[Info-vax] Processes hang on USB-access

[abrsvc]

On Thursday, February 17, 2022 at 4:49:44 PM UTC-5, Dave Froble wrote:
> On 2/17/2022 2:27 PM, abrsvc wrote: 
> > On Thursday, February 17, 2022 at 2:07:29 PM UTC-5, Dave Froble wrote: 
> >> On 2/17/2022 4:15 AM, Joukj wrote: 
> >>> Hi all, 
> >>> 
> >>> I have an 2 TB USB-disk attached to a DPWS500 running OpenVMS Alpha V8.4-2L1. 
> >>> After a couple of hours to couple of days the process reading/writing to the 
> >>> disk seems to hang on the disk access. When some other process tries to access 
> >>> this disk in this state, it also hangs. A normal reboot of the machine is not 
> >>> possible anymore since also the dismount of the disk hangs. I have to reset the 
> >>> machine. 
> >>> 
> >>> No errors are logged for the device. 
> >>> 
> >>> What could be the problem and how can I avoid processes hanging on access to 
> >>> this disk? 
> >>> 
> >>> I already changed the USB-disk for a brand new one and I changed the machine to 
> >>> which it is connected. I think I also tried a XP1000. 
> >>> 
> >>> 
> >>> regards 
> >>> Jouk 
> >> I seem to recall that the USB implementation on older systems from DEC was less 
> >> than great. You may discover the problem with a system dump and SDA, but, then 
> >> what? 
> >> 
> >> -- 
> >> David Froble Tel: 724-529-0450 
> >> Dave Froble Enterprises, Inc. E-Mail: da... at tsoft-inc.com 
> >> DFE Ultralights, Inc. 
> >> 170 Grimplin Road 
> >> Vanderbilt, PA 15486 
> > 
> > As with any "problem", knowing what exactly is wrong yields a potential solution. With a nebulous report of "hangs", there is nothing to go on here. Is the process waiting for an I/O completion that was missed or never occurred? An examination of the dump would answer this question. Perhaps there is a code path that results in a deadlock and that is what happened here. Who knows without additional investigation? 
> >
> Correct Dan, but then what? Re-write the driver, and add to VMS? Perhaps a 
> non-USB SSD would be an easier "fix"? 
> 
> It could also be the USB hardware.
> -- 
> David Froble Tel: 724-529-0450 
> Dave Froble Enterprises, Inc. E-Mail: da... at tsoft-inc.com 
> DFE Ultralights, Inc. 
> 170 Grimplin Road 
> Vanderbilt, PA 15486

That is true.  But until the problem is understood, it is pure speculation on the  solution.  It could be a really simple fix.  I recall a problem that was presented to me as a Cobol program that would "skip" a record in an indexed file using the START statement followed by a READ NEXT.  Took almost a month to finally fix the problem that turned out to be a missing dot "." in the code for a CMPC3 instruction for Alpha.  The problem would only be seen when the operands spanned an alpha page boundary and a few other things I no longer recall.  Difficult to find, simple to fix.  Could be the case here too.