I'm close to convinced that there is something broken in the bowels of
the interrupt handling on MacPPC that drops interrupts.   I reported a
problem a while back where my system would go to a slow crawl when
trying to copy between two ATA drives when each was on a different
channel on a two-channel ATA card.  I put the drives on the *same* channel,
and the problem went away.

For me, the log would also get messages complaining about:

Mar 11 23:27:40 grace /netbsd:  type: ata tc_bcount: 8192 tc_skip: 0
Mar 11 23:27:49 grace /netbsd: pciide0:0:0: lost interrupt
Mar 11 23:27:49 grace /netbsd:  type: ata tc_bcount: 8192 tc_skip: 0
Mar 11 23:27:50 grace /netbsd: pciide0:1:0: lost interrupt
Mar 11 23:27:50 grace /netbsd:  type: ata tc_bcount: 8192 tc_skip: 0
Mar 11 23:28:31 grace /netbsd: pciide0:0:0: lost interrupt
Mar 11 23:28:31 grace /netbsd:  type: ata tc_bcount: 65536 tc_skip: 0
Mar 11 23:28:31 grace /netbsd: pciide0:1:0: lost interrupt
Mar 11 23:28:31 grace /netbsd:  type: ata tc_bcount: 65536 tc_skip: 0
Mar 11 23:28:40 grace /netbsd: pciide0:0:0: lost interrupt
Mar 11 23:28:40 grace /netbsd:  type: ata tc_bcount: 65536 tc_skip: 0

There is a thread in this list about this... "Re: Oops: pciide0:0:0: lost interrupt"

I have what looks like a related problem with my Cyclades PCI 8=port
serial card, and I have had to put in an ugly "polling" kludge to
make it work.  Same issue, lost interrupts.

-dgl-

At 4:39 PM -0400 5/8/04, John Klos wrote:
>Hi,
>
>I've just noticed something strange today. After removing my 4 drive
>RAIDframe set (interactive performance under disk load is not good), I put
>in two large IDE disks, each on its own channel on a PCI card (PCI0648)
>and two SCSI disks.
>
>When copying between the two IDE disks, I saw network transfers drop from
>around 1.2 megs a second (between my machine and ftp.netbsd.org) to less
>than 20k a sec! I stopped and restarted the copy several times, tried
>nicing it, but it's definitely happening. When copying between SCSI and
>IDE, the speed only drops to around 700k - 800k sec. SCSI to SCSI shows no
>network slowdown (as expected).
>
>I know that IDE has more overhead than SCSI, but this is crazy! Does
>anyone else run multiple IDE devices without problems? If so, please tell
>me what kind of IDE card you're using and how you've tested it.
>
>More information:
>The system is a PM 9600, 400 MHz 604ev, lots of memory, running 1.6.2.
>IDE:
>pciide0 at pci1 dev 13 function 0: CMD Technology PCI0648 (rev. 0x01)
>pciide0: bus-master DMA support present
>pciide0: primary channel configured to native-PCI mode
>pciide0: using irq 27 for native-PCI interrupt
>SCSI:
>esiop0 at pci0 dev 13 function 0: Symbios Logic 53c896 (ultra2-wide scsi)
>esiop0: using on-board RAM
>esiop0: interrupting at irq 23
>scsibus0 at esiop0: 16 targets, 8 luns per target
>esiop1 at pci0 dev 13 function 1: Symbios Logic 53c896 (ultra2-wide scsi)
>esiop1: using on-board RAM
>esiop1: interrupting at irq 23
>Both IDE drives are running at Ultra-66 speed:
>wd0 at pciide0 channel 0 drive 0: <Maxtor 6Y120P0>
>wd0: drive supports 16-sector PIO transfers, LBA addressing
>wd0: 114 GB, 238216 cyl, 16 head, 63 sec, 512 bytes/sect x 240121728 sectors
>wd0: drive supports PIO mode 4, DMA mode 2, Ultra-DMA mode 6 (Ultra/133)
>wd0(pciide0:0:0): using PIO mode 4, Ultra-DMA mode 4 (Ultra/66) (using DMA data transfers)
>pciide0: secondary channel configured to native-PCI mode
>wd1 at pciide0 channel 1 drive 0: <Maxtor 7Y250P0>
>wd1: drive supports 16-sector PIO transfers, LBA48 addressing
>wd1: 233 GB, 486344 cyl, 16 head, 63 sec, 512 bytes/sect x 490234752 sectors
>wd1: drive supports PIO mode 4, DMA mode 2, Ultra-DMA mode 6 (Ultra/133)
>wd1(pciide0:1:0): using PIO mode 4, Ultra-DMA mode 4 (Ultra/66) (using DMA data transfers)
>
>Thanks,
>John Klos