From: Michael Walle mwalle@kernel.org
[ Upstream commit 09737cb80b8686ffca4ed1805fee745d5c85604d ]
of_get_mac_address() might fetch the MAC address from NVMEM and that driver might not have been loaded. In that case, -EPROBE_DEFER is returned. Right now, this will trigger an immediate fallback to am65_cpsw_am654_get_efuse_macid() possibly resulting in a random MAC address although the MAC address is stored in the referenced NVMEM.
Fix it by handling the -EPROBE_DEFER return code correctly. This also means that the creation of the MDIO device has to be moved to a later stage as -EPROBE_DEFER must not be returned after child devices are created.
Signed-off-by: Michael Walle mwalle@kernel.org Reviewed-by: Andrew Lunn andrew@lunn.ch Link: https://patch.msgid.link/20250414084336.4017237-3-mwalle@kernel.org Signed-off-by: Jakub Kicinski kuba@kernel.org Signed-off-by: Sasha Levin sashal@kernel.org ---
Based on my analysis of this commit and comparison with the similar commits provided, here is my assessment: **YES** This commit should be backported to stable kernel trees for the following reasons: ## Analysis of Code Changes 1. **Proper EPROBE_DEFER Handling**: The commit adds explicit handling for `-EPROBE_DEFER` from `of_get_mac_address()` by checking `if (ret == -EPROBE_DEFER)` and properly propagating this error code up the call stack. This matches the pattern seen in the similar commits marked as "YES" for backporting. 2. **Critical Timing Issue Fix**: The code reordering addresses a fundamental probe ordering problem. Previously, the MDIO device creation happened before MAC address acquisition, but the commit message explicitly states that "-EPROBE_DEFER must not be returned after child devices are created." This reordering moves: - `am65_cpsw_nuss_get_ver(common)` - `am65_cpsw_nuss_init_host_p(common)` - `am65_cpsw_nuss_init_slave_ports(common)` (which contains the MAC address handling) Before the MDIO device creation, ensuring proper defer handling. 3. **Functional Correctness**: Without this fix, when NVMEM backing the MAC address isn't ready, the driver would immediately fall back to `am65_cpsw_am654_get_efuse_macid()` and potentially use a random MAC address, even though a proper MAC address exists in NVMEM but just isn't available yet. ## Comparison with Similar Commits This commit follows the exact same pattern as the "YES" backport commits: - **dm9000**: Added `else if (PTR_ERR(mac_addr) == -EPROBE_DEFER) return ERR_CAST(mac_addr);` - **mv643xx_eth**: Added `ret = of_get_mac_address(pnp, ppd.mac_addr); if (ret) return ret;` - **mtk_eth_soc**: Added `if (err == -EPROBE_DEFER) return err;` ## Stability Assessment 1. **Minimal Risk**: The changes are minimal and surgical - only adding proper error handling for a specific case 2. **No Architectural Changes**: No new features or major architectural modifications 3. **Confined to Single Driver**: Changes are isolated to the TI am65-cpsw driver 4. **Clear Error Path**: The error handling follows established kernel patterns 5. **No Side Effects**: The reordering doesn't change functional behavior, only ensures proper defer semantics ## Backport Suitability Criteria Met ✅ **Fixes Important Bug**: Prevents incorrect MAC address assignment when NVMEM provider isn't ready ✅ **Small and Contained**: Only 13 insertions, 11 deletions across 24 lines ✅ **No New Features**: Pure bugfix for existing functionality ✅ **Minimal Risk**: Well-established pattern used in other drivers ✅ **Critical Subsystem**: Network driver functionality ✅ **Clear Problem Statement**: Commit message clearly explains the issue and solution The commit addresses a real-world scenario where MAC addresses stored in NVMEM (common in embedded systems) would be ignored due to probe timing, leading to random MAC addresses and potential network configuration issues.
drivers/net/ethernet/ti/am65-cpsw-nuss.c | 24 +++++++++++++----------- 1 file changed, 13 insertions(+), 11 deletions(-)
diff --git a/drivers/net/ethernet/ti/am65-cpsw-nuss.c b/drivers/net/ethernet/ti/am65-cpsw-nuss.c index a21e7c0afbfdc..ddda5274f5b2c 100644 --- a/drivers/net/ethernet/ti/am65-cpsw-nuss.c +++ b/drivers/net/ethernet/ti/am65-cpsw-nuss.c @@ -2693,7 +2693,9 @@ static int am65_cpsw_nuss_init_slave_ports(struct am65_cpsw_common *common) goto of_node_put;
ret = of_get_mac_address(port_np, port->slave.mac_addr); - if (ret) { + if (ret == -EPROBE_DEFER) { + goto of_node_put; + } else if (ret) { am65_cpsw_am654_get_efuse_macid(port_np, port->port_id, port->slave.mac_addr); @@ -3586,6 +3588,16 @@ static int am65_cpsw_nuss_probe(struct platform_device *pdev) return ret; }
+ am65_cpsw_nuss_get_ver(common); + + ret = am65_cpsw_nuss_init_host_p(common); + if (ret) + goto err_pm_clear; + + ret = am65_cpsw_nuss_init_slave_ports(common); + if (ret) + goto err_pm_clear; + node = of_get_child_by_name(dev->of_node, "mdio"); if (!node) { dev_warn(dev, "MDIO node not found\n"); @@ -3602,16 +3614,6 @@ static int am65_cpsw_nuss_probe(struct platform_device *pdev) } of_node_put(node);
- am65_cpsw_nuss_get_ver(common); - - ret = am65_cpsw_nuss_init_host_p(common); - if (ret) - goto err_of_clear; - - ret = am65_cpsw_nuss_init_slave_ports(common); - if (ret) - goto err_of_clear; - /* init common data */ ale_params.dev = dev; ale_params.ale_ageout = AM65_CPSW_ALE_AGEOUT_DEFAULT;