NetBSD Problem Report #10059
Received: (qmail 1550 invoked from network); 6 May 2000 23:25:28 -0000
Message-Id: <200005062324.BAA23716@vader.runit.sintef.no>
Date: Sun, 7 May 2000 01:24:57 +0200 (MEST)
From: he@runit.no (Havard Eidnes)
Reply-To: he@runit.no (Havard Eidnes)
To: gnats-bugs@gnats.netbsd.org
Subject: synetbsd-INSTALL.gz fails with "panic: kernel fault"
X-Send-Pr-Version: 3.95
>Number: 10059
>Category: port-sparc
>Synopsis: synetbsd-INSTALL.gz fails with "panic: kernel fault"
>Confidential: no
>Severity: serious
>Priority: medium
>Responsible: port-sparc-maintainer
>State: closed
>Class: sw-bug
>Submitter-Id: net
>Arrival-Date: Sat May 06 23:26:00 +0000 2000
>Closed-Date: Sun Jul 09 22:38:59 +0000 2000
>Last-Modified: Sun Jul 09 22:38:59 +0000 2000
>Originator: Havard Eidnes
>Release: 1.4.2
>Organization:
RUNIT AS
>Environment:
System: SPARCsystem 600MP Series (2 X CY605C)
Trying to boot NetBSD 1.4.2
>Description:
Tried to boot this system using the network and the boot
blocks from 1.4.2 resulted in the following panic/traceback:
<#0> ok boot net
Resetting ...
SPARCsystem 600MP Series (2 X CY605C), No Keyboard
ROM Rev. 2.5, 128 MB memory installed, Serial #4196356.
Ethernet address 8:0:20:e:f7:94, Host ID: 71400804.
Rebooting with command: net
Boot device: /iommu/sbus/lebuffer@f,40000/le@f,60000 File and args:
13c00
>How-To-Repeat:
See above.
>Fix:
Sorry; fixing at this level is beyond me...
I'll however gladly test any modifications.
For the record: booting the normal 1.4.2 GENERIC_SCSI3 kernel
using the same setup as was used above works.
>Release-Note:
>Audit-Trail:
From: Havard.Eidnes@runit.sintef.no
To: he@runit.no
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: port-sparc/10059: synetbsd-INSTALL.gz fails with "panic: kernel fault"
Date: Sun, 07 May 2000 13:14:31 +0200
> >Number: 10059
> >Category: port-sparc
> >Synopsis: synetbsd-INSTALL.gz fails with "panic: kernel fault"
> ...
> >> NetBSD/sparc Secondary Boot, Revision 1.8
> >> (toddpw@threepio.toddpw.net, Fri Mar 3 17:51:31 PST 2000)
> boot: client IP address: 129.241.249.77
> boot: client name: anker.runit.no
> root addr=129.241.249.70 path=/var/tmp
> Booting netbsd @ 0x4000
> 1056768+1738008+146404+[62268+71473]=0x2f2b71
> OBP version 3, revision 2.5 (plugin rev 2)
> [ preserving 133748 bytes of netbsd a.out symbol table ]
> console is ttya
> Copyright (c) 1996, 1997, 1998, 1999, 2000
> The NetBSD Foundation, Inc. All rights reserved.
> Copyright (c) 1982, 1986, 1989, 1991, 1993
> The Regents of the University of California. All rights reserved.
>
> data fault: pc=0xf00b8f9c addr=0x0 sfsr=1a6<PERR=0,LVL=1,AT=5,FT=1,FAV,OW>
> panic: kernel fault
> Stopped in at _Debugger+0x4: jmpl [%o7 + 0x8], %g0
> db>
Hm, it appears you can't trust "trace" to give correct data in
this case. The PC given above points into uvm_pagealloc_strat,
at uvm_pagealloc_strat+416:
(gdb) i li 937
Line 937 of "../../../../uvm/uvm_page.c"
starts at address 0x1850 <uvm_pagealloc_strat+380>
and ends at 0x1878 <uvm_pagealloc_strat+420>.
