kernel family filogic - update edge to 6.18 #9275
Conversation
|
Important Review skippedDraft detected. Please check the settings in the CodeRabbit UI or the You can disable this status message by setting the 📝 WalkthroughWalkthroughEdge kernel configuration for the filogic family is updated to target Linux 6.18 and pushes a filogic-6.18 patch archive containing many kernel patches: DT additions/overlays, new drivers (PCS/USXGMII, rng, mtdrw, clocks, pinctrl, etc.), kernel build script tweaks, and multiple device-tree updates for BPI-R4 / MT7988/MT7987 platforms. Changes
Sequence Diagram(s)sequenceDiagram
participant Build as Armbian Build
participant Repo as Kernel Repo (frank-w / frank-w or tabrisnet)
participant Patch as Patch Archive (filogic-6.18)
participant DT as DTS/Overlay Handler
participant Kernel as Kernel Build System
participant Artifacts as Output (DTBs, images)
Build->>Repo: fetch kernel source (KERNELSOURCE/KERNELBRANCH)
Build->>Patch: load 0000.patching_config.yaml and series
Patch->>Repo: apply patches (series -> individual patch files)
Patch->>DT: copy dts & overlays per patching config (dts/overlays -> tree)
DT->>Repo: integrate DTB Makefile changes and new dtbs/dtbo
Build->>Kernel: invoke build (make -j${numproc}, bindeb-pkg)
Kernel->>Artifacts: produce kernel images, DTBs, deb packages
Estimated code review effort🎯 5 (Critical) | ⏱️ ~120 minutes Possibly related PRs
Suggested reviewers
Poem
🚥 Pre-merge checks | ✅ 3✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
github-actions
bot
added
size/medium
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 4
Note
Due to the large number of review comments, Critical severity comments were prioritized as inline comments.
🤖 Fix all issues with AI agents
In
`@patch/kernel/archive/filogic-6.18/frank-w/0050-net-dsa-mxl862xx-add-ds-mux.patch`:
- Around line 253-271: The of_node_put(mux_np) is being called inside the
for_each_available_child_of_node loop which releases the parent node reference
mid-iteration (use-after-free); move the of_node_put(mux_np) call to immediately
after the loop (outside and after the for_each_available_child_of_node block) so
the parent reference is held for the whole iteration over children (ensure
mux_np is checked for NULL before the loop and always put after the loop
completes, including when ds_add_mux returns an error).
In
`@patch/kernel/archive/filogic-6.18/frank-w/0053-net-phy-as21-try-the-driver-from-mtk-sdk.patch`:
- Around line 1298-1326: Replace C stdint types and missing headers in
as21xxx.h: change all uint8_t to kernel type u8 in struct downshift_cfg, struct
an_mdi_cfg and struct as21xxx_priv; add the appropriate kernel headers at the
top of the file—include <linux/types.h> for u8, <linux/mutex.h> for struct
mutex, and <linux/dcache.h> (or <linux/fs.h> if preferred) for struct dentry—so
the definitions for ipc_lock and debugfs_root are properly declared.
In
`@patch/kernel/archive/filogic-6.18/frank-w/0058-dt-bindings-net-pcs-mediatek-sgmiisys-add-phys-and-r.patch`:
- Around line 37-45: The YAML schema for the resets property is incorrectly
using a nested "items: - items:" structure; update the "resets" entry (symbol:
resets) to describe a phandle-array/flat list instead of nested objects—either
change it to a single items list where each entry has a single description like
"Phandle to reset controller with reset ID" or use the "phandle-array" type if
supported, so each array element maps to a phandle+specifier tuple rather than
an object with nested items.
In
`@patch/kernel/archive/filogic-6.18/frank-w/0078-net-mtk-eth-mux-fix-rcu_lock-in-mux.patch`:
- Around line 69-120: The phylink creation error path jumps to out_unlock
causing an unmatched rtnl_unlock() because rtnl_lock() wasn't taken there;
change the failure branch in the mtk_mux_create_phylink error handling to "goto
reschedule" instead of "goto out_unlock", then remove the out_unlock label and
its rtnl_unlock() so the only rtnl_unlock()/rtnl_lock() pairs remain around the
mtk_open() call; ensure references to mux->data[new_channel]->phylink,
mtk_mux_create_phylink, reschedule, and rtnl_lock/unlock are updated
accordingly.
🟠 Major comments (22)
patch/kernel/archive/filogic-6.18/frank-w/0050-net-dsa-mxl862xx-add-ds-mux.patch-172-179 (1)
172-179: Uninitializedsfp_present_channelif DT property is missing.
of_property_read_u32return value is not checked. If thesfp-present-channelproperty is absent,mux->sfp_present_channelremains uninitialized, causing undefined behavior on line 178.Proposed fix
+ mux->sfp_present_channel = 0; /* default */ of_property_read_u32(np, "sfp-present-channel", &mux->sfp_present_channel);patch/kernel/archive/filogic-6.18/frank-w/0050-net-dsa-mxl862xx-add-ds-mux.patch-52-70 (1)
52-70: Race condition: state updates after lock release.
mux->channelandmux->initializedare updated (lines 69-70) afterrtnl_unlock(). If another thread or reschedule reads these fields before they're written, it may see stale values. Move the state updates inside the critical section.Proposed fix
gpiod_set_value_cansleep(mux->chan_sel_gpio, new_channel); + mux->channel = new_channel; + mux->initialized = true; + rtnl_unlock(); - mux->channel = new_channel; - mux->initialized = true; - reschedule:patch/kernel/archive/filogic-6.18/frank-w/0050-net-dsa-mxl862xx-add-ds-mux.patch-181-188 (1)
181-188: Incomplete cleanup on channel add failure.If
ds_add_mux_channelfails for the second child, the first channel'smux->data[0](including its phylink) is leaked. The error path should destroy any already-added channels.Proposed fix
for_each_child_of_node(np, child) { err = ds_add_mux_channel(mux, child); if (err) { dev_err(priv->dev, "failed to add ds_mux\n"); of_node_put(child); - goto err_put_chan_sel; + goto err_release_channels; } } INIT_DELAYED_WORK(&mux->sfp_monitor_work, sfp_monitor_work_func); mod_delayed_work(system_wq, &mux->sfp_monitor_work, msecs_to_jiffies(3000)); return 0; +err_release_channels: + for (int i = 0; i < 2; i++) { + if (mux->data[i]) { + if (mux->data[i]->phylink) + phylink_destroy(mux->data[i]->phylink); + kfree(mux->data[i]); + } + } err_put_chan_sel:patch/kernel/archive/filogic-6.18/frank-w/0092-net-pcs-mtk_lynxi-add-mt7987-support.patch-22-27 (1)
22-27: Resource leak:of_parse_phandle()reference not released.
of_parse_phandle()returns a reference-counteddevice_node*that must be released withof_node_put(). This code never releases the reference, creating a resource leak.Additionally, there's a kernel coding style violation: if one branch of an if-else has braces, both should.
