[9fans] Re: [V9fs-developer] playing with Tim's python 9p implementation

[9fans] Re: [V9fs-developer] playing with Tim's python 9p implementation

[David Leimbach]

On 10/4/05, erik quanstrom <quanstro@quanstro.net> wrote:
> that number is (unsigned long)-1L on a 32bit machine.
>
> i'd bet (sight unseen) that the python code is using unsigned longs for the fids
> and everybody else is using a long with -1 as an error code.
>

I figured it was -1 :).  I just didn't think that it was an error code.

Still interesting.  When I have an error trying to mount the python 9p
server from inferno I don't get that error... only with v9fs.

Dave

> erik
>
> David Leimbach <leimy2k@gmail.com> writes
>
> |
> | I can mount it from Plan 9 but when I try to mount it with v9fs I get
> | errors on the server side:
> |
> | On Mac OS X I do:
> | "sudo  python srv.py -r /Users/dave/Desktop dave homework.net"
> |
> |
> |
> | On plan 9 it's just
> | "9fs fuji"
> |
> | and things work well.
> |
> | On Linux I do:
> | "mount -oname=dave,noextend -t 9P 192.168.1.100 /mnt/9"
> |
> | and I get:
> | "mount: permission denied" in linux
> |
> | and the 9p in python gives the following:
> |
> | "serving: '192.168.1.81',35556
> | Error fid 4294967295 not in use
> | client EOF"
> |
> | Not clear what's happening there.
> |
> | Before I used noextend I got errors on the server side about not
> | understanding 9p2000.u.  Totally expected :).
> |
> | I'm currently using the Python 9p as a quick replacement for testing
> | something like u9fs now that I can't really build it on anything but
> | linux.
> |
> | I'm not terribly interested in hacking on u9fs right now either.  I've
> | got bigger fish to fry and this Python 9p server may do exactly what I
> | want for a particular project I have in mind.
> |
> | Dave
>

Re: [9fans] Re: [V9fs-developer] playing with Tim's python 9p implementation

[Eric Van Hensbergen]

If you turn on the debug=1 during mount you should get the error in
your syslog.  Or if you really want a flood of debug go for debug=255.
 Either way it'll let us know what's going on.  It probably has to do
with auth, it may also have to do with what tid is sent during
version/attach, there is some variation between the various systems on
what is expected.

       -eric

On 10/4/05, David Leimbach <leimy2k@gmail.com> wrote:
> On 10/4/05, erik quanstrom <quanstro@quanstro.net> wrote:
> > that number is (unsigned long)-1L on a 32bit machine.
> >
> > i'd bet (sight unseen) that the python code is using unsigned longs for the fids
> > and everybody else is using a long with -1 as an error code.
> >
>
> I figured it was -1 :).  I just didn't think that it was an error code.
>
> Still interesting.  When I have an error trying to mount the python 9p
> server from inferno I don't get that error... only with v9fs.
>
> Dave
>
> > erik
> >
> > David Leimbach <leimy2k@gmail.com> writes
> >
> > |
> > | I can mount it from Plan 9 but when I try to mount it with v9fs I get
> > | errors on the server side:
> > |
> > | On Mac OS X I do:
> > | "sudo  python srv.py -r /Users/dave/Desktop dave homework.net"
> > |
> > |
> > |
> > | On plan 9 it's just
> > | "9fs fuji"
> > |
> > | and things work well.
> > |
> > | On Linux I do:
> > | "mount -oname=dave,noextend -t 9P 192.168.1.100 /mnt/9"
> > |
> > | and I get:
> > | "mount: permission denied" in linux
> > |
> > | and the 9p in python gives the following:
> > |
> > | "serving: '192.168.1.81',35556
> > | Error fid 4294967295 not in use
> > | client EOF"
> > |
> > | Not clear what's happening there.
> > |
> > | Before I used noextend I got errors on the server side about not
> > | understanding 9p2000.u.  Totally expected :).
> > |
> > | I'm currently using the Python 9p as a quick replacement for testing
> > | something like u9fs now that I can't really build it on anything but
> > | linux.
> > |
> > | I'm not terribly interested in hacking on u9fs right now either.  I've
> > | got bigger fish to fry and this Python 9p server may do exactly what I
> > | want for a particular project I have in mind.
> > |
> > | Dave
> >
>

Re: [9fans] Re: [V9fs-developer] playing with Tim's python 9p implementation

[Ronald G Minnich]

Eric Van Hensbergen wrote:
> If you turn on the debug=1 during mount you should get the error in
> your syslog.  Or if you really want a flood of debug go for debug=255.
>  Either way it'll let us know what's going on.  It probably has to do
> with auth, it may also have to do with what tid is sent during
> version/attach, there is some variation between the various systems on
> what is expected.
>

I've been through this exact thing, and it has to do with auth. I'll try 
to find my modified python server to see what I did to hack around it. 
Also, newsham did put in a fix for it -- you have to invoke the server a 
little differently -- more or less turn off any auth.

I need to learn how to use factotum on linux, however.

ron

Re: [9fans] Re: [V9fs-developer] playing with Tim's python 9p implementation

[David Leimbach]

> I need to learn how to use factotum on linux, however.

likewise.

