From d2ccde1c8e90d38cee87a1b0309ad2827f3fd30d Mon Sep 17 00:00:00 2001
From: hc <hc@nodka.com>
Date: Mon, 11 Dec 2023 02:45:28 +0000
Subject: [PATCH] add boot partition size
---
kernel/drivers/gpu/drm/amd/amdgpu/vega10_ih.c | 618 +++++++++++++++++++++++++++++++++++++++----------------
1 files changed, 433 insertions(+), 185 deletions(-)
diff --git a/kernel/drivers/gpu/drm/amd/amdgpu/vega10_ih.c b/kernel/drivers/gpu/drm/amd/amdgpu/vega10_ih.c
index 21bc12e..407c609 100644
--- a/kernel/drivers/gpu/drm/amd/amdgpu/vega10_ih.c
+++ b/kernel/drivers/gpu/drm/amd/amdgpu/vega10_ih.c
@@ -20,7 +20,9 @@
* OTHER DEALINGS IN THE SOFTWARE.
*
*/
-#include <drm/drmP.h>
+
+#include <linux/pci.h>
+
#include "amdgpu.h"
#include "amdgpu_ih.h"
#include "soc15.h"
@@ -31,7 +33,7 @@
#include "soc15_common.h"
#include "vega10_ih.h"
-
+#define MAX_REARM_RETRY 10
static void vega10_ih_set_interrupt_funcs(struct amdgpu_device *adev);
@@ -48,8 +50,47 @@
ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, RB_ENABLE, 1);
ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, ENABLE_INTR, 1);
- WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL, ih_rb_cntl);
+ if (amdgpu_sriov_vf(adev)) {
+ if (psp_reg_program(&adev->psp, PSP_REG_IH_RB_CNTL, ih_rb_cntl)) {
+ DRM_ERROR("PSP program IH_RB_CNTL failed!\n");
+ return;
+ }
+ } else {
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL, ih_rb_cntl);
+ }
adev->irq.ih.enabled = true;
+
+ if (adev->irq.ih1.ring_size) {
+ ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING1);
+ ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL_RING1,
+ RB_ENABLE, 1);
+ if (amdgpu_sriov_vf(adev)) {
+ if (psp_reg_program(&adev->psp, PSP_REG_IH_RB_CNTL_RING1,
+ ih_rb_cntl)) {
+ DRM_ERROR("program IH_RB_CNTL_RING1 failed!\n");
+ return;
+ }
+ } else {
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING1, ih_rb_cntl);
+ }
+ adev->irq.ih1.enabled = true;
+ }
+
+ if (adev->irq.ih2.ring_size) {
+ ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING2);
+ ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL_RING2,
+ RB_ENABLE, 1);
+ if (amdgpu_sriov_vf(adev)) {
+ if (psp_reg_program(&adev->psp, PSP_REG_IH_RB_CNTL_RING2,
+ ih_rb_cntl)) {
+ DRM_ERROR("program IH_RB_CNTL_RING2 failed!\n");
+ return;
+ }
+ } else {
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING2, ih_rb_cntl);
+ }
+ adev->irq.ih2.enabled = true;
+ }
}
/**
@@ -65,12 +106,103 @@
ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, RB_ENABLE, 0);
ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, ENABLE_INTR, 0);
- WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL, ih_rb_cntl);
+ if (amdgpu_sriov_vf(adev)) {
+ if (psp_reg_program(&adev->psp, PSP_REG_IH_RB_CNTL, ih_rb_cntl)) {
+ DRM_ERROR("PSP program IH_RB_CNTL failed!\n");
+ return;
+ }
+ } else {
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL, ih_rb_cntl);
+ }
+
/* set rptr, wptr to 0 */
WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR, 0);
WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR, 0);
adev->irq.ih.enabled = false;
adev->irq.ih.rptr = 0;
+
+ if (adev->irq.ih1.ring_size) {
+ ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING1);
+ ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL_RING1,
+ RB_ENABLE, 0);
+ if (amdgpu_sriov_vf(adev)) {
+ if (psp_reg_program(&adev->psp, PSP_REG_IH_RB_CNTL_RING1,
+ ih_rb_cntl)) {
+ DRM_ERROR("program IH_RB_CNTL_RING1 failed!\n");
