Re: [ECOS] Re: not able to bring link up !!

[Gary Thomas <gary at mlbassoc dot com>]

On Fri, 2003-08-22 at 08:51, Nick Garnett wrote:
> "mohanlal jangir" <mohanlaljangir@hotmail.com> writes:
> 
> > > You just have to put a bunch of diag_printf() calls into the
> > > initialization code. Comparing the values that these give with the
> > > documentation may give you a clue as to what is happening.
> > Hi Nick,
> >    There is a little improvement. I compiled the eCos source without -O2
> > option and also set back HAL_PCI_ALLOC_BASE_MEMORY to 0. Thing are going
> > better. I am able to see BAR[0] = 0x00020000 even after setting
> > HAL_PCI_ALLOC_BASE_MEMORY to 0. And also link is up. This is confirmed by
> > msg "i82559_start 0 flg 47 Link = Up, 100 Mbps, Full Duplex". Although I am
> > not able to send any packet out but I was able to receive some packets :-).
> > Mostly received packets were broadcasted ARP packets and _ONE_ packet was
> > multicast, unicast packets are not received at all. But receiving of packets
> > is very random. Sometimes it receives and then stops receiving. Also when I
> > run the image without any breakpoint, I receive messages like "config update
> > failed" and "muliticase setup failed", but if I single step through these
> > functions, these functions work perfectly. There is no hang as timer in my
> > code is timing out correctly and corrosponding timer handler is called on
> > every time out.
> > Any clues, where should I look now?
> 
> That taking out -O2 changed things is weird. Maybe there are timing
> problems on the PCI bus. Or maybe there are caching issues.
> 

Another possibility is a missing "volatile".  I know this is tried&true
code, but new compilers always seem to bring new problems in this area.

-- 
Gary Thomas <gary@mlbassoc.com>
MLB Associates


-- 
Before posting, please read the FAQ: http://sources.redhat.com/fom/ecos
and search the list archive: http://sources.redhat.com/ml/ecos-discuss