[9fans] 9vx wrong up? or wrong pgrp?

[9fans] 9vx wrong up? or wrong pgrp?

[erik quanstrom]

any ideas appreciated.

when running plan 9 ip stack under 9vx, i get
a curious situation where the mount table goes
missing, but only for the namec called from
etherbind.  it seems that up->pg isn't right.
the pgrpid is 2 for the call from namec but
3 for the call from the shell.  both seem
wrong.

here's what happens, annotated
with some debugging.

; ip/ipconfig
netifwalk 0 [ipifc]
gen 0 0 00	80.3 ether0
gen 0 0 01	-> -1 (top)
netifwalk 0 [ipifc]
gen 0 0 00	80.3 ether0
gen 0 0 01	-> -1 (top)
netifwalk 0 [ipifc clone]
gen 0 0 00	80.3 ether0
gen 0 0 01	-> -1 (top)
etherbind
chandial /net/ether0!0x800
who you gonna call? /net/ether0/clone
namec /net/ether0/clone
slash 829f860
dumpmount-> pgrpid=2
<-dumpmount
parsename /net: 1 4 11 17
walk error 2: does not exist
error: '/net/ether0' does not exist
ipconfig: /net/ipifc/clone: bind ether /net/ether0: '/net/ether0' does not exist

however, if i inspect the same path from
the shell,

; ls -l /net/ether0/clone
slash 829f860
dumpmount-> pgrpid=3
head: 829d8b0: / 0x0.0 / 0 ->
	82b2000: / (umh 0) (path 00000000 dev / 0)
	82b2438: /root (umh 82b0af8) (path 0000000b dev / 0)
	829e2e0: #Zplan9 (umh 0) (path 100000002c213 dev Z 2)
[...]

one more tidbit, cat /proc/2/ns ->
; cat /proc/2/ns
9vx panic: kernel fault: signo=11 addr=f8[6830c0f8] 1 eip=807385f esp=93988960
aborting, to dump core.
0x807385f is in procread (9vx/a/devproc.c:897).
892				return 0;
893			}
894			mntscan(mw, p);
895			if(mw->mh == 0){
896				mw->cddone = 1;
897				i = snprint(a, n, "cd %s\n", p->dot->path->s);

- erik

Re: [9fans] 9vx wrong up? or wrong pgrp?

[erik quanstrom]

sorry to reply to myself, but the up in the kernel
belongs to an *io*, not the calling process.

- erik

Re: [9fans] 9vx wrong up? or wrong pgrp?

[Devon H. O'Dell]

2008/12/22 erik quanstrom <quanstro@quanstro.net>:
> sorry to reply to myself, but the up in the kernel
> belongs to an *io*, not the calling process.

That's weird, the up in the kernel that I was seeing belonged to
*init*. If you go further down into:

walk error 2: does not exist

You will see the error comes from domount() returning false in
rootdevgen, I believe.

> - erik

--dho

Re: [9fans] 9vx wrong up? or wrong pgrp?

[erik quanstrom]

cf. the email about the comment
that fixes this.

- erik