Use
of_property_present()instead, which returns a bool without requiring reference management:Proposed fix
- if (of_parse_phandle(np->parent, "resets", 0)) { + if (of_property_present(np->parent, "resets")) { mpcs->rstc = of_reset_control_get_shared(np->parent, NULL); if (IS_ERR(mpcs->rstc)) return PTR_ERR(mpcs->rstc); - } else + } else { mpcs->rstc = NULL; + }patch/kernel/archive/filogic-6.18/frank-w/0059-dt-bindings-net-pcs-add-bindings-for-MediaTek-USXGMI.patch-45-69 (1)
45-69: Add#pcs-cellsto the USXGMII PCS binding.
additionalProperties: falseplus the DTS adding#pcs-cells = <0>will fail schema validation unless the binding explicitly allows it.✅ Proposed fix
properties: $nodename: pattern: "^pcs@[0-9a-f]+$" compatible: const: mediatek,mt7988-usxgmiisys + "#pcs-cells": + const: 0 + reg: maxItems: 1patch/kernel/archive/filogic-6.18/frank-w/0006-kbuild-deb-pkg-add-kernels-for-bpi-boards-to-deb-pac.patch-20-41 (1)
20-41: Setsource_image_path/installed_image_pathfor bpi‑r64 and bpi‑r2proThese branches fall back to
./$board.itb, causing the finalcp "${source_image_path}"to fail since kernel .itb files are not generated for these boards. For bpi‑r2pro you already copyImage.gztoextlinux, so the paths must match that; for bpi‑r64 you have the sameuImagestructure as bpi‑r2 but forgot to set the paths.🛠️ Proposed fix
bpi-r64) mkdir -p "${pdir}/boot/bananapi/$board/linux/dtb" DTBFILE=arch/arm64/boot/dts/mediatek/mt7622-bananapi-bpi-r64.dtb cp ${srctree}/uImage_nodt "${pdir}/boot/bananapi/$board/linux/uImage-${KERNELRELEASE}_nodt" + source_image_path="${srctree}/uImage" + installed_image_path="boot/bananapi/$board/linux/uImage-${KERNELRELEASE}" ;; bpi-r2pro) DTBFILE=arch/arm64/boot/dts/rockchip/rk3568-bpi-r2-pro.dtb mkdir -p ${pdir}/boot/extlinux/ - cp arch/arm64/boot/Image.gz "${pdir}/boot/extlinux/Image-${KERNELRELEASE}.gz" + source_image_path="${srctree}/arch/arm64/boot/Image.gz" + installed_image_path="boot/extlinux/Image-${KERNELRELEASE}.gz"For bpi‑r3/bpi‑r4, verify that
./$board.itbis actually generated as a kernel artifact during the build; if those boards only use U-Boot .itb, set explicit paths there as well.patch/kernel/archive/filogic-6.18/frank-w/0053-net-phy-as21-try-the-driver-from-mtk-sdk.patch-342-436 (1)
342-436: IPC receive can wrap errors + IPC transaction is no longer serialized.
aeon_ipc_rcv_msgstores negative reads into an unsignedsize, which then returns a huge positive value. Also,ipc_locknow covers send and receive separately, so another thread can interleave a command between them. That can corrupt IPC responses.🐛 Proposed fix for error wrapping
-static int aeon_ipc_rcv_msg(struct phy_device *phydev, - u16 ret_sts, u16 *data) +static int aeon_ipc_rcv_msg(struct phy_device *phydev, + u16 ret_sts, u16 *data) { struct as21xxx_priv *priv = phydev->priv; - unsigned int size; + int size; int ret; int i; @@ - if (ret < 0) { - size = ret; - goto out; - } + if (ret < 0) { + size = ret; + goto out; + } data[i] = ret; } out: mutex_unlock(&priv->ipc_lock); return size; }Please also ensure callers serialize the entire IPC transaction (send + receive) with
ipc_lockor move the lock to a wrapper that spans both steps to prevent interleaving.patch/kernel/archive/filogic-6.18/frank-w/0053-net-phy-as21-try-the-driver-from-mtk-sdk.patch-290-333 (1)
290-333: Fix IPC polling error handling and signedness.
