {
+               if (ih[i]->ring_size) {
+                       r = vega20_ih_toggle_ring_interrupts(adev, ih[i], enable);
+                       if (r)
+                               return r;
+               }
+       }
+
+       return 0;
+}
+
+static uint32_t vega20_ih_rb_cntl(struct amdgpu_ih_ring *ih, uint32_t ih_rb_cntl)
+{
+       int rb_bufsz = order_base_2(ih->ring_size / 4);
+
+       ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+                                  MC_SPACE, ih->use_bus_addr ? 1 : 4);
+       ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+                                  WPTR_OVERFLOW_CLEAR, 1);
+       ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+                                  WPTR_OVERFLOW_ENABLE, 1);
+       ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, RB_SIZE, rb_bufsz);
+       /* Ring Buffer write pointer writeback. If enabled, IH_RB_WPTR register
+        * value is written to memory
+        */
+       ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+                                  WPTR_WRITEBACK_ENABLE, 1);
+       ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, MC_SNOOP, 1);
+       ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, MC_RO, 0);
+       ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, MC_VMID, 0);
+
+       return ih_rb_cntl;
+}
+
+static uint32_t vega20_ih_doorbell_rptr(struct amdgpu_ih_ring *ih)
+{
+       u32 ih_doorbell_rtpr = 0;
+
+       if (ih->use_doorbell) {
+               ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr,
+                                                IH_DOORBELL_RPTR, OFFSET,
+                                                ih->doorbell_index);
+               ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr,
+                                                IH_DOORBELL_RPTR,
+                                                ENABLE, 1);
+       } else {
+               ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr,
+                                                IH_DOORBELL_RPTR,
+                                                ENABLE, 0);
+       }
+       return ih_doorbell_rtpr;
+}
+
+/**
+ * vega20_ih_enable_ring - enable an ih ring buffer
+ *
+ * @adev: amdgpu_device pointer
+ * @ih: amdgpu_ih_ring pointer
+ *
+ * Enable an ih ring buffer (VEGA20)
+ */
+static int vega20_ih_enable_ring(struct amdgpu_device *adev,
+                                struct amdgpu_ih_ring *ih)
+{
+       struct amdgpu_ih_regs *ih_regs;
+       uint32_t tmp;
+
+       ih_regs = &ih->ih_regs;
+
+       /* Ring Buffer base. [39:8] of 40-bit address of the beginning of the ring buffer*/
+       WREG32(ih_regs->ih_rb_base, ih->gpu_addr >> 8);
+       WREG32(ih_regs->ih_rb_base_hi, (ih->gpu_addr >> 40) & 0xff);
+
+       tmp = RREG32(ih_regs->ih_rb_cntl);
+       tmp = vega20_ih_rb_cntl(ih, tmp);
+       if (ih == &adev->irq.ih)
+               tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RPTR_REARM, !!adev->irq.msi_enabled);
+       if (ih == &adev->irq.ih1) {
+               tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_ENABLE, 0);
+               tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_FULL_DRAIN_ENABLE, 1);
+       }
+       if (amdgpu_sriov_vf(adev)) {
+               if (psp_reg_program(&adev->psp, ih_regs->psp_reg_id, tmp)) {
+                       dev_err(adev->dev, "PSP program IH_RB_CNTL failed!\n");
+                       return -ETIMEDOUT;
+               }
+       } else {
+               WREG32(ih_regs->ih_rb_cntl, tmp);
+       }
+
+       if (ih == &adev->irq.ih) {
+               /* set the ih ring 0 writeback address whether it's enabled or not */
+               WREG32(ih_regs->ih_rb_wptr_addr_lo, lower_32_bits(ih->wptr_addr));
+               WREG32(ih_regs->ih_rb_wptr_addr_hi, upper_32_bits(ih->wptr_addr) & 0xFFFF);
+       }
+
+       /* set rptr, wptr to 0 */
+       WREG32(ih_regs->ih_rb_wptr, 0);
+       WREG32(ih_regs->ih_rb_rptr, 0);
+
+       WREG32(ih_regs->ih_doorbell_rptr, vega20_ih_doorbell_rptr(ih));
+
+       return 0;
+}
+
+/**
+ * vega20_ih_irq_init - init and enable the interrupt ring
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Allocate a ring buffer for the interrupt controller,
+ * enable the RLC, disable interrupts, enable the IH
+ * ring buffer and enable it (VI).
+ * Called at device load and reume.
+ * Returns 0 for success, errors for failure.
