From: Tan Siewert <tan.siewert@9elements.com>
To: Prasanth Kumar Padarthi <prasanth.padarthi10@gmail.com>,
srini@kernel.org
Cc: robh@kernel.org, krzk+dt@kernel.org, conor+dt@kernel.org,
joel@jms.id.au, andrew@codeconstruct.com.au,
ryan_chen@aspeedtech.com, billy_tsai@aspeedtech.com,
neal_liu@aspeedtech.com, devicetree@vger.kernel.org,
linux-arm-kernel@lists.infradead.org,
linux-aspeed@lists.ozlabs.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2 2/2] nvmem: aspeed: Add read-only AST2600 OTP support
Date: Fri, 20 Nov 2026 20:05:03 +0100 [thread overview]
Message-ID: <f70eee05-1d17-4060-b1ee-f4845b74e3e6@9elements.com> (raw)
In-Reply-To: <20261120174440.123487-3-prasanth.padarthi10@gmail.com>
On 11/20/26 6:44 PM, Prasanth Kumar Padarthi wrote:
> Provide root-only NVMEM access for userspace diagnostics comparing OTP
> ranges with vendor-derived reference data during BMC bring-up. The
> consumer uses raw offsets rather than named cells or a private ioctl.
>
> Initialize read mode once at probe and mask the fixed secure prefix using
> an NVMEM keepout. Recheck SEC54 on each controller read and reject requests
> into an extended protected prefix. Programming protection does not block
> reads. Serialize transactions and stop issuing commands after a timeout.
> Use a 1-us initial polling delay and a generous 100-ms timeout.
>
This is describing the code...
Most of it belongs in the cover letter or below the --- line.
Please explain what the *hardware* does and *why* the driver is designed
this way, not what the code is doing!
> Adapt the read sequence from ASPEED's downstream driver, preserving its
> attribution, and add the driver to the ASPEED MAINTAINERS entry.
>
> Tested on ASRock Rack B650D4U (AST2600 A3) with Linux 7.4-rc1 and
> OpenBMC Linux 6.18.24. Repeated reads, cross-pair reads and four concurrent
> readers (40 reads) passed. The selected 12-byte range matched a saved
> ASPEED-derived reference. The 7.4-rc1 run used a RAM-booted kernel with
> the working OpenBMC board DTB and initramfs; its final-pair check also
> passed.
This part should be outside of the commit message, like I explained earlier.
>
> Link: https://github.com/AspeedTech-BMC/linux/blob/769f62b7baa84d6998723b0ea60280e380183553/drivers/soc/aspeed/ast2600-otp.c
> Link: https://lore.kernel.org/linux-aspeed/KL1PR0601MB3986A356BD06F98D0FF9AD5E80792@KL1PR0601MB3986.apcprd06.prod.outlook.com/
> Co-developed-by: Neal Liu <neal_liu@aspeedtech.com>
> Signed-off-by: Neal Liu <neal_liu@aspeedtech.com>
> Signed-off-by: Prasanth Kumar Padarthi <prasanth.padarthi10@gmail.com>
>
> ---
> MAINTAINERS | 1
> drivers/nvmem/Kconfig | 13 ++
> drivers/nvmem/Makefile | 1
> drivers/nvmem/nvmem-aspeed-otp.c | 297 ++++++++++++++++++++++++++++++++++++++
> 4 files changed, 312 insertions(+)
>
> diff --git a/MAINTAINERS b/MAINTAINERS
> index 93abbb315..5415e2be4 100644
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -2813,6 +2813,7 @@ F: Documentation/devicetree/bindings/arm/aspeed/
> F: arch/arm/boot/dts/aspeed/
> F: arch/arm/mach-aspeed/
> F: arch/arm64/boot/dts/aspeed/
> +F: drivers/nvmem/nvmem-aspeed-otp.c
The matcher here *may* be unnecessary given that we have `N`.
> N: aspeed
>
> ARM/AXIADO ARCHITECTURE
> diff --git a/drivers/nvmem/Kconfig b/drivers/nvmem/Kconfig
> index d2f5bb236..e6445262b 100644
> --- a/drivers/nvmem/Kconfig
> +++ b/drivers/nvmem/Kconfig
> @@ -164,6 +164,19 @@ config NVMEM_APPLE_SPMI
> This driver can also be built as a module. If so, the module
> will be called apple-nvmem-spmi.
>
> +config NVMEM_ASPEED_AST2600_OTP
> + tristate "ASPEED AST2600 OTP support"
The file is now named generic. Please name the kconfig
"NVMEM_ASPEED_OTP" so it doesn't need to be renamed once AST2700 support
is implemented.
