Project Home
Project Home
Wiki
Wiki
Discussion Forums
Discussions
Project Information
Project Info
Forum Topic - custom io-net driver failing to start in io-pkt - suggestions where to start looking ?: (7 Items)
   
custom io-net driver failing to start in io-pkt - suggestions where to start looking ?  
Customer has a custom io-net driver on PPC that works just fine in 6.3.x. Might be as old as 6.3.0 SP1

Now, they want to migrate it to 6.4.0, so they try to start it, but io-pkt cores right away:

I realize it could be *anything* but are there some specific, known areas to look for ? I didn't see anything that 
talked about this in the wiki already - if it's there, a pointer would be great.

Thanks

Dave

With DL_DEBUG enabled, they see something like this:

# io-pkt-v4 -v -d mydrvr deviceindex=1,phy_addr=2,mac_incr=1
dlopen("NULL",2048)
dlsym(0xfe381234,output_for_percent_s_NULL)=0xfe37e550
dlopen("cache-ppcbe-8541CDS.so",2)
load_object: attempt load of cache-ppcbe-8541CDS.so
dlopen: Library cannot be found
dlopen("NULL",2048)
dlsym(0xfe381234,iofunc_notify_trigger_strict)=0xfe33c290
dlsym(0xfe381234,iofunc_notify_remove_strict)=0xfe33c18c
dlopen("devnp-mydrvr .so",3072)
load_object: attempt load of devnp-mydrvr .so
dlopen: Library cannot be found
dlopen("devn-mydrvr .so",3072)
load_object: attempt load of devn-mydrvr .so
load_elf32: loaded lib at addr fe382000(text) fe38e000(data)
dlsym(0x480db060,_btext)=0x4804d9ec
Library loaded; type 'add-sym devn-mydrvr .so 0x4804d9ec' in gdb to
load symbols
dlsym(0x480db060,iopkt_lsm_entry)=NULL
dlsym(0x480db060,iopkt_drvr_entry)=NULL
dlsym(0x480db060,io_net_dll_entry)=0xfe38e504
dlopen("devnp-shim.so",3072)
load_object: attempt load of devnp-shim.so
load_elf32: loaded lib at addr fe382000(text) fe38a648(data)
dlsym(0x480db060,_btext)=0x4804d9ec
Library loaded; type 'add-sym devnp-shim.so 0x4804d9ec' in gdb to load
symbols
dlsym(0x480db060,iopkt_drvr_entry)=0xfe38a724
dlopen("devn-mydrvr .so",1024)
load_object: attempt load of devn-mydrvr .so
load_elf32: loaded lib at addr fe38b000(text) fe397000(data)
dlsym(0x480db080,_btext)=0xfe38d580
Library loaded; type 'add-sym devn-mydrvr .so 0xfe38d580' in gdb to
load symbols
dlsym(0x480db080,io_net_dll_entry)=0xfe397504
Memory fault (core dumped)
Re: custom io-net driver failing to start in io-pkt - suggestions where to start looking ?  
On Mon, Jun 08, 2009 at 04:50:41PM -0400, Dave Bott wrote:
> Customer has a custom io-net driver on PPC that works just fine in 6.3.x. Might be as old as 6.3.0 SP1
> 
> Now, they want to migrate it to 6.4.0, so they try to start it, but io-pkt cores right away:
> 
> I realize it could be *anything* but are there some specific, known areas to look for ? I didn't see anything that 
talked about this in the wiki already - if it's there, a pointer would be great.

Doesn't ring a bell.  The usual routine applies: start
with a core file and a backtrace...

Regards,

-seanb
RE: custom io-net driver failing to start in io-pkt - suggestions where to start looking ?  
And command-line/mount parameters? You can play on your vmware
"io-pkt-v4-hc -d pcnet lan4", you'll see your io-pkt-v4-hc crashes
immediately. While, on 6.3.x, "io-net -p tcpip -d pcnet lan4" only shows
a warning and the interface looks working. ... This indicates that shim
layer is a kind of simulation and a simulation means it is not real ;-).


Thanks
Weijie
 
-----Original Message-----
From: Sean Boudreau [mailto:community-noreply@qnx.com] 
Sent: June 8, 2009 4:53 PM
To: drivers-networking
Subject: Re: custom io-net driver failing to start in io-pkt -
suggestions where to start looking ?

On Mon, Jun 08, 2009 at 04:50:41PM -0400, Dave Bott wrote:
> Customer has a custom io-net driver on PPC that works just fine in
6.3.x. Might be as old as 6.3.0 SP1
> 
> Now, they want to migrate it to 6.4.0, so they try to start it, but
io-pkt cores right away:
> 
> I realize it could be *anything* but are there some specific, known
areas to look for ? I didn't see anything that talked about this in the
wiki already - if it's there, a pointer would be great.

Doesn't ring a bell.  The usual routine applies: start
with a core file and a backtrace...

Regards,

-seanb

_______________________________________________
Networking Drivers
http://community.qnx.com/sf/go/post31164
RE: custom io-net driver failing to start in io-pkt - suggestions where to start looking ?  
OK, but given that it is *our* code, couldn't we either make it more 'real' or have it be more helpful when 'bad' 
drivers do 'bad' things ? Or both ? Or am I crazy ?
 