valisunsigned int, but you compare it to< 0and use negative error codes. That check never triggers, and polling failures are logged but not returned.🐛 Proposed fix
- unsigned int val; + int val; @@ - ret = read_poll_timeout(aeon_cl45_read, val, + ret = read_poll_timeout(aeon_cl45_read, val, (FIELD_GET(AEON_IPC_STS_PARITY, val) == curr_parity && (val & AEON_IPC_STS_STATUS) != AEON_IPC_STS_STATUS_RCVD && (val & AEON_IPC_STS_STATUS) != AEON_IPC_STS_STATUS_PROCESS && (val & AEON_IPC_STS_STATUS) != AEON_IPC_STS_STATUS_BUSY) || (val < 0), 10000, 2000000, false, phydev, MDIO_MMD_VEND1, VEND1_IPC_STS); - if (val < 0) - ret = val; - - if (ret) - phydev_err(phydev, "%s fail to polling status failed: %d\n", __func__, ret); + if (ret) + return ret; + if (val < 0) + return val; *ret_sts = val;patch/kernel/archive/filogic-6.18/frank-w/0053-net-phy-as21-try-the-driver-from-mtk-sdk.patch-450-476 (1)
450-476: Ensure firmware version string is bounded and NUL‑terminated.Line 475 logs
ret_dataas a string without ensuring termination. That can read past the buffer and leak memory.🐛 Proposed fix
- u16 ret_data[8], data[1]; - u16 ret_sts; + u16 ret_data[8], data[1]; + u16 ret_sts; + char fw_version[AEON_IPC_DATA_MAX + 1]; @@ - ret = aeon_ipc_rcv_msg(phydev, ret_sts, ret_data); + ret = aeon_ipc_rcv_msg(phydev, ret_sts, ret_data); if (ret < 0) return ret; - - phydev_info(phydev, "Firmware Version: %s\n", (char *)ret_data); + memcpy(fw_version, ret_data, min_t(int, ret, AEON_IPC_DATA_MAX)); + fw_version[min_t(int, ret, AEON_IPC_DATA_MAX)] = '\0'; + phydev_info(phydev, "Firmware Version: %s\n", fw_version);patch/kernel/archive/filogic-6.18/frank-w/0053-net-phy-as21-try-the-driver-from-mtk-sdk.patch-170-228 (1)
170-228: Propagate failures fromaeon_set_default_value.Line 186/192 set
reton error, but Line 227 always returns 0, silently masking failures. This hides allocation/overflow problems and can leave the device in a partial state.🐛 Proposed fix
remaining = byte_count - pos; if (remaining > 0) { - if (wdata_count + 2 <= MAX_WDATA_SIZE) { + if (wdata_count + 2 <= MAX_WDATA_SIZE) { // Here we just need padded_bytes once, otherwise we need to read from mem memcpy(padded_bytes, &bytebuf[pos], remaining); wdata[wdata_count++] = le16_to_cpu(*(unsigned short *)&padded_bytes[0]); wdata[wdata_count++] = le16_to_cpu(*(unsigned short *)&padded_bytes[2]); + } else { + pr_err("wdata array overflow\n"); + ret = -ENOSPC; + goto cleanup; } } @@ cleanup: kfree(wdata); - return 0; + return ret;patch/kernel/archive/filogic-6.18/frank-w/0035-net-pcs-pcs-mtk-lynxi-add-platform-driver-for-MT7988.patch-173-210 (1)
173-210: Add error handling and rollback tomtk_pcs_lynxi_enable().The function ignores return values from
clk_prepare_enable()andphy_power_on(), both of which can fail with negative errno codes in the kernel API. Failures are silently masked, potentially leaving the PCS partially enabled. Proper error propagation and rollback are needed.🛠️ Suggested fix with error propagation and rollback
static int mtk_pcs_lynxi_enable(struct phylink_pcs *pcs) { struct mtk_pcs_lynxi *mpcs = pcs_to_mtk_pcs_lynxi(pcs); + int ret; if (mpcs->sgmii_tx && mpcs->sgmii_rx) { - clk_prepare_enable(mpcs->sgmii_rx); - clk_prepare_enable(mpcs->sgmii_tx); + ret = clk_prepare_enable(mpcs->sgmii_rx); + if (ret) + return ret; + ret = clk_prepare_enable(mpcs->sgmii_tx); + if (ret) { + clk_disable_unprepare(mpcs->sgmii_rx); + return ret; + } } - if (mpcs->xfi_tphy) - phy_power_on(mpcs->xfi_tphy); + if (mpcs->xfi_tphy) { + ret = phy_power_on(mpcs->xfi_tphy); + if (ret) { + if (mpcs->sgmii_tx && mpcs->sgmii_rx) { + clk_disable_unprepare(mpcs->sgmii_tx); + clk_disable_unprepare(mpcs->sgmii_rx); + } + return ret; + } + } return 0; }patch/kernel/archive/filogic-6.18/frank-w/0042-net-dsa-add-tag-formats-for-MaxLinear-switches.patch-213-225 (1)
213-225: Fix magic-tag validation logic (&&→||).The current check only rejects packets when both bytes are wrong. It should reject when either byte mismatches.
🐛 Proposed fix
- if ((mxl862_tag[0] != 0x88) && (mxl862_tag[1] != 0xc3)) { + if ((mxl862_tag[0] != 0x88) || (mxl862_tag[1] != 0xc3)) {patch/kernel/archive/filogic-6.18/frank-w/0046-net-dsa-mxl862xx-Support-serdes1-as-a-lan-port.patch-755-771 (1)
755-771: Port translation error handling returns valid port number.Both
mxl862_dsa_port_to_tag_port()andmxl862_tag_port_to_dsa_port()return0on error, which could be confused with a valid port. In the RX path, this could cause packets to be attributed to the wrong interface. Consider returning a negative value or using a clearly invalid sentinel:static int mxl862_dsa_port_to_tag_port(const int in_dsa_port) { if (in_dsa_port < 15) return in_dsa_port + 1; dev_err_ratelimited(NULL, "%s Wrong in_dsa_port value: %d\n", __FILE__, in_dsa_port); - return 0; + return -EINVAL; } static int mxl862_tag_port_to_dsa_port(const int in_hw_port) { if (in_hw_port >= 1 && in_hw_port <= 15) return in_hw_port - 1; dev_err_ratelimited(NULL, "%s Wrong in_hw_port value: %d\n", __FILE__, in_hw_port); - return 0; + return -EINVAL; }The callers would then need to check for negative return values. Alternatively, if the error case truly should never happen in production, a
WARN_ON_ONCEmight be more appropriate to catch bugs during development.patch/kernel/archive/filogic-6.18/frank-w/0033-net-phylink-support-late-PCS-provider-attach.patch-46-52 (1)
46-52: Missing notifier unregistration risks UAF on phylink teardown.Each phylink instance registers