> + depends on ARCH_ASPEED || COMPILE_TEST
> + depends on HAS_IOMEM
> + help
> + Say Y here to enable read-only access to the OTP data region
> + on ASPEED AST2600 BMCs through the NVMEM interface.
> +
> + This driver can also be built as a module. If so, the module
> + will be called nvmem-aspeed-otp.
> +
> + If unsure, say N.
> +
> config NVMEM_BCM_OCOTP
> tristate "Broadcom On-Chip OTP Controller support"
> depends on ARCH_BCM_IPROC || COMPILE_TEST
*snip*
> diff --git a/drivers/nvmem/nvmem-aspeed-otp.c b/drivers/nvmem/nvmem-aspeed-otp.c
> new file mode 100644
> --- /dev/null
> +++ b/drivers/nvmem/nvmem-aspeed-otp.c
> @@ -0,0 +1,297 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Copyright (C) ASPEED Technology Inc.
> + *
> + * ASPEED AST2600 OTP Controller driver
> + */
> +
> +#include <linux/bitfield.h>
> +#include <linux/cleanup.h>
> +#include <linux/io.h>
> +#include <linux/iopoll.h>
> +#include <linux/mod_devicetable.h>
> +#include <linux/module.h>
> +#include <linux/mutex.h>
> +#include <linux/nvmem-provider.h>
> +#include <linux/platform_device.h>
> +#include <linux/unaligned.h>
> +
> +#define AST2600_OTP_PROTECT_KEY 0x00
> +#define AST2600_OTP_CMD 0x04
> +#define AST2600_OTP_ADDR 0x10
> +#define AST2600_OTP_STATUS 0x14
> +#define AST2600_OTP_COMPARE_1 0x20
> +#define AST2600_OTP_COMPARE_2 0x24
> +#define AST2600_OTP_EXTRA_READ_PROTECT 0x54
> +#define AST2600_OTP_EXTRA_READ_RANGE GENMASK(5, 0)
Please keep register offsets and field masks visually separate.
Usually you put the fields directly below their register, indented with
an extra space. For example:
#define AST2600_OTP_EXTRA_READ_PROTECT 0x54
#define AST2600_OTP_EXTRA_READ_RANGE GENMASK(5, 0)
> +
> +#define AST2600_OTP_CMD_READ 0x23b1e361
> +#define AST2600_OTP_CMD_WRITE_MODE 0x23b1e362
> +#define AST2600_OTP_UNLOCK_KEY 0x349fe38a
> +#define AST2600_OTP_STATUS_DONE GENMASK(2, 1)
> +/* Normal reads take less than 1 us; allow a generous bounded wait. */
> +#define AST2600_OTP_TIMEOUT_US 100000
> +
> +#define AST2600_OTP_DATA_WORDS 0x800
> +#define AST2600_OTP_SIZE (AST2600_OTP_DATA_WORDS * sizeof(u32))
> +#define AST2600_OTP_CONF0_ADDR 0x800
> +#define AST2600_OTP_CONF0_SEC_SIZE GENMASK(21, 16)
> +#define AST2600_OTP_SEC_SIZE_UNIT (32 * sizeof(u32))
The alignment is somewhat inconsistent.
> +
> +struct ast2600_otp {
> + struct device *dev;
> + void __iomem *base;
> + /* Serialize the complete unlock, read and relock sequence. */
> + struct mutex lock;
> + struct nvmem_keepout secure_keepout;
> + bool failed;
> +};
> +
*snip*
> +static int ast2600_otp_init_read_mode(struct ast2600_otp *priv)
> +{
> + static const u32 mode_addresses[] = { 0x3000, 0x5000, 0x1000 };
Please give these registers names instead of using an anonymous array.
> + int i, ret;
> +
> + ret = ast2600_otp_wait_complete(priv);
> + if (ret)
> + return ret;
> +
> + /*
> + * Restore read mode once at probe; this driver never programs fuses.
> + * These writes target volatile mode registers, not OTP fuse cells.
> + * No fuse programming command or programming timing is used.
> + */
All the sentences say the same in slightly different ways, but none of
them says *why* the step is needed. Maybe: "Linux never programs the
OTP, but it might be left in it before, so restore the read mode once"?