+ return;
+ }
+ } else {
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING1, ih_rb_cntl);
+ }
+ /* set rptr, wptr to 0 */
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR_RING1, 0);
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR_RING1, 0);
+ adev->irq.ih1.enabled = false;
+ adev->irq.ih1.rptr = 0;
+ }
+
+ if (adev->irq.ih2.ring_size) {
+ ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING2);
+ ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL_RING2,
+ RB_ENABLE, 0);
+ if (amdgpu_sriov_vf(adev)) {
+ if (psp_reg_program(&adev->psp, PSP_REG_IH_RB_CNTL_RING2,
+ ih_rb_cntl)) {
+ DRM_ERROR("program IH_RB_CNTL_RING2 failed!\n");
+ return;
+ }
+ } else {
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING2, ih_rb_cntl);
+ }
+
+ /* set rptr, wptr to 0 */
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR_RING2, 0);
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR_RING2, 0);
+ adev->irq.ih2.enabled = false;
+ adev->irq.ih2.rptr = 0;
+ }
+}
+
+static uint32_t vega10_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 vega10_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;
}
/**
@@ -86,68 +218,117 @@
*/
static int vega10_ih_irq_init(struct amdgpu_device *adev)
{
+ struct amdgpu_ih_ring *ih;
+ u32 ih_rb_cntl, ih_chicken;
int ret = 0;
- int rb_bufsz;
- u32 ih_rb_cntl, ih_doorbell_rtpr;
u32 tmp;
- u64 wptr_off;
/* disable irqs */
vega10_ih_disable_interrupts(adev);
- adev->nbio_funcs->ih_control(adev);
+ adev->nbio.funcs->ih_control(adev);
+
+ ih = &adev->irq.ih;
+ /* Ring Buffer base. [39:8] of 40-bit address of the beginning of the ring buffer*/
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_BASE, ih->gpu_addr >> 8);
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_BASE_HI, (ih->gpu_addr >> 40) & 0xff);
ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL);
- /* Ring Buffer base. [39:8] of 40-bit address of the beginning of the ring buffer*/
- if (adev->irq.ih.use_bus_addr) {
- WREG32_SOC15(OSSSYS, 0, mmIH_RB_BASE, adev->irq.ih.rb_dma_addr >> 8);
- WREG32_SOC15(OSSSYS, 0, mmIH_RB_BASE_HI, ((u64)adev->irq.ih.rb_dma_addr >> 40) & 0xff);
- ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, MC_SPACE, 1);
+ ih_rb_cntl = vega10_ih_rb_cntl(ih, ih_rb_cntl);
+ ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, RPTR_REARM,
+ !!adev->irq.msi_enabled);
+ if (amdgpu_sriov_vf(adev)) {
+ if (psp_reg_program(&adev->psp, PSP_REG_IH_RB_CNTL, ih_rb_cntl)) {
+ DRM_ERROR("PSP program IH_RB_CNTL failed!\n");
+ return -ETIMEDOUT;
+ }
} else {
- WREG32_SOC15(OSSSYS, 0, mmIH_RB_BASE, adev->irq.ih.gpu_addr >> 8);
- WREG32_SOC15(OSSSYS, 0, mmIH_RB_BASE_HI, (adev->irq.ih.gpu_addr >> 40) & 0xff);
- ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, MC_SPACE, 4);
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL, ih_rb_cntl);
}
- rb_bufsz = order_base_2(adev->irq.ih.ring_size / 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);
- if (adev->irq.msi_enabled)
- ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, RPTR_REARM, 1);
-
- WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL, ih_rb_cntl);
+ if ((adev->asic_type == CHIP_ARCTURUS &&
+ adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) ||
+ adev->asic_type == CHIP_RENOIR) {
+ 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);
+ }
/* set the writeback address whether it's enabled or not */