fwnode_pcs_nbbut there’s no corresponding unregister on destroy, leaving a dangling notifier entry that can call into freed memory on later provider-add events. Please add an unregister API and invoke it inphylink_destroy(and any failure paths after registration).🛠️ Suggested fix
diff --git a/drivers/net/pcs/pcs.c b/drivers/net/pcs/pcs.c @@ int register_fwnode_pcs_notifier(struct notifier_block *nb) { return blocking_notifier_chain_register(&fwnode_pcs_notify_list, nb); } EXPORT_SYMBOL_GPL(register_fwnode_pcs_notifier); + +int unregister_fwnode_pcs_notifier(struct notifier_block *nb) +{ + return blocking_notifier_chain_unregister(&fwnode_pcs_notify_list, nb); +} +EXPORT_SYMBOL_GPL(unregister_fwnode_pcs_notifier); diff --git a/include/linux/pcs/pcs.h b/include/linux/pcs/pcs.h @@ int register_fwnode_pcs_notifier(struct notifier_block *nb); +int unregister_fwnode_pcs_notifier(struct notifier_block *nb); @@ static int register_fwnode_pcs_notifier(struct notifier_block *nb) { return -EOPNOTSUPP; } +static int unregister_fwnode_pcs_notifier(struct notifier_block *nb) +{ + return -EOPNOTSUPP; +} diff --git a/drivers/net/phy/phylink.c b/drivers/net/phy/phylink.c @@ void phylink_destroy(struct phylink *pl) { + if (!phy_interface_empty(pl->config->pcs_interfaces)) + unregister_fwnode_pcs_notifier(&pl->fwnode_pcs_nb); ... }Also applies to: 173-176
patch/kernel/archive/filogic-6.18/frank-w/0033-net-phylink-support-late-PCS-provider-attach.patch-71-92 (1)
71-92: Release fwnode references while scanning PCS handles.Each loop iteration acquires a
pcsspec.fwnodereference but never releases it when the handle doesn’t match, leaking refs. Addfwnode_handle_put()before continuing/breaking.🛠️ Suggested fix
while (true) { ret = fwnode_parse_pcsspec(fwnode, i, NULL, &pcsspec); if (ret) break; if (pcsspec.fwnode == pcs_fwnode) - break; + goto out_put; +out_put: + fwnode_handle_put(pcsspec.fwnode); i++; } return fwnode_pcs_get(fwnode, i);patch/kernel/archive/filogic-6.18/frank-w/0031-net-phylink-add-phylink_release_pcs-to-externally-re.patch-46-92 (1)
46-92: Fix lifecycle issue:phylink_release_pcs()cannot be called afterphylink_stop().When
phylink_stop()setspcs->phylink = NULLfor all PCS in the list without removing them, a subsequent call tophylink_release_pcs()performs an early return sincepl = pcs->phylinkis now NULL. Thelist_del(&pcs->list)never executes, leaving the stale PCS entry inpl->pcs_list. On the nextphylink_start(), the PCS is re-attached via the list iteration, violating the release operation's contract.Either ensure
phylink_release_pcs()is always called beforephylink_stop(), or track the PCS-phylink association independently of thepcs->phylinkpointer so list removal is always possible.patch/kernel/archive/filogic-6.18/frank-w/0036-net-pcs-add-driver-for-MediaTek-USXGMII-PCS.patch-396-403 (1)
396-403: Propagate PHY power‑on failures.
phy_power_on()returns anint(0 on success, negative errno on failure), but the return value is ignored, allowing silent failures. Themtk_usxgmii_pcs_enable()function signature supports returning errors; propagate them to phylink.Suggested fix
static int mtk_usxgmii_pcs_enable(struct phylink_pcs *pcs) { struct mtk_usxgmii_pcs *mpcs = pcs_to_mtk_usxgmii_pcs(pcs); + int ret; - phy_power_on(mpcs->xfi_tphy); + ret = phy_power_on(mpcs->xfi_tphy); + if (ret) + return ret; return 0; }patch/kernel/archive/filogic-6.18/frank-w/0001-add-mtd-rw-driver.patch-505-582 (1)
505-582: Fix unlocked bitmask shifts for partitions ≥ 32.Line 554 and Line 579 use
1 << iwithiup to 63. This is undefined for 32-bit shifts and will mis-track unlocked partitions beyond 31. UseBIT_ULL(i)or1ULL << i.🛠️ Suggested fix
@@ - unlocked |= (1 << i); + unlocked |= (1ULL << i); @@ - if (unlocked & (1 << i)) { + if (unlocked & (1ULL << i)) {patch/kernel/archive/filogic-6.18/frank-w/0002-add-regs-command-from-mtk-sdk-and-add-crosscompile-s.patch-300-387 (1)
300-387: HardenregsCLI validation and munmap size.Line 366 onward reads
argv[4]without ensuringargc >= 5, soregs r <addr> <startbit>(argc=4) will read past argv. Also, Line 386 unmaps onlyMAP_SIZEdespite mapping2 * MAP_SIZEat Line 328.🛠️ Suggested fix
@@ - case 'r': - if (argc == 3) + case 'r': + if (argc == 3) reg_read_bits((uint32_t *)virt_addr, (uint32_t *)map_base, (uint32_t *)(offset & ~MAP_MASK), 0, 32); - else { + else if (argc >= 5) { startbit = strtoul(argv[3], 0, 10); if (startbit > 32) PRINT_ERROR; datalen = strtoul(argv[4], 0, 10); if (datalen > 32) PRINT_ERROR; reg_read_bits((uint32_t *)virt_addr, (uint32_t *)map_base, (uint32_t *)(offset & ~MAP_MASK), startbit, datalen); + } else { + usage(); } goto out; @@ -out: - if(munmap(map_base, MAP_SIZE) == -1) +out: + if(munmap(map_base, 2 * MAP_SIZE) == -1) PRINT_ERROR;patch/kernel/archive/filogic-6.18/frank-w/0001-add-mtd-rw-driver.patch-33-42 (1)
33-42: Gate mtdrw build on CONFIG_MTD_RW (currently unconditional).Lines 41-42 add
mtdrw/unconditionally and the subdir Makefile usesobj-mwith onlyCONFIG_MTDgating. This builds the module even whenCONFIG_MTD_RW=n, which breaks Kconfig intent.🛠️ Suggested fix (align build with Kconfig)
diff --git a/drivers/mtd/Makefile b/drivers/mtd/Makefile @@ -obj-y += mtdrw/ +obj-$(CONFIG_MTD_RW) += mtdrw/ diff --git a/drivers/mtd/mtdrw/Makefile b/drivers/mtd/mtdrw/Makefile @@ -ifdef CONFIG_MTD -obj-m += mtd-rw.o -endif +obj-$(CONFIG_MTD_RW) += mtd-rw.opatch/kernel/archive/filogic-6.18/frank-w/0077-net-ethernet-mtk_eth_soc-support-ethernet-passive-mu.patch-252-270 (1)
252-270: Moveof_node_put(mux_np)out of the child loop.
