a few more words on cgfourteen

A customer asked me if he’ll ever gets a chance to watch HD movies on an emulated SPARC machine. At first, I was going to say never ever, but then I looked at the specs of some ancient adapters and got surprised.  The SGI machines were commonly known for good graphics, but it never occurred to me that Sun machines were really powerful as well.

The cgfourteen (cg14) adapter could show Full-HD (1920x1080) pictures back in 1993!

Since the manufacturers sometimes claim just the theoretic abilities of the chipset, I wanted to check that the software has been really aware of such graphic modes in early nineties.

So, I gave a spin to the old cgfourteen patch from Bob Breuer. And here we go.

 

SPARCstation 20 OBP supports 1920x1080

Overall the cgfourteen adapter was really powerful. Michael Lorenz who added the hardware acceleration support for cgfourteen to NetBSD, wrote “SX has plenty of registers (128 - eight of them have special functions, the rest is free for all) and every instruction takes a count to operate on several subsequent registers or memory locations (ALU ops can use up to 16, memory accesses up to 32). SX supports some parallelism too - the ALUs can do up to two 16bit multiplications and two other arithmetic or logical ops per clock cycle (32bit multiplications use both ALUs).

The bad news is that Bob’s patch doesn’t implement the SX rendering engine. Also, the Solaris (or rather OpenWindows) driver is very much bound to sun4m SRMMU, so it won’t work on sun4u machines.

Would really like to have the emulation of a graphic card which would be better than cgthree, and ideally would work on both sun4m and sun4u machines, but it’s not easy.

Right now, I’m choosing between cgsix and creator ffb, but none of them has public documentation.

P.S. A funny fact about the Solaris SX driver: there are lots of symbols which have the word “SPAM” in a name. It can spamify and unspamify.  It’s really hard to google what it could mean back in nineties. “M” is probably for “Memory”.

NetBSD vfs and MMU emulation

Since Solaris works perfectly, I turned again to NetBSD for the further test cases.

root on md0a dumps on md0b

root file system type: ffs

WARNING: clock gained 1005 days

WARNING: CHECK AND RESET THE DATE!

exec /sbin/init: error 8

init: trying /sbin/oinit

exec /sbin/oinit: error 2

init: trying /sbin/init.bak

exec /sbin/init.bak: error 2

init: not found panic: no init

cpu0: kdb breakpoint at 1362e60

Stopped in pid 1.1 (init) at 0x1362e64: nop

db>

Hmm what would be a reason for such a behavior? Sounds familiar, right? I hear you saying CPU math bug? Wrong. :)
The previous time I saw the message it was really a math bug. This time it's a MMU problem. Unlike SPARC v8, the v9 has a NFO mode for page mappings, which means No Fault Only. The 'only' part is crucial: all other page loads should fault.

Solaris knows what pages has been marked NFO (after all they can be marked by the OS only) and uses only the appropriate loads.

On the other side NetBSD vfs implementation explicitly provokes such faults to load RAM pages from a file. Had to read the NetBSD documentation to understand how its vfs works. The sources are not that good documented.

After fixing the bug (a sort of a counterpart to the one Tsuneo Saito mentioned on the mailing list), NetBSD gets a bit further:

Kernelized RAIDframe activated md0:

internal 5120 KB image area

root on md0a dumps on md0b

root file system type: ffs

WARNING: clock gained 1005 days

WARNING: CHECK AND RESET THE DATE!

erase ^?, werase ^W, kill ^U, intr ^C

The only part is missing here is the '#' prompt after the message. :) Working on it.

hsfs_putpage:birthday gift

I think I've fixed the problem with the dirty pages. This is my birthday gift to me.
The bug is really simple: if we fail before modifying a RAM page, we don't really get the page dirty.

Submitted the patch upstream.

MMU & irq fixes

Submitted mmu and irq fixes upstream and updated the solaris/sparc under qemu howto. Now all the fixes I had are in the vanilla qemu (git). Don't have anything else up my sleeve.

So please test everything you can and please send reports. Here and to the qemu-devel mailing list.

Happy 2010

I'm back from skiing. Happy 2010 everyone!

Updated the Solaris under qemu how-to, added launching instructions for the Solaris versions prior 2.6.

Meanwhile I'm working on the MMU emulation problems. It's harder than it looked. There is a documentation on the SuperSPARC Multi Cache Controller, which describes, what MMU does in a case of a double fault differently than it is currently implemented in qemu. Unfortunately it looks like either it describes it wrong, or I don't get what is written there (yes, I've seen it says "Subject to Change Without Notice" in the footer). At least I can not confirm the described behavior on the real SS-20.

So, there are 3 variants of the MMU behavior: qemu's, described, and the correct one. I'm exploring the last two to fix the first one.