Re: newfs seems stuck in loop

[Rhialto <rhialto%falu.nl@localhost>]

On Sun 02 May 2010 at 15:47:09 +0200, Johnny Billquist wrote:
> Sounds more like a broken disk to me. The fact that newfs sits still
> in userland don't say anything. It might be waiting for somegthing
> just sitting in userland as well. Any other processes on the machine
> that might be relevant?

Before this, I read and wrote the whole disk with dd, which went at
reasonable pace (reading about 1100 KB/sec, writing about 800 KB/sec,
iirc) and without any obvious errors.

Newfs isn't just waiting, it is actively using CPU time.

I don't think there were any other significant processes, certainly not
with the -current install. At most sysinst with whatever it starts.

On the 2.0F install, I tried attaching gdb to newfs, but that first gave
me some error/warning from gdb, and then landed me in KDB. So that
wasn't very informative. First starting gdb and then running newfs gave
consistent stack traces when I interrupted newfs several times:

#0 0x0001643c in isblock (189220, 175930, 0)
#1 0x00015b14 in alloc (2048, 16877)
#2 ...........in fsinit (2147471224, 1023, 0, 0)
#3 ...........in mkfs (171672, 171788, 5, 6, 1023, 0, 0)
#4 ...........in main (8, ...)

Two times it interrupted in alloc (2048, 16877). alloc() seems to be
called from fsinit to create the root directory only (at least in
-current sources) and calls isblock() in a loop to find a free block in
the first cylinder group (only). Doesn't sounds like it should take too
long, and no I/O seems to be done there...

-Olaf.
-- 
___ Olaf 'Rhialto' Seibert    -- You author it, and I'll reader it.
\X/ rhialto/at/xs4all.nl      -- Cetero censeo "authored" delendum esse.