TP-Link Archer C2600: rcu_sched detected stalls on CPUs/tasks

A TP-Link Archer C2600 running OpenWrt 18.06.1 r7258-5eb055306f (official image) stopped responding to LuCI and other requests after six days of uptime. The logs show the following:

Dec 18 17:28:52 kernel: [534990.381013] INFO: rcu_sched detected stalls on CPUs/tasks:
Dec 18 17:28:52 kernel: [534990.381068]         1-...: (0 ticks this GP) idle=216/140000000000000/0 softirq=3851165/3851165 fqs=1050
Dec 18 17:28:52 kernel: [534990.385405]         (detected by 0, t=2102 jiffies, g=3271918, c=3271917, q=58)
Dec 18 17:28:52 kernel: [534990.394438] Sending NMI from CPU 0 to CPUs 1:
Dec 18 17:28:52 kernel: [535000.405370] ath10k_pci 0000:01:00.0: SWBA overrun on vdev 0, skipped old beacon
Dec 18 17:28:52 kernel: [535000.405418] ath10k_pci 0000:01:00.0: SWBA overrun on vdev 0, skipped old beacon
Dec 18 17:28:52 kernel: [535000.411910] ath10k_pci 0000:01:00.0: SWBA overrun on vdev 0, skipped old beacon
Dec 18 17:28:52 kernel: [535000.419233] ath10k_pci 0000:01:00.0: SWBA overrun on vdev 0, skipped old beacon
Dec 18 17:28:52 kernel: [535000.426689] ath10k_pci 0000:01:00.0: SWBA overrun on vdev 0, skipped old beacon
Dec 18 17:28:52 kernel: [535000.434073] ath10k_pci 0000:01:00.0: SWBA overrun on vdev 0, skipped old beacon
Dec 18 17:28:52 kernel: [535000.441436] ath10k_pci 0000:01:00.0: SWBA overrun on vdev 0, skipped old beacon
Dec 18 17:28:52 kernel: [535000.448748] ath10k_pci 0000:01:00.0: SWBA overrun on vdev 0, skipped old beacon
Dec 18 17:28:52 kernel: [535000.456194] ath10k_pci 0000:01:00.0: SWBA overrun on vdev 0, skipped old beacon
Dec 18 17:28:52 kernel: [535000.463563] ath10k_pci 0000:01:00.0: SWBA overrun on vdev 0, skipped old beacon
Dec 18 17:29:14 kernel: [535021.901784] INFO: rcu_sched detected expedited stalls on CPUs/tasks: { 1-... } 2106 jiffies s: 2353 root: 0x2/.
Dec 18 17:29:14 kernel: [535021.901846] blocking rcu_node structures:
Dec 18 17:29:14 kernel: [535021.911046] Task dump for CPU 1:
Dec 18 17:29:14 kernel: [535021.915073] sh              R  running task        0 18914  18912 0x00000002
Dec 18 17:29:14 kernel: [535021.918349] Function entered at [<c07bbf18>] from [<befff000>]
Dec 18 17:29:55 kernel: [535053.432426] INFO: rcu_sched detected stalls on CPUs/tasks:
Dec 18 17:29:55 kernel: [535053.432480]         1-...: (0 ticks this GP) idle=216/140000000000000/0 softirq=3851165/3851165 fqs=3699
Dec 18 17:29:55 kernel: [535053.436816]         (detected by 0, t=8407 jiffies, g=3271918, c=3271917, q=94)
Dec 18 17:29:55 kernel: [535053.445848] Sending NMI from CPU 0 to CPUs 1:
Dec 18 17:29:55 kernel: [535063.456546] ath10k_warn: 72 callbacks suppressed

The ath10k_pci messages might be a result of the problem, not a cause, but I'm not sure. There were no signs of trouble before the rcu_sched messages began.

On startup, the router runs the following to improve throughput:

echo 35 > /sys/devices/system/cpu/cpufreq/ondemand/up_threshold
echo 10 > /sys/devices/system/cpu/cpufreq/ondemand/sampling_down_factor

Otherwise, the setup is pretty standard as a dumb AP.

Does anyone have some suggestions on how to fix the problem or diagnostics to try?

Familiar message, but not one that I recall running into except here

https://forum.openwrt.org/search?q=rcu_sched