Type 15: Data TLB Miss Fault/Data Page Fault

[Wilder Mellotto]

Hi

We had a problem with a rp4440 running hpux 11.11 64 bits. The server rebooted after one panic and create /var/adm/crash/crash.0

I ran the crashinfo tool, but I can't understand what happened. See attached file (crash.txt).

Thanks.

[Don Morris_1 · Accepted solution]

This is the relevant section:

Trap information
================

Type 15: Data TLB Miss Fault/Data Page Fault

Interruption Instruction Register:
IIR = 0x42750022

Interruption Space and Offset Registers:
ISR.IOR = 0x0.0x11

Interruption Instruction Address Queue:
PCSQ.PCOQ = 0x0.0x1c0f34 = TPI_discon_ind+0x24

Interrupt Instruction at TPI_discon_ind+0x24:
ldb 0x11(r19),r21

Note: The trapping address is in kernel space at a low offset
This is probably a null pointer de-reference

------

Its right, too. The function TPI_discon_ind() in the kernel attempted to dereference the kernel virtual address 0x11. That's pretty much guaranteed to be a case of treating a NULL pointer as a pointer to a valid structure and then asking for a field within that structure. Usually due to the function taking an error path and not accounting for uninitialized arguments or a race with other kernel threads using common data where we've had a buffer overrun or the original pointer was freed then cleared while the panic'ing thread thought it was still valid.

Seaching around a bit... it looks like this was addressed on 11.0 as PHNE_32041, superseded by PHNE_35729. That's the 11.0 cumulative ARPA transport patch... the 11.11 current cumulative ARPA transport patch is PHNE_35183. Hidden within there is the defect description for SR: 8606180160, CR: JAGad49831 -- which looks like your case.

Please check if you have at least PHNE_24211, since it looks like that flavor should contain the fix... though since that has a Warning against it, I'd go right to PHNE_35183 (latest fixes, it has a 2-star rating so its been out for a while and has a lot of folks using it).

If you have PHNE_35183, then most certainly contact your HP support rep so they can determine why this happened again with the resolving patch (did the fix miss a case, etc.).