Serial Port (very old card)

[wblock]

Received the card, and after the normal fight with serial connections and null-modem conversion, verified that it works with stock FreeBSD 11.1. This was just connecting one port to another and running two copies of cu(1), one on each port:

cu -l /dev/cuau2 -s 115200 in one console
Then selecting another console with Alt-F2, then running cu -l /dev/cuau3 -s 115200

Installing FreeNAS 11.0, I was able to see /dev/cuau0 through /dev/cua03. The first two are the motherboard ports, the last two are the Lava card.

The same test with cu worked on the FreeNAS system, no additional kernel modules required. I'll install an earlier version of FreeNAS to see whether this was recently added.

Tested with FreeNAS 11.0-RELEASE, and the test between the two ports worked.

 

[wblock]

The card, for reference:

 

[Ericloewe]

What a convoluted design. The TI chip is a fairly standard two serial+one parallel controller, but it needs glue logic to interface with basically any system. So, the manufacturer of the board went with their own ASIC, which is actually just a PCI parallel port adapter. So, apparently, they must program it with a different ID and then the driver does all the work of interfacing with the actual serial port controller.

I hope that their PCIe products aren't just this hacky solution behind a PCIe-PCI bridge IC...

 

[NASbox]

Received the card, and after the normal fight with serial connections and null-modem conversion, verified that it works with stock FreeBSD 11.1. This was just connecting one port to another and running two copies of cu(1), one on each port:

cu -l /dev/cuau2 -s 115200 in one console
Then selecting another console with Alt-F2, then running cu -l /dev/cuau3 -s 115200

Installing FreeNAS 11.0, I was able to see /dev/cuau0 through /dev/cua03. The first two are the motherboard ports, the last two are the Lava card.

The same test with cu worked on the FreeNAS system, no additional kernel modules required. I'll install an earlier version of FreeNAS to see whether this was recently added.

Tested with FreeNAS 11.0-RELEASE, and the test between the two ports worked.

Thanks very much for this... yes this is the same card, except that mine is a much earlier vintage-the ASIC on my card is physically larger, but fulfuills the same . I have contacted the company (and to their credit, they said they would check to see if they had some early documentation to determine if the cards were compatable).

For reference here is a photo.

What a convoluted design. The TI chip is a fairly standard two serial+one parallel controller, but it needs glue logic to interface with basically any system. So, the manufacturer of the board went with their own ASIC, which is actually just a PCI parallel port adapter. So, apparently, they must program it with a different ID and then the driver does all the work of interfacing with the actual serial port controller.

I hope that their PCIe products aren't just this hacky solution behind a PCIe-PCI bridge IC...

This card dates way back, and I suspect that they use almost exactly the same design for the dual serial card as the serial/paralell card. You can see the similarities between the old/new card--just more "up-to-date" parts - i.e. drivers are in 2 chips instead of 4, ASIC is smaller, UART is a later revision of the old 16550. Given how limited a market there is for this type of card, I'd suspect they wanted to keep the engineering costs to an absolute minimum. I've had this card sitting on the shelf for about 10-15 years, and only now that it's only FreeNAS that I have any use for a "real" serial port. Other than a small instrumentation or linux/freebsd installations where simplicity/low cost are wanted and short distances are involved would serial ports make any sense today.

The only reasons that I'm looking at a serial port are:

1. The hope that a serial port becomes active soon enough into the boot process to be able to see a lot of the boot messages on a terminal.
   
2. Cheap out of band management to enable intervention if the IP stack or SSH daemon stops working.

I'm assuming that a dumb serial port should come up a lot faster than SSH/Network services.

I know I won't get the whole boot process, but I'd assume that I'd get the last half or so.

Can anyone advise/comment? Thanks in advance.

 

[Ericloewe]

Well, that's a rather different design. It uses the same serial/parallel interface IC (well, a compatible part anyway), but has an FPGA for the glue logic function (which needs that itty-bitty PROM in the corner for its configuration). And the RS-232 drivers are split into two ICs...
Holy crap, I don't even want to know how much one these cost compared to the newer model. No wonder they felt that throwing their parallel port ASIC at the problem was a good idea.

 

[NASbox]

Well, that's a rather different design. It uses the same serial/parallel interface IC (well, a compatible part anyway), but has an FPGA for the glue logic function (which needs that itty-bitty PROM in the corner for its configuration). And the RS-232 drivers are split into two ICs...
Holy crap, I don't even want to know how much one these cost compared to the newer model. No wonder they felt that throwing their parallel port ASIC at the problem was a good idea.

I'll bet that design dates from the 1990s (whenever PCI took over from ISA) - when we actually used a lot more serial cards back in those days, network hardware cost a fortune, ICs cost a lot more than they do today, and electronics was a lot more expensive and limited.

 

[NASbox]

Thanks to the kindness of wblock, for exchanging cards with me, I now have a more up-to-date card that is recognized by FreeBSD. In the interest of creating a useful trail for anyone with a similar problem that doesn't force them to wade through tons of irrelevant posts, I'm going to consider this thread closed since I am now working with a different card. If wblock wants to post something regarding my old card, naturally he is free to continue this thread.