phy: add support for airoha en8811h 2.5G PHY by Yuzhii0718 · Pull Request #422 · hanwckf/immortalwrt-mt798x

Yuzhii0718 · 2026-02-28T08:44:07Z

In the current repository, the driver for en8811h depends on target/linux/mediatek/patches-5.4/9999-mt7986a-bananapi-bpi-r3-air_en8811-2.5G-phy.patch, which doesn't work well. I updated to version v1.2.5 of airoha, and it works correctly.

change kmod-phy-air-en8811h to kmod-phy-airoha-en8811h

Copilot

Pull request overview

This PR replaces the previous EN8811H PHY kernel-tree patch approach with an updated Airoha EN8811H (v1.2.5) out-of-tree driver package, and updates MT7986 device images to depend on the renamed kernel module package.

Changes:

Remove the old mediatek kernel patch that injected air_en8811h into the 5.4 kernel tree and drop the old kmod-phy-air-en8811h kernel module definition.
Add a new package/airoha/drivers/an8811 kernel module package containing the EN8811H driver sources and firmware assets.
Update MT7986 (BPI-R3MINI*) images to use kmod-phy-airoha-en8811h.

Reviewed changes

Copilot reviewed 13 out of 16 changed files in this pull request and generated 14 comments.

Show a summary per file

File	Description
target/linux/mediatek/patches-5.4/9999-mt7986a-bananapi-bpi-r3-air_en8811-2.5G-phy.patch	Removes the old in-tree kernel patch for EN8811H.
target/linux/mediatek/image/mt7986.mk	Switches device package dependency to `kmod-phy-airoha-en8811h`.
package/kernel/linux/modules/netdevices.mk	Removes the old `KernelPackage/phy-air-en8811h` definition.
package/airoha/drivers/an8811/Makefile	Introduces a new OpenWrt kernel module package and installs firmware files.
package/airoha/drivers/an8811/README.md	Adds English documentation for the new package.
package/airoha/drivers/an8811/document/README.sc.md	Adds Simplified Chinese documentation for the new package.
package/airoha/drivers/an8811/src/mtk/Makefile	Adds out-of-tree module build rules for `air_en8811h`.
package/airoha/drivers/an8811/src/mtk/Kconfig	Adds module-local Kconfig options (currently not integrated into the kernel Kconfig tree).
package/airoha/drivers/an8811/src/mtk/air_en8811h_main.c	Adds the v1.2.5 EN8811H PHY driver main implementation.
package/airoha/drivers/an8811/src/mtk/air_en8811h_api.c	Adds EN8811H MDIO/PBUS access helpers and optional debugfs tooling (contains several correctness issues).
package/airoha/drivers/an8811/src/mtk/air_en8811h_api.h	Adds API declarations and debugfs constants.
package/airoha/drivers/an8811/src/mtk/air_en8811h.h	Adds shared defines/structs and driver version update to v1.2.5.
package/airoha/drivers/an8811/src/mtk/air_en8811h_fw.h	Adds embedded firmware arrays used by the driver.
package/airoha/drivers/an8811/src/mtk/EthMD32.dm.bin	Adds DM firmware binary asset.
package/airoha/drivers/an8811/src/mtk/EthMD32.DSP.bin	Adds DSP firmware binary asset.
package/airoha/drivers/an8811/src/mtk/EthMD32_PM_CheckSum.txt	Adds a firmware checksum reference file.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-02-28T08:52:10Z

package/airoha/drivers/an8811/src/mtk/air_en8811h_api.c

+		if (val < 0 || val > 1) {
+			dev_err(dev,
+				   "airoha,surge value %u out of range.",
+				   val);
+			return -EINVAL;


airoha,surge is read into a u32 val, so the val < 0 part of the range check is ineffective. The check should only validate the upper bound (e.g., val > 1) and/or use a signed type if negatives are meaningful.

