gc@vax135.UUCP (Eugene Cristofor) (03/08/85)
There is a bug in the 4.2 disk driver: bad sector forwarding does not work for EAGLEs with System Industries controller. Problem is that the SI controller clobbers the MBA registers - one must use the drive registers `hpdc' and `hpda' to find where the BSE error has occurred. A fix is available - all changes are in vaxmba/hp.c gene cristofor vax135!gc
ed@mtxinu.UUCP (Ed Gould) (03/22/85)
> There is a bug in the 4.2 disk driver: bad sector forwarding > does not work for EAGLEs with System Industries controller. > Problem is that the SI controller clobbers the MBA registers - > one must use the drive registers `hpdc' and `hpda' to find > where the BSE error has occurred. The problem is in the SI controller, not the driver. Various of us have been complaining to SI for *years* to get their controllers to emulate the DEC controllers correctly. > A fix is available - all changes are in vaxmba/hp.c A workaround, perhaps. The fix would need to be in the controller itself. SI has a driver that they'll supply. It has code in it that comes from Berkeley, from me, from Si Field, and others to deal with their controller. Their driver is about 70% different, on a line-by line basis (counting lines with differences), than the 4.2 driver. I doubt that it will correctly work with either DEC or Emulex controllers so for some of us, it's not an option. > gene cristofor > vax135!gc -- Ed Gould mt Xinu, 739 Allston Way, Berkeley, CA 94710 USA {ucbvax,decvax}!mtxinu!ed +1 415 644 0146
dfk@unido.UUCP (03/23/85)
About the same thing happened when we got one of the first SC7000s from Emulex. The MBA bytecount was off after hitting a bad block (BSE). This doesn't happen on a SC750 we have in another machine. We checked with another site that had an Eagle/SC7000 pair just installed: Same symptoms. So we assumed a firmware bug, put a quick fix in hp.c and got on the phone to emulex. We were soon talking to someone very competent at their London office. Although there were some relaying problems Dortmund <-> London <-> USA they made very sure to understand the problem. Not one word about driver kludges and "it works with VMS". If it didn't work like DEC they wouldn't argue. About six weeks later (remember this is Europe) we had new firmware and now it works. -Daniel Karrenberg <dfk@unido.uucp> NB: I am not affiliated with Emulex, just a satisfied customer.
jeff@voder.UUCP (Jeff Gilliam) (04/06/85)
There seems to be some confusion regarding the problems the 4.2 hp disk driver has when used with Systems Industries controllers. The difficulty is that the SI controller may buffer up to 16 sectors internally. When a BSE or ECC error occurs the mba byte count register is correct, but not relevant to where the *error* is. The driver can be changed to calculate the error location from the hpdc and hpda registers; this method works for both SI controllers and DEC controllers. The total changes (plus another fix to increase throughput by disabling the lookahead calculation) are only about a dozen lines. Send email if you're interested. -- Jeff Gilliam {ucbvax,ihnp4!nsc}!voder!jeff