It's called inside the loop (line 268), releasing the parent node reference on each iteration while still iterating over its children. Eachof_get_child_by_name()call acquires one reference that must be released exactly once. Put it after the loop completes.Proposed fix
for_each_available_child_of_node(mux_np, child) { if (err) dev_err(&pdev->dev, "failed to add mux\n"); - - of_node_put(mux_np); - }; + } + of_node_put(mux_np); }patch/kernel/archive/filogic-6.18/frank-w/0077-net-ethernet-mtk_eth_soc-support-ethernet-passive-mu.patch-161-234 (1)
161-234: Validatesfp-present-channelbounds and clean up partial channel allocations on error.The
sfp_present_channelvalue is used as an index intomux->data[2](a 2-element array) at line 48 of the patch (mux->data[new_channel]), but there is no validation to ensure it's in the valid range [0, 1]. An invalid value causes out-of-bounds access.Additionally, if
mtk_add_mux_channel()fails after allocating one or more channels, the previously allocatedmtk_mux_datastructures and their embeddedphylinkobjects leak, as the error path skips the per-channel cleanup thatmtk_release_mux()provides elsewhere in the patch.🐛 Proposed fix
- struct mtk_mux *mux; - unsigned int id; - int err; + struct mtk_mux *mux; + unsigned int id; + int err, i; @@ - mux = kmalloc(sizeof(struct mtk_mux), GFP_KERNEL); + mux = kzalloc(sizeof(*mux), GFP_KERNEL); @@ - of_property_read_u32(np, "sfp-present-channel", - &mux->sfp_present_channel); + err = of_property_read_u32(np, "sfp-present-channel", + &mux->sfp_present_channel); + if (err || mux->sfp_present_channel > 1) { + dev_err(eth->dev, "invalid sfp-present-channel\n"); + err = -EINVAL; + goto err_put_chan_sel; + } @@ if (err) { dev_err(eth->dev, "failed to add mtk_mux\n"); of_node_put(child); - goto err_put_chan_sel; + goto err_free_channels; } } @@ +err_free_channels: + for (i = 0; i < 2; i++) { + if (mux->data[i]) { + if (mux->data[i]->phylink) + phylink_destroy(mux->data[i]->phylink); + kfree(mux->data[i]); + } + } err_put_chan_sel: gpiod_put(mux->chan_sel_gpio);
🟡 Minor comments (9)
patch/kernel/archive/filogic-6.18/frank-w/0094-net-pcs-mtk_lynxi-make-xfi-tphy-optional.patch-23-28 (1)
23-28: Missingof_node_put()for phandle reference.
of_parse_phandle()returns a node with an incremented reference count. The returned pointer should be released withof_node_put()to avoid a reference leak.Suggested fix
- if (of_parse_phandle(dev->of_node, "phys", 0)) { + struct device_node *phy_np = of_parse_phandle(dev->of_node, "phys", 0); + if (phy_np) { + of_node_put(phy_np); mpcs->xfi_tphy = devm_of_phy_get(mpcs->dev, dev->of_node, NULL); if (IS_ERR(mpcs->xfi_tphy)) return PTR_ERR(mpcs->xfi_tphy); - } else + } else { mpcs->xfi_tphy = NULL; + }patch/kernel/archive/filogic-6.18/frank-w/0050-net-dsa-mxl862xx-add-ds-mux.patch-141-144 (1)
141-144: Unsigned comparison always false.
idis declared asunsigned inton line 133, so the checkid < 0on line 142 is always false. Either changeidtoint(as done inds_add_mux_channel) or remove the negative check.Proposed fix
- unsigned int id; + int id; int err; if (!_id) {patch/kernel/archive/filogic-6.18/frank-w/0082-dt-bindings-pinctrl-add-binding-for-MT7987-SoC.patch-148-152 (1)
148-152: Function name mismatch between enum and conditional.The function enum at line 151 uses
eth, but the conditional at lines 181-184 checks forconst: ethernet. The documentation table (lines 103-145) also uses"ethernet". This inconsistency means:
- The example using
function = "eth"will pass enum validation- But the conditional enforcing allowed groups for
mdc_mdiowon't trigger since it checks forethernetEither change the enum to include
ethernetinstead ofeth, or update the conditional and documentation to useeth.Suggested fix
properties: function: description: A string containing the name of the function to mux to the group. - enum: [eth, flash, i2c, i2s, led, pcie, pcm, pwm, jtag, spi, uart, usb, wdt] + enum: [ethernet, flash, i2c, i2s, led, pcie, pcm, pwm, jtag, spi, uart, usb, wdt]And update the example at line 410:
mdio0_pins: mdio0-pins { mux { - function = "eth"; + function = "ethernet"; groups = "mdc_mdio"; };patch/kernel/archive/filogic-6.18/frank-w/0017-arm64-dts-mediatek-mt7988-Add-devicetree-for-BananaP.patch-100-105 (1)
100-105: Model string missing "Pro" designation.The
modelproperty says "Bananapi BPI-R4" but this is the BPI-R4 Pro variant (4E). The compatible strings correctly includebananapi,bpi-r4-pro-4e, but the model string should match to avoid confusion with the standard BPI-R4.The same issue exists in the 8x variant at lines 122-127.
📝 Suggested fix
For mt7988a-bananapi-bpi-r4-pro-4e.dts:
/ { - model = "Bananapi BPI-R4"; + model = "Bananapi BPI-R4 Pro 4E"; compatible = "bananapi,bpi-r4-pro-4e",For mt7988a-bananapi-bpi-r4-pro-8x.dts:
/ { - model = "Bananapi BPI-R4"; + model = "Bananapi BPI-R4 Pro 8X"; compatible = "bananapi,bpi-r4-pro-8x",patch/kernel/archive/filogic-6.18/frank-w/0053-net-phy-as21-try-the-driver-from-mtk-sdk.patch-1066-1075 (1)
1066-1075: Don’t swallow PHY ID read errors.If
aeon_read_pid()fails, returning 0 treats it as “not a match” and hides real MDIO errors.🐛 Proposed fix
/* AEONSEMI get pid. */ phydev->phy_id = aeon_read_pid(phydev); + if ((int)phydev->phy_id < 0) + return (int)phydev->phy_id; if (phydev->phy_id != PHY_ID_AS21XXX) return 0;patch/kernel/archive/filogic-6.18/frank-w/0046-net-dsa-mxl862xx-Support-serdes1-as-a-lan-port.patch-403-417 (1)
403-417: Potential use of uninitializedcpu_portvariable.If the device tree has no port marked as CPU port,
cpu_portwill be used uninitialized in thedev_info()on line 416. While there's a validation check later in probe (lines 662-666), thisdev_infoexecutes first. Consider initializingcpu_portto an invalid sentinel value:static int mxl862xx_setup(struct dsa_switch *ds) { struct mxl862xx_priv *priv = ds->priv; - unsigned int cpu_port, j; + unsigned int cpu_port = UINT_MAX, j; int ret; u8 i;This way, if no CPU port is found, the log will show an obviously invalid value rather than garbage.
patch/kernel/archive/filogic-6.18/frank-w/0032-net-pcs-implement-Firmware-node-support-for-PCS-driv.patch-238-262 (1)
238-262: Propagate parse errors instead of masking as -ENOENT.
