Well, that too, but also nobody expects that a bland, run-of-the mill Novell NE2000 NDIS driver would crash/hang just because it runs on 486 or later CPUs.
I wanted to try the “basic” DOS redirector shipped with Microsoft’s LAN Manager 2.0 (1990) and more or less by a flip of a coin I decided to use the NE2000 NDIS driver that came with the package. Previously I had no trouble with Microsoft’s NE2000.DOS driver shipped with LAN Manager 2.1 and Microsoft’s Network Client 2.0.
But the old LAN Manager NE2000.DOS driver (16,342 bytes, dated 11-19-90, calls itself version 0.31) loaded and then promptly hung as soon as Netbind was started:
Netbind hangs with LAN Manager 2.0 NE2000 driver
At first I naturally suspected some problem with the card configuration or the NIC hardware, but what I found was much more surprising.
The reason the driver hung actually wasn’t related to networking at all. The driver hung in a routine that was clearly trying to detect the CPU type. How can someone screw something so simple so badly? Well…
Over the last few months I have been on and off digging into the history of early PC networking products, especially Ethernet-based ones. In that context, it is impossible to miss the classic NE2000 adapter with all its offshoots and clones. Especially in the Linux community, the NE2000 seems to have had rather bad reputation that was in part understandable but in part based on claims that simply make no sense upon closer examination.
A genuine Novell NE2000 card (1992) with DP83901
First let’s recap a bit. In late 1986, National Semiconductor introduced the DP8390/91/92 chip set including a complete Ethernet controller, encoder/decoder, and a transceiver. The DP8390 NIC was a relatively simple design, not as advanced as the Intel 82586 or AMD LANCE, but significantly more capable and cheaper than the low-end offering of the era, the 3Com 3C501 EtherLink.
For R&D purposes, I would very much like to get my hands on the circa 1991 Microsoft LAN Manager Network Device Driver Kit which was meant to support the development of NDIS 2.0 drivers. While it is obvious that some kind of development kit for NDIS 2.0 drivers must have existed, the exact name is actually known thanks to the Q80562 KB Article.
That same KB Article also mentions MTTOOL, a test tool that sounds very useful, but unfortunately I’ve not been able to find it anywhere. The tool itself would be helpful even without the rest of the kit.
The closest thing I could find is a 1993 NDDK (Network Device Development Kit) that supports NDIS 3.0 drivers for the Windows for Workgroups 3.11 environment. While the NDDK is valuable on its own, it is quite different and not immediately useful because it is oriented towards Windows NT and 32-bit environments, unlike the 16-bit NDIS 2.0 which supported DOS and 16-bit OS/2.
The old LAN Manager NDDK seems to have fallen through the cracks of the post-IBM-divorce chaos at Microsoft. It wasn’t documented in the older Microsoft Programmer’s Library and by the time MSDN was rolled out, the NDIS 3.0 NDDK was current. And because OS/2 had been disowned by then, Microsoft probably saw no need to widely distribute the older NDIS 2.0 kit.
Which is ironic because although NDIS 2.0 development might be finally dead now, it was not a few years ago, with Intel offering an updated DOS network driver package as late as 2019. The newest NDIS 2.0 driver in the set is dated December 28, 2015, which means NDIS 2.0 survived even past the Windows XP era, leave alone Windows 9x or Windows 3.x!
Update (June 23, 2021): Less than 3 months later, the 30+ year old NDDK popped up. It is now available here.
The other day I was trying to fill a couple of gaps in my understanding of the Intel 8237A DMA controller documentation. I wrote a small testcase that performed a dummy transfer and modified the base address and count registers in various ways, and then examined what happens to the current address and count registers.
I ended up with printing out the current DMA address and count at the beginning and end of the test. I noticed that the current address changed between test runs, which was quite unexpected. No one else should have been using the DMA channel and the current address can’t just randomly change.
8237A DMA is… fun?
The change itself wasn’t random at all: The current address was being set to the base address. That happens when the base address register is written, but I was pretty sure no one was doing that.
After much head scratching, I realized that my own code was triggering the change. I had some trivial code in place to save and restore the channel’s DMA page register, and it was restoring the page register that caused the current base address to change after the last state printout. That was definitely not expected to happen. So why was it happening?