> + for (i = 0; i < ARRAY_SIZE(mode_addresses); i++) {
> + writel(mode_addresses[i], priv->base + AST2600_OTP_ADDR);
> + writel(0, priv->base + AST2600_OTP_COMPARE_1);
> + writel(AST2600_OTP_CMD_WRITE_MODE, priv->base + AST2600_OTP_CMD);
> + ret = ast2600_otp_wait_complete(priv);
> + if (ret)
> + return ret;
> + }
> +
> + return 0;
> +}
> +
> +static int ast2600_otp_read_pair(struct ast2600_otp *priv,
> + unsigned int aligned_word,
> + u32 *word0, u32 *word1)
> +{
> + int ret;
> +
> + /* The hardware reads two words starting at an even word index. */
> + writel(aligned_word, priv->base + AST2600_OTP_ADDR);
> +
> + writel(AST2600_OTP_CMD_READ,
> + priv->base + AST2600_OTP_CMD);
> +
> + ret = ast2600_otp_wait_complete(priv);
> + if (ret)
> + return ret;
> +
> + *word0 = readl(priv->base + AST2600_OTP_COMPARE_1);
> + *word1 = readl(priv->base + AST2600_OTP_COMPARE_2);
> +
> + return 0;
> +}
> +
> +static int ast2600_otp_init_keepout(struct ast2600_otp *priv)
The is misleading, because it also restores read mode.
> +{
> + u32 conf0, unused;
> + int ret;
> +
> + guard(mutex)(&priv->lock);
> +
> + writel(AST2600_OTP_UNLOCK_KEY, priv->base + AST2600_OTP_PROTECT_KEY);
> + if (readl(priv->base + AST2600_OTP_PROTECT_KEY) != 1) {
> + ret = -EACCES;
> + goto out_lock;
> + }
Mixing goto and scope-based cleanup in the same function isn't allowed
(see the big comment in include/linux/cleanup.h). The same applies to
ast2600_otp_read().
The unlock/relock sequence is also duplicated between both functions.
Please move this into helpers (or a DEFINE_GUARD()) so it doesn't need
labels.
Also, -EIO might be better here given that it's a hardware problem, and
not a permission problem.
> +
> + ret = ast2600_otp_init_read_mode(priv);
> + if (ret)
> + goto out_lock;
> +
> + ret = ast2600_otp_read_pair(priv, AST2600_OTP_CONF0_ADDR,
> + &conf0, &unused);
> + if (ret)
> + goto out_lock;
> +
> + priv->secure_keepout.start = 0;
> + priv->secure_keepout.end =
> + FIELD_GET(AST2600_OTP_CONF0_SEC_SIZE, conf0) *
> + AST2600_OTP_SEC_SIZE_UNIT;
> + priv->secure_keepout.value = 0;
priv comes from devm_kzalloc(), so they're are already zero.
> +
> +out_lock:
> + writel(0, priv->base + AST2600_OTP_PROTECT_KEY);
> + /* Complete the lock write before returning and releasing the mutex. */
> + readl(priv->base + AST2600_OTP_PROTECT_KEY);
Again (as in v1), nothing else accesses the device during probe. This
read is useless.
> + return ret;
> +}
> +
> +static int ast2600_otp_read(void *context, unsigned int offset,
> + void *val, size_t bytes)
> +{
> + struct ast2600_otp *priv = context;
> + u8 *buf = val;
> + unsigned int word_index;
> + unsigned int aligned_word;
> + unsigned int secure_size;
> + u32 extra_read;
> + u32 word0, word1;
> + int ret;
> +
> + if (!bytes)
> + return 0;
> +
> + if (offset % sizeof(u32) || bytes % sizeof(u32))
> + return -EINVAL;
> +
> + if (offset >= AST2600_OTP_SIZE ||
> + bytes > AST2600_OTP_SIZE - offset)
> + return -EINVAL;
> +
> + guard(mutex)(&priv->lock);
> +
> + if (priv->failed)
> + return -EIO;
> +
> + /* SEC54 can grow after probe; keep the registered keepout immutable. */
Please describe the hardware behaviour instead, e.g. "SEC54[5:0] can
only grow until the ARM cores reset."
> + extra_read = readl(priv->base + AST2600_OTP_EXTRA_READ_PROTECT);
> + secure_size = FIELD_GET(AST2600_OTP_EXTRA_READ_RANGE, extra_read) *
> + AST2600_OTP_SEC_SIZE_UNIT;
> + secure_size = max(priv->secure_keepout.end, secure_size);
> + if (offset < secure_size)
> + return -EACCES;
> +
> + writel(AST2600_OTP_UNLOCK_KEY, priv->base + AST2600_OTP_PROTECT_KEY);
> + if (readl(priv->base + AST2600_OTP_PROTECT_KEY) != 1) {
> + ret = -EACCES;
> + goto out_lock;
> + }
> +
> + ret = ast2600_otp_wait_complete(priv);
> + if (ret)
> + goto out_lock;
> +
> + word_index = offset / sizeof(u32);
> + while (bytes) {
> + aligned_word = word_index & ~1U;
> +
> + ret = ast2600_otp_read_pair(priv, aligned_word,
> + &word0, &word1);
> + if (ret)
> + goto out_lock;
> +
> + if (!(word_index & 1)) {
> + put_unaligned_le32(word0, buf);
> + buf += sizeof(u32);
> + bytes -= sizeof(u32);
> + word_index++;
> + }
> + if (bytes) {
> + put_unaligned_le32(word1, buf);
> + buf += sizeof(u32);
> + bytes -= sizeof(u32);
> + word_index++;
> + }
Why not reading the pair into an array? That'd remove the duplicates.