(gdb) x/i
0x1850 <uvm_pagealloc_strat+380>: ld [ %i0 ], %o1
(gdb)
0x1854 <uvm_pagealloc_strat+384>: cmp %o1, 0
(gdb)
0x1858 <uvm_pagealloc_strat+388>: be 0x1868 <uvm_pagealloc_strat+404>
(gdb)
0x185c <uvm_pagealloc_strat+392>: ld [ %i0 + 4 ], %o0
(gdb)
0x1860 <uvm_pagealloc_strat+396>: b 0x186c <uvm_pagealloc_strat+408>
(gdb)
0x1864 <uvm_pagealloc_strat+400>: st %o0, [ %o1 + 4 ]
(gdb)
0x1868 <uvm_pagealloc_strat+404>: st %o0, [ %o2 + 4 ]
(gdb)
0x186c <uvm_pagealloc_strat+408>: ld [ %i0 + 4 ], %o1
(gdb)
0x1870 <uvm_pagealloc_strat+412>: ld [ %i0 ], %o0
(gdb)
0x1874 <uvm_pagealloc_strat+416>: st %o0, [ %o1 ]
(gdb)
0x1878 <uvm_pagealloc_strat+420>: mov %l1, %o3
(gdb)
This points to the TAILQ_REMOVE in the below section:
(gdb) l 937
932 panic("uvm_pagealloc_strat: bad strat %d", strat);
933 /* NOTREACHED */
934 }
935
936 gotit:
937 TAILQ_REMOVE(freeq, pg, pageq);
938 uvmexp.free--;
939
940 uvm_unlock_fpageq(); /* unlock free page queue */
941 splx(s);
(gdb)
Now, *why* this happens is still a mystery to me.
Regards,
- Havard
From: Havard.Eidnes@runit.sintef.no
To: gnats-bugs@gnats.netbsd.org
Cc:
Subject: Re: port-sparc/10059: synetbsd-INSTALL.gz fails with "panic: kernel fault"
Date: Sun, 07 May 2000 20:00:39 +0200
More info:
Created a kernel with the following test added:
gotit:
#if 1
if (pg->pageq.tqe_prev == NULL)
panic("uvm_pagealloc_strat: freeq busted");
#endif
TAILQ_REMOVE(freeq, pg, pageq);
uvmexp.free--;
and booting a kernel (stripped -- more about that below) results
in this panic triggering:
<#0> ok boot net
Resetting ...
SPARCsystem 600MP Series (2 X CY605C), No Keyboard
ROM Rev. 2.5, 128 MB memory installed, Serial #4196356.
Ethernet address 8:0:20:e:f7:94, Host ID: 71400804.
Rebooting with command: net
Boot device: /iommu/sbus/lebuffer@f,40000/le@f,60000 File and args:
13c00
>> NetBSD/sparc Secondary Boot, Revision 1.8
>> (he@ken.runit.sintef.no, Fri Mar 3 19:42:53 MET 2000)
boot: client IP address: 129.241.249.77
boot: client name: anker.runit.no
root addr=129.241.249.70 path=/var/tmp
Booting netbsd @ 0x4000
1056768+1885464+146404=0x2f60fc
OBP version 3, revision 2.5 (plugin rev 2)
[ no symbols available ]
console is ttya
Copyright (c) 1996, 1997, 1998, 1999, 2000
The NetBSD Foundation, Inc. All rights reserved.
Copyright (c) 1982, 1986, 1989, 1991, 1993
The Regents of the University of California. All rights reserved.
