Re: CFT: vr(4)

On Sun, Feb 17, 2008 at 03:17:26PM +0100, Milan Obuch wrote:
 > On Sunday 17 February 2008, Milan Obuch wrote:
 > > On Sunday 17 February 2008, Robert Watson wrote:
 > > > On Sat, 16 Feb 2008, Milan Obuch wrote:
 > > > >> I have tested 4-port Rhine III(6105LOM) and I never seen this hangs.
 > > > >> Does this also happen on other network interface too? When the system
 > > > >> hangs, would you break into DDB and show me the output of 'show
 > > > >> alllocks' and 'ps'?
 > > > >
 > > > > I need some tests with another box. I was not able to break into DDB.
 > > > > Ctrl-Alt-Esc worked until hang. Hang was really hard, Ctrl-Alt-Esc did
 > > > > not invoke DDB. I do not feel if_vr is culprit here, more probably PCI
 > > > > bus got somehow locked, but I have no idea how could I test it
 > > > > currently.
 > > >
 > > > FYI, there are known issues with the effectiveness of syscons
 > > > ctrl-alt-esc to get into the debugger -- if you haven't tried with a
 > > > serial break on a serial console, you might want to do so.  This is
 > > > because syscons's interrupt handler acquires the Giant lock in an
 > > > ithread, requiring a lot more things to be happy to succeed.  In
 > > > constrast, sio (and friends) use fast interrupt handlers and no Giant
 > > > lock on the way to processing the break request.  It may well be that a
 > > > serial break doesn't get into DDB for you, but it's worth trying...
 > > >
 > > > Robert N M Watson
 > > > Computer Laboratory
 > > > University of Cambridge
 > >
 > > You are right, after rebuilding kernel with BREAK_TO_DEBUGGER I am able to
 > > get requested data. BTW, I can't find just now what ALT_BREAK_TO_DEBUGGER
 > > means...
 > >
 > > After kldload if_vr, dhclient vr0 and ping -f <someaddress> in a minute
 > > system freezes.
 > >
 > 
 > Correction... as I moved cable from on-board re0 to vr0, system remembered arp 
 > table entry and did not actually try to use vr0. Now I repeated it, but boot 
 > system with cable in vr0. This time it hangs right after ping <>, and break 
 > does not enter debugger... hard hang. I tried it twice, the same result...
 > 
 > So no data to check, actually, only hard hang.
 > 

I've put up a new version under the old URL.
It's not tested in sparc64 but it seems to work on i386.
Would you give it spin?

-- 
Regards,
Pyun YongHyeon