Dave

________________________________

From: Weijie Zhang [mailto:community-noreply@qnx.com]
Sent: Mon 08/06/2009 2:23 PM
To: drivers-networking
Subject: RE: custom io-net driver failing to start in io-pkt - suggestions where to start looking ?



And command-line/mount parameters? You can play on your vmware
"io-pkt-v4-hc -d pcnet lan4", you'll see your io-pkt-v4-hc crashes
immediately. While, on 6.3.x, "io-net -p tcpip -d pcnet lan4" only shows
a warning and the interface looks working. ... This indicates that shim
layer is a kind of simulation and a simulation means it is not real ;-).


Thanks
Weijie

-----Original Message-----
From: Sean Boudreau [mailto:community-noreply@qnx.com]
Sent: June 8, 2009 4:53 PM
To: drivers-networking
Subject: Re: custom io-net driver failing to start in io-pkt -
suggestions where to start looking ?

On Mon, Jun 08, 2009 at 04:50:41PM -0400, Dave Bott wrote:
> Customer has a custom io-net driver on PPC that works just fine in
6.3.x. Might be as old as 6.3.0 SP1
>
> Now, they want to migrate it to 6.4.0, so they try to start it, but
io-pkt cores right away:
>
> I realize it could be *anything* but are there some specific, known
areas to look for ? I didn't see anything that talked about this in the
wiki already - if it's there, a pointer would be great.

Doesn't ring a bell.  The usual routine applies: start
with a core file and a backtrace...

Regards,

-seanb

_______________________________________________
Networking Drivers
http://community.qnx.com/sf/go/post31164


_______________________________________________
Networking Drivers
http://community.qnx.com/sf/go/post31169



RE: custom io-net driver failing to start in io-pkt - suggestions where to start looking ?  
I searched around the przilla, Looks like our driver developers are
making progress of smoothing the pain, look at the pr63288 for example. 

How about the customer to collect information. Such as the command line,
the driver (hope it then is able to be reproduced in house) and the core
(will be helpful if can't reproduce in house ...) etc? 

Thanks
Weijie
 

-----Original Message-----
From: Dave Bott [mailto:community-noreply@qnx.com] 
Sent: June 8, 2009 8:03 PM
To: drivers-networking
Subject: RE: custom io-net driver failing to start in io-pkt -
suggestions where to start looking ?

OK, but given that it is *our* code, couldn't we either make it more
'real' or have it be more helpful when 'bad' drivers do 'bad' things ?
Or both ? Or am I crazy ?
 
Dave

________________________________

From: Weijie Zhang [mailto:community-noreply@qnx.com]
Sent: Mon 08/06/2009 2:23 PM
To: drivers-networking
Subject: RE: custom io-net driver failing to start in io-pkt -
suggestions where to start looking ?



And command-line/mount parameters? You can play on your vmware
"io-pkt-v4-hc -d pcnet lan4", you'll see your io-pkt-v4-hc crashes
immediately. While, on 6.3.x, "io-net -p tcpip -d pcnet lan4" only shows
a warning and the interface looks working. ... This indicates that shim
layer is a kind of simulation and a simulation means it is not real ;-).


Thanks
Weijie

-----Original Message-----
From: Sean Boudreau [mailto:community-noreply@qnx.com]
Sent: June 8, 2009 4:53 PM
To: drivers-networking
Subject: Re: custom io-net driver failing to start in io-pkt -
suggestions where to start looking ?

On Mon, Jun 08, 2009 at 04:50:41PM -0400, Dave Bott wrote:
> Customer has a custom io-net driver on PPC that works just fine in
6.3.x. Might be as old as 6.3.0 SP1
>
> Now, they want to migrate it to 6.4.0, so they try to start it, but
io-pkt cores right away:
>
> I realize it could be *anything* but are there some specific, known
areas to look for ? I didn't see anything that talked about this in the
wiki already - if it's there, a pointer would be great.

Doesn't ring a bell.  The usual routine applies: start
with a core file and a backtrace...

Regards,

-seanb

_______________________________________________
Networking Drivers
http://community.qnx.com/sf/go/post31164


_______________________________________________
Networking Drivers
http://community.qnx.com/sf/go/post31169





_______________________________________________
Networking Drivers
http://community.qnx.com/sf/go/post31171
Re: custom io-net driver failing to start in io-pkt - suggestions where to start looking ?  
On Mon, Jun 08, 2009 at 08:02:32PM -0400, Dave Bott wrote:
> OK, but given that it is *our* code, couldn't we either make it more 'real' or have it be more helpful when 'bad' 
drivers do 'bad' things ? Or both ? Or am I crazy ?

We'll try within reason.  PR 63723 / 63288 is out for CI
but it doesn't sound like what's happening here.

-seanb
RE: custom io-net driver failing to start in io-pkt - suggestions where to start looking ?  
Whenever you have problems with a networking driver,
try starting it with "verbose" or maybe even "verbose=100"
then look at the sloginfo output for clues.

--
aboyd