I directed the Armbian guys (Werner) to your patch and they included it into their master branch. So I got to compile an Armbian kernel in order to test the patch in my setup with the EspressoBin Board and the AS-Media SATA-controller chips.
I've tested it and it works flawlessly :-)
Thanks!
Am 21.03.2021 16:09 schrieb Rötti:
I organized a T60 Thinkpad, pulled out the Wificard (MiniPCIE) and plugged in the Marvell SATA-Controller card. Good news is that you're right, the DevCap MaxPayload is 128 bytes, so I couldn't reproduce that error on that thinkpad. I tried two different Marvell controller cards. Wierd thing is, that both cards did not sho up in the lspci -nn -vv command. So I'm not sure if these got recognized.
With these patches supplied (@thank you very much Marek & Björn) is there a build server I can download a nightly version of armbian I can test for you? Is there any way I can support?
Thank you very much in advance!
Am 19.03.2021 20:02 schrieb Pali Rohár:
On Wednesday 17 March 2021 18:03:55 Bjorn Helgaas wrote:
On Wed, Mar 17, 2021 at 11:55:44PM +0100, Pali Rohár wrote:
On Wednesday 17 March 2021 17:45:49 Bjorn Helgaas wrote:
This quirk suggests that there's a hardware defect in the ASMedia ASM1062. But if that's really the case, we should see reports on lots of platforms, and I'm only aware of these two.
Do you have platform which support MPS of 512 bytes? Because I have not seen any x86 / Intel PCIe controller with such support on ordinary laptop and desktop.
These two (A3720 and CN9130) are the only which has support for it.
Has somebody else PCIe controller which Root Bridge supports MPS of 512 bytes?
Maybe they are in servers, but then such "cheap" SATA controllers are not used in servers. So this is probably reason why nobody else reported such issue.
I have no idea. My laptop only supports 512 (except for an ASMedia USB controller). If the device advertises it, I would expect the vendor to test it. Obviously it still could be a device defect. They should publish an erratum if that's the case so people know to avoid it. So I would try to get ASMedia to say "no, that's tested and should work" or "oh, sorry, here's an erratum and we'll fix it in the next round."
I doubt that ASMedia publish something...
But has somebody contact to ASMedia? I can try it.
Basically these ASMedia SATA controller chips are present on more "noname" mPCIe-form cards and I guess ASMedia is not going to support them.
Note that we have also tested Marvell PCIe-based SATA controllers which support MPS of 512 bytes too and there were no problem with them on A3720 nor CN9130.