+ */
+static int vega20_ih_irq_init(struct amdgpu_device *adev)
+{
+       struct amdgpu_ih_ring *ih[] = {&adev->irq.ih, &adev->irq.ih1, &adev->irq.ih2};
+       u32 ih_chicken;
+       int ret;
+       int i;
+       u32 tmp;
+
+       /* disable irqs */
+       ret = vega20_ih_toggle_interrupts(adev, false);
+       if (ret)
+               return ret;
+
+       adev->nbio.funcs->ih_control(adev);
+
+       if (adev->asic_type == CHIP_ARCTURUS &&
+           adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
+               ih_chicken = RREG32_SOC15(OSSSYS, 0, mmIH_CHICKEN);
+               if (adev->irq.ih.use_bus_addr) {
+                       ih_chicken = REG_SET_FIELD(ih_chicken, IH_CHICKEN,
+                                                  MC_SPACE_GPA_ENABLE, 1);
+               } else {
+                       ih_chicken = REG_SET_FIELD(ih_chicken, IH_CHICKEN,
+                                                  MC_SPACE_FBPA_ENABLE, 1);
+               }
+               WREG32_SOC15(OSSSYS, 0, mmIH_CHICKEN, ih_chicken);
+       }
+
+       for (i = 0; i < ARRAY_SIZE(ih); i++) {
+               if (ih[i]->ring_size) {
+                       ret = vega20_ih_enable_ring(adev, ih[i]);
+                       if (ret)
+                               return ret;
+               }
+       }
+
+       tmp = RREG32_SOC15(OSSSYS, 0, mmIH_STORM_CLIENT_LIST_CNTL);
+       tmp = REG_SET_FIELD(tmp, IH_STORM_CLIENT_LIST_CNTL,
+                           CLIENT18_IS_STORM_CLIENT, 1);
+       WREG32_SOC15(OSSSYS, 0, mmIH_STORM_CLIENT_LIST_CNTL, tmp);
+
+       tmp = RREG32_SOC15(OSSSYS, 0, mmIH_INT_FLOOD_CNTL);
+       tmp = REG_SET_FIELD(tmp, IH_INT_FLOOD_CNTL, FLOOD_CNTL_ENABLE, 1);
+       WREG32_SOC15(OSSSYS, 0, mmIH_INT_FLOOD_CNTL, tmp);
+
+       pci_set_master(adev->pdev);
+
+       /* enable interrupts */
+       ret = vega20_ih_toggle_interrupts(adev, true);
+       if (ret)
+               return ret;
+
+       return 0;
+}
+
+/**
+ * vega20_ih_irq_disable - disable interrupts
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Disable interrupts on the hw (VEGA20).
+ */
+static void vega20_ih_irq_disable(struct amdgpu_device *adev)
+{
+       vega20_ih_toggle_interrupts(adev, false);
+
+       /* Wait and acknowledge irq */
+       mdelay(1);
+}
+
+/**
+ * vega20_ih_get_wptr - get the IH ring buffer wptr
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Get the IH ring buffer wptr from either the register
+ * or the writeback memory buffer (VEGA20).  Also check for
+ * ring buffer overflow and deal with it.
+ * Returns the value of the wptr.
+ */
+static u32 vega20_ih_get_wptr(struct amdgpu_device *adev,
+                             struct amdgpu_ih_ring *ih)
+{
+       u32 wptr, tmp;
+       struct amdgpu_ih_regs *ih_regs;
+
+       wptr = le32_to_cpu(*ih->wptr_cpu);
+       ih_regs = &ih->ih_regs;
+
+       if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW))
+               goto out;
+
+       /* Double check that the overflow wasn't already cleared. */
+       wptr = RREG32_NO_KIQ(ih_regs->ih_rb_wptr);
+       if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW))
+               goto out;
+
+       wptr = REG_SET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW, 0);
+
+       /* When a ring buffer overflow happen start parsing interrupt
+        * from the last not overwritten vector (wptr + 32). Hopefully
+        * this should allow us to catchup.
+        */
+       tmp = (wptr + 32) & ih->ptr_mask;
+       dev_warn(adev->dev, "IH ring buffer overflow "
+                "(0x%08X, 0x%08X, 0x%08X)\n",
+                wptr, ih->rptr, tmp);
+       ih->rptr = tmp;
+
+       tmp = RREG32_NO_KIQ(ih_regs->ih_rb_cntl);
+       tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1);
+       WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
+
+out:
+       return (wptr & ih->ptr_mask);
+}
+
+/**
+ * vega20_ih_irq_rearm - rearm IRQ if lost
+ *
+ * @adev: amdgpu_device pointer
+ *
+ */
+static void vega20_ih_irq_rearm(struct amdgpu_device *adev,
+                              struct amdgpu_ih_ring *ih)
+{
+       uint32_t v = 0;
+       uint32_t i = 0;
+       struct amdgpu_ih_regs *ih_regs;
+
+       ih_regs = &ih->ih_regs;
+
+       /* Rearm IRQ / re-wwrite doorbell if doorbell write is lost */
+       for (i = 0; i < MAX_REARM_RETRY; i++) {
+               v = RREG32_NO_KIQ(ih_regs->ih_rb_rptr);
+               if ((v < ih->ring_size) && (v != ih->rptr))
+                       WDOORBELL32(ih->doorbell_index, ih->rptr);
+               else
+                       break;
+       }
+}
+
+/**
+ * vega20_ih_set_rptr - set the IH ring buffer rptr
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Set the IH ring buffer rptr.
