dfk@unido.UUCP (01/20/86)
I'm posting this because Snoopy's site has news problems. PLEASE RESPOND TO HIM, NOT TO ME! -Daniel ============================================================================== Dear Net, I am currently attempting to put together a MicroVAX2 with lots of mixed hardware. I have one problem doing this to do with the eagle disc drive. Mega-Disclaimer: I am a 4.2 BSD person. This is the first time I have ever laid my hands on an ULTRIX driven keyboard. Especially the new disc partitioning scheme where you can repartition discs quickly (*very, very nice*) is entirely new to me. Perhaps I am doing something blatantly silly, but I don't know about it. So here is the problem: the MV2 comes with Ultrix 1.1 (binary only) and what I have tried to do is bring up this new Eagle disc with the Emulex controller. Here is the config file: # # MicroVAX 2 configured and butchered by snoopy # machine mvax cpu "MVAX" ident qtecvax timezone -1 dst 4 maxusers 16 maxuprc 25 physmem 16 config vmunix root on ra0 swap on ra0 controller uba0 at nexus ? controller uda1 at uba0 csr 0160344 vector udintr controller uda0 at uba0 csr 0172150 vector udintr disk ra0 at uda0 drive 0 disk ra1 at uda0 drive 1 disk ra2 at uda0 drive 2 disk ra3 at uda0 drive 3 disk ra4 at uda1 drive 4 device dhu0 at uba0 csr 0160440 flags 0xff vector dhurint dhuxint device dhu1 at uba0 csr 0160460 flags 0xff vector dhurint dhuxint controller zs0 at uba0 csr 0172520 vector tsintr device ts0 at zs0 drive 0 pseudo-device pty ##############end of config file So my problem is: the new eagle has to be an ra disc (for the right driver to be used). Ra81's are not the right size (too small). What I have therefore tried to do was to redeclare the sizes of the ra81 disc in /etc/disktab (which is where I think the driver gets the information from) to follow the eagle partitions mentioned in hp(4). Here is the bit of the disktab file: ra81|RA81|RA81 with eagle values :\ :ty=winchester:ns#48:nt#20:nc#842:rm#3961:\ :pa#15884:ba#8192:fa#1024:\ :pb#66880:pc#808320:\ :pd#15884:bd#4096:fd#1024:\ :pe#55936:pf#109248:pg#432768:\ :be#4096:fe#1024:bf#4096:ff#1024:\ :bg#4096:fg#1024:\ :ph#291346:bh#8192:fh#1024: I then used /etc/chpt (change partitions) to bung those values into the new discs superblock. Here is the output of the /etc/chpt -q /dev/rra4a: /dev/rra4a Current partition table: partition bottom top size overlap a 0 15883 15884 c b 16320 83199 66880 c c 0 808319 808320 a,b,d,e,f,g,h d 375360 391243 15884 c,g e 391680 447615 55936 c,g f 698880 808126 109248 c,g g 375360 808126 432768 c,d,e,f h 83520 374865 291346 c HERE IT COMES: the eagle disc works fine up to a point. I can make a file system on partitions a,b,c,d,e,h. When I try to make a filestore (using /etc/newfs /dev/rra4x ra81) on the F or G partitions I get an error message: Write failed sn xxxxx on /dev/rra4y where y is the partition concerned and N.B. xxxxx is a number which is ALWAYS THE PARTITION SIZE MINUS ONE ! This is true even if I change the partition sizes. (I tried decreasing them by updating disktab and using chpt again). Questions: It seems that ULTRIX will not let me write into the last tracks of the disc. All the other partitions work fine. I have a faint idea that there might a problem with space being reserved by the driver for the bad block table. However the partitions are declared finished at 808126 and Eagles are 808320 large. Whats going wrong ? Whats happening ? APOLOGY: Please forgive the length. I have tried to state the problems with background info as detailed as possible so that you know whats going on. THANKS: for your help if you can give me some. I would really like to use that machine soon. PLEASE HELP NOW. Lots of love Seb "There are three types of people: - those who make things happen - those who watch things happening - and those who wonder what happened" ...\!mcvax\!unido\!ecrcvax\!snoopy /* N.B. valid csh address */
