bus error IRQ

bus error IRQ

[Ulf Carlsson]

Hi,

Got this little message which crashed the machine:

Got a bus error IRQ, shouldn't happen yet
$0 : 00000000 1004fc01 88008048 00000000
$4 : 88008000 88008000 fffffc18 00000001
$8 : 88009fe0 3004fc01 8803d2f8 00000003
$12: 00000038 000003e2 88350a08 89f5be18
$16: 00000000 00000000 8800a16c 00000f00
$20: a8747310 9fc45da0 00000000 9fc45da0
$24: 00000001 2ab0c110
$28: 88008000 88009e90 9fc45f0c 88013650
epc   : 880262b8
Status: 1004fc03
Cause : 00004000
Spinning...

That's in 'schedule'

    88026298:   03c0e821        move    $sp,$s8
    8802629c:   8fbf0040        lw      $ra,64($sp)
    880262a0:   8fbe003c        lw      $s8,60($sp)
    880262a4:   8fb40038        lw      $s4,56($sp)
    880262a8:   8fb30034        lw      $s3,52($sp)
    880262ac:   8fb20030        lw      $s2,48($sp)
    880262b0:   8fb1002c        lw      $s1,44($sp)
    880262b4:   8fb00028        lw      $s0,40($sp)
    880262b8:   03e00008        jr      $ra
    880262bc:   27bd0048        addiu   $sp,$sp,72

00000000880262c0 <__wake_up>:
    880262c0:   27bdfff8        addiu   $sp,$sp,-8
    880262c4:   afbe0000        sw      $s8,0($sp)

Do you know anything about this Ralf? Maybe it's fixed in some version I don't
have yet?

- Ulf

Re: bus error IRQ

[ralf]

On Mon, Aug 17, 1998 at 08:45:00PM +0200, Ulf Carlsson wrote:

> Got this little message which crashed the machine:
> 
> Got a bus error IRQ, shouldn't happen yet
> $0 : 00000000 1004fc01 88008048 00000000
> $4 : 88008000 88008000 fffffc18 00000001
> $8 : 88009fe0 3004fc01 8803d2f8 00000003
> $12: 00000038 000003e2 88350a08 89f5be18
> $16: 00000000 00000000 8800a16c 00000f00
> $20: a8747310 9fc45da0 00000000 9fc45da0
> $24: 00000001 2ab0c110
> $28: 88008000 88009e90 9fc45f0c 88013650
> epc   : 880262b8
> Status: 1004fc03
> Cause : 00004000
> Spinning...
> 
> That's in 'schedule'
> 
>     88026298:   03c0e821        move    $sp,$s8
>     8802629c:   8fbf0040        lw      $ra,64($sp)
>     880262a0:   8fbe003c        lw      $s8,60($sp)
>     880262a4:   8fb40038        lw      $s4,56($sp)
>     880262a8:   8fb30034        lw      $s3,52($sp)
>     880262ac:   8fb20030        lw      $s2,48($sp)
>     880262b0:   8fb1002c        lw      $s1,44($sp)
>     880262b4:   8fb00028        lw      $s0,40($sp)
>     880262b8:   03e00008        jr      $ra
>     880262bc:   27bd0048        addiu   $sp,$sp,72
> 
> 00000000880262c0 <__wake_up>:
>     880262c0:   27bdfff8        addiu   $sp,$sp,-8
>     880262c4:   afbe0000        sw      $s8,0($sp)
> 
> Do you know anything about this Ralf? Maybe it's fixed in some version I don't
> have yet?

No, I have no idea what might be causing this.  I myself got bus errors
now and then but for me they have disappeared.  Looks like I fixed them
en passer.

The bad thing with a bus error is that it may be delayed for a very long
time thus resulting in a useless program counter.  What happens is that
the CPU writes to some invalid address but the write access over the
system bus is delayed because the writeback cache policy is being used.
Later, maybe even much later, when the cacheline gets written back to
memory for some reason the system board signals a bus error interrupt.
At this point the program counter may already be completly useless.

I'm shure this has happened in your case as well.

  Ralf

Re: bus error IRQ

[William J. Earl]

ralf@uni-koblenz.de writes:
...
 > The bad thing with a bus error is that it may be delayed for a very long
 > time thus resulting in a useless program counter.  What happens is that
 > the CPU writes to some invalid address but the write access over the
 > system bus is delayed because the writeback cache policy is being used.
 > Later, maybe even much later, when the cacheline gets written back to
 > memory for some reason the system board signals a bus error interrupt.
 > At this point the program counter may already be completly useless.
...

     You cannot get a delayed bus error on a cached write, unless
you do a "create dirty exclusive" cache operation to validate the line
before writing.  You can get delayed bus errors on uncached writes,
as to device control registers.  Since any K1SEG address is uncached,
it is not too hard to generate a bus error with a bad pointer value.

Re: bus error IRQ

[ralf]

On Tue, Aug 18, 1998 at 07:31:05PM -0700, William J. Earl wrote:

> ralf@uni-koblenz.de writes:
> ...
>  > The bad thing with a bus error is that it may be delayed for a very long
>  > time thus resulting in a useless program counter.  What happens is that
>  > the CPU writes to some invalid address but the write access over the
>  > system bus is delayed because the writeback cache policy is being used.
>  > Later, maybe even much later, when the cacheline gets written back to
>  > memory for some reason the system board signals a bus error interrupt.
>  > At this point the program counter may already be completly useless.
> ...
> 
>      You cannot get a delayed bus error on a cached write, unless
> you do a "create dirty exclusive" cache operation to validate the line
> before writing.

