[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <3b578b1b-53d5-4f2d-a2b8-8483a4051a24@linaro.org>
Date: Thu, 5 Sep 2024 13:34:48 +0100
From: Bryan O'Donoghue <bryan.odonoghue@...aro.org>
To: quic_dikshita@...cinc.com, Vikash Garodia <quic_vgarodia@...cinc.com>,
Abhinav Kumar <quic_abhinavk@...cinc.com>,
Mauro Carvalho Chehab <mchehab@...nel.org>, Rob Herring <robh@...nel.org>,
Krzysztof Kozlowski <krzk+dt@...nel.org>, Conor Dooley
<conor+dt@...nel.org>, Philipp Zabel <p.zabel@...gutronix.de>
Cc: linux-media@...r.kernel.org, linux-arm-msm@...r.kernel.org,
devicetree@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v3 08/29] media: iris: implement boot sequence of the
firmware
On 27/08/2024 11:05, Dikshita Agarwal via B4 Relay wrote:
> From: Dikshita Agarwal <quic_dikshita@...cinc.com>
>
> Set memory region to firmware and implement boot sequence.
>
> Signed-off-by: Dikshita Agarwal <quic_dikshita@...cinc.com>
> ---
> drivers/media/platform/qcom/iris/Makefile | 1 +
> drivers/media/platform/qcom/iris/iris_core.c | 7 ++
> .../platform/qcom/iris/iris_platform_common.h | 1 +
> .../platform/qcom/iris/iris_platform_sm8550.c | 3 +
> drivers/media/platform/qcom/iris/iris_vpu_common.c | 87 ++++++++++++++++++++++
> drivers/media/platform/qcom/iris/iris_vpu_common.h | 13 ++++
> 6 files changed, 112 insertions(+)
>
> diff --git a/drivers/media/platform/qcom/iris/Makefile b/drivers/media/platform/qcom/iris/Makefile
> index ddd4c994a0b9..95f4e92fe085 100644
> --- a/drivers/media/platform/qcom/iris/Makefile
> +++ b/drivers/media/platform/qcom/iris/Makefile
> @@ -8,5 +8,6 @@ iris-objs += iris_core.o \
> iris_probe.o \
> iris_resources.o \
> iris_vidc.o \
> + iris_vpu_common.o \
>
> obj-$(CONFIG_VIDEO_QCOM_IRIS) += iris.o
> diff --git a/drivers/media/platform/qcom/iris/iris_core.c b/drivers/media/platform/qcom/iris/iris_core.c
> index 8c7d53c57086..5ad66ac113ae 100644
> --- a/drivers/media/platform/qcom/iris/iris_core.c
> +++ b/drivers/media/platform/qcom/iris/iris_core.c
> @@ -6,6 +6,7 @@
> #include "iris_core.h"
> #include "iris_firmware.h"
> #include "iris_state.h"
> +#include "iris_vpu_common.h"
>
> void iris_core_deinit(struct iris_core *core)
> {
> @@ -39,10 +40,16 @@ int iris_core_init(struct iris_core *core)
> if (ret)
> goto error_queue_deinit;
>
> + ret = iris_vpu_boot_firmware(core);
> + if (ret)
> + goto error_unload_fw;
> +
> mutex_unlock(&core->lock);
>
> return 0;
>
> +error_unload_fw:
> + iris_fw_unload(core);
> error_queue_deinit:
> iris_hfi_queues_deinit(core);
> error:
> diff --git a/drivers/media/platform/qcom/iris/iris_platform_common.h b/drivers/media/platform/qcom/iris/iris_platform_common.h
> index 9c919367f9d7..47fdebd8135c 100644
> --- a/drivers/media/platform/qcom/iris/iris_platform_common.h
> +++ b/drivers/media/platform/qcom/iris/iris_platform_common.h
> @@ -45,6 +45,7 @@ struct iris_platform_data {
> const char *fwname;
> u32 pas_id;
> struct tz_cp_config *tz_cp_config_data;
> + u32 core_arch;
> };
>
> #endif
> diff --git a/drivers/media/platform/qcom/iris/iris_platform_sm8550.c b/drivers/media/platform/qcom/iris/iris_platform_sm8550.c
> index 1bb34c3e6e18..a559e095fefc 100644
