You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
logread kernel crash - general protection fault in lib.so #4346
Lot's of these still showing up in the console and logread:
Sun Aug 9 13:47:00 2026 kern.info kernel: [509378.242848] traps: logread[5580] general protection fault ip:74833773 sp:7ffcf5fa6332 error:0 in libc.so[69846,5f8321ecb000+54000]
Does this mean that logging is broken in this version?
@ioogithub The kernel isn't crashing, logread is, and the bad data comes from logd.
The trap is musl's asctime_r() stopping on purpose (a hlt, at offset 0x69846 of your build's libc.so) because logread asked it to format a garbage timestamp. Once logd's ring buffer has wrapped around, the pointer to its oldest message can run into stale data, and logd then hands out entries with random sizes and timestamps. That's OpenWrt's FS#4118 (openwrt/openwrt#9100), "logread crashes sometimes", open since 2021. OMR runs logread -l 2 every minute, so it shows up here a lot. Logging itself keeps working: it's that one logread call that dies.
I reproduced it in QEMU with 20 minutes of logread loops plus a logger flood. It happened on the May 6.18 snapshot, on v0.64.2 and on a current develop build. With the fix, a 30-minute run had none: 26,583 logread -l 2 and no trap. The develop build without the fix had 3 in 20 minutes.
The fix for logd is proposed upstream in openwrt/ubox#9, and carried in the build repo in #4391.
Activity
ioogithub commentedon Aug 9, 2026
Lot's of these still showing up in the console and logread:
Sun Aug 9 13:47:00 2026 kern.info kernel: [509378.242848] traps: logread[5580] general protection fault ip:74833773 sp:7ffcf5fa6332 error:0 in libc.so[69846,5f8321ecb000+54000]
Does this mean that logging is broken in this version?
pylaterreur commentedon Oct 7, 2026
@ioogithub The kernel isn't crashing,
logreadis, and the bad data comes fromlogd.The trap is musl's
asctime_r()stopping on purpose (ahlt, at offset 0x69846 of your build's libc.so) becauselogreadasked it to format a garbage timestamp. Once logd's ring buffer has wrapped around, the pointer to its oldest message can run into stale data, and logd then hands out entries with random sizes and timestamps. That's OpenWrt's FS#4118 (openwrt/openwrt#9100), "logread crashes sometimes", open since 2021. OMR runslogread -l 2every minute, so it shows up here a lot. Logging itself keeps working: it's that onelogreadcall that dies.I reproduced it in QEMU with 20 minutes of
logreadloops plus aloggerflood. It happened on the May 6.18 snapshot, on v0.64.2 and on a current develop build. With the fix, a 30-minute run had none: 26,583logread -l 2and no trap. The develop build without the fix had 3 in 20 minutes.The fix for logd is proposed upstream in openwrt/ubox#9, and carried in the build repo in #4391.