Inferno has factotum it seems but it doesn't behave the same way I'd
expect.  In fact the man pages suggest that the factotum that is there
works with plan 9 but it has it's own "/mnt/factotum" complete with
the "ctl" file.... so maybe it's more complete than the documentation
implies.

I'll bug Charles about it.

Dave

>
> ron
>

Re: [9fans] Re: [V9fs-developer] playing with Tim's python 9p implementation

[Eric Van Hensbergen]

On 10/4/05, Ronald G Minnich <rminnich@lanl.gov> wrote:
>
> I need to learn how to use factotum on linux, however.
>

It's actually pretty straightforward if you have a Plan 9 Ports
environment installed and Lucho's amount program.  There was some talk
about getting it added into the plan9port collective, but also some
concern as it is pretty Linux specific at this point.  Comments Russ?

I'd like to get a more-pure Linux auth infrastructure in place -- at
the very least being able to host auth servers under Linux, either
using Plan 9 ports or by porting them (IIRC: there were some unix auth
tools, but maybe I'm just thinking of netkey).  There's also the
option of trying for a Kerberos or SASL auth mechanism for Linux-only
environments.  This stuff just hasn't been a priority for me because
its not a concern in my environment, but its something we clearly need
a better story on.

    -eric

Re: [9fans] Re: [V9fs-developer] playing with Tim's python 9p implementation

[David Leimbach]

> I'd like to get a more-pure Linux auth infrastructure in place -- at
> the very least being able to host auth servers under Linux, either
> using Plan 9 ports or by porting them (IIRC: there were some unix auth
> tools, but maybe I'm just thinking of netkey).  There's also the
> option of trying for a Kerberos or SASL auth mechanism for Linux-only
> environments.  This stuff just hasn't been a priority for me because
> its not a concern in my environment, but its something we clearly need
> a better story on.
>

Yeah, auth would be cool, but I don't need it for what I'm planning to
do with v9fs either.

Dave

>     -eric
>

Re: [9fans] Re: [V9fs-developer] playing with Tim's python 9p implementation

[Tim Newsham]

> I've been through this exact thing, and it has to do with auth. I'll try to 
> find my modified python server to see what I did to hack around it. Also, 
> newsham did put in a fix for it -- you have to invoke the server a little 
> differently -- more or less turn off any auth

My fixes earlier where on the client side.  The current bug is about
the server side (srv.py) not playing nicely when it gets an unexpected
version (and then later again when its doing auth, I believe).

Fix #1 - change _srvTversion to not error out if the version
does not match.  It should instead return its own version.

Fix #2 - Not sure, but the code in the write() method of AuthFs
likely needs to be less restrictive.

If someone tweaks this and tests it please let me know the fix.
(I dont have v9fs installed anywhere).

> ron

Tim Newsham
http://www.lava.net/~newsham/

Re: [9fans] Re: [V9fs-developer] playing with Tim's python 9p implementation

[Tim Newsham]

> I've been through this exact thing, and it has to do with auth. I'll try to 
> find my modified python server to see what I did to hack around it. Also, 
> newsham did put in a fix for it -- you have to invoke the server a little 
> differently -- more or less turn off any auth

My fixes earlier where on the client side.  The current bug is about
the server side (srv.py) not playing nicely when it gets an unexpected
version (and then later again when its doing auth, I believe).

Fix #1 - change _srvTversion to not error out if the version
does not match.  It should instead return its own version.

Fix #2 - Not sure, but the code in the write() method of AuthFs
likely needs to be less restrictive.

If someone tweaks this and tests it please let me know the fix.
(I dont have v9fs installed anywhere).

> ron

Tim Newsham
http://www.lava.net/~newsham/

Re: [9fans] Re: [V9fs-developer] playing with Tim's python 9p implementation

[C H Forsyth]

>>works with plan 9 but it has it's own "/mnt/factotum" complete with
>>the "ctl" file.... so maybe it's more complete than the documentation
>>implies.

i'd left that version in just for testing; now both factotum(4) and factotum(2)
have changed to reflect its existence.  of course, you haven't been able to see that
because i didn't do a distribution pack in august or september, mainly because
i was trying to get a few more things ready for distribution, on and off
(including the new update scheme).

Re: [9fans] Re: [V9fs-developer] playing with Tim's python 9p implementation

[David Leimbach]

On 10/5/05, C H Forsyth <forsyth@vitanuova.com> wrote:
> >>works with plan 9 but it has it's own "/mnt/factotum" complete with
> >>the "ctl" file.... so maybe it's more complete than the documentation
> >>implies.
>
> i'd left that version in just for testing; now both factotum(4) and factotum(2)
> have changed to reflect its existence.  of course, you haven't been able to see that
> because i didn't do a distribution pack in august or september, mainly because
> i was trying to get a few more things ready for distribution, on and off
> (including the new update scheme).
>

Ah, so is it true then that mount -9 requires a plan 9 factotum?  If
not then I'd say the man pages are a bit outdated.  auth/feedkey seems
to be pretty neat though :-)

Dave