Re: [PATCH] net: hinic: avoid kernel hung in hinic_get_stats64()

From: Paolo Abeni
Date: Thu Jun 30 2022 - 05:57:01 EST


On Wed, 2022-06-29 at 15:28 +0800, Qiao Ma wrote:
> When using hinic device as a bond slave device, and reading device stats of
> master bond device, the kernel may hung.
>
> The kernel panic calltrace as follows:
> Kernel panic - not syncing: softlockup: hung tasks
> Call trace:
> native_queued_spin_lock_slowpath+0x1ec/0x31c
> dev_get_stats+0x60/0xcc
> dev_seq_printf_stats+0x40/0x120
> dev_seq_show+0x1c/0x40
> seq_read_iter+0x3c8/0x4dc
> seq_read+0xe0/0x130
> proc_reg_read+0xa8/0xe0
> vfs_read+0xb0/0x1d4
> ksys_read+0x70/0xfc
> __arm64_sys_read+0x20/0x30
> el0_svc_common+0x88/0x234
> do_el0_svc+0x2c/0x90
> el0_svc+0x1c/0x30
> el0_sync_handler+0xa8/0xb0
> el0_sync+0x148/0x180
>
> And the calltrace of task that actually caused kernel hungs as follows:
> __switch_to+124
> __schedule+548
> schedule+72
> schedule_timeout+348
> __down_common+188
> __down+24
> down+104
> hinic_get_stats64+44 [hinic]
> dev_get_stats+92
> bond_get_stats+172 [bonding]
> dev_get_stats+92
> dev_seq_printf_stats+60
> dev_seq_show+24
> seq_read_iter+964
> seq_read+220
> proc_reg_read+164
> vfs_read+172
> ksys_read+108
> __arm64_sys_read+28
> el0_svc_common+132
> do_el0_svc+40
> el0_svc+24
> el0_sync_handler+164
> el0_sync+324
>
> When getting device stats from bond, kernel will call bond_get_stats().
> It first holds the spinlock bond->stats_lock, and then call
> hinic_get_stats64() to collect hinic device's stats.
> However, hinic_get_stats64() calls `down(&nic_dev->mgmt_lock)` to
> protect its critical section, which may schedule current task out.
> And if system is under high pressure, the task cannot be woken up
> immediately, which eventually triggers kernel hung panic.
>
> Fixes: edd384f682cc ("net-next/hinic: Add ethtool and stats")
> Signed-off-by: Qiao Ma <mqaio@xxxxxxxxxxxxxxxxx>

Side note: it looks like that after this patch every section protected
by the mgmt_lock is already under rtnl lock protection, so you could
probably remove the hinic specific lock (in a separate, net-next,
patch).

Please double check the above as I skimmed upon that quickly.

Thanks,

Paolo