GH-CHUNKS problem - system failed to boot up

[test_46]

Dear all,

We have an Alphaserver ES40 (with 4 CPUs & 8GB memory), running with Tru64 5.1B & patchkit 3. We tried to tune the gh-chunks to reserve 2 to 3GB of memory for SGA. After updating the file /etc/sysconfigtab, (set gh-chunks as 512 or 768) we failed to boot up the machine. We need to boot to CDROM and restore the sysconfigtab from backup.

We faced this problem on patchkit 2 and attempted to fix it by applying patchkit 3. However, problem persists. If patchkit 3 is OK for tuning gh-chunks, we prefer to using PK3 as we have experience of crashing on another machine with PK4.

Following is the error message when machine is starting up:

Loading vmunix ...
Loading at 0xfffffc0000230000

Sizes:
text = 7061696
data = 1345216
bss = 1908944
Starting at 0xfffffc0000242df0

waiting for pza0.0.0.4004.1 to start...
Loading vmunix symbol table ... [1719688 bytes]
sysconfigtab: attribute num-of-sems not in subsystem ipc
could not reserve memory for region 1 in RAD 0
panic (cpu 0): gh_reserve
DUMP: Warning: no disk available for dump.DUMP: first crash dump failed: attempting memory dump...DUMP: Warning: no indicators for memory dump.DUMP: third crash dump ignored: giving up.
trap: invalid memory read access from kernel mode

faulting virtual address: 0x00000000100000e0
pc of faulting instruction: 0xfffffc00006b63e8
ra contents at time of fault: 0xfffffc00006b4eac
sp contents at time of fault: 0xffffffffffff7660


trap: invalid memory read access from kernel mode

faulting virtual address: 0x00000000100000e0
pc of faulting instruction: 0xfffffc00006b63e8
ra contents at time of fault: 0xfffffc00006b4eac
sp contents at time of fault: 0xffffffffffff7220


trap: invalid memory read access from kernel mode

faulting virtual address: 0x00000000100000e0
pc of faulting instruction: 0xfffffc00006b63e8
ra contents at time of fault: 0xfffffc00006b4eac
sp contents at time of fault: 0xffffffffffff6de0


trap: invalid memory read access from kernel mode

faulting virtual address: 0x00000000100000e0
pc of faulting instruction: 0xfffffc00006b63e8
ra contents at time of fault: 0xfffffc00006b4eac
sp contents at time of fault: 0xffffffffffff69a0


trap: invalid memory read access from kernel mode

faulting virtual address: 0x00000000100000e0
pc of faulting instruction: 0xfffffc00006b63e8
ra contents at time of fault: 0xfffffc00006b4eac
sp contents at time of fault: 0xffffffffffff6560


trap: invalid memory read access from kernel mode

faulting virtual address: 0x00000000100000e0
pc of faulting instruction: 0xfffffc00006b63e8
ra contents at time of fault: 0xfffffc00006b4eac
sp contents at time of fault: 0xffffffffffff6120


trap: invalid memory read access from kernel mode

faulting virtual address: 0x00000000100000e0
pc of faulting instruction: 0xfffffc00006b63e8
ra contents at time of fault: 0xfffffc00006b4eac
sp contents at time of fault: 0xffffffffffff5ce0


trap: invalid memory read access from kernel mode

faulting virtual address: 0x00000000100000e0
pc of faulting instruction: 0xfffffc00006b63e8
ra contents at time of fault: 0xfffffc00006b4eac
sp contents at time of fault: 0xffffffffffff58a0


trap: invalid memory read access from kernel mode

faulting virtual address: 0x00000000100000e0
pc of faulting instruction: 0xfffffc00006b63e8
ra contents at time of fault: 0xfffffc00006b4eac
sp contents at time of fault: 0xffffffffffff5460


trap: invalid memory read access from kernel mode

faulting virtual address: 0x00000000100000e0
pc of faulting instruction: 0xfffffc00006b63e8
ra contents at time of fault: 0xfffffc00006b4eac
sp contents at time of fault: 0xffffffffffff5020


trap: invalid memory read access from kernel mode

faulting virtual address: 0x00000000100000e0
pc of faulting instruction: 0xfffffc00006b63e8
ra contents at time of fault: 0xfffffc00006b4eac
sp contents at time of fault: 0xffffffffffff4be0


trap: invalid memory read access from kernel mode

faulting virtual address: 0x00000000100000e0
pc of faulting instruction: 0xfffffc00006b63e8
ra contents at time of fault: 0xfffffc00006b4eac
sp contents at time of fault: 0xffffffffffff47a0


halted CPU 0

halt code = 2
kernel stack not valid halt
PC = fffffc00002af0f8
P00>>>
P00>>>

Appreciate if there is any idea. It's quite urgent.

Thanks a lot.

Rgds,

LKY

[Ralf Puchner · Accepted solution]

first boot the machine without reading the current sysconfigtab. It is a simple bootflag:

>>> boot -fl c

after machine coming up ask your software vendor about the right formula for your kernel parameters. Changing parameters by playing around with values is not very wise.

[Hein van den Heuvel]

Hmmm, new to me.
Sounds like a support question to me.

Anyway on V5.1(B) you are supposed to set 'rad_gh_regions' (in 1 MB units) instead of gh_chunks (in 4 MB units)

However, I would suggest to use "vm_bigpg_enabled = 1 " instead of GH.
Much more flexible, much more useable (SHM, malloc, code...), equally effective.

Check recent posts here for more comments and a tool to verify effectiveness.

Hein.

[Ravi_8]

Hi

SGA will occupy shared memory (SHMMAX)why do you tune gh-chunks, this creates problems.

if shared memory is not sufficient oracle gives you the error, you can tune the shmmax that time instead of landing in problems like this.

[test_46]

Thanks a lot. I checked the session/event log again and noted patch ID 1414 was skipped. I re-applied it again and problem is gone.

Thanks,

LKY