Re: -current userland working for others?

[Michael <macallan%netbsd.org@localhost>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

On Feb 6, 2011, at 12:53 PM, Jeff Rizzo wrote:

> Since the PPC changes brought in a couple weeks ago, I have not been  
> able to run a -current userland on my G4/400.
>
> There was a problem where the kernel would not start, which was  
> fixed by matt@ after a few days, and I can successfully boot a  
> GENERIC from the latest autobuilds.  However, when I unpack userland  
> (base.tgz, in this case), basically all binaries except those in / 
> rescue stop working.  I get these kinds of messages from the kernel:
>
> trap: pid 694.1 (ls): user read DSI trap @ 0x407eb46c by 0xeff0f2dc  
> (DSISR 0x40000000, err=14)
> trap: pid 541.1 (dmesg): user ISI trap @ 0x2000 (SRR1=0x4000f032)
> trap: pid 167.1 (dmesg): user ISI trap @ 0x2000 (SRR1=0x4000f032)
>
>
> ...once I back out to userland from 20110111, things start working  
> again.
>
> Does anyone have any clue what's going on here?  This is the box  
> that I was using to run ATF tests on, and it's basically been idle  
> for three weeks because of this.

So dynamically linked binaries bork? On a 7400? Smells like missing  
dcache syncs / icache invalidates on freshly written PLTs. We had this  
problem before, for some reason it doesn't seem to bite ( as much ) on  
other PowerPCs, IIRC PowerPCs aren't required to keep data and  
instruction caches coherent but apparently at least some do it anyway  
so they don't get bitten by this.

have fun
Michael

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)

iQEVAwUBTU9WY8pnzkX8Yg2nAQKQjwf/SeddnJVIV9CjN9TxynYX2mMOk8NLyOBw
bRZ3811lOuIs3B6T77FObiVt9BcDN6znjO/4j0rr3REY8WCX7OXurUAcF622SRAF
vioApbMMazV0cQddObA6XEFnJabGtCD1wyQwG8QPwWEFUrGaD5Va7oq0LEbofxGv
5j6ORuOvpbsmVevm0tho/F/6Cb9E7BkGnsHYiXT2QBHrL+APnS1XIyCeC9eMvbTT
Fw9WEzu/KOUeeRayGgN/l3gBP7pdFWpSB2Ii6T95gi0IfQTaQKhsYVuXeALzlSY1
HMA7UY3S0EKfL0ncJdzP7dxZ+1Qgo/1CQXq6qcZfyRxqlHjaq9Hkxg==
=B3ZG
-----END PGP SIGNATURE-----