nanobsd g_vfs_write md1 error, /var memory exhaust

[eyebone]

Hello folks,

During testing my nanobsd image with 8.2 kernel on i386 I recognized that filling the /var with data
[cmd=]dd if=/dev/urandom of=/var/test bs=64k[/cmd]
exhausts the system, it is not usable any more and just prints out:

g_vfs_done():md1[WRITE(offset=72990720, length=131072)]error = 28
g_vfs_done():md1[WRITE(offset=73121792, length=131072)]error = 28
g_vfs_done():md1[WRITE(offset=45363200, length=2048)]error = 28

Problematic as well is that the /var path could be filled as well normally by syslogd messages for instance.

Anyone know this bug or how to mitigate it? Did I miss any important nanobsd config?

best regards,
eyebone

 

[aragon]

I've filled my /var occasionally, but I've never run into those kernel errors you're seeing. On NanoBSD, /var should be a malloc backed md(4) device. Can't imagine how filesystem errors could occur there...