linux-kernel - Re: [PATCH v2] media: iris: vpu3x: Add MNoC low power handshake during hardware power-off

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] [day] [month] [year] [list]

Message-ID: <bb919b4f-b51f-41c5-aed8-a3809316f7fb@linaro.org>
Date: Wed, 13 Aug 2025 21:20:35 +0100
From: Bryan O'Donoghue <bryan.odonoghue@...aro.org>
To: Dikshita Agarwal <quic_dikshita@...cinc.com>,
 Vikash Garodia <quic_vgarodia@...cinc.com>,
 Abhinav Kumar <abhinav.kumar@...ux.dev>,
 Mauro Carvalho Chehab <mchehab@...nel.org>
Cc: linux-media@...r.kernel.org, linux-arm-msm@...r.kernel.org,
 linux-kernel@...r.kernel.org, Neil Armstrong <neil.armstrong@...aro.org>
Subject: Re: [PATCH v2] media: iris: vpu3x: Add MNoC low power handshake
 during hardware power-off

On 13/08/2025 08:53, Dikshita Agarwal wrote:
> +	/* set MNoC to low power */
> +	writel(REQ_POWER_DOWN_PREP, core->reg_base + AON_WRAPPER_MVP_NOC_LPI_CONTROL);
> +
> +	do {
> +		value = readl(core->reg_base + AON_WRAPPER_MVP_NOC_LPI_STATUS);
> +
> +		handshake_done = value & NOC_LPI_STATUS_DONE;
> +		handshake_busy = value & (NOC_LPI_STATUS_DENY | NOC_LPI_STATUS_ACTIVE);
> +
> +		if (handshake_done || !handshake_busy)
> +			break;
> +
> +		writel(0, core->reg_base + AON_WRAPPER_MVP_NOC_LPI_CONTROL);
> +
> +		udelay(15);
> +
> +		writel(REQ_POWER_DOWN_PREP, core->reg_base + AON_WRAPPER_MVP_NOC_LPI_CONTROL);
> +	} while (++count < 1000);

So if this loop executes 999 times but succeeds on the 1000th try, your 
test would fail because your final write never gets evaluated in a 
subsequent loop.

Wouldn't this be more logical like this

do {
	/* issue command */
	writel(REQ_POWER_DOWN_PREP, core->reg_base +
	       AON_WRAPPER_MVP_NOC_LPI_CONTROL);

	/* wait for command to take effect */
	udelay(15);

	/* read back status */
	value = readl(core->reg_base + AON_WRAPPER_MVP_NOC_LPI_STATUS);

	handshake_done = value & NOC_LPI_STATUS_DONE;
	handshake_busy = value & (NOC_LPI_STATUS_DENY |
				  NOC_LPI_STATUS_ACTIVE);

	if (handshake_done || !handshake_busy)
		break;

	/* power down? the mnoc */
	writel(0, core->reg_base + AON_WRAPPER_MVP_NOC_LPI_CONTROL);

	/* Let that command take effect ? */
	udelay(15);

} while (++count < 1000);

That way you exit the loop with the mnoc state in a known state, instead 
of as your loop currently is perhaps having succeeded but incorrectly 
signalling failure.

Also why 1000 ? Thats 1000 x 1.5 microseconds - 1.5 milliseconds 
potentially in your submitted patch now nearly 3 milliseconds assuming 
the power-down command similarly requires a grace period.

Please at least fix the loop so that the last power-on command should it 
succeed at the termination of your loop can be captured.

> +	if (!handshake_done && handshake_busy)
> +		dev_err(core->dev, "LPI handshake timeout\n");

---
bod