fwnode_phylink_pcs_count()treats any error fromfwnode_property_get_reference_args()as end-of-list, which can silently hide malformedpcs-handledata. Consider propagating non--ENOENTerrors.🛠️ Suggested fix
while (true) { ret = fwnode_property_get_reference_args(fwnode, "pcs-handle", "#pcs-cells", -1, index, &out_args); - /* We expect to reach an -ENOENT error while counting */ - if (ret) - break; + if (ret == -ENOENT) + break; + if (ret) + return ret; fwnode_handle_put(out_args.fwnode); index++; }patch/kernel/archive/filogic-6.18/frank-w/0025-net-ethernet-mtk_eth_soc-Add-RSS-support.patch-388-455 (1)
388-455: Confirm PDMA IRQ naming is complete for MTK_PDMA_INT SoCs.The driver now requires named IRQs
pdma0..pdma3for SoCs with theMTK_PDMA_INTcapability (MT7985, MT7986, MT7987, MT7988). Patch 0106 (MT7987) includes these names, but MT7988 DTS patches (0016, 0017) do not add them to the base ethernet node. Probe will fail unless the MT7988 ethernet node in the kernel's basemt7988a.dtsialready defines these interrupt names; confirm this is the case or add them to patch 0017.patch/kernel/archive/filogic-6.18/frank-w/0077-net-ethernet-mtk_eth_soc-support-ethernet-passive-mu.patch-35-72 (1)
35-72: Handle GPIO read errors before using the value.
gpiod_get_value_cansleep()returns the GPIO logical value (0 or 1) on success, or a negative errno on failure. The current code doesn't check for errors, so a negative return would be treated as truthy in the ternary operator on line 48, potentially triggering an unintended channel switch.Proposed fix
sfp_present = gpiod_get_value_cansleep(mux->mod_def0_gpio); + if (sfp_present < 0) + goto reschedule; new_channel = sfp_present ? mux->sfp_present_channel : !mux->sfp_present_channel;
🧹 Nitpick comments (21)
patch/kernel/archive/filogic-6.18/frank-w/0052-dbg-net-dsa-mxl862xx-extend-information-on-write-err.patch (1)
15-20: Enhanced error logging improves debuggability.The additional diagnostic fields (hardware port, serial port ID, speed, and return code) are useful for troubleshooting interface setup failures.
Minor style note: kernel coding style typically uses spaces after commas in function arguments. This is a nitpick for an edge-branch patch.
Optional: Add spaces for style consistency
- dev_err(dp->ds->dev, "failed to set intf on port %d (hw:%d,ser_port_id:%d) to %d (ret:%d)\n", dp->index,hw_port,ser_intf.port_id,ser_intf.speed,ret); + dev_err(dp->ds->dev, + "failed to set intf on port %d (hw:%d, ser_port_id:%d) to %d (ret:%d)\n", + dp->index, hw_port, ser_intf.port_id, ser_intf.speed, ret);patch/kernel/archive/filogic-6.18/frank-w/0054-net-phy-as21x-disable-the-dcp_ra-call.patch (1)
18-40: Consider using#if 0or preprocessor guards instead of mixed comment styles.The patch uses
/* */block comments for the function definition (lines 18-28) and//single-line comments for the call site (lines 38-39). For temporarily disabling code in kernel sources,#if 0 ...#endif`` is the conventional approach—it's easier to grep for, toggle, and avoids nested comment issues.That said, for an edge branch with experimental code, this is a minor style concern.
💡 Alternative using preprocessor guards
-/*static int aeon_dpc_ra_enable(struct phy_device *phydev) +#if 0 /* DCP RA disabled - see patch 0054 */ +static int aeon_dpc_ra_enable(struct phy_device *phydev) { u16 data[2]; u16 ret_sts; @@ -709,7 +709,7 @@ static int aeon_dpc_ra_enable(struct phy_device *phydev) return aeon_ipc_send_msg(phydev, IPC_CMD_CFG_PARAM, data, sizeof(data), &ret_sts); -}*/ +} +#endifAnd for the call site:
- //if (phydev->interface == PHY_INTERFACE_MODE_USXGMII) - // ret = aeon_dpc_ra_enable(phydev); +#if 0 /* DCP RA disabled */ + if (phydev->interface == PHY_INTERFACE_MODE_USXGMII) + ret = aeon_dpc_ra_enable(phydev); +#endifpatch/kernel/archive/filogic-6.18/frank-w/0008-hwrng-add-driver-for-MediaTek-TRNG-SMC.patch (1)
27-42: Minor typo in Kconfig help text.Line 38 in the patch: "choose M here. the module" — the sentence-initial "the" should be capitalized to "The".
📝 Suggested fix
- To compile this driver as a module, choose M here. the + To compile this driver as a module, choose M here. The module will be called mtk-rng-v2.patch/kernel/archive/filogic-6.18/frank-w/0014-build.sh-Build-DEB-packages-in-parallel.patch (1)
18-19: Guardnumprocbefore using-jto avoid unbounded parallelism.If
numprocis unset/empty,make -jruns unlimited jobs and can overwhelm build hosts. Please confirmnumprocis always initialized before this call, or add a local default and quote it. Line 19 is the only place that assumesnumprocis set.♻️ Suggested hardening
- LOCALVERSION="${gitbranch}" board="$baseboard" KDEB_COMPRESS=gzip make -j${numproc} bindeb-pkg + : "${numproc:=$(nproc)}" + LOCALVERSION="${gitbranch}" board="$baseboard" KDEB_COMPRESS=gzip make -j"${numproc}" bindeb-pkgpatch/kernel/archive/filogic-6.18/frank-w/0038-net-ethernet-mtk_eth_soc-add-more-DMA-monitor-for-MT.patch (1)
123-123: Macro parameterishould be parenthesized.The macro
MTK_GDM_RX_FC_OFFSET(eth, i)usesiwithout parentheses. If a caller passes an expression likeidx + 1, the multiplication binds tighter than addition, causing incorrect evaluation.Suggested fix
-#define MTK_GDM_RX_FC_OFFSET(eth, i) (i * (mtk_is_netsys_v3_or_greater(eth) ? MTK_STAT_OFFSET_V3 : MTK_STAT_OFFSET) + MTK_GDM_RX_FC) +#define MTK_GDM_RX_FC_OFFSET(eth, i) ((i) * (mtk_is_netsys_v3_or_greater(eth) ? MTK_STAT_OFFSET_V3 : MTK_STAT_OFFSET) + MTK_GDM_RX_FC)patch/kernel/archive/filogic-6.18/frank-w/0095-net-pcs-mtk-lynxi-add-phya-tx-rx-clock-path.patch (1)
57-69: Consider documenting the TRX buffer threshold magic numbers.The values
0x3112(default) and0x2111(PHYA clock) are hardware-specific constants from the MediaTek SDK. Adding a brief comment explaining their purpose or defining named constants would improve maintainability.Example documentation
- unsigned int rgc3, sgm_mode, bmcr = 0, trxbuf_thr = 0x3112; + unsigned int rgc3, sgm_mode, bmcr = 0; + /* TRX buffer threshold: default 0x3112, PHYA clock path 0x2111 */ + unsigned int trxbuf_thr = 0x3112;patch/kernel/archive/filogic-6.18/frank-w/0012-firmware-add-firmware-for-mt7987-BPI-R4Lite-phy.patch (1)
6-11: Firmware blobs are redistributable under MediaTek license; consider documenting source for reference.The MT7987 firmware files (i2p5ge-phy-pmb.bin and i2p5ge-phy-DSPBitTb.bin) are included in upstream linux-firmware under the redistributable LICENCE.mediatek and are sourced from MediaTek's OpenWrt feed. No licensing compliance issue exists. For audit clarity, optionally reference the upstream source or firmware version in the patch header or PR notes.