- if (adev->irq.ih.use_bus_addr)
- wptr_off = adev->irq.ih.rb_dma_addr + (adev->irq.ih.wptr_offs * 4);
- else
- wptr_off = adev->wb.gpu_addr + (adev->irq.ih.wptr_offs * 4);
- WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR_ADDR_LO, lower_32_bits(wptr_off));
- WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR_ADDR_HI, upper_32_bits(wptr_off) & 0xFFFF);
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR_ADDR_LO,
+ lower_32_bits(ih->wptr_addr));
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR_ADDR_HI,
+ upper_32_bits(ih->wptr_addr) & 0xFFFF);
/* set rptr, wptr to 0 */
- WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR, 0);
WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR, 0);
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR, 0);
- ih_doorbell_rtpr = RREG32_SOC15(OSSSYS, 0, mmIH_DOORBELL_RPTR);
- if (adev->irq.ih.use_doorbell) {
- ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr, IH_DOORBELL_RPTR,
- OFFSET, adev->irq.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);
+ WREG32_SOC15(OSSSYS, 0, mmIH_DOORBELL_RPTR,
+ vega10_ih_doorbell_rptr(ih));
+
+ ih = &adev->irq.ih1;
+ if (ih->ring_size) {
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_BASE_RING1, ih->gpu_addr >> 8);
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_BASE_HI_RING1,
+ (ih->gpu_addr >> 40) & 0xff);
+
+ ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING1);
+ ih_rb_cntl = vega10_ih_rb_cntl(ih, ih_rb_cntl);
+ ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+ WPTR_OVERFLOW_ENABLE, 0);
+ ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+ RB_FULL_DRAIN_ENABLE, 1);
+ if (amdgpu_sriov_vf(adev)) {
+ if (psp_reg_program(&adev->psp, PSP_REG_IH_RB_CNTL_RING1,
+ ih_rb_cntl)) {
+ DRM_ERROR("program IH_RB_CNTL_RING1 failed!\n");
+ return -ETIMEDOUT;
+ }
+ } else {
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING1, ih_rb_cntl);
+ }
+
+ /* set rptr, wptr to 0 */
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR_RING1, 0);
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR_RING1, 0);
+
+ WREG32_SOC15(OSSSYS, 0, mmIH_DOORBELL_RPTR_RING1,
+ vega10_ih_doorbell_rptr(ih));
}
- WREG32_SOC15(OSSSYS, 0, mmIH_DOORBELL_RPTR, ih_doorbell_rtpr);
- adev->nbio_funcs->ih_doorbell_range(adev, adev->irq.ih.use_doorbell,
- adev->irq.ih.doorbell_index);
+
+ ih = &adev->irq.ih2;
+ if (ih->ring_size) {
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_BASE_RING2, ih->gpu_addr >> 8);
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_BASE_HI_RING2,
+ (ih->gpu_addr >> 40) & 0xff);
+
+ ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING2);
+ ih_rb_cntl = vega10_ih_rb_cntl(ih, ih_rb_cntl);
+
+ if (amdgpu_sriov_vf(adev)) {
+ if (psp_reg_program(&adev->psp, PSP_REG_IH_RB_CNTL_RING2,
+ ih_rb_cntl)) {
+ DRM_ERROR("program IH_RB_CNTL_RING2 failed!\n");
+ return -ETIMEDOUT;
+ }
+ } else {
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING2, ih_rb_cntl);
+ }
+
+ /* set rptr, wptr to 0 */
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR_RING2, 0);
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR_RING2, 0);
+
+ WREG32_SOC15(OSSSYS, 0, mmIH_DOORBELL_RPTR_RING2,
+ vega10_ih_doorbell_rptr(ih));
+ }
tmp = RREG32_SOC15(OSSSYS, 0, mmIH_STORM_CLIENT_LIST_CNTL);
tmp = REG_SET_FIELD(tmp, IH_STORM_CLIENT_LIST_CNTL,
@@ -191,115 +372,58 @@
* ring buffer overflow and deal with it.
* Returns the value of the wptr.
