Re: [PATCH net-next v13 00/13] Introduce PHY listing and link_topology tracking

Hello Marc,

Thanks for giving this a test.

On Wed, 26 Jun 2024 11:47:52 +0200
Marc Kleine-Budde [off-list ref] wrote:

quoted

# ethtool --show-phys *  

This creates the following warning for me:

[   51.877429] ------------[ cut here ]------------
[   51.882094] WARNING: CPU: 0 PID: 333 at lib/refcount.c:31 ref_tracker_free+0x1ac/0x254
[   51.890222] refcount_t: decrement hit 0; leaking memory.
[   51.895611] Modules linked in: mcp251xfd flexcan imx_sdma can_dev spi_imx
[   51.902493] CPU: 0 PID: 333 Comm: ethtool Not tainted 6.10.0-rc4+ #327
[   51.909056] Hardware name: Freescale i.MX6 Quad/DualLite (Device Tree)
[   51.915603] Call trace: 
[   51.915623] [<c0d2cbd0>] (unwind_backtrace) from [<c0109bcc>] (show_stack+0x10/0x14)
[   51.925979] [<c0109bcc>] (show_stack) from [<c0d4744c>] (dump_stack_lvl+0x50/0x64)
[   51.933605] [<c0d4744c>] (dump_stack_lvl) from [<c0d2d2ec>] (__warn+0x88/0xc0)
[   51.940877] [<c0d2d2ec>] (__warn) from [<c0120ba0>] (warn_slowpath_fmt+0x1b4/0x1c4)
[   51.948590] [<c0120ba0>] (warn_slowpath_fmt) from [<c0697f74>] (ref_tracker_free+0x1ac/0x254)
[   51.957176] [<c0697f74>] (ref_tracker_free) from [<c0ae4b7c>] (ethnl_phy_done+0x24/0x54)
[   51.965318] [<c0ae4b7c>] (ethnl_phy_done) from [<c0acda68>] (genl_done+0x3c/0x88)
[   51.972845] [<c0acda68>] (genl_done) from [<c0ac9b6c>] (netlink_dump+0x2d8/0x3d0)
[   51.980387] [<c0ac9b6c>] (netlink_dump) from [<c0aca2fc>] (__netlink_dump_start+0x1f4/0x2c4)
[   51.988889] [<c0aca2fc>] (__netlink_dump_start) from [<c0acd7bc>] (genl_family_rcv_msg+0x140/0x328)
[   51.997989] [<c0acd7bc>] (genl_family_rcv_msg) from [<c0acd9e8>] (genl_rcv_msg+0x44/0x88)
[   52.006204] [<c0acd9e8>] (genl_rcv_msg) from [<c0acc554>] (netlink_rcv_skb+0xb8/0x118)
[   52.014157] [<c0acc554>] (netlink_rcv_skb) from [<c0acd038>] (genl_rcv+0x20/0x34)
[   52.021673] [<c0acd038>] (genl_rcv) from [<c0acbd24>] (netlink_unicast+0x23c/0x3d0)
[   52.029367] [<c0acbd24>] (netlink_unicast) from [<c0acc044>] (netlink_sendmsg+0x18c/0x3d4)
[   52.037667] [<c0acc044>] (netlink_sendmsg) from [<c0a4b30c>] (__sys_sendto+0xd4/0x128)
[   52.045626] [<c0a4b30c>] (__sys_sendto) from [<c0100080>] (ret_fast_syscall+0x0/0x54)
[   52.053496] Exception stack(0xc3967fa8 to 0xc3967ff0)
[   52.058576] 7fa0:                   b6f1130c 0000000c 00000003 015dd238 00000018 00000000
[   52.066780] 7fc0: b6f1130c 0000000c b6fb6700 00000122 00571000 00000001 0052a2f8 015dd190
[   52.074978] 7fe0: 00000122 bec7cf38 b6ea847d b6e1fe86
[   52.080184] ---[ end trace 0000000000000000 ]---

Hmm I've never seen this one before, but I could be missing some debug
options. Looks like my ethnl_parse_header_dev_put() doesn't belong in
the ethnl_phy_done() callback. I'll see if I can reproduce this error
on some setups, and address that in the next iteration.

While a "ethtool --show-phys lan0" works w/o problems.

Thanks a lot for the test,

Maxime