Hi Simon,
On 05/11/2024 15:30, Simon Horman wrote:
> On Fri, Nov 01, 2024 at 12:18:51PM +0200, Roger Quadros wrote:
>> flow->irq is initialized to 0 which is a valid IRQ. Set it to -EINVAL
>> in error path of am65_cpsw_nuss_init_rx_chns() so we do not try
>> to free an unallocated IRQ in am65_cpsw_nuss_remove_rx_chns().
>>
>> If user tried to change number of RX queues and am65_cpsw_nuss_init_rx_chns()
>> failed due to any reason, the warning will happen if user tries to change
>> the number of RX queues after the error condition.
>>
>> root@am62xx-evm:~# ethtool -L eth0 rx 3
>> [ 40.385293] am65-cpsw-nuss 8000000.ethernet: set new flow-id-base 19
>> [ 40.393211] am65-cpsw-nuss 8000000.ethernet: Failed to init rx flow2
>> netlink error: Invalid argument
>> root@am62xx-evm:~# ethtool -L eth0 rx 2
>> [ 82.306427] ------------[ cut here ]------------yes.
>> [ 82.311075] WARNING: CPU: 0 PID: 378 at kernel/irq/devres.c:144 devm_free_irq+0x84/0x90
>> [ 82.469770] Call trace:
>> [ 82.472208] devm_free_irq+0x84/0x90
>> [ 82.475777] am65_cpsw_nuss_remove_rx_chns+0x6c/0xac [ti_am65_cpsw_nuss]
>> [ 82.482487] am65_cpsw_nuss_update_tx_rx_chns+0x2c/0x9c [ti_am65_cpsw_nuss]
>> [ 82.489442] am65_cpsw_set_channels+0x30/0x4c [ti_am65_cpsw_nuss]
>> [ 82.495531] ethnl_set_channels+0x224/0x2dc
>> [ 82.499713] ethnl_default_set_doit+0xb8/0x1b8
>> [ 82.504149] genl_family_rcv_msg_doit+0xc0/0x124
>> [ 82.508757] genl_rcv_msg+0x1f0/0x284
>> [ 82.512409] netlink_rcv_skb+0x58/0x130
>> [ 82.516239] genl_rcv+0x38/0x50
>> [ 82.519374] netlink_unicast+0x1d0/0x2b0
>> [ 82.523289] netlink_sendmsg+0x180/0x3c4
>> [ 82.527205] __sys_sendto+0xe4/0x158
>> [ 82.530779] __arm64_sys_sendto+0x28/0x38
>> [ 82.534782] invoke_syscall+0x44/0x100
>> [ 82.538526] el0_svc_common.constprop.0+0xc0/0xe0
>> [ 82.543221] do_el0_svc+0x1c/0x28
>> [ 82.546528] el0_svc+0x28/0x98
>> [ 82.549578] el0t_64_sync_handler+0xc0/0xc4
>> [ 82.553752] el0t_64_sync+0x190/0x194
>> [ 82.557407] ---[ end trace 0000000000000000 ]---
>>
>> Fixes: da70d184a8c3 ("net: ethernet: ti: am65-cpsw: Introduce multi queue Rx")
>
> Hi Roger,
>
> I wonder if the problem predates the cited commit and was, rather,
> introduced by:
>
> Fixes: 24bc19b05f1f ("net: ethernet: ti: am65-cpsw: Add suspend/resume support")
Partly yes. But at that commit AM65_CPSW_MAX_RX_FLOWS is 1 so there is no
opportunity for user to change the number of RX queues.
--
cheers,
-roger