*/
-static u32 vega10_ih_get_wptr(struct amdgpu_device *adev)
+static u32 vega10_ih_get_wptr(struct amdgpu_device *adev,
+ struct amdgpu_ih_ring *ih)
{
- u32 wptr, tmp;
+ u32 wptr, reg, tmp;
- if (adev->irq.ih.use_bus_addr)
- wptr = le32_to_cpu(adev->irq.ih.ring[adev->irq.ih.wptr_offs]);
+ wptr = le32_to_cpu(*ih->wptr_cpu);
+
+ if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW))
+ goto out;
+
+ /* Double check that the overflow wasn't already cleared. */
+
+ if (ih == &adev->irq.ih)
+ reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_WPTR);
+ else if (ih == &adev->irq.ih1)
+ reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_WPTR_RING1);
+ else if (ih == &adev->irq.ih2)
+ reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_WPTR_RING2);
else
- wptr = le32_to_cpu(adev->wb.wb[adev->irq.ih.wptr_offs]);
+ BUG();
- if (REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW)) {
- wptr = REG_SET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW, 0);
+ wptr = RREG32_NO_KIQ(reg);
+ if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW))
+ goto out;
- /* 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) & adev->irq.ih.ptr_mask;
- dev_warn(adev->dev, "IH ring buffer overflow (0x%08X, 0x%08X, 0x%08X)\n",
- wptr, adev->irq.ih.rptr, tmp);
- adev->irq.ih.rptr = tmp;
+ wptr = REG_SET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW, 0);
- tmp = RREG32_NO_KIQ(SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL));
- tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1);
- WREG32_NO_KIQ(SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL), tmp);
- }
- return (wptr & adev->irq.ih.ptr_mask);
-}
-
-/**
- * vega10_ih_prescreen_iv - prescreen an interrupt vector
- *
- * @adev: amdgpu_device pointer
- *
- * Returns true if the interrupt vector should be further processed.
- */
-static bool vega10_ih_prescreen_iv(struct amdgpu_device *adev)
-{
- u32 ring_index = adev->irq.ih.rptr >> 2;
- u32 dw0, dw3, dw4, dw5;
- u16 pasid;
- u64 addr, key;
- struct amdgpu_vm *vm;
- int r;
-
- dw0 = le32_to_cpu(adev->irq.ih.ring[ring_index + 0]);
- dw3 = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]);
- dw4 = le32_to_cpu(adev->irq.ih.ring[ring_index + 4]);
- dw5 = le32_to_cpu(adev->irq.ih.ring[ring_index + 5]);
-
- /* Filter retry page faults, let only the first one pass. If
- * there are too many outstanding faults, ignore them until
- * some faults get cleared.
+ /* When a ring buffer overflow happen start parsing interrupt
+ * from the last not overwritten vector (wptr + 32). Hopefully
+ * this should allow us to catchup.
*/
- switch (dw0 & 0xff) {
- case SOC15_IH_CLIENTID_VMC:
- case SOC15_IH_CLIENTID_UTCL2:
- break;
- default:
- /* Not a VM fault */
- return true;
- }
+ 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;
- pasid = dw3 & 0xffff;
- /* No PASID, can't identify faulting process */
- if (!pasid)
- return true;
+ if (ih == &adev->irq.ih)
+ reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL);
+ else if (ih == &adev->irq.ih1)
+ reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL_RING1);
+ else if (ih == &adev->irq.ih2)
+ reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL_RING2);
+ else
+ BUG();
- /* Not a retry fault, check fault credit */
- if (!(dw5 & 0x80)) {
- if (!amdgpu_vm_pasid_fault_credit(adev, pasid))
- goto ignore_iv;
- return true;
- }
+ tmp = RREG32_NO_KIQ(reg);
+ tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1);
+ WREG32_NO_KIQ(reg, tmp);
- addr = ((u64)(dw5 & 0xf) << 44) | ((u64)dw4 << 12);
- key = AMDGPU_VM_FAULT(pasid, addr);
- r = amdgpu_ih_add_fault(adev, key);
-
- /* Hash table is full or the fault is already being processed,
- * ignore further page faults
- */
- if (r != 0)
- goto ignore_iv;
-
- /* Track retry faults in per-VM fault FIFO. */
- spin_lock(&adev->vm_manager.pasid_lock);
- vm = idr_find(&adev->vm_manager.pasid_idr, pasid);
- if (!vm) {
- /* VM not found, process it normally */
- spin_unlock(&adev->vm_manager.pasid_lock);
- amdgpu_ih_clear_fault(adev, key);
- return true;
- }
- /* No locking required with single writer and single reader */
- r = kfifo_put(&vm->faults, key);
- if (!r) {
- /* FIFO is full. Ignore it until there is space */
- spin_unlock(&adev->vm_manager.pasid_lock);
- amdgpu_ih_clear_fault(adev, key);
- goto ignore_iv;
- }
- spin_unlock(&adev->vm_manager.pasid_lock);
-
- /* It's the first fault for this address, process it normally */
- return true;
-
-ignore_iv:
- adev->irq.ih.rptr += 32;
- return false;
+out:
+ return (wptr & ih->ptr_mask);
}
/**
@@ -311,20 +435,21 @@
* position and also advance the position.