For reference and for comparison, here’s the sketchy one:
A likely fake Adaptec 39160 SCSI HBA
The PCB is not quite the same (ASSY 1817206-00 vs. 1817206-01) but it’s close enough. The real one has none of the sketchy labeling—it says “CH 2/B” and not “CH 27B” and so on.
Last week the OS/2 Museum received a classic red NetWare box with all sorts of junk inside: PCI and ISA network cards (most Ethernet, one ArcNET), BNC cabling, one or two manuals, and over a 100 floppies, mostly NetWare but also a handful of 3Com driver disks.
There was a mix of 5.25″ and 3.5″ NetWare floppies, the 5.25″ ones in three original NetWare boxes but most of the 3.5″ disks just more or less loose in the big red crate. As I tried to organize the floppies, I quickly realized that it’s not that simple.
At a quick glance, there were floppies from several NetWare sets:
NetWare 2.2 on 5.25″ floppies
NetWare 3.11 on 5.25″ floppies
NetWare 3.11 on 3.5″ floppies, two sets
NetWare 3.12 on 3.5″ floppies
Now the difficulty with NetWare is that unlike, say, Microsoft or IBM, Novell didn’t just label all the disks in a box “NetWare 3.11”. There was in fact significant overlap and e.g. many disks were identical between NetWare 2.2 and 3.11, and later between 3.12 and 4.0. Related to that, NetWare didn’t refresh all disks for each update; only the disks that actually needed updating were changed. It was thus standard for a NetWare disk set to contain floppies with several different revisions.
Same or different?
That gets really complicated if you have a pile of disks and no easy way to tell which sets belong together. And it didn’t end there either: Novell shipped various add-on products with their systems, such as Macintosh clients, OS/2 Requesters, backup and mail software, and so on. And again, these were not labeled as belonging to a specific release, because they were to some extent independent.
The other day someone asked how hard it would be to modify the Open Watcom linker, wlink, to properly support exports from IOPL segments in OS/2 LX modules. Not terribly hard it turned out, all it needed was to emit a different “bundle type” for exports from IOPL segments. Rather than a regular 32-bit export, it needs to be a special 16-bit call gate export.
A very slight complication is that the 16-bit call gate export is naturally limited to 16-bit offsets, so the linker needs to error out if an attempt is made to export an entry at an offset 65,536 bytes or more into the segment/object. That is easy enough. There are also complications when calling into IOPL segments from the same module, but that’s a separate topic.
When I tried to do rudimentary testing of whether the linker now produces sensible output, I ran into an unexpected problem. For calling into any IOPL segment, in both NE and LX modules, OS/2 uses call gates. The Intel 286/386 call gate mechanism has a provision for the CPU to automatically copy a certain number of parameters (either words or dwords, depending on whether the call gate is 16-bit or 32-bit) when switching stacks. In OS/2, these call gates are always 16-bit, so there are parameter words optionally copied.
The NE and LX executable formats both use the top 5 bits of a flag byte to specify the number of parameter words (those five bits get copied into the CPU-defined call gate descriptor, which also has 5 bits for the parameter count). The Microsoft and IBM linkers (LINK/LINK386), sensibly enough, accept the number of parameter words through the EXPORTS directive.
Now, the Watcom linker already does that too… but things were not adding up. I realized that the Watcom IOPL_WORD_SHIFT macro is not defined as 3 as I expected, but rather as 2. So the number or parameter words in the resulting executable is wrong. Except… hang on a sec.
Several weeks ago I bought this Adaptec 39160 64-bit PCI SCSI HBA in order to experiment with different HBAs:
Adaptec 39160 SCSI HBA, perhaps 2007
The motivation was that although I’ve been a happy user of LSI HBAs (SCSI and SAS, PCI and PCIe) based on the MPT Fusion architecture, the Linux driver for the LSI SCSI HBAs is dumb, broken, or buggy, and ignores certain firmware settings. I have several older SCSI drives (early 1990s) which badly fail if the host tries to negotiate wide and/or fast transfers; the drive in some cases resets but usually hangs and must be power cycled.
