RED State Exception with Netra T1 105 and FreeBSD 5.3 (RC /RELEASE or even -p5)

[Marius Strobl]

On Fri, Mar 18, 2005 at 09:59:32AM +0300, Andrew Belashov wrote:
> Xavier Beaudouin wrote:
> > [...]
> > 
> >>> RED State Exception
> >>> TL=0000.0000.0000.0005 TT=0000.0000.0000.0080
> >>>    TPC=0000.0000.c003.4208 TnPC=0000.0000.c003.420c 
> >>> TSTATE=0000.0044.5800.1504
> >>> TL=0000.0000.0000.0004 TT=0000.0000.0000.0010
> >>>    TPC=0000.0000.c003.4218 TnPC=0000.0000.c003.421c 
> >>> TSTATE=0000.0044.5800.1504
> >>> TL=0000.0000.0000.0003 TT=0000.0000.0000.0010
> >>>    TPC=0000.0000.c003.4218 TnPC=0000.0000.c003.421c 
> >>> TSTATE=0000.0044.5800.1503
> >>> TL=0000.0000.0000.0002 TT=0000.0000.0000.0010
> >>>    TPC=0000.0000.c003.8f80 TnPC=0000.0000.c003.8f84 
> >>> TSTATE=0000.0044.5800.1402
> >>> TL=0000.0000.0000.0001 TT=0000.0000.0000.0063
> >>>    TPC=0000.0000.4056.f250 TnPC=0000.0000.4056.f254 
> >>> TSTATE=0000.0044.0000.1201
> >>
> >>
> >> Trap level (TL) = 1 (kernel):
> >> Trap Type (TT) = 0x63 - implementation_dependent_exception_3 (impl. 
> >> dep. #35)
> >> - corrected_ECC_error for UltraSPARC-IIi Processor
> > 
> > 
> > This is ... what I was afraid of.... ;/
> > 
> >> Trap Program Counter (TPC) = 0x4056f250
> >>
> >> You may try to disassemble kernel code 0x4056f250:
> >> (kgdb) x/10i 0x4056f250
> >>
> >> Trap level (TL) = 2 (exception in kernel routines):
> >> Trap Type (TT) = 0x10 - illegal_instruction
> >> Trap Program Counter (TPC) = 0xc0038f80
> >>
> >> See "The SPARC Architecture Manual, Version 9" and "UltraSPARC-IIi 
> >> User's Manual".
> >>
> >> This is harware problem?
> > 
> > 
> > I dunno, but this errors is comming less frequently when I turn off 
> > softupdates and
> > when I mount all disk with -o sync.
> > 
> > Do you think this is RAM that is going to be faulty ?
> 
> First, try to clean contacts (or simply reinsert) RAM and CPU modules.
> 
> SAFETY WARNING: Always follow ELECTROSTATIC DISCHARGE DAMAGE prevention
> procedures when removing and replacing components.
> 
> > 
> > /Xavier
> 
> To ALL:
> 
> I have looked the source code (src/sys/sparc64/sparc64/exception.S,
> src/sys/i386/i386/trap.c) and has found, that the kernel go to panic,
> when corrected ECC error exception occured.
> 
> This is correct?
> 

As I read them FreeBSD would cause a panic when a corrected ECC
exception occurs while in user mode and just do nothing while in
kernel. But FreeBSD shouldn't see such exceptions in the first
place as it never enables the CEEN bit in ASI_ESTATE_ERROR_EN_REG.
Anyway, if a corrected ECC error causes a RED state exception
than there's probably something going on outside the scope of
the OS, e.g. a hardware problem.

Marius