Hiroo Ono
2006-Jun-06 10:11 UTC
loader and ifconfig_*_name problem in 6.1-STABLE of 5 June
I upgraded via build/installworld from 6.1-PRERELEASE as of 27 Mars to 6.1-STABLE 5 June, and encounterd 2 problems. a) loader and kernel has some problem that lead to instant reboot upon loading the kernel. the problem was as follows: 1) booting the new kernel with the old loader was fine. I did buildkernel and installkernel and rebooted in single user mode. It had no problems. 2) loading the new kernel with the new loader made the machine reboot. After step (1), I did installworld and rebooted. It resulted in rebooting just after loader menu. I could escape to loader prompt and found that doing > unload > load kernel makes the machine reboot. 3) loading the old kernel with the new loader can boot in single user mode. > unload > set module_path=/boot/kernel.old > load kernel (and some other necessary modules) > boot -s was fine (though, exiting in multi-user mode had problem around mountd and panic'ed with the new userland). I copied the old /boot/loader.old to /boot/loader as a workaround. I have CFLAGS= -O2 -pipe COPTFLAGS= -O2 -pipe and no CPUTYPE setting in /etc/make.conf, so I wonder what caused the problem. b) pre-loading ng_ether.ko module and doing ifconfig name may have problem with ppp. I put the following in /etc/rc.conf ifconfig_sis0_name="net0" and changed device setting in /etc/ppp/ppp.conf as follows: set device PPPoE:net0 This did not work and had the following error message in /var/log/messages: ppp[258]: tun0: Warning: net0: Cannot send a netgraph message: No such file or directory changing device setting to set device PPPoE:sis0 or doing ngctl name sis0: net0 could make ppp work. The problem may be caused by loading ng_ether.ko in loader.conf (I do not remember why I did this, though). If I did not do that, ng_ether module was loaded when ppp was invoked after interface rename and might have no problem. I did not verify this yet, though.
Michael Butler
2006-Jun-06 10:17 UTC
loader and ifconfig_*_name problem in 6.1-STABLE of 5 June
Hiroo Ono wrote:> I upgraded via build/installworld from 6.1-PRERELEASE as of 27 Mars > to 6.1-STABLE 5 June, and encounterd 2 problems.[ .. snip .. ]> I copied the old /boot/loader.old to /boot/loader as a workaround. > I have > CFLAGS= -O2 -pipe > COPTFLAGS= -O2 -pipePlease remove these and re-test. The default is "-O2 -fno-strict-aliasing" which generates slightly different code in some instances, -- Michael Butler, CISSP Security Architect Protected Networks http://www.protected-networks.net
Hiroo Ono
2006-Jun-06 15:44 UTC
loader and ifconfig_*_name problem in 6.1-STABLE of 5 June
At Tue, 06 Jun 2006 13:17:38 -0400, Michael Butler wrote:> > Hiroo Ono wrote: > > I upgraded via build/installworld from 6.1-PRERELEASE as of 27 Mars > > to 6.1-STABLE 5 June, and encounterd 2 problems. > > [ .. snip .. ] > > > I copied the old /boot/loader.old to /boot/loader as a workaround. > > I have > > CFLAGS= -O2 -pipe > > COPTFLAGS= -O2 -pipe > > Please remove these and re-test. The default is "-O2 -fno-strict-aliasing" which generates slightly different code in some instances,Thank you. I commented out above two lines from /etc/make.conf and did build/installkernel, build/installworld again (a little superfluous maybe). So, the loader and the kernel must be built with default options. The result was same. Loading new kernel with the new loader reboots the machine. The CPU is an athlon (CPU lines from dmesg were as follows). I do not remember what was the chipset now; CPU: AMD Athlon(tm) XP 1600+ (1403.19-MHz 686-class CPU) Origin = "AuthenticAMD" Id = 0x662 Stepping = 2 Features=0x383fbff<FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CM OV,PAT,PSE36,MMX,FXSR,SSE> AMD Features=0xc0480800<SYSCALL,MP,MMX+,3DNow+,3DNow>