+ */
+static void vega20_ih_set_rptr(struct amdgpu_device *adev,
+                              struct amdgpu_ih_ring *ih)
+{
+       struct amdgpu_ih_regs *ih_regs;
+
+       if (ih->use_doorbell) {
+               /* XXX check if swapping is necessary on BE */
+               *ih->rptr_cpu = ih->rptr;
+               WDOORBELL32(ih->doorbell_index, ih->rptr);
+
+               if (amdgpu_sriov_vf(adev))
+                       vega20_ih_irq_rearm(adev, ih);
+       } else {
+               ih_regs = &ih->ih_regs;
+               WREG32(ih_regs->ih_rb_rptr, ih->rptr);
+       }
+}
+
+/**
+ * vega20_ih_self_irq - dispatch work for ring 1 and 2
+ *
+ * @adev: amdgpu_device pointer
+ * @source: irq source
+ * @entry: IV with WPTR update
+ *
+ * Update the WPTR from the IV and schedule work to handle the entries.
+ */
+static int vega20_ih_self_irq(struct amdgpu_device *adev,
+                             struct amdgpu_irq_src *source,
+                             struct amdgpu_iv_entry *entry)
+{
+       uint32_t wptr = cpu_to_le32(entry->src_data[0]);
+
+       switch (entry->ring_id) {
+       case 1:
+               *adev->irq.ih1.wptr_cpu = wptr;
+               schedule_work(&adev->irq.ih1_work);
+               break;
+       case 2:
+               *adev->irq.ih2.wptr_cpu = wptr;
+               schedule_work(&adev->irq.ih2_work);
+               break;
+       default: break;
+       }
+       return 0;
+}
+
+static const struct amdgpu_irq_src_funcs vega20_ih_self_irq_funcs = {
+       .process = vega20_ih_self_irq,
+};
+
+static void vega20_ih_set_self_irq_funcs(struct amdgpu_device *adev)
+{
+       adev->irq.self_irq.num_types = 0;
+       adev->irq.self_irq.funcs = &vega20_ih_self_irq_funcs;
+}
+
+static int vega20_ih_early_init(void *handle)
+{
+       struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+       vega20_ih_set_interrupt_funcs(adev);
+       vega20_ih_set_self_irq_funcs(adev);
+       return 0;
+}
+
+static int vega20_ih_sw_init(void *handle)
+{
+       struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+       int r;
+
+       r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_IH, 0,
+                             &adev->irq.self_irq);
+       if (r)
+               return r;
+
+       r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 256 * 1024, true);
+       if (r)
+               return r;
+
+       adev->irq.ih.use_doorbell = true;
+       adev->irq.ih.doorbell_index = adev->doorbell_index.ih << 1;
+
+       r = amdgpu_ih_ring_init(adev, &adev->irq.ih1, PAGE_SIZE, true);
+       if (r)
+               return r;
+
+       adev->irq.ih1.use_doorbell = true;
+       adev->irq.ih1.doorbell_index = (adev->doorbell_index.ih + 1) << 1;
+
+       r = amdgpu_ih_ring_init(adev, &adev->irq.ih2, PAGE_SIZE, true);
+       if (r)
+               return r;
+
+       adev->irq.ih2.use_doorbell = true;
+       adev->irq.ih2.doorbell_index = (adev->doorbell_index.ih + 2) << 1;
+
+       /* initialize ih control registers offset */
+       vega20_ih_init_register_offset(adev);
+
+       r = amdgpu_irq_init(adev);
+
+       return r;
+}
+
+static int vega20_ih_sw_fini(void *handle)
+{
+       struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+       amdgpu_irq_fini(adev);
+       amdgpu_ih_ring_fini(adev, &adev->irq.ih2);
+       amdgpu_ih_ring_fini(adev, &adev->irq.ih1);
+       amdgpu_ih_ring_fini(adev, &adev->irq.ih);
+
+       return 0;
+}
+
+static int vega20_ih_hw_init(void *handle)
+{
+       int r;
+       struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+       r = vega20_ih_irq_init(adev);
+       if (r)
+               return r;
+
+       return 0;
+}
+
+static int vega20_ih_hw_fini(void *handle)
+{
+       struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+       vega20_ih_irq_disable(adev);
+
+       return 0;
+}
+
+static int vega20_ih_suspend(void *handle)
+{
+       struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+       return vega20_ih_hw_fini(adev);