dan@rna.UUCP (Dan Ts'o) (01/22/86)
In article <17100006@unido.UUCP> dfk@unido.UUCP writes: > >I'm posting this because Snoopy's site has news problems. >PLEASE RESPOND TO HIM, NOT TO ME! > >-Daniel > >============================================================================== > >Dear Net, > >I am currently attempting to put together a MicroVAX2 with lots >of mixed hardware. I have one problem doing this to do >with the eagle disc drive. > >Mega-Disclaimer: I am a 4.2 BSD person. This is the first time >I have ever laid my hands on an ULTRIX driven keyboard. >Especially the new disc partitioning scheme where you can >repartition discs quickly (*very, very nice*) is entirely new >to me. Perhaps I am doing something blatantly silly, but I >don't know about it. > >So here is the problem: the MV2 comes with Ultrix 1.1 (binary >only) and what I have tried to do is bring up this new Eagle >disc with the Emulex controller. Here is the config file: Maybe I don't know something, but I believe that Ultrix 1.1 is 4.2BSD, even in its disk partitioning software. I am not aware of any major changes that allow you to repartition the disks any quicker... In the following reply, I will assume you are using an Eagle with an Emulex SC03/MS. If that is not the case, this stuff may be way of base... >controller uda1 at uba0 csr 0160344 vector udintr Incidentally, to bring it up fast, you can ADB /vmunix, _ubminit and patch in the right address instead of rebuilding the kernel. > >So my problem is: the new eagle has to be an ra disc (for the >right driver to be used). Ra81's are not the right size (too >small). What I have therefore tried to do was to redeclare the Wrong, RA81's are too large... >Questions: >It seems that ULTRIX will not let me write into the last tracks >of the disc. All the other partitions work fine. I have a faint >idea that there might a problem with space being reserved by >the driver for the bad block table. However the partitions are >declared finished at 808126 and Eagles are 808320 large. You are assuming that the SC03/MS gives you the full Eagle. But the manual says you only get 722,770 (I think). One way to test it is to use dd to read the C partition til the end. The driver asks the controller what the size is. The sectors/track and tracks/cylinder specs in /etc/disktab have no use on a MSCP disk. I believe the (ridiculous) numbers given for the real RA81 are just numbers which happen to multiply out correctly to the real size. BTW, Emulex just told me that a firmware fix is available (Rev D) to allow the SC03/MS to boot Ultrix 1.1. Previous, the SC03/MS could only be a secondary (well, non-bootable) disk (although MicroVMS booted fine...) I hope this helps. Cheers, Dan Ts'o Dept. Neurobiology Rockefeller Univ. 1230 York Ave. NY, NY 10021 212-570-7671 ...cmcl2!rna!dan rna!dan@cmcl2.arpa
jsdy@hadron.UUCP (Joseph S. D. Yao) (01/22/86)
In article <465@rna.UUCP> dan@rna.UUCP (Dan Ts'o) writes: >In article <17100006@unido.UUCP> dfk@unido.UUCP writes: >> ... >>I'm posting this because Snoopy's site has news problems. >> ... >>So here is the problem: the MV2 comes with Ultrix 1.1 (binary >>only) and what I have tried to do is bring up this new Eagle >>disc with the Emulex controller. Here is the config file: > Maybe I don't know something, but I believe that Ultrix 1.1 >is 4.2BSD, even in its disk partitioning software. I am not aware of >any major changes that allow you to repartition the disks any quicker... Sorry, Dan, this is one of the two major changes DEC made to the Ultrix hp driver. You know where it has said in the hp driver "this should be read in from disk"? DEC went and did it. >>So my problem is: the new eagle has to be an ra disc (for the >>right driver to be used). Ra81's are not the right size (too >>small). What I have therefore tried to do was to redeclare the Actually, I don't understand why it has to look like an RA81. For