patch/kernel/archive/filogic-6.18/frank-w/0096-net-ethernet-mtk_eth_soc-add-mt7987-support.patch (2)
109-116: Consider using a dedicated capability bit instead of direct caps comparison.The check
eth->soc->caps != MT7988_CAPSworks but is fragile—it implicitly excludes MT7988 by comparing the entire capability bitmask. If additional netsys_v3 SoCs are added in the future, this logic may need revisiting.A cleaner approach would be to introduce a dedicated capability bit (e.g.,
MTK_QTX_SCH_V3) to explicitly indicate which rate register layout to use. However, given this is an edge kernel patch from an upstream vendor tree, the current approach is acceptable.
244-246: Commented-out RSS configuration.The
//rss_num = 4line appears to be development leftover. Note thatMT7987_CAPSalready includesMTK_RSS, so RSS capability is enabled. Consider either removing this commented line or uncommenting it if RSS is intended to be active with 4 queues.patch/kernel/archive/filogic-6.18/frank-w/0083-pinctrl-add-pinctrl-driver-for-mt7987-from-sdk.patch (1)
66-73: Remove duplicatePIN_FIELD_BASEdefinition.
It’s defined twice identically; this is redundant and can trigger -Wmacro-redefined in some toolchains.♻️ Proposed cleanup
-#define PIN_FIELD_BASE(_s_pin, _e_pin, _i_base, _s_addr, _x_addrs, _s_bit, \ - _x_bits) \ - PIN_FIELD_CALC(_s_pin, _e_pin, _i_base, _s_addr, _x_addrs, _s_bit, \ - _x_bits, 32, 0) -#define PIN_FIELD_BASE(_s_pin, _e_pin, _i_base, _s_addr, _x_addrs, _s_bit, \ - _x_bits) \ - PIN_FIELD_CALC(_s_pin, _e_pin, _i_base, _s_addr, _x_addrs, _s_bit, \ - _x_bits, 32, 0) +#define PIN_FIELD_BASE(_s_pin, _e_pin, _i_base, _s_addr, _x_addrs, _s_bit, \ + _x_bits) \ + PIN_FIELD_CALC(_s_pin, _e_pin, _i_base, _s_addr, _x_addrs, _s_bit, \ + _x_bits, 32, 0)patch/kernel/archive/filogic-6.18/frank-w/0018-arm64-dts-mediatek-mt7988a-bpi-r4-pro-Add-PCIe-overl.patch (1)
48-48: Unused include in cn15 overlay.The
dt-bindings/gpio/gpio.hheader is included but no GPIO macros (likeGPIO_ACTIVE_HIGH/LOW) are used in this overlay file. The same applies to cn18.dtso at line 74. This is harmless but could be cleaned up.♻️ Suggested cleanup
-#include <dt-bindings/gpio/gpio.h> - / { compatible = "bananapi,bpi-r4-pro", "mediatek,mt7988a"; };patch/kernel/archive/filogic-6.18/frank-w/0017-arm64-dts-mediatek-mt7988-Add-devicetree-for-BananaP.patch (1)
321-326: Duplicate fan properties.The
&fanreference re-specifiespinctrl-0,pinctrl-names, andpwmsthat are already defined in the originalfan: pwm-fannode (lines 166-174). This duplication is unnecessary since the values are identical.♻️ Suggested simplification
&fan { - pinctrl-0 = <&pwm0_pins>; - pinctrl-names = "default"; - pwms = <&pwm 0 50000>; status = "okay"; };patch/kernel/archive/filogic-6.18/frank-w/0071-arm64-dts-r4pro-add-mxl-switch.patch (1)
145-161: Minor formatting: missing space afterreg=.The switchphy nodes have
reg= <x>instead of the standardreg = <x>formatting with a space before the equals sign.♻️ Suggested formatting fix
- switchphy0:switchphy@0 { - reg= <0>; + switchphy0: switchphy@0 { + reg = <0>; }; - switchphy1:switchphy@1 { - reg= <1>; + switchphy1: switchphy@1 { + reg = <1>; };(Apply similar fixes for switchphy2-4)
patch/kernel/archive/filogic-6.18/frank-w/0070-arm64-dts-r4pro-add-gmac-1-2-and-phys.patch (1)
33-33: Leftover debug comment.The comment
//change 5appears to be a development note that should be removed or clarified before merging.♻️ Suggested fix
- reset-gpios = <&pio 83 GPIO_ACTIVE_LOW>; //change 5 + reset-gpios = <&pio 83 GPIO_ACTIVE_LOW>;patch/kernel/archive/filogic-6.18/frank-w/0035-net-pcs-pcs-mtk-lynxi-add-platform-driver-for-MT7988.patch (1)
302-307: Prefer devm-managed reset control acquisition.
