fsset@bach.lerc.nasa.gov (Scott E. Townsend) (07/19/90)
I have another plea for insight regarding the Personal Iris VME implementation. (I posted before about system "freezes", now I have more data) I'll be glad to RTFM, if someone will tell me what FM I need to read -- "Guide To Writing Device Drivers For Silicon Graphics IRIS-4D Computer Systems" was interesting, but a bit sparse regarding the VME adapter implementation. Sorry for the length of this posting, I'm hoping the data will provide someone with a clue as to what's going on. The scenario is this: a Personal Iris has an HVE Engineering bus extender attached to it. The other end of the bus extander is attached to a VME-based 88000 system. Infrequently, but enough to be annoying and unreliable, the Iris will completely freeze -- no cursor movement, no life signs at the monitor at all. I have been able to perform runs all day without a hitch, only to come in the next day and freeze 3 times in the first hour. I have not been able to characterize what causes the freeze other than to say that reasonable to heavy user program loads seem to make it more frequent. Access to the remote VME system's memory is via the mmap() call. No DMA or interrupts are being passed from the 88K system to the Iris. (Known fact, the interrupt jumpers were removed and the 88K doesn't have any idea where the Iris memory is) (Bad assumption? could the 88K, as VME master, somehow blast into the Iris's memory & corrupt things so subtly that the only result is the trace below? I would expect wild & terrible things if I was scribbling on the kernel!) Here's a bus trace showing the problem. It starts out with the 88K owning the bus. Next, the Iris makes a bus request. Once it has it, the Iris never performs a bus transaction. Eventually the 88K requests the bus again. Once it completes its transaction, the Iris gets the bus back, but still never performs any bus transaction. This will continue until the Iris is reset. (The Iris is at bus requester level 0, the 88K at requester level 2. This trace is in the expansion chassis. Bus grant lines aren't shown active because the tracer is downstream of both masters) V M E B U S T R A C E Sampling mode : ASYNC 16MHz ! BUS ADDRESS DATA WRITE AS BERR BR* BG* BBSY BCLR IACK AM ! LEVEL * * * 3210 3210 * * * -----!--------------------------------------------------------------------- TRIG ! 0 FFE80002 FFFFFFFF 1 0 1 1111 1111 0 1 0 0D 0001 ! 0 FFE80002 FFFFFFFF 1 0 1 1111 1111 0 1 0 0D 0002 ! 0 FFE80002 FFFFFFFF 1 0 1 1111 1111 0 1 0 0D 0003 ! 0 FFE80002 FFFFFFFF 1 0 1 1111 1111 0 1 0 0D 0004 ! 0 FFE80002 FFFFFFFF 1 0 1 1111 1111 0 1 0 0D 0005 ! 0 FFE80002 FFFFFF1E 1 0 1 1110 1111 0 1 0 0D 0006 ! 0 FFE80002 FFFFFF00 1 0 1 1110 1111 0 1 0 0D 0007 ! 0 FFE80002 FFFFFF00 1 0 1 1110 1111 0 1 0 0D 0008 ! 0 FFE80002 FFFFFF00 1 0 1 1110 1111 0 1 0 0D 0009 ! 0 FFE80002 FFFFFF00 1 0 1 1110 1111 0 1 0 0D 0010 ! 0 FFE80002 FFFFFF00 1 1 1 1110 1111 0 1 0 0D 0011 ! 0 FFFFFFFE FFFFFFFF 1 1 1 1110 1111 1 1 1 3F 0012 ! 0 FFFFFFFE FFFFFFFF 1 1 1 1110 1111 1 1 1 3F 0013 ! 0 FFFFFFFE FFFFFFFF 1 1 1 1110 1111 1 1 1 3F 0014 ! 0 FFFFFFFE FFFFFFFF 1 1 1 1110 1111 1 1 1 3F 0015 ! 0 FFFFFFFE FFFFFFFF 1 1 1 1110 1111 1 1 1 3F 0016 ! 0 FFFFFFFE FFFFFFFF 1 1 1 1111 1111 0 1 1 3F 0017 ! 0 0FFFFFFE FFFFFFFF 1 1 1 1111 1111 0 1 1 3F 0018 ! 0 00FFFFFE FFFFFFFF 1 1 1 1111 1111 0 1 1 3F 0019 ! 0 00FFFFFE FFFFFFFF 1 1 1 1111 1111 0 1 1 3F 0020 ! 0 00FFFFFE FFFFFFFF 1 1 1 1111 1111 0 1 1 3F Note: all Iris accesses are in A24 space, so the address lines shown here prove that the Iris has recognized that it owns the bus & is asserting its bus drivers. (address & BBSY at least) (this pattern continues unchanged thru the next 57 samples) 0078 ! 0 00FFFFFE FFFFFFFF 1 1 1 1111 1111 0 1 1 3F 0079 ! 