Re: Long downloads, network frozen

[David Albert <david_albert%comcast.net@localhost>]

I was in the middle of an scp operation on a raq2+ running
NetBSD 3.99.7 and it died in the usual manner:

mmm_1_1_78.hex                                100%  287KB 286.7KB/s   00:00
reader_1_1_78.bin                             100%   94KB  94.0KB/s   00:00
reader_1_1_78.hex                               0%    0     0.0KB/s   
--:-- ETAR
ead from remote host raq2: Connection reset by peer

The serial console showed:
db> bt
801ca724+214 (8fffe000,0,bc800000,d) ra 80148b6c sz 0
panic+190 (8fffe000,d,bc800000,74) ra 80121b68 sz 40
80121930+238 (8fffe000,d,bc800000,74) ra 0 sz 0
User-level: curlwp NULL
db> show registers
at          0x802a0000  saupcall_pool+0x50
v0          0x1
v1          0x63061
a0          0x8fffe000
a1          0
a2          0xbc800000
a3          0xd
t0          0xcc42d5fc
t1          0
t2          0x4
t3          0x7c
t4          0x276c
t5          0x4f4fd19e
t6          0x6060a0c
t7          0xd762b5d7
s0          0x8023d388  ksyms_cdevsw+0x340
s1          0x80296448  fpu_id+0x16ff0
s2          0
s3          0
s4          0xfffffffe
s5          0
s6          0
s7          0
t8          0x5cacfe3b
t9          0x7d8e9190
k0          0
k1          0
gp          0x80287440  fpu_id+0x7fe8
sp          0xcc42d5f8
fp          0x100
ra          0x80148b6c  panic+0x190
sr          0xa403
mdlo        0x4b
mdhi        0x5e0
bad         0
cs          0
pc          0x801ca938  cpu_Debugger+0x4
netbsd:cpu_Debugger+0x4:        jr      ra
               bdslot: nop
db> show pool
POOL (null): size 2353174832, align 0, ioff 2353135736, roflags 0x00000000
       alloc 0xac68000c
       minitems 0, minpages 608436384, maxpages 2353332224, npages 
2353397764
       itemsperpage 2353463304, nitems 2892300288, nout 2892365828, 
hardlimit 2
892431368

       nget 2359427088, nfail 0, nput 2898395156
       npagealloc 2359492628, npagefree 0, hiwat 2898460696, nidle 
2359427096
db> show object
OBJECT 0x801ca938: locked=65011720, pgops=0x0, npages=1007058986, 
refs=616967376
db> show buf
 vp 0xac650000 lblkno 0x8ca30004 blkno 0x244400a000000000 dev 0x8c429930
 error 1007058986 flags 0x30840001<AGE,PHYS,WANTED>
 bufsize 0x10800014 bcount 0xc01821 resid 0x3c028028
 data 0x0 saveaddr 0xac670008 dep 0xac660004
 iodone 0x8c470008
db> show map
MAP 0x80296390: [0xc0000000->0xd1929000]
       #ent=40, sz=205824000, ref=1, version=111467, flags=0x1
       pmap=0x802a2f70(resident=2152, wired=114)
db> show sched_qs
6      3556.1 (ssh)


Hubert Feyrer wrote:
> On Wed, 5 Mar 2008, David Albert wrote:
> > I have also noted lock-ups on a Raq2+ with 256MB RAM
> > on large file transfers.
>
> Can you break into ddb from the serial console and see where it hangs?
> ddb may be "+++" or so, then run "bt" - the stack backtrace may help 
> to tell what's wrong.
>
>
>  - Hubert