All that can be avoided if the LSI HBA is set up to not use wide transfers and not negotiate fast transfers for the given SCSI ID. That works well with the LSI HBAs at POST time, but the Linux driver then just ignores the firmware settings and hangs the drive anyway—unlike the LSI Windows driver. So far I have not found any way to convince the Linux kernel drivers not to do that, and after perusing the source code I’m pretty sure there isn’t.
At any rate, I wanted to try an Adaptec SCSI HBA to check if it has the same problem in Linux (spoiler: it doesn’t) and ordered the 39160 pictured above off eBay. This is a nice dual channel HBA and (for an U160 SCSI controller) somewhat unusual for having both 68-pin and 50-pin connectors on the first channel; that feature looked useful.
The HBA I ordered was sold as new and at 10.50 Euro, it was a bargain. It arrived very promptly, in an original-looking Adaptec box, sealed in an anti-stat bag, and included a printed installation brochure as well as a driver CD.
The adapter works fine, and although it exhibits this extremely annoying error in my Intel Stormville board, I have no reason to think that this particular specimen is defective, especially given that the error does not show up on a Supermicro X7DWN+ board.
Whenever I get a new piece of hardware, I like to have a close look at it simply out of curiosity. When it was made? What chips were used? That’s when I noticed that the HBA just does not look like new.
While discussing Intel desktops with DDR2 memory using 2 Gbit technology (4 GB UDIMMs), the question of Intel’s next generation and 4 Gbit DDR3 (8 GB UDIMMs) came up. It’s more or less the next iteration of exactly the same problem.
I decided to focus on the LGA1366 platform which was the first post-Core 2 generation and Intel’s first platform that used exclusively DDR3 memory (recap: later Core 2 chipsets supported both DDR2 and DDR3). Reviewing Intel’s documentation turned out to be… confusing.
The datasheet for the first generation Bloomfield Core i7 (45nm Core i7 LGA1366, Nehalem microarchitecture) says that DDR3 “512Mb, 1Gb, 2Gb, technologies/densities are supported”, 24 GB RAM maximum. The memory controller in the CPUs could handle 6 DIMMs, and with 2 Gbit technology/4 GB UDIMMs, that obviously results in 24 GB maximum.
Can this 2008 desktop CPU support 48 GB RAM? Intel says no…
The datasheet for the follow-up Gulftown processors (32nm Core i7 LGA1366, Westmere microarchitecture) does not state what technologies or memory limits are supported. It was not hard to find reports of people using Westmere generation LGA1366 CPUs with 48 GB RAM so I started digging further.
I only have one desktop LGA1366 board that could possibly take 48 GB RAM, Intel’s own DX58SO2 (Smack Over 2). That’s where it gets interesting. The DX58SO2 Technical Product Specification clearly says 24 GB maximum. But the DX58SO2 Product Brief just as clearly says “Maximum system memory up to 48 GB using 8 GB double-sided DIMMs.” Who do you trust, Intel or Intel? Could they be both both wrong?
I spent a bit of time recently putting together technical documentation for Intel’s 4-series chipsets, partly motivated by research into Intel’s support of 4 GB DDR2 memory modules, partly driven by idle curiosity about one of Intel’s many hyped up and failed projects, Turbo Memory.
There are plenty of mentions of the Intel GM47 chipset. And that Turbo Memory article explicitly says that Intel’s Turbo Memory 2.0 is available on the Cantiga GM47 chipset as announced by Intel on July 15, 2008. Although the phrasing in the Wikipedia article may be misleading in that Turbo Memory 2.0 was probably available on the Cantiga (Mobile 4 Series) chipset and not just GM47. of course there’s no reference given for the statement… so who knows.
At any rate, there were certainly rumors that GM47 was to be released in 2008, and then in Q1 2009. But then 2009 came and there was no GM47 chipset. It appears to have gotten Intel’s typical second-class burial—a decree comes from up on high, the victim product is no longer mentioned, must not be spoken of, and references to it are scrubbed. Thus GM47 is notable only by its absence, say, here on Intel’s ark, or in the Mobile 4 Series Chipset Specification Update.
Perhaps the somewhat confusing status of the GM47 chipset’s existence is what led to curious statements like the one in this Italian Wikipedia article. “Cantiga (the 4-series mobile chipset) is marketed in three variants: GM47, GM45, GS45, PM45.” Now which one of the four listed was not actually marketed?
