Steve Fan

I have researched a little bit and found out that in order to build ttm.ko, you need the kernel to support these two functions. ``` [ 9719.731180] ttm: Unknown symbol wake_up_state (err 0) [ 9719.736362] ttm: Unknown symbol perf_trace_buf_prepare (err 0) ``` This means it would be a little bit impossible for now

Result: Successfully - Bootloader version and model: ARPL v1.1 beta2a with DS1621+ setup - DSM version: DSM 7.1.1-42962 Update 5 - Additional modules extra.lzma: Configured to select all modules in ARPL menu - Hardware solution: Motherboard: AsRock X570M Pro4 CPU: Ryzen 5600G, configured to run in 35W mode Memory: 2x16GB Corsair Vengeance PLX + 2x8GB random offbrand RAM running in DDR4-3000MHz mode Total Cost: $3000 HKD base (~$399 USD), $5000 HKD total with drives. But all of the parts are from few years ago so with depreciation rate should be around $2600 HKD total with drives today. What works: 1. QuickConnect (but you need to BYOC) 2. VM Manager (with VirtualDSM support out of the box) 3. Model recognition works 4. Hard drive bay and identification. X570M Pro4 has 8 SATA ports and 6 of them are coalesced together that sits vertically (so the SATA ports aims into the front of the case where the front panel fans are) so it is pretty easy to recognize and remap in Synology What does not work: 1. Vega iGPU, there is no /dev/dri nor is there any driver loaded for the iGPU PCI device. So Emby and Plex probably won't work at its best possible way. Working to try and compile the amdgpu kernel module but I'm still not sure if software support is right on spot. This is because the CPU of DS1621+ itself (V1500B) does not have an iGPU either so Synology won't bother supporting it What needs to be tested: 1. SR-IOV, so far I don't have installed my CX3 Pro yet 2. Drive sense event. I've seen there are ATA critical events in dmesg log but Synology did not report that as a critical error which may need further debugging. But I do see that it can be a power supply issue or maybe the drive is faulty after 5 years of service, and I do see that none of the drives are gone in fdisk -l 3. High availability and NAS cluster, but I think this is well beyond home use. 4. Extra SATA SSD installation. Recall that X570M Pro4 has 8 SATA ports, so we can use the remaining two SATA ports for SSD Cache (in this case you don't care about speed but IOPS acceleration which SATA SSDs still works fine). However I don't think that would work out of the box because I don't remember DS1621+ has the option to install extra internal SATA SSDs so this is a mystery. I do have a 960GB SATA SSD that will be tried and installed soon. 5. PCIE Bifurcation. I see that I only have a PCIE x8 to x4x4 split option in BIOS while in the manufacturer site it claims I can have PCIE x16 on PCIE1 slot. This implies I only got PCIE x8 in reality if I used an APU, so the information right now is pretty confusing. Looking to contact AsRock with this matter. 6. NVME drives. I don't have spare NVME parts on hand since two of them are pending for another system install. I can also try U.2. too

I want to use my Vega GPU on the 5600G APU so that I can leverage Emby's video decoding support. Right now it seems like I can spot the PCI device using lspci but there is apparently no driver loaded for it. No, I don't think even amgpu driver is bundled in so I would need to manually compile it myself. I wonder what is the guide for doing that in the context of ARPL? I would like to make it available for everyone of course.

Hi, I want to compile my own kernel with respect to V1000 branch, since my Proxmox VM host uses Ryzen anyway and I want to ensure the best performance as possible. I have enabled more VirtIO modules as built-in so that I don't have to go through copying all the modules once again. However I cannot build it correctly with Synology config that they have apparently made some roadblocks along the way, one of which is a mysterious MY_ABC_HERE macro definition which hindered init/initramfs.c compilation and as a workaround I stripped it, but then I see some Synology related LED constant and I'm pretty there are more little snitches like this ahead, so rather than painfully see through the build logs and enable the right config, I'd like to see if anyone did build the kernel instead. And also I see that synoboot didn't really accept bzImage I guess since it showed itself to be zImage...? But then I can't make it.

Would you like to provide a kernel thats running on DS1821? I have a Ryzen 1600 AF NAS that could benefit a lot, or let me know about the instructions to build such kernel, thank you @jun!

Yes I followed the order of loading module and now it somehow works, at least the virtio net driver does, meanwhile balloon and the rest doesn't: Also interestingly I checked lsmod to see the deps: But I don't know why it is not built-in since Linux kernel should have had included these out of the box. I also don't see why Synology will cherry pick these modules since it is not really impacting any performance, plus by the inclusion of virtio they can utilize the VM station more efficiently thus bringing in more value to the customers. Silly choice for Synology. Also it seems like there is a problem with virtio scsi? I have attached an excerpt of the log from dmesg and you can check it out.

That doesn't seems possible, since most of these cloud providers use virtio, including NIC and disk to boost VM performance, and Xpenology unfortunately does not support virtio out of the box, and this mean unless the guest can host nested VM (that is, VT-X is passed though to the VM), and this alone is already a rare thing so this mean trying to virtualize Xpenology at a cost of slight performance loss is also rare. But with Ryzen/EPYC hosting however I think this could change as I see most of the Ryzen host providers allowed passthough of SVM. This can potentially be a great combo to get Xpenology worked on the cloud, especially when Synology is going to kick off their first series of Ryzen Embedded line of products, so that we will be finally getting optimized Ryzen kernel!

Hey would you like to fix and provide a driver extension pack for this? I tried editing /etc/rc.modules in extra{,2}.lzma but in 7-Zip this is a read only archive so I can't do it myself easily. You have the source to reproduce the pack. I think the kernel version is not matching. I want to use virtio so that I can squeeze most of the performance as much as possible so that it can go production. I'm running DSM off of my Proxmox NAS box running atop of ZFS right now and I think DSM and Proxmox can complement each other well I have the commands to load virtio in my log.txt, you can take a look. There is some symbols indeed not found in the kernel and maybe I think I need to change the kernel too or swap to DS361x log.txt

In a typical server to support iSCSI you will need (i)PXE I guess, then you will have a virtualized drive in BIOS. Then you could operate with the iSCSI target transparently without much intervention from DSM. Not sure if installing the extra firmware will make it work with the iSCSI disk though.