*/
static void vega10_ih_decode_iv(struct amdgpu_device *adev,
- struct amdgpu_iv_entry *entry)
+ struct amdgpu_ih_ring *ih,
+ struct amdgpu_iv_entry *entry)
{
/* wptr/rptr are in bytes! */
- u32 ring_index = adev->irq.ih.rptr >> 2;
+ u32 ring_index = ih->rptr >> 2;
uint32_t dw[8];
- dw[0] = le32_to_cpu(adev->irq.ih.ring[ring_index + 0]);
- dw[1] = le32_to_cpu(adev->irq.ih.ring[ring_index + 1]);
- dw[2] = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]);
- dw[3] = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]);
- dw[4] = le32_to_cpu(adev->irq.ih.ring[ring_index + 4]);
- dw[5] = le32_to_cpu(adev->irq.ih.ring[ring_index + 5]);
- dw[6] = le32_to_cpu(adev->irq.ih.ring[ring_index + 6]);
- dw[7] = le32_to_cpu(adev->irq.ih.ring[ring_index + 7]);
+ dw[0] = le32_to_cpu(ih->ring[ring_index + 0]);
+ dw[1] = le32_to_cpu(ih->ring[ring_index + 1]);
+ dw[2] = le32_to_cpu(ih->ring[ring_index + 2]);
+ dw[3] = le32_to_cpu(ih->ring[ring_index + 3]);
+ dw[4] = le32_to_cpu(ih->ring[ring_index + 4]);
+ dw[5] = le32_to_cpu(ih->ring[ring_index + 5]);
+ dw[6] = le32_to_cpu(ih->ring[ring_index + 6]);
+ dw[7] = le32_to_cpu(ih->ring[ring_index + 7]);
entry->client_id = dw[0] & 0xff;
entry->src_id = (dw[0] >> 8) & 0xff;
@@ -340,9 +465,40 @@
entry->src_data[2] = dw[6];
entry->src_data[3] = dw[7];
-
/* wptr/rptr are in bytes! */
- adev->irq.ih.rptr += 32;
+ ih->rptr += 32;
+}
+
+/**
+ * vega10_ih_irq_rearm - rearm IRQ if lost
+ *
+ * @adev: amdgpu_device pointer
+ *
+ */
+static void vega10_ih_irq_rearm(struct amdgpu_device *adev,
+ struct amdgpu_ih_ring *ih)
+{
+ uint32_t reg_rptr = 0;
+ uint32_t v = 0;
+ uint32_t i = 0;
+
+ if (ih == &adev->irq.ih)
+ reg_rptr = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_RPTR);
+ else if (ih == &adev->irq.ih1)
+ reg_rptr = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_RPTR_RING1);
+ else if (ih == &adev->irq.ih2)
+ reg_rptr = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_RPTR_RING2);
+ else
+ return;
+
+ /* Rearm IRQ / re-wwrite doorbell if doorbell write is lost */
+ for (i = 0; i < MAX_REARM_RETRY; i++) {
+ v = RREG32_NO_KIQ(reg_rptr);
+ if ((v < ih->ring_size) && (v != ih->rptr))
+ WDOORBELL32(ih->doorbell_index, ih->rptr);
+ else
+ break;
+ }
}
/**
@@ -352,18 +508,62 @@
*
* Set the IH ring buffer rptr.