panic: uvm_pagealloc_strat: freeq busted
Stopped in at 0xf00fd068: jmpl [%o7 + 0x8], %g0
db> trace
?(0xf051e000, 0x6ac000, 0x0, 0x0, 0x0, 0x0) at 0xf00b8fac
?(0xf02d2ed8, 0x6ac000, 0x1, 0x0, 0x0, 0xf00f4510) at 0xf00b2828
?(0xf06ac000, 0xfe000000, 0xf02edfe0, 0x1, 0xf06ab7a4, 0xf02edfa8) at 0xf002c54c
?(0xf0110400, 0x1, 0xf0002000, 0xf010df20, 0xf010df18, 0x8) at 0xf00b1c8c
?(0x0, 0xf02ece70, 0xf00021b8, 0xf010e0a3, 0xffef201c, 0x0) at 0xf00262a8
?(0x1, 0x30c800, 0x301ee0, 0x0, 0x0, 0x8) at 0xf0007694
db>
Getting the corresponding symbols:
(gdb) x/i 0xf00b8fac
0xf00b8fac <uvm_pagealloc_strat+400>: call 0xf0038564 <panic>
(gdb) x/i 0xf00b2828
0xf00b2828 <uvm_km_alloc1+244>: call 0xf00b8e1c <uvm_pagealloc_strat>
(gdb) x/i 0xf002c54c
0xf002c54c <kmeminit+16>: call 0xf00b2734 <uvm_km_alloc1>
(gdb) x/i 0xf00b1c8c
0xf00b1c8c <uvm_init+104>: call 0xf002c53c <kmeminit>
(gdb) x/i 0xf00262a8
0xf00262a8 <main+52>: call 0xf00b1c24 <uvm_init>
(gdb) x/i 0xf0007694
0xf0007694 <startmap_done+460>: call 0xf0026274 <main>
(gdb)
I tried generating a sysinst-on-ramdisk kernel *with* debugging
symbols, but the resulting kernel failed to load with
Rebooting with command: net
Boot device: /iommu/sbus/lebuffer@f,40000/le@f,60000 File and args:
13c00
>> NetBSD/sparc Secondary Boot, Revision 1.8
>> (he@ken.runit.sintef.no, Fri Mar 3 19:42:53 MET 2000)
boot: client IP address: 129.241.249.77
boot: client name: anker.runit.no
root addr=129.241.249.70 path=/var/tmp
Booting netbsd @ 0x4000
1056768+1885464+146404+[62268Memory Address not Aligned
Type help for more information
<#0> ok
<#0> ok .locals
0 1 2 3 4 5 6 7
IN: b0ef 5020000 f00d593c b0f8 5020000 f008ac28 b104 9000000
LOC: 7010000 f010e2a8 b0df 5020000 f0026274 b0e5 5020000 f00eee2c
OUT: 400 400 2d8138 2ffd00 400 2d8138 2ffd70 3048f8
<#0> ok .registers
%g0 %g1 %g2 %g3 %g4 %g5 %g6 %g7
0 8000000 0 0 1 fff 199b2 4000
PC nPC Y PSR WIM TBR
304920 304924 8000 11001be2 8 ffefa070
<#0> ok 304918 dis
304918 beq 304944
30491c or %g0, %o4, %o1
304920 ldd [%i0 + 8], %o2
304924 sra %o4, 1f, %o0
304928 addcc %o3, %o1, %o3
30492c addx %o2, %o0, %o2
304930 sub %i2, %o4, %i2
304934 std %o2, [%i0 + 8]
304938 subcc %i2, 0, %g0
30493c bgt 3048dc
304940 add %i1, %o4, %i1
304944 subcc %i3, 0, %g0
304948 bne 304950 annuled
30494c st %i2, [%i3]
304950 or %g0, 0, %i0
304954 jmp %i7, 8, %g0
304958 restore %g0, %g0, %g0
<#0> ok
This also happens with a kernel which has not been touched by
mdsetimage (i.e. the plain INSTALL image).
(Hmm, might look like the kernel symbol table is overrunning the
default boot code, which I think gets loaded at 0x300000, and the
kernel loads at 0x4000, so.... Ouch!)
Regards,
- Havard
From: Havard.Eidnes@runit.sintef.no
To: gnats-bugs@gnats.netbsd.org
Cc:
Subject: Re: port-sparc/10059: synetbsd-INSTALL.gz fails with "panic: kernel fault"
Date: Sun, 07 May 2000 20:36:54 +0200
Still more info:
after generating a new boot.net based on boot.700000, the boot
proceeds as it should, and I get no crash/panic.
Thus, it would seem that this isn't a UVM problem per se, but
rather the boot code interfering with the kernel, either through
the symbol table (as in my last and tangential case), or in some
other way with the kernel.
Regards,
- H=E5vard
State-Changed-From-To: open->closed
State-Changed-By: pk
State-Changed-When: Sun Jul 9 15:36:47 PDT 2000
State-Changed-Why:
Bootblock link address moved up to 0x3E0000 to avoid overwriting
with kernel image (back-ported to 1.4.3).