0 00FFFFFE FFFFFFFF 1 1 1 1011 1111 0 0 1 3F 0080 ! 0 00FFFFFE FFFFFFFF 1 1 1 1011 1111 0 0 1 3F 0081 ! 0 00FFFFFE FFFFFFFF 1 1 1 1011 1111 0 0 1 3F 0082 ! 0 00FFFFFE FFFFFFFF 1 1 1 1011 1111 0 0 1 3F 0083 ! 0 00FFFFFE FFFFFFFF 1 1 1 1011 1111 0 0 1 3F 0084 ! 0 00FFFFFE FFFFFFFF 1 1 1 1011 1111 0 0 1 3F 0085 ! 0 00FFFFFE FFFFFFFF 1 1 1 1011 1111 0 0 1 3F 0086 ! 0 00FFFFFE FFFFFFFF 1 1 1 1011 1111 0 0 1 3F 0087 ! 0 FFFFFFFE FFFFFFFF 1 1 1 1011 1111 1 0 1 3F 0088 ! 0 FFFFFFFE FFFFFFFF 1 1 1 1011 1111 1 1 1 3F 0089 ! 0 FFFFFFFE FFFFFFFF 1 1 1 1011 1111 1 1 1 3F 0090 ! 0 FFFFFFFE FFFFFFFF 1 1 1 1011 1111 1 1 1 2F 0091 ! 0 FFFF2008 FFFFFFFF 0 0 1 1111 1111 0 1 1 2D 0092 ! 0 FFFF2009 FFFF4040 0 0 1 1111 1111 0 1 1 2D 0093 ! 0 FFFF2009 FFFF4040 0 0 1 1111 1111 0 1 1 2D 0094 ! 0 FFFF2009 FFFF4040 0 0 1 1111 1111 0 1 1 2D 0095 ! 0 FFFF2009 FFFF4040 0 0 1 1111 1111 0 1 1 2D 0096 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0097 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0098 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0099 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0100 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0101 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0102 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0103 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0104 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0105 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0106 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0107 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0108 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0109 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0110 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0111 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0112 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0113 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0114 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0115 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0116 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0117 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0118 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0119 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0120 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0121 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0122 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0123 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0124 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0125 ! 0 FFFF2009 FFFF4040 0 0 1 1110 1111 0 1 1 2D 0126 ! 0 40404040 FFFFFFFF 0 1 1 1110 1111 0 1 1 0F 0127 ! 0 FFFFFFFE FFFFFFFF 1 1 1 1110 1111 1 1 1 3F 0128 ! 0 FFFFFFFE FFFFFFFF 1 1 1 1110 1111 1 1 1 3F 0129 ! 0 FFFFFFFE FFFFFFFF 1 1 1 1110 1111 1 1 1 3F 0130 ! 0 FFFFFFFE FFFFFFFF 1 1 1 1110 1111 1 1 1 3F 0131 ! 0 FFFFFFFE FFFFFFFF 1 1 1 1111 1111 0 1 1 3F 0132 ! 0 0FFFFFFE FFFFFFFF 1 1 1 1111 1111 0 1 1 3F 0133 ! 0 00FFFFFE FFFFFFFF 1 1 1 1111 1111 0 1 1 3F 0134 ! 0 00FFFFFE FFFFFFFF 1 1 1 1111 1111 0 1 1 3F 0135 ! 0 00FFFFFE FFFFFFFF 1 1 1 1111 1111 0 1 1 3F 0136 ! 0 00FFFFFE FFFFFFFF 1 1 1 1111 1111 0 1 1 3F 0137 ! 0 00FFFFFE FFFFFFFF 1 1 1 1111 1111 0 1 1 3F 0138 ! 0 00FFFFFE FFFFFFFF 1 1 1 1111 1111 0 1 1 3F 0139 ! 0 00FFFFFE FFFFFFFF 1 1 1 1111 1111 0 1 1 3F (this pattern repeated until the 88K wants the bus again, at this point that doesn't happen for over 200 samples) Any responses welcome, even if all you have is something improbable to try! -- ------------------------------------------------------------------------ Scott Townsend | Phone: 216-433-8101 NASA Lewis Research Center | Mail Stop: 5-11 Cleveland, Ohio 44135 | Email: fsset@bach.lerc.nasa.gov ------------------------------------------------------------------------