*/
-static void vega10_ih_set_rptr(struct amdgpu_device *adev)
+static void vega10_ih_set_rptr(struct amdgpu_device *adev,
+ struct amdgpu_ih_ring *ih)
{
- if (adev->irq.ih.use_doorbell) {
+ if (ih->use_doorbell) {
/* XXX check if swapping is necessary on BE */
- if (adev->irq.ih.use_bus_addr)
- adev->irq.ih.ring[adev->irq.ih.rptr_offs] = adev->irq.ih.rptr;
- else
- adev->wb.wb[adev->irq.ih.rptr_offs] = adev->irq.ih.rptr;
- WDOORBELL32(adev->irq.ih.doorbell_index, adev->irq.ih.rptr);
- } else {
- WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR, adev->irq.ih.rptr);
+ *ih->rptr_cpu = ih->rptr;
+ WDOORBELL32(ih->doorbell_index, ih->rptr);
+
+ if (amdgpu_sriov_vf(adev))
+ vega10_ih_irq_rearm(adev, ih);
+ } else if (ih == &adev->irq.ih) {
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR, ih->rptr);
+ } else if (ih == &adev->irq.ih1) {
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR_RING1, ih->rptr);
+ } else if (ih == &adev->irq.ih2) {
+ WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR_RING2, ih->rptr);
}
+}
+
+/**
+ * vega10_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 vega10_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 vega10_ih_self_irq_funcs = {
+ .process = vega10_ih_self_irq,
+};
+
+static void vega10_ih_set_self_irq_funcs(struct amdgpu_device *adev)
+{
+ adev->irq.self_irq.num_types = 0;
+ adev->irq.self_irq.funcs = &vega10_ih_self_irq_funcs;
}
static int vega10_ih_early_init(void *handle)
@@ -371,28 +571,40 @@
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
vega10_ih_set_interrupt_funcs(adev);
+ vega10_ih_set_self_irq_funcs(adev);
return 0;
}
static int vega10_ih_sw_init(void *handle)
{
- int r;
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+ int r;
- r = amdgpu_ih_ring_init(adev, 256 * 1024, true);
+ 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 = AMDGPU_DOORBELL64_IH << 1;
+ adev->irq.ih.doorbell_index = adev->doorbell_index.ih << 1;
- adev->irq.ih.faults = kmalloc(sizeof(*adev->irq.ih.faults), GFP_KERNEL);
- if (!adev->irq.ih.faults)
- return -ENOMEM;
- INIT_CHASH_TABLE(adev->irq.ih.faults->hash,
- AMDGPU_PAGEFAULT_HASH_BITS, 8, 0);
- spin_lock_init(&adev->irq.ih.faults->lock);
- adev->irq.ih.faults->count = 0;
+ 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;
r = amdgpu_irq_init(adev);
@@ -404,10 +616,9 @@
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
amdgpu_irq_fini(adev);
- amdgpu_ih_ring_fini(adev);
-
- kfree(adev->irq.ih.faults);
- adev->irq.ih.faults = NULL;
+ 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;
}
@@ -466,10 +677,49 @@
return 0;
}
+static void vega10_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;
+ /**
+ * Vega10 does not have IH_RETRY_INT_CAM_MEM_CLK_SOFT_OVERRIDE
+ * and IH_BUFFER_MEM_CLK_SOFT_OVERRIDE field.
+ */
+ if (adev->asic_type > CHIP_VEGA10) {
+ 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 vega10_ih_set_clockgating_state(void *handle,
enum amd_clockgating_state state)
{
+ struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+ vega10_ih_update_clockgating_state(adev,
+ state == AMD_CG_STATE_GATE);
return 0;
+
}
static int vega10_ih_set_powergating_state(void *handle,
@@ -497,15 +747,13 @@
static const struct amdgpu_ih_funcs vega10_ih_funcs = {
.get_wptr = vega10_ih_get_wptr,
- .prescreen_iv = vega10_ih_prescreen_iv,
.decode_iv = vega10_ih_decode_iv,
.set_rptr = vega10_ih_set_rptr
};
static void vega10_ih_set_interrupt_funcs(struct amdgpu_device *adev)
{
- if (adev->irq.ih_funcs == NULL)
- adev->irq.ih_funcs = &vega10_ih_funcs;
+ adev->irq.ih_funcs = &vega10_ih_funcs;
}
const struct amdgpu_ip_block_version vega10_ih_ip_block =
--
Gitblit v1.6.2