of_reset_control_get_shared()isn’t devm‑managed and there’s no correspondingreset_control_put()on remove/error paths. Considerdevm_reset_control_get_shared()(or_optional_sharedif reset is optional).♻️ Suggested refactor
- mpcs->rstc = of_reset_control_get_shared(np->parent, NULL); + mpcs->rstc = devm_reset_control_get_shared(dev, NULL); if (IS_ERR(mpcs->rstc)) return PTR_ERR(mpcs->rstc);patch/kernel/archive/filogic-6.18/frank-w/0046-net-dsa-mxl862xx-Support-serdes1-as-a-lan-port.patch (3)
124-140: Tag protocol parsing function is functional, minor style nits.The cast on line 128 is unnecessary—
dsa_to_port()already returnsstruct dsa_port *. Also,of_property_read_string()could be cleaner thanof_get_property()for string retrieval. These are optional refinements:- struct dsa_port *dp = (struct dsa_port *)dsa_to_port(ds, port); - const char *user_protocol = NULL; + struct dsa_port *dp = dsa_to_port(ds, port); + const char *user_protocol; - if (dp != NULL) - user_protocol = of_get_property(dp->dn, "dsa-tag-protocol", NULL); - if (user_protocol != NULL) { + if (dp && !of_property_read_string(dp->dn, "dsa-tag-protocol", &user_protocol)) {
227-242: Safety check added for block_id, minor style issue.The added
&& block_idcheck on line 231 prevents attempting to deactivate VLAN filter entries when the block is not allocated (block_id = 0). This is a good defensive fix.Minor style: line 239 is missing spaces around the minus operator.
- block_info->final_filters_idx = block_info->filters_max-1; + block_info->final_filters_idx = block_info->filters_max - 1;
591-615: Mirror state preservation fix is important.Lines 595-596 now properly preserve the existing mirror enable states before calling the API. This fixes a likely bug where only the changed mirror direction was set, potentially disabling the other direction unintentionally.
The brace placement on lines 604-612 doesn't follow kernel coding style (opening brace should be on the same line as
for):- for (i = 0; i < max_ports; i++) - { + for (i = 0; i < max_ports; i++) { if (dsa_is_unused_port(ds, i)) continue;patch/kernel/archive/filogic-6.18/frank-w/0028-net-ethernet-mtk_eth_soc-add-paths-and-SerDes-modes-.patch (1)
122-145: Validate mac_id before defaulting to GMAC3 paths.The ternary fallback routes any unexpected
mac_idto GMAC3, which can silently misconfigure the mux. A small switch with a default-EINVALmakes failures explicit.♻️ Suggested defensive update
int mtk_gmac_usxgmii_path_setup(struct mtk_eth *eth, int mac_id) { u64 path; - path = (mac_id == MTK_GMAC1_ID) ? MTK_ETH_PATH_GMAC1_USXGMII : - (mac_id == MTK_GMAC2_ID) ? MTK_ETH_PATH_GMAC2_USXGMII : - MTK_ETH_PATH_GMAC3_USXGMII; + switch (mac_id) { + case MTK_GMAC1_ID: + path = MTK_ETH_PATH_GMAC1_USXGMII; + break; + case MTK_GMAC2_ID: + path = MTK_ETH_PATH_GMAC2_USXGMII; + break; + case MTK_GMAC3_ID: + path = MTK_ETH_PATH_GMAC3_USXGMII; + break; + default: + return -EINVAL; + } /* Setup proper MUXes along the path */ return mtk_eth_mux_setup(eth, path); } int mtk_gmac_sgmii_path_setup(struct mtk_eth *eth, int mac_id) { u64 path; - path = (mac_id == MTK_GMAC1_ID) ? MTK_ETH_PATH_GMAC1_SGMII : - (mac_id == MTK_GMAC2_ID) ? MTK_ETH_PATH_GMAC2_SGMII : - MTK_ETH_PATH_GMAC3_SGMII; + switch (mac_id) { + case MTK_GMAC1_ID: + path = MTK_ETH_PATH_GMAC1_SGMII; + break; + case MTK_GMAC2_ID: + path = MTK_ETH_PATH_GMAC2_SGMII; + break; + case MTK_GMAC3_ID: + path = MTK_ETH_PATH_GMAC3_SGMII; + break; + default: + return -EINVAL; + } /* Setup proper MUXes along the path */ return mtk_eth_mux_setup(eth, path); }patch/kernel/archive/filogic-6.18/frank-w/0078-net-mtk-eth-mux-fix-rcu_lock-in-mux.patch (2)
135-149: Remove the commented-outphylink_createblock.
If creation is intentionally deferred, keep only the live path to avoid dead code noise.♻️ Cleanup suggestion
- /*phylink = phylink_create(&mux->mac->phylink_config, - of_fwnode_handle(np), - phy_mode, &mtk_phylink_ops); - if (IS_ERR(phylink)) { - dev_err(eth->dev, "failed to create phylink structure\n"); - err = PTR_ERR(phylink); - goto err_free_data; - }*/ @@ - data->phylink = NULL;//phylink; + data->phylink = NULL;
157-171: Alignmux->channelinit with the comment’s intent.
The comment says the initial value should be invalid to force a first switch, but it’s set to0(valid). Either set an out‑of‑range sentinel or drop the comment.
patch/kernel/archive/filogic-6.18/frank-w/0050-net-dsa-mxl862xx-add-ds-mux.patch
Show resolved
Hide resolved
patch/kernel/archive/filogic-6.18/frank-w/0053-net-phy-as21-try-the-driver-from-mtk-sdk.patch
Show resolved
Hide resolved
...archive/filogic-6.18/frank-w/0058-dt-bindings-net-pcs-mediatek-sgmiisys-add-phys-and-r.patch
Show resolved
Hide resolved
patch/kernel/archive/filogic-6.18/frank-w/0078-net-mtk-eth-mux-fix-rcu_lock-in-mux.patch
Show resolved
Hide resolved
tabrisnet
force-pushed
the
tabrisnet_bpir4-6.18
branch
from
4b0bc30 to
0c4f5a0
Compare
…t use them directly
tabrisnet
force-pushed
the
tabrisnet_bpir4-6.18
branch
from
0c4f5a0 to
07f5014
Compare
1 participant
Description
Rebased tree from frank-w's. Note, not recommended for production use as
eth1ð2are missing.eth0's child interfaces are functional.How Has This Been Tested?
Please describe the tests that you ran to verify your changes. Please also note any relevant details for your test configuration.
Checklist:
Please delete options that are not relevant.
Summary by CodeRabbit
Chores
New Features
Bug Fixes
✏️ Tip: You can customize this high-level summary in your review settings.