linux-kernel - Re: [PATCH v2 12/34] media: iris: add video processing unit(VPU) specific register handling

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <f9c841f0-9839-47b7-b026-d0cd2138af11@kernel.org>
Date: Mon, 18 Dec 2023 17:19:36 +0100
From: Krzysztof Kozlowski <krzk@...nel.org>
To: Dikshita Agarwal <quic_dikshita@...cinc.com>,
 linux-media@...r.kernel.org, linux-kernel@...r.kernel.org,
 stanimir.k.varbanov@...il.com, quic_vgarodia@...cinc.com, agross@...nel.org,
 andersson@...nel.org, konrad.dybcio@...aro.org, mchehab@...nel.org,
 bryan.odonoghue@...aro.org
Cc: linux-arm-msm@...r.kernel.org, quic_abhinavk@...cinc.com
Subject: Re: [PATCH v2 12/34] media: iris: add video processing unit(VPU)
 specific register handling

On 18/12/2023 12:32, Dikshita Agarwal wrote:
> Registers are defined differently for different VPUs.
> Define ops for VPU specific handling to accommodate
> different VPUs. Implement boot sequence of firmware and interrupt
> programming.
> 

...

> +
> +int write_register(struct iris_core *core, u32 reg, u32 value)
> +{
> +	void __iomem *base_addr;
> +	int ret;
> +
> +	ret = check_core_lock(core);
> +	if (ret)
> +		return ret;
> +
> +	base_addr = core->reg_base;
> +	base_addr += reg;
> +	writel_relaxed(value, base_addr);
> +
> +	/* Make sure value is written into the register */
> +	wmb();

Just don't use relaxed method. The same applies to other places like that.

> +
> +	return ret;
> +}
> +
> +int read_register(struct iris_core *core, u32 reg, u32 *value)
> +{
> +	void __iomem *base_addr;
> +
> +	base_addr = core->reg_base;
> +
> +	*value = readl_relaxed(base_addr + reg);
> +
> +	/* Make sure value is read correctly from the register */
> +	rmb();
> +
> +	return 0;
> +}
> +
> +static const struct compat_handle compat_handle[] = {
> +	{
> +		.compat                  = "qcom,sm8550-iris",
> +		.init                    = init_iris3,

Uh...

> +	},
> +};
> +
> +int init_vpu(struct iris_core *core)
> +{
> +	struct device *dev = NULL;
> +	int i, ret = 0;
> +
> +	dev = core->dev;
> +
> +	for (i = 0; i < ARRAY_SIZE(compat_handle); i++) {
> +		if (of_device_is_compatible(dev->of_node, compat_handle[i].compat)) {
> +			ret = compat_handle[i].init(core);


This does not look good. Use flags, quirks, type, pointer ops in
structures. Just look at existing code in Linux kernel. Do not
reimplement driver match data.



Best regards,
Krzysztof