Linux uses Create_Dirty_Excl_D as optimization where possible, so the
probability for this to happen is relativly high.  Linux however should
never use Create_Dirty_Excl_D or Create_Dirty_Excl_SD on R4[04]00SC
CPUs, have to verify this.

>                  You can get delayed bus errors on uncached writes,
> as to device control registers.  Since any K1SEG address is uncached,
> it is not too hard to generate a bus error with a bad pointer value.

  Ralf

bus error IRQ

[Joachim Schmitz]

Hi list

I'm new to this list and currently trying to setup my Indy for Linux.
For a start I tried just to boot vmlinux-970916-efs and vmlinux-2.1.67
... without success. Both boots ends up with:

boot vmlinux-2.1.67
PROMLIB: SGI ARCS firmware Version 1 Revision 10
PROMLIB: Total free ram 64376832 bytes (62868k,61MB)
ARCH: SGI-IP22
CPU: MIPS-R4600 FPU<MIPS-R4600FPC> ICACHE DCACHE
Loading R4000 MMU routines.
CPU revision is: 00002020
Primary ICACHE 16k (linesize 32 bytes)
Primary DCACHE 16k (linesize 32 bytes)
R4600/R5000 SCACHE size 0K linesize 32 bytes
MC: SGI memory controller Revision 3
calculating r4coff... 000a8fe8(692200)
GFX INIT: SHMIQ setup
usemaclone misc device registered (minor: 151)
Video screen size is 00004c88 at 883e5d48
Console: 16 point font, 992 scans
Console: color NEWPORT 158x62, 1 virtual console (max 63)
Calibrating delay loop.. ok - 138.04 BogoMIPS
Memory: 60304k/196196k available (1164k kernel code, 2880k data)
Swansea University Computer Society NET3.039 for Linux 2.1
NET3: Unix domain sockets 0.16 for Linux NET3.038.
Swansea University Computer Society TCP/IP for NET3.037
IP Protocols: IGMP, ICMP, UDP, TCP
Checking for 'wait' instruction... available.
Linux version 2.1.67 (ralf@indy) (gcc version 2.7.2) #1 Sat Dec 6 12:48:52 PST 1997
POSIX conformance testing by UNIFIX
Starting kswapd v 1.23
SGI Zilog8530  serial driver version 1.00
tty00 at 0xbfdb9838 (irq=21) is a Zilog8530
tty01 at 0xbfdb9830 (irq=21) is a Zilog8530
loop: registered device at major 7
Got a bus error IRQ, shouldn't happen yet
$0 : 00000000 1000fc01 88130000 00000000
$4 : 88174274 8812cb10 883fdcb8 00000001
$8 : 1000fc03 00000201 0000bf81 8813de68
$12: 00000001 00000001 00000001 fffffffc
$16: 0000c000 8bf81000 00000000 00000000
$20: a87ffc20 a8746d10 9fc55064 00000000
$24: 1000fc01 0000000f
$28: 881da370 883fdcb8 00000001 8800b0e8
epc   : 880359f8
Status: 1000fc03
Cause : 00004000
Spinning...


boot vmlinux-970916-efs
...
calculating r4coff... 000a6cc0(683200)
...
usema device registered (major 86)
Video screensize is 00004c88 at 883a5c60
...
Calibrating delay loop.. ok - 136.40 BogoMIPS
Memory: 60560k/196196k available (1020 kernel code, 2768k data)
...
Linux version 2.1.55 (shaver@neon.ingenia.ca) (gcc version 2.7.2) #121 Tue Sep 16 16:30:52 EDT 1997
...
$0 : 00000000 00000000 00030000 80004000
$4 : 00000080 8bf84000 1000fc01 80003fe0
$8 : 00000000 00000000 0000bf83 881194d8
$12: 00000001 00000001 00000001 fffffffc
$16: 0bf83000 8bf8r300 00000000 00000000
$20: a87ffc20 a8746d10 9fc55064 00000000
$24: 1000fc01 0000000f
$28: 881da370 883bfda8 00000001 880d5910
epc   : 880261bc
Status: 1000fc03
Cause : 00004000
Spinning...

(I hope I didn't make to much mistakes keying in the above)
At this point it hangs. Only pulling the power cord helps.
I have no linux-root filesystem (yet), but I don't think
this is the reason for this behavoir, right?

-- 
Bye,	Jojo

email:
work SCHMITZ_JOACHIM@Tandem.COM
home Joachim_Schmitz@D.maus.de

Re: bus error IRQ

[Ralf Baechle]

On Tue, Dec 16, 1997 at 11:36:08AM +0100, Joachim Schmitz wrote:

> I'm new to this list and currently trying to setup my Indy for Linux.
> For a start I tried just to boot vmlinux-970916-efs and vmlinux-2.1.67
> ... without success. Both boots ends up with:
> 
> boot vmlinux-2.1.67
> PROMLIB: SGI ARCS firmware Version 1 Revision 10
> PROMLIB: Total free ram 64376832 bytes (62868k,61MB)
> ARCH: SGI-IP22
> CPU: MIPS-R4600 FPU<MIPS-R4600FPC> ICACHE DCACHE
> Loading R4000 MMU routines.
> CPU revision is: 00002020
> Primary ICACHE 16k (linesize 32 bytes)
> Primary DCACHE 16k (linesize 32 bytes)
> R4600/R5000 SCACHE size 0K linesize 32 bytes
                          ^^
That's the problem, there is a kernel bug which prevents Linux from
working on R4600/R5000 Indys without second level cache.  My next
kernel release will fix that.

  Ralf