+}
+
+static int vega20_ih_resume(void *handle)
+{
+       struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+       return vega20_ih_hw_init(adev);
+}
+
+static bool vega20_ih_is_idle(void *handle)
+{
+       /* todo */
+       return true;
+}
+
+static int vega20_ih_wait_for_idle(void *handle)
+{
+       /* todo */
+       return -ETIMEDOUT;
+}
+
+static int vega20_ih_soft_reset(void *handle)
+{
+       /* todo */
+
+       return 0;
+}
+
+static void vega20_ih_update_clockgating_state(struct amdgpu_device *adev,
+                                              bool enable)
+{
+       uint32_t data, def, field_val;
+
+       if (adev->cg_flags & AMD_CG_SUPPORT_IH_CG) {
+               def = data = RREG32_SOC15(OSSSYS, 0, mmIH_CLK_CTRL);
+               field_val = enable ? 0 : 1;
+               data = REG_SET_FIELD(data, IH_CLK_CTRL,
+                                    IH_RETRY_INT_CAM_MEM_CLK_SOFT_OVERRIDE, field_val);
+               data = REG_SET_FIELD(data, IH_CLK_CTRL,
+                                    IH_BUFFER_MEM_CLK_SOFT_OVERRIDE, field_val);
+               data = REG_SET_FIELD(data, IH_CLK_CTRL,
+                                    DBUS_MUX_CLK_SOFT_OVERRIDE, field_val);
+               data = REG_SET_FIELD(data, IH_CLK_CTRL,
+                                    OSSSYS_SHARE_CLK_SOFT_OVERRIDE, field_val);
+               data = REG_SET_FIELD(data, IH_CLK_CTRL,
+                                    LIMIT_SMN_CLK_SOFT_OVERRIDE, field_val);
+               data = REG_SET_FIELD(data, IH_CLK_CTRL,
+                                    DYN_CLK_SOFT_OVERRIDE, field_val);
+               data = REG_SET_FIELD(data, IH_CLK_CTRL,
+                                    REG_CLK_SOFT_OVERRIDE, field_val);
+               if (def != data)
+                       WREG32_SOC15(OSSSYS, 0, mmIH_CLK_CTRL, data);
+       }
+}
+
+static int vega20_ih_set_clockgating_state(void *handle,
+                                         enum amd_clockgating_state state)
+{
+       struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+       vega20_ih_update_clockgating_state(adev,
+                               state == AMD_CG_STATE_GATE);
+       return 0;
+
+}
+
+static int vega20_ih_set_powergating_state(void *handle,
+                                         enum amd_powergating_state state)
+{
+       return 0;
+}
+
+const struct amd_ip_funcs vega20_ih_ip_funcs = {
+       .name = "vega20_ih",
+       .early_init = vega20_ih_early_init,
+       .late_init = NULL,
+       .sw_init = vega20_ih_sw_init,
+       .sw_fini = vega20_ih_sw_fini,
+       .hw_init = vega20_ih_hw_init,
+       .hw_fini = vega20_ih_hw_fini,
+       .suspend = vega20_ih_suspend,
+       .resume = vega20_ih_resume,
+       .is_idle = vega20_ih_is_idle,
+       .wait_for_idle = vega20_ih_wait_for_idle,
+       .soft_reset = vega20_ih_soft_reset,
+       .set_clockgating_state = vega20_ih_set_clockgating_state,
+       .set_powergating_state = vega20_ih_set_powergating_state,
+};
+
+static const struct amdgpu_ih_funcs vega20_ih_funcs = {
+       .get_wptr = vega20_ih_get_wptr,
+       .decode_iv = amdgpu_ih_decode_iv_helper,
+       .set_rptr = vega20_ih_set_rptr
+};
+
+static void vega20_ih_set_interrupt_funcs(struct amdgpu_device *adev)
+{
+       adev->irq.ih_funcs = &vega20_ih_funcs;
+}
+
+const struct amdgpu_ip_block_version vega20_ih_ip_block =
+{
+       .type = AMD_IP_BLOCK_TYPE_IH,
+       .major = 4,
+       .minor = 2,
+       .rev = 0,
+       .funcs = &vega20_ih_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/vega20_ih.h b/drivers/gpu/drm/amd/amdgpu/vega20_ih.h
new file mode 100644 (file)
index 0000000..7af6d87
--- /dev/null
@@ -0,0 +1,30 @@
+/*
+ * Copyright 2020 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __VEGA20_IH_H__
+#define __VEGA20_IH_H__
+
+extern const struct amd_ip_funcs vega20_ih_ip_funcs;
+extern const struct amdgpu_ip_block_version vega20_ih_ip_block;
+
+#endif