Copilot · 2026-02-28T08:52:10Z

package/airoha/drivers/an8811/src/mtk/air_en8811h_api.c

+			"[debugfs] = /sys/kernel/debug/mdio-bus\':[phy_addr]\n"
+			"echo [tx polarity] [rx polarity] > /[debugfs]/polarity\n"
+			"option: tx_normal, tx_reverse, rx_normal, rx_revers\n");
+}
+
+static int air_set_polarity(struct phy_device *phydev, int tx_rx)
+{
+	int ret = 0;
+	unsigned int pbus_data = 0;
+
+	pr_debug("\nPolarit %s\n", tx_rx_string[tx_rx]);


The debugfs help text has typos that could confuse users (rx_revers, Polarit). Please correct these strings (e.g., rx_reverse, Polarity).

Copilot · 2026-02-28T08:52:10Z

package/airoha/drivers/an8811/src/mtk/air_en8811h_api.c

+	if (!dir) {
+		dev_err(phydev_dev(phydev), "%s:err at %d\n",
+					 __func__, __LINE__);
+		ret = -ENOMEM;


If debugfs_create_dir() fails, the function sets ret = -ENOMEM but still continues creating files with dir == NULL and stores priv->debugfs_root = dir. This can lead to unexpected placement under the debugfs root and/or NULL derefs later. Return immediately on failure (or gate file creation on dir).

Suggested change

ret = -ENOMEM;

ret = -ENOMEM;

return ret;

Copilot · 2026-02-28T08:52:10Z

package/airoha/drivers/an8811/README.md

+### 1) Firmware
+
+The driver loads firmware using `request_firmware_direct()`:
+
+- `EthMD32.dm.bin`
+- `EthMD32.DSP.bin`
+
+Ensure these files are present in the target system’s `/lib/firmware/`.


The README states the driver loads firmware via request_firmware_direct() from /lib/firmware/, but the current driver code embeds firmware arrays via air_en8811h_fw.h and calls air_mdio_write_buf(...) (no request_firmware* usage). Please update the documentation to match the implementation (or switch the driver back to runtime firmware loading if that’s the goal).

Copilot · 2026-02-28T08:52:11Z

package/airoha/drivers/an8811/Makefile

+	KCONFIG:= \
+	CONFIG_AIROHA_EN8811H_PHY=m \
+	CONFIG_AIROHA_EN8811H_PHY_DEBUGFS=n


The KCONFIG entries set CONFIG_AIROHA_EN8811H_PHY*, but these symbols don’t exist in the kernel Kconfig tree (only in the external module’s unused src/mtk/Kconfig). As a result, these settings won’t reliably affect compilation (e.g., enabling debugfs). Consider controlling module features via ccflags-y/EXTRA_CFLAGS or keying off existing kernel symbols like CONFIG_DEBUG_FS instead of non-existent ones.

Suggested change

KCONFIG:= \

CONFIG_AIROHA_EN8811H_PHY=m \

CONFIG_AIROHA_EN8811H_PHY_DEBUGFS=n

Copilot · 2026-02-28T08:52:11Z

package/airoha/drivers/an8811/src/mtk/air_en8811h_api.c

+	memset(buf, 0, 64);
+	if (copy_from_user(buf, buffer, count))
+		return -EFAULT;


airphy_debugfs_pbus() copies count bytes from userspace into a fixed 64-byte stack buffer without validating count first. This is a stack overflow risk. Add the same if (count > sizeof(buf) - 1) return -EINVAL; guard used in airphy_debugfs_buckpbus() (and consider NUL-terminating before parsing).

Suggested change

memset(buf, 0, 64);

if (copy_from_user(buf, buffer, count))

return -EFAULT;

memset(buf, 0, 64);

if (count > sizeof(buf) - 1)

return -EINVAL;

if (copy_from_user(buf, buffer, count))

return -EFAULT;

buf[count] = '\0';

Copilot · 2026-02-28T08:52:12Z

package/airoha/drivers/an8811/src/mtk/air_en8811h_api.c

+	if (buf[0] == 'w') {
+		if (sscanf(buf, "w %x %x", &reg, &val) == -1)
+			return -EFAULT;
+


Several debugfs handlers use sscanf(...) == -1 to validate input (e.g., here for the write path). sscanf returns the number of successfully parsed items, so malformed input can return 0/1/2 (not -1) and leave reg/val uninitialized, leading to unintended register reads/writes. Validate the exact expected count (e.g., != 2).

