accel/ivpu: Use cached buffers for FW loading

Create buffers with cache coherency on the CPU side (write-back) while
disabling snooping on the VPU side. These buffers require an explicit
cache flush after each CPU-side modification.

Configuring pages as write-combined may introduce significant delays,
potentially taking hundreds of milliseconds for 64 MB buffers.

Added internal DRM_IVPU_BO_NOSNOOP mask which disables snooping on the
VPU side. Allocate FW runtime memory buffer (64 MB) as cached with
snooping-disabled.

This fixes random long FW loading times and boot params memory
corruption on warmboot (due to missed wmb).

Fixes: 02d5b0aacd05 ("accel/ivpu: Implement firmware parsing and booting")
Signed-off-by: Karol Wachowski <karol.wachowski@linux.intel.com>
Reviewed-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com>
Reviewed-by: Jeffrey Hugo <quic_jhugo@quicinc.com>
Signed-off-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230926120943.GD846747@linux.intel.com

diff --git a/drivers/accel/ivpu/ivpu_fw.c b/drivers/accel/ivpu/ivpu_fw.c
index 9827ea4..0191cf8 100644 (file)
--- a/drivers/accel/ivpu/ivpu_fw.c
+++ b/drivers/accel/ivpu/ivpu_fw.c
@@ -220,7 +220,8 @@ static int ivpu_fw_mem_init(struct ivpu_device *vdev)
        if (ret)
                return ret;
 
-       fw->mem = ivpu_bo_alloc_internal(vdev, fw->runtime_addr, fw->runtime_size, DRM_IVPU_BO_WC);
+       fw->mem = ivpu_bo_alloc_internal(vdev, fw->runtime_addr, fw->runtime_size,
+                                        DRM_IVPU_BO_CACHED | DRM_IVPU_BO_NOSNOOP);
        if (!fw->mem) {
                ivpu_err(vdev, "Failed to allocate firmware runtime memory\n");
                return -ENOMEM;
@@ -330,7 +331,7 @@ int ivpu_fw_load(struct ivpu_device *vdev)
                memset(start, 0, size);
        }
 
-       wmb(); /* Flush WC buffers after writing fw->mem */
+       clflush_cache_range(fw->mem->kvaddr, fw->mem->base.size);
 
        return 0;
 }
@@ -432,6 +433,7 @@ void ivpu_fw_boot_params_setup(struct ivpu_device *vdev, struct vpu_boot_params
        if (!ivpu_fw_is_cold_boot(vdev)) {
                boot_params->save_restore_ret_address = 0;
                vdev->pm->is_warmboot = true;
+               clflush_cache_range(vdev->fw->mem->kvaddr, SZ_4K);
                return;
        }
 
@@ -493,7 +495,7 @@ void ivpu_fw_boot_params_setup(struct ivpu_device *vdev, struct vpu_boot_params
        boot_params->punit_telemetry_sram_size = ivpu_hw_reg_telemetry_size_get(vdev);
        boot_params->vpu_telemetry_enable = ivpu_hw_reg_telemetry_enable_get(vdev);
 
-       wmb(); /* Flush WC buffers after writing bootparams */
+       clflush_cache_range(vdev->fw->mem->kvaddr, SZ_4K);
 
        ivpu_fw_boot_params_print(vdev, boot_params);
 }

diff --git a/drivers/accel/ivpu/ivpu_gem.h b/drivers/accel/ivpu/ivpu_gem.h
index 6b0ceda..f413058 100644 (file)
--- a/drivers/accel/ivpu/ivpu_gem.h
+++ b/drivers/accel/ivpu/ivpu_gem.h
@@ -8,6 +8,8 @@
 #include <drm/drm_gem.h>
 #include <drm/drm_mm.h>
 
+#define DRM_IVPU_BO_NOSNOOP       0x10000000
+
 struct dma_buf;
 struct ivpu_bo_ops;
 struct ivpu_file_priv;
@@ -83,6 +85,9 @@ static inline u32 ivpu_bo_cache_mode(struct ivpu_bo *bo)
 
 static inline bool ivpu_bo_is_snooped(struct ivpu_bo *bo)
 {
+       if (bo->flags & DRM_IVPU_BO_NOSNOOP)
+               return false;
+
        return ivpu_bo_cache_mode(bo) == DRM_IVPU_BO_CACHED;
 }