> + }
> +
> + ret = 0;
> +
> +out_lock:
> + writel(0, priv->base + AST2600_OTP_PROTECT_KEY);
> + /* Complete the lock write before returning and releasing the mutex. */
> + readl(priv->base + AST2600_OTP_PROTECT_KEY);
> + return ret;
> +}
> +
> +static int ast2600_otp_probe(struct platform_device *pdev)
> +{
> + struct nvmem_config config = {
> + .name = "ast2600-otp",
> + .id = NVMEM_DEVID_AUTO,
> + .owner = THIS_MODULE,
> + .type = NVMEM_TYPE_OTP,
> + .read_only = true,
> + .root_only = true,
> + .word_size = sizeof(u32),
> + .stride = sizeof(u32),
> + .size = AST2600_OTP_SIZE,
> + .reg_read = ast2600_otp_read,
> + };
> + struct nvmem_device *nvmem;
> + struct ast2600_otp *priv;
> + struct device *dev = &pdev->dev;
> + int ret;
> +
> + priv = devm_kzalloc(dev, sizeof(*priv), GFP_KERNEL);
> + if (!priv)
> + return -ENOMEM;
> +
> + priv->dev = dev;
> + priv->base = devm_platform_ioremap_resource(pdev, 0);
> + if (IS_ERR(priv->base))
> + return PTR_ERR(priv->base);
> +
> + ret = devm_mutex_init(dev, &priv->lock);
> + if (ret)
> + return ret;
> +
> + ret = ast2600_otp_init_keepout(priv);
> + if (ret)
> + return dev_err_probe(dev, ret,
> + "Failed to initialize OTP read mode and secure prefix\n");
> +
> + config.dev = dev;
> + config.priv = priv;
> + if (priv->secure_keepout.end) {
> + config.keepout = &priv->secure_keepout;
> + config.nkeepout = 1;
> + }
> + nvmem = devm_nvmem_register(dev, &config);
> + return PTR_ERR_OR_ZERO(nvmem);
nit: return PTR_ERR_OR_ZERO(devm_nvmem_register(dev, &config));
> +}
> +
> +static const struct of_device_id ast2600_otp_of_match[] = {
> + { .compatible = "aspeed,ast2600-sbc" },
> + { }
> +};
> +MODULE_DEVICE_TABLE(of, ast2600_otp_of_match);
> +
> +static struct platform_driver ast2600_otp_driver = {
> + .probe = ast2600_otp_probe,
> + .driver = {
> + .name = "ast2600-otp",
If the plan is to add AST2700 later, then please use "aspeed-otp" and
aspeed_otp_* names.
Last: Please run your next build with W=1 and do checkpatch.pl +
dt_binding_check...
Tan
next prev parent reply other threads:[~2026-11-20 19:05 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-13 3:42 [PATCH v1 0/1] " Prasanth Kumar Padarthi
2026-11-13 3:42 ` [PATCH v1 1/1] " Prasanth Kumar Padarthi
2026-11-13 13:05 ` Tan Siewert
2026-11-14 13:21 ` Prasanth
2026-11-20 17:44 ` [PATCH v2 0/2] " Prasanth Kumar Padarthi
2026-11-20 17:44 ` [PATCH v2 1/2] dt-bindings: arm: aspeed: sbc: Describe OTP NVMEM provider Prasanth Kumar Padarthi
2026-11-20 17:44 ` [PATCH v2 2/2] nvmem: aspeed: Add read-only AST2600 OTP support Prasanth Kumar Padarthi
2026-11-20 19:05 ` Tan Siewert [this message]
2026-11-20 18:58 ` [PATCH v2 0/2] " Tan Siewert
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=f70eee05-1d17-4060-b1ee-f4845b74e3e6@9elements.com \
--to=tan.siewert@9elements.com \
--cc=andrew@codeconstruct.com.au \
--cc=billy_tsai@aspeedtech.com \
--cc=conor+dt@kernel.org \
--cc=devicetree@vger.kernel.org \
--cc=joel@jms.id.au \
--cc=krzk+dt@kernel.org \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-aspeed@lists.ozlabs.org \
--cc=linux-kernel@vger.kernel.org \
--cc=neal_liu@aspeedtech.com \
--cc=prasanth.padarthi10@gmail.com \
--cc=robh@kernel.org \
--cc=ryan_chen@aspeedtech.com \
--cc=srini@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®