one thing, RA81's have no hp-type interface: they are MSCP drives. The size of an RA81 is: :ty=winchester:ns#51:nt#14:nc#1248:pc#891072: The size of an Eagle is: :ty=winchester:ns#48:nt#20:nc#842:rm#3961:pc#808320: Interestingly, in our ISI m68k disktab, I also find: beagle|Snoopy| Brain-Damaged Eagle (47 sectors courtesy of SC03):\ :ty=winchester:ns#47:nt#20:nc#842:rm#3961:pc#791480: If this is what Snoopy is using, that certainly explains why it loses! The entire entry is at the end of this posting. Although this wasn't going to be my original suggestion, this would certainly explain why he can't read the last sector of something: note that, as opposed to a real Eagle, this has 47 sectors/track! (why, i don't know, but the tab does say 'B-D'.) I was originally going to mention that the 4.2 and Ultrix hp drivers have at least two serious bugs. (These are changes that DEC did not make.) One is that the driver promises to support Emulex controllers with an RM02 drive type ("Impossible, so ask me what my cyl/trk/sec parameters really are"), but crashes any system that tries it. The other is that attempts to read or write the raw disk using a number of blocks that does not evenly divide the size of the file system will return a read or write error on the last (partial) read. This can really mess up your file system. Fixes for Ultrix have been posted to DEC (decuac, and soon decvax), and fixes for 4.2 will be posted to bugs-4bsd as soon as I can retroport and get some red tape untangled. Oh, let me not forget to append: beagle|Snoopy| Brain-Damaged Eagle (47 sectors courtesy of SC03):\ :ty=winchester:ns#47:nt#20:nc#842:rm#3961:\ :pa#15884:ba#8192:fa#1024:\ :pb#66880:pc#791480:\ :pd#15884:bd#8192:fd#1024:\ :pe#307200:be#4096:fe#1024:\ :pf#92872:bf#8192:ff#1024:\ :pg#416232:bg#4096:fg#1024:\ :ph#291346:bh#8192:fh#1024: [and just fer what it's worth -- echhh:] erp06|Extended rp06 for Eagle:\ :ty=winchester:ns#44:nt#19:nc#1772:\ :pa#15884:ba#8192:fa#1024:\ :pb#66880:\ :pc#1481392:\ :pd#15884:bd#8192:fd#1024:\ :pe#307200:be#4096:fe#512:\ :pf#783156:bf#4096:ff#1024:\ :pg#1106688:bg#4096:fg#1024:\ :ph#291346:bh#4096:fh#512: -- Joe Yao hadron!jsdy@seismo.{CSS.GOV,ARPA,UUCP}
dan@rna.UUCP (Dan Ts'o) (01/27/86)
>>Actually, I don't understand why it has to look like an RA81. For >one thing, RA81's have no hp-type interface: they are MSCP drives. > >The size of an RA81 is: > :ty=winchester:ns#51:nt#14:nc#1248:pc#891072: > >The size of an Eagle is: > :ty=winchester:ns#48:nt#20:nc#842:rm#3961:pc#808320: > >Interestingly, in our ISI m68k disktab, I also find: >beagle|Snoopy| Brain-Damaged Eagle (47 sectors courtesy of SC03):\ > :ty=winchester:ns#47:nt#20:nc#842:rm#3961:pc#791480: I believe, Joe, you're thinking of an SC03, not an SC03/MS, which IS an MSCP controller. Thus the hp driver is irrelevant. I believe your partitions are fine for an SC03, which is like a Qbus SC750, SC7000, etc. However, the SC03/MS looks like a UDA-50. An Eagle on the SC03/MS is 722770 sectors long. MSCP disk software doesn't care about sectors/track tracks/cylinder. It is a bit crazy that the SC03/MS loses so much of the Eagle, > 10%. I can understand using some for bad sector replacement, but... Until recently, the SC03/MS wouldn't even boot under Ultrix-32m. That meant/means it might be reasonable to actually use a SC03 and get more of the Eagle (the SC03 won't boot either since the MVAX II doesn't have the boot ROMS for an HP disk, although you could write one, and ...). If you're thinking of going that route, I would suggest looking into another Qbus - HP-type controller, from Spectralogics. They will get much higher performance from the Eagle since they can do 1:1 sector interleaving (I think the SC03 is 3:1) and they have block-mode DMA's, which the SC03 doesn't. Cheers, Dan Ts'o Dept. Neurobiology Rockefeller Univ. 1230 York Ave. NY, NY 10021 212-570-7671 ...cmcl2!rna!dan rna!dan@cmcl2.arpa