Mojave, Thunderbolt-3 and SeriTek/1S2 (1SE2)

[(S)ATAman]

Does work with the new driver, but be aware: there are two PCI-PCIe bridges.

After having a disastrous experience with reverse bridging of AHCI controller on G4 (with a common bridge found on eBay), same thing applies with forward bridging.

The cheap $10-$25 bridge boards use ASMedia 1083 bridge. That bridge is know to be faulty, so the manufacturer can buy it very cheap - and make a very cheap, faulty product. There is still a profit even if it is sold at $10. Check it on eBay!

None of them, be it PCIe -> PCI or PCI -> PCI-e does work.

The PCIe -> PCI "produces" a MacsBug screen on "9" and instant kernel panic (once you address the bridge) under every version of macOS-X.
(Chose, what do you like more!).

The PCI -> PCIe will make you feel, it works... for few minutes. The problem is that ASMedia 1083 will screw up the DMA totally.
This is documented in Linux, but I am not going do fix the PCIe driver on macOS for a bridge, which made people like Tejun or Linus mad on their life.

Luckily, the products sold under Startech name DO WORK.
I tried both bridges, so far it looks good.

Did see SeriTek under Mojave today the first time properly.

The performance was not that great, just acceptable (for 3112). 3114 is not acceptable in any form, wonder, why SImage made it - but it's going to be supported, just make sure, your expectations aren't higher, than ca. 63 MB/Sec - while 3112 is clearly over 100 MB/Sec.

I think, we need to demand a good solution from manufacturers without bridges known to be bad.

[IIO]

ich hör immer "mojave", es ist wohl mal zeit für eine ohrenreinigung.

[(S)ATAman]

ich hör immer "mojave", es ist wohl mal zeit für eine ohrenreinigung.

Bad builds, the quality since Snow Leopard is on constant decline.
But it is important to see the modern LLVM compilers cope with old code.

The LLDB debugging is better, than of GDB, dealing with GDB is a pain in the rear.
It is far easier to debug the code in Mojave than, say, in Jaguar or Panther (and forget earlier ones).

The first macOS-X where debugging is started to be serious is Tiger.
Unfortunately Tiger -> Leopard -> Snow Leopard did mean certain upgrades (64-bit) and certain downgrades (no classic, no Appletalk if you have older printer).

After Snow Leopard things went sharply downstairs - due mismanagement first, than due simply bad engineering.
Bad managers won't attract good engineers.

What I did hear, I am certainly not looking forward to WWDC this year.

But on the other side I should not complain. The way Apple behaves in few years even Atto will be gone.
But I have no choice, I am going to stay.


In any case, testing and debugging the old code and old hardware with the latest OS makes engineering sense, same as backwards (adopting new hardware to the old versions of macOS).

As far as I see, the Moser bridges are good, reasonably priced and symmetric. There is even a 4xPCI-e <--> 5V/3.3V tolerant PCI-X bridge.

https://www.diodes.com/products/connectivity-and-timing/pcie-packet-switchbridges/pcie-pci-bridges

[jt]

Very interested in your thoughts on using a DIODES bridge to intertwine PCIe on Raspberry Pi's Compute Module 4 with the likes of the 6400/6500/TAM generation under OS9. The project would be similar to RaSCSI development, but targeting the PCI bus. Implementing CM4's EMMc as solid state memory on the PCI bus a/o WiFi function would be the primary purpose. If either might fly, other functions may follow?

Insane WAG would be that WiFi on CM4 could be done on the CSII interface as it's specific to Apple's 10bT NIC. Dunno about doing mass storage on that interface though?   

https://www.mwrf.com/technologies/systems/article/21145143/its-here-the-raspberry-pi-cm4-with-no-edge-connector