port-amiga/41456: "mfcs0: fifo overflow" when receiving serial data, caused by Picasso IV scrolling (?)

[frank%phoenix.owl.de@localhost]

>Number:         41456
>Category:       port-amiga
>Synopsis:       "mfcs0: fifo overflow" when receiving serial data, caused by 
>Picasso IV scrolling (?)
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    port-amiga-maintainer
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Mon May 18 21:25:00 +0000 2009
>Originator:     Frank Wille
>Release:        NetBSD/amiga 5.0 GENERIC
>Organization:
>Environment:
NetBSD aquila 5.0 NetBSD 5.0 (GENERIC) #0: Sun Apr 26 08:44:17 UTC 2009  
builds%b3.netbsd.org@localhost:/home/builds/ab/netbsd-5-0-RELEASE/amiga/200904260229Z-obj/home/builds/ab/netbsd-5-0-RELEASE/src/sys/arch/amiga/compile/GENERIC
 amiga
>Description:
At some point after the 4.0.1 release the serial interface on a Multiface II 
board can no longer receive streams of data, while displaying them on a 
PicassoIV-screen, without a FIFO overflow.

It will show up in /var/log/messages as:
May 15 16:57:37 aquila /netbsd: mfcs0: fifo overflow 
May 15 16:58:06 aquila last message repeated 83 times
May 15 16:59:42 aquila last message repeated 240 times

The problem might have something to do with the PicassoIV scroll routine. When 
displaying the received serial data over a telnet-connection there is no 
overflow. Also, when I clear the screen and no scrolling is required to display 
all received lines, there is also no overflow! Only when output reaches the 
bottom of the screen and scrolling is required.

Sending data over the MFC is no problem (perhaps because there is no scrolling 
required when sending?).

I have reproduced the overflow with 4.99.63, 5.0 and 5.99.11.
The problem does not occur with 4.0 and 4.0.1.

Full dmesg:

Amiga 4000 (68060 rev.5 CPU/MMU/FPU)
total memory = 97792 KB
avail memory = 92520 KB
memory segment 0 at 08000000 size 05f80000
memory segment 1 at 07000000 size 01000000
memory segment 2 at 00000000 size 00200000
timecounter: Timecounters tick every 10.000 msec
mainbus0 (root)
clock0 at mainbus0: CIA B system hz 100 hardware hz 709379
timecounter: Timecounter "CIA B" frequency 709379 Hz -- Insufficient hz, needs 
at least 190
a34kbbc0 at mainbus0
ser0 at mainbus0: input fifo 512 output fifo 32
par0 at mainbus0
kbd0 at mainbus0: CIA A type Amiga
ms0 at mainbus0
grfcc0 at mainbus0
grf0 at grfcc0: width 640 height 400 colors 4
ite0 at grf0: rows 0 cols 0 repeat at (30/100)s next at (10/100)s has keyboard
fdc0 at mainbus0: dmabuf pa 0x1e12c0: dmabuf ka 0x5ab2c0
fd0 at fdc0 unit 0: 3.5dd 80 cyl, 2 head, 11 sec [9 sec], 512 bytes/sec
wdc0 at mainbus0
atabus0 at wdc0 channel 0
aucc0 at mainbus0
audio0 at aucc0: half duplex
zbus0 at mainbus0: i/o size 0x02000000
grfcl0 at zbus0 pa 0x40000000 man/pro 2167/24
grf3 at grfcl0: width 640 height 480 colors 16
ite3 at grf3: rows 60 cols 80 repeat at (30/100)s next at (10/100)s has keyboard
grfcl: 4MB Picasso IV being used
le0 at zbus0 pa 0xe90000 man/pro 2167/201: address 00:60:30:00:20:38
le0: 16 receive buffers, 4 transmit buffers
mfc0 at zbus0 pa 0xea0000 man/pro 2092/18
mfcs0 at mfc0: input fifo 1024 output fifo 128
mfcs1 at mfc0: input fifo 1024 output fifo 128
cbiiisc0 at zbus0 pa 0xf0102c man/pro 8512/100
cbiiisc0:  no SCSI termination, host adapter deactivated.
scsibus0 at cbiiisc0: 0 targets, 8 luns per target
timecounter: Timecounter "clockinterrupt" frequency 100 Hz quality 0
scsibus0: waiting 2 seconds for devices to settle...
wd0 at atabus0 drive 0: <QUANTUM FIREBALL_TM3200A>
wd0: drive supports 16-sector PIO transfers, LBA addressing
wd0: 3067 MB, 6232 cyl, 16 head, 63 sec, 512 bytes/sect x 6281856 sectors
wd0: drive supports PIO mode 4, DMA mode 2
wd1 at atabus0 drive 1: <FUJITSU MPC3032AT>
wd1: drive supports 16-sector PIO transfers, LBA addressing
wd1: 3093 MB, 6704 cyl, 15 head, 63 sec, 512 bytes/sect x 6335280 sectors
wd1: drive supports PIO mode 4, DMA mode 2, Ultra-DMA mode 2 (Ultra/33)
4 views configured
Kernelized RAIDframe activated
root on wd0a dumps on wd0b
root file system type: ffs

>How-To-Repeat:
You need a Zorro3-Amiga with MultifaceII and PicassoIV board.

Connect a second system (in this example a NetBSD/i386 system) via a null-modem 
cable to the first serial port of the MFC.

On the MFC-system make an entry for the MFC in /etc/remote with 9600bps:
mfc0:dv=/dev/ttyA0:br#9600:pa=none:dc:

Then start the connection while the cursor is at the bottom of the PicassoIV 
console:
# tip mfc0

Send a short text file on the second system, e.g.:
# cat /etc/hosts > /dev/tty00

/var/log/messages will immediately contain "fifo overflow", while there are 
characters missing in the transfered data. Example:
#      $NetBSD: hosts,v 1.6.4.1 2004/08/30 10:08:02 tron Exp $
# Hotabase
# Thi file should contain the addresses and aliases
# for local hots that share this file.
[...]

>Fix:
Unknown.

I tried to revert amiga/dev/grf_cl.c to 1.39, but that didn't change anything.