>Unformatted:
>> NetBSD/sparc Secondary Boot, Revision 1.8
>> (toddpw@threepio.toddpw.net, Fri Mar 3 17:51:31 PST 2000)
boot: client IP address: 129.241.249.77
boot: client name: anker.runit.no
root addr=129.241.249.70 path=/var/tmp
Booting netbsd @ 0x4000
1056768+1738008+146404+[62268+71473]=0x2f2b71
OBP version 3, revision 2.5 (plugin rev 2)
[ preserving 133748 bytes of netbsd a.out symbol table ]
console is ttya
Copyright (c) 1996, 1997, 1998, 1999, 2000
The NetBSD Foundation, Inc. All rights reserved.
Copyright (c) 1982, 1986, 1989, 1991, 1993
The Regents of the University of California. All rights reserved.
data fault: pc=0xf00b8f9c addr=0x0 sfsr=1a6<PERR=0,LVL=1,AT=5,FT=1,FAV,OW>
panic: kernel fault
Stopped in at _Debugger+0x4: jmpl [%o7 + 0x8], %g0
db> trace
_mem_access_fault4m(0x0, 0x1a6, 0x0, 0xf010dcd0, 0xf0396874, 0xf0002000) at _me
m_access_fault4m+0x308
memfault_sun4m(0x0, 0x0, 0xf02cfa74, 0x7a51, 0xf02cf800, 0xffffffff) at memfaul
t_sun4m+0xe4
_uvm_map(0xf051e000, 0x6ac000, 0x0, 0x0, 0x0, 0x0) at _uvm_map+0x478
_uvm_km_alloc1(0xf02aeed8, 0x6ac000, 0x1, 0x0, 0x0, 0xf00f44e8) at _uvm_km_allo
c1+0xf4
_kmeminit(0xf06ac000, 0xfe000000, 0xf02c9fe0, 0x1, 0xf06ab874, 0xf02c9fa8) at _
kmeminit+0x10
_uvm_init(0xf0110400, 0x1, 0xf0002000, 0xf010df20, 0xf010df18, 0x8) at _uvm_ini
t+0x68
_main(0x0, 0xf02c8e70, 0xf00021b8, 0xf010e0a3, 0xffef201c, 0x0) at _main+0x34
startmap_done(0x1, 0x30c800, 0x301ee0, 0x0, 0x0, 0x8) at startmap_done+0x1cc
db>
If my ability to disassemble "-g"-compiled .o files is up to
par, it would seem that this is happening near the end of
uvm_map():
(gdb) i li 689
Line 689 of "../../../../uvm/uvm_map.c" starts at address 0xc08 <uvm_map+1136>
and ends at 0xc18 <uvm_map+1152>.
(gdb) x/i 0xc08
0xc08 <uvm_map+1136>: mov %l3, %o0
(gdb) x/i
0xc0c <uvm_map+1140>: mov 6, %o1
(gdb)
0xc10 <uvm_map+1144>: call 0xc10 <uvm_map+1144>
(gdb)
0xc14 <uvm_map+1148>: clr %o2
(gdb)
0xc18 <uvm_map+1152>: clr %i0
(gdb) l 689
684 if ((map->first_free == prev_entry) &&
685 (prev_entry->end >= new_entry->start))
686 map->first_free = new_entry;
687
688 UVMHIST_LOG(maphist,"<- done!", 0, 0, 0, 0);
689 vm_map_unlock(map);
690 return(KERN_SUCCESS);
691 }
692
693 /*
(gdb)
and 1144 == 0x478 (see traceback above)
/sys/vm/vm_map.h defines vm_map_unlock as:
#define vm_map_unlock(map) \
lockmgr(&(map)->lock, LK_RELEASE, (void *)0)
It *does* however sound a little nuts that it should fault on
a call instruction, and the preceding ones also look
"innocent" enough to me...
(Contact us)
$NetBSD: query-full-pr,v 1.51 2026/08/10 02:28:17 riastradh Exp $
$NetBSD: gnats_config.sh,v 1.10 2026/05/13 22:00:09 riastradh Exp $
Copyright © 1994-2026
The NetBSD Foundation, Inc. ALL RIGHTS RESERVED.