Copilot · 2026-02-28T08:52:12Z

package/airoha/drivers/an8811/src/mtk/air_en8811h_api.c

+	seq_printf(seq, "| Boot mode            : %s\n",
+		((val & BIT(24)) >> 24) ? "Flash" : "Download Code");
+	seq_printf(seq, "| EthMD32.dm.bin  CRC32: %08x\n",
+				priv->dm_crc32);
+	seq_printf(seq, "| EthMD32.DSP.bin CRC32: %08x\n",
+				priv->dsp_crc32);
+	val = air_buckpbus_reg_read(phydev, 0x3b3c);


airphy_info_show() prints priv->dm_crc32 and priv->dsp_crc32, but these fields are never assigned anywhere in the driver (so they’ll always read as 0/garbage). Either compute/store the CRC32 values during firmware load/init or remove these fields from the debugfs output.

Copilot · 2026-02-28T08:52:12Z

package/airoha/drivers/an8811/src/mtk/air_en8811h_main.c

+	struct device *dev = phydev_dev(phydev);
+	int ret = 0;
+	u32 pbus_value = 0;
+	struct en8811h_priv *priv = phydev->priv;


en8811h_load_firmware() declares struct en8811h_priv *priv = phydev->priv; but never uses it. Please remove the unused variable (or use it to populate fields like firmware CRCs if that was the intent) to avoid warnings and keep the function clear.

Suggested change

struct en8811h_priv *priv = phydev->priv;

Copilot · 2026-02-28T08:52:12Z

package/airoha/drivers/an8811/document/README.sc.md

+### 1) 固件
+
+驱动通过 `request_firmware_direct()` 加载：
+
+- `EthMD32.dm.bin`
+- `EthMD32.DSP.bin`
+
+确保它们在目标系统 `/lib/firmware/` 中可见。
+


文档写明驱动通过 request_firmware_direct() 从 /lib/firmware/ 加载固件，但当前实现是通过 air_en8811h_fw.h 内嵌固件数据并直接写入（代码中无 request_firmware* 调用）。请同步更新说明（或改回运行时加载固件）。

Copilot AI review requested due to automatic review settings February 28, 2026 08:44

Copilot started reviewing on behalf of Yuzhii0718 February 28, 2026 08:44 View session

Copilot AI reviewed Feb 28, 2026

View reviewed changes

Yuzhii0718 force-pushed the pr-an8811-1.2.5 branch 4 times, most recently from 59aafcd to b64b93c Compare February 28, 2026 13:12

phy: add support for airoha an8811 2.5G PHY

6b03a81

Yuzhii0718 force-pushed the pr-an8811-1.2.5 branch from b64b93c to 6b03a81 Compare February 28, 2026 15:17

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

phy: add support for airoha en8811h 2.5G PHY#422

phy: add support for airoha en8811h 2.5G PHY#422
Yuzhii0718 wants to merge 1 commit intohanwckf:openwrt-21.02from
Yuzhii0718:pr-an8811-1.2.5

Yuzhii0718 commented Feb 28, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Feb 28, 2026

Uh oh!

Copilot AI Feb 28, 2026

Uh oh!

Copilot AI Feb 28, 2026

Uh oh!

Copilot AI Feb 28, 2026

Uh oh!

Copilot AI Feb 28, 2026

Uh oh!

Copilot AI Feb 28, 2026

Uh oh!

Copilot AI Feb 28, 2026

Uh oh!

Copilot AI Feb 28, 2026

Uh oh!

Copilot AI Feb 28, 2026

Uh oh!

Copilot AI Feb 28, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

	KCONFIG:= \
	CONFIG_AIROHA_EN8811H_PHY=m \
	CONFIG_AIROHA_EN8811H_PHY_DEBUGFS=n

Conversation

Yuzhii0718 commented Feb 28, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Feb 28, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 28, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 28, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 28, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 28, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 28, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 28, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 28, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 28, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 28, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants