Description of problem:
When performing an upgrade from RHEL 3.8 to 3.9-re20070516.0 and observing the
result on serial console it is flooded with text messages of the type:
CPU 4: now < last_tick (now=0x27c6a5de28,last_tick=0x2f35c57f80)!
CPU 10: now < last_tick (now=0x27c8f2668c,last_tick=0x2f35c57f80)!
CPU 4: now < last_tick (now=0x27c6df6260,last_tick=0x2f3612c626)!
CPU 1: now < last_tick (now=0x27c5da5794,last_tick=0x2f3612c626)!
CPU 4: now < last_tick (now=0x27c7188733,last_tick=0x2f365a1b0e)!
CPU 4: now < last_tick (now=0x27c7366a0b,last_tick=0x2f3683b740)!
CPU 7: now < last_tick (now=0x27c8a1e970,last_tick=0x2f36e2d320)!
Perform an installation of RHEL 3.8 and upgrade to 3.9 using a directory as a
repository. Observer the process over serial console. I used cyclades interface
over ssh. When the upgrade is complete and the system is rebooted the text
messages start to appear rendering the console unusable.
I found a patch that fixes probably a similar issue. I don't know if it is
related but the kernel error message is the same.
After retesting I got:
installing rhel3.8:AS everything with kernel-2.4.21-47.EL, reboot.
The error messages start to appear over the serial console and dmesg as well.
I am able to login using ssh into the system.
Created attachment 155120 [details]
the dmesg file from the machine
After being logged on the machine via ssh for about 2/3 hours it stopped
responding. Opening vi or running ls started to take very long.
At the end I was not able to ping it.
Created attachment 155255 [details]
Upstream patch that fixes this issue
This patch resolves this issue by noting that the check in
gettimeoffset/itc_get_offset is bogus. The "real" check to ensure that time
does not go backwards should be synchronized in the caller to this function as
is done in the case of do_gettimeofday.
Upstream here in 2.5.
Created attachment 155283 [details]
RHEL3 fix for this issue
Backport to RHEL3.