> --- a/drivers/media/platform/qcom/iris/iris_platform_sm8550.c
> +++ b/drivers/media/platform/qcom/iris/iris_platform_sm8550.c
> @@ -8,6 +8,8 @@
> #include "iris_platform_common.h"
> #include "iris_resources.h"
>
> +#define VIDEO_ARCH_LX 1
> +
> static const struct icc_info sm8550_icc_table[] = {
> { "cpu-cfg", 1000, 1000 },
> { "video-mem", 1000, 15000000 },
> @@ -48,4 +50,5 @@ struct iris_platform_data sm8550_data = {
> .fwname = "qcom/vpu/vpu30_p4.mbn",
> .pas_id = IRIS_PAS_ID,
> .tz_cp_config_data = &tz_cp_config_sm8550,
> + .core_arch = VIDEO_ARCH_LX,
> };
> diff --git a/drivers/media/platform/qcom/iris/iris_vpu_common.c b/drivers/media/platform/qcom/iris/iris_vpu_common.c
> new file mode 100644
> index 000000000000..df87b1b719a9
> --- /dev/null
> +++ b/drivers/media/platform/qcom/iris/iris_vpu_common.c
> @@ -0,0 +1,87 @@
> +// SPDX-License-Identifier: GPL-2.0-only
> +/*
> + * Copyright (c) 2022-2024 Qualcomm Innovation Center, Inc. All rights reserved.
> + */
> +
> +#include <linux/iopoll.h>
> +
> +#include "iris_core.h"
> +#include "iris_vpu_common.h"
> +
> +#define CPU_BASE_OFFS 0x000A0000
> +
> +#define CPU_CS_BASE_OFFS (CPU_BASE_OFFS)
> +
> +#define CTRL_INIT (CPU_CS_BASE_OFFS + 0x48)
> +#define CTRL_STATUS (CPU_CS_BASE_OFFS + 0x4C)
> +
> +#define CTRL_ERROR_STATUS__M 0xfe
> +
> +#define QTBL_INFO (CPU_CS_BASE_OFFS + 0x50)
> +#define QTBL_ADDR (CPU_CS_BASE_OFFS + 0x54)
> +#define CPU_CS_SCIACMDARG3 (CPU_CS_BASE_OFFS + 0x58)
> +#define SFR_ADDR (CPU_CS_BASE_OFFS + 0x5C)
> +#define UC_REGION_ADDR (CPU_CS_BASE_OFFS + 0x64)
> +#define UC_REGION_SIZE (CPU_CS_BASE_OFFS + 0x68)
> +
> +#define CPU_CS_H2XSOFTINTEN (CPU_CS_BASE_OFFS + 0x148)
> +#define CPU_CS_X2RPMH (CPU_CS_BASE_OFFS + 0x168)
> +
> +static void iris_vpu_setup_ucregion_memory_map(struct iris_core *core)
> +{
> + u32 queue_size, value;
> +
> + /* Iris hardware requires 4K queue alignment */
> + queue_size = ALIGN(sizeof(struct iris_hfi_queue_table_header) +
> + (IFACEQ_QUEUE_SIZE * IFACEQ_NUMQ), SZ_4K);
> +
> + value = (u32)core->iface_q_table_daddr;
> + writel(value, core->reg_base + UC_REGION_ADDR);
> +
> + /* Iris hardware requires 1M queue alignment */
> + value = ALIGN(SFR_SIZE + queue_size, SZ_1M);
> + writel(value, core->reg_base + UC_REGION_SIZE);
> +
> + value = (u32)core->iface_q_table_daddr;
> + writel(value, core->reg_base + QTBL_ADDR);
> +
> + writel(0x01, core->reg_base + QTBL_INFO);
A general comment I have is instead of writing hard-coded values to
registers we should define at a minimum the bit-fields we use if not the
entire set of bits-fields for the register.
The only exception to this is when we don't know what those values are -
for example receiving a magic write sequence for a camera sensor.
In this case though we have full access to enumerate the register
bit-fields.
Without looking at the register descriptions I guess this bit is an
enable or a startup bit =>
#define QTBL_INFO_EN BIT(0)
I'll not go through this series reiterating this comment but, it
certainly applies to any bit-field/register in the same => please define
at least the bits used if not the full set of bits for register writes
instead of using magic numbers.
---
bod
Powered by blists - more mailing lists