Develop and refine the DS3622xs+ loader

[smilenkovski]

9 hours ago, Green said:

I think the answer is PLeX Transcoding needs a IGPU for hardware transcoding. Celerons & i3,i5,i7 have an IGPU. Xeon CPU's do not and therefor can not do the hardware transcoding you are thinking of. 

The question is eventho the DS3622xs+ has a xeon cpu ,But since it a newer cpu than what is in the 918+ is the PLeX transcoding ability  in the  pat file like it should be in the 918+ pat file, since the 918+ does Plex transcoding.  So, if setting up Xpenology mainly for plex with an newer i7 cpu should we use the 918+ or DS3622xs+ setup ?

About transcoding in Plex, in dsm 7.-1 918+ is not working at the moment on newer intel cpu's (my is 10 gen). I've tried different methods using plex forums (docker and native Plex media sever together workaround on original Synology) with no success. Maybe IG-88 will find solution modifying i915 from previous dsm. 

[IG-88]

13 hours ago, ed_co said:

Does DS3622xs+ have tranconding like the 918+? Thanks!!

it does not come with a i915 driver ootb so it's a not for intel quick sync based hardware transcoding (iGPU), it will transcode based on cpu if there is not support for intel quick sync, depending on the material and cpu it can be challenging to transcode one 4k movie based on cpu power - its often possible to just stream in source quality and let the client device handle it, as the client device only has to decode and resize (no encoding as on the server) and it usually has hardware based decoding its often no problem to just stream in source quality

the only thing where transcoding on the server might be needed is when there is a bandwidth limit between server and client (wifi or powerline in some home environment or when accessing dsm over the internet and having "slower" upload rates to the internet from dsm over a asymetric broadband connection)

 

but it might be possible to compile a i915 driver and add it - problem might be that there could be something missing in synology's kernel we use as base (protection of dsm and missing source to build own kernel)

with 3.10 kernel on 3615/17 on 6.2.3 it was not possible to use i915 driver from kernel 3.10 but maybe it does now with kernel 4.4, i plan to try this later when i have a few things up and running

[IG-88]

36 minutes ago, smilenkovski said:

About transcoding in Plex, in dsm 7.-1 918+ is not working at the moment on newer intel cpu's (my is 10 gen). I've tried different methods using plex forums (docker and native Plex media sever together workaround on original Synology) with no success. Maybe IG-88 will find solution modifying i915 from previous dsm. 

i already compiled a i915 driver for 918+ having all gpu id's in it up to 12th gen (as long as it still is UHD xxx)

not tried it yet on my skylake test system and i can't on 10th or up as i dont have that (my main system is just coffee lake and that works ootb with 918+ default driver)

the main problem would be if the driver halt's the system on loading (as seen with some systems on 6.2.3 and the patched driver), then you will need to remove the new i915.ko in some way, its its done the dump and easy way as i suggested in the 6.2.3 thread you would have to test that on a non productive environment and just kill it if it goes wrong (can be a productive hardware if you just use a new usb and a empty disk) or you would need to access it as if you would do a recovery like booting up a recue linux assemble the raid1 (system) with mdadm, access the system partition and delete the i915.ko

 

it would need to be a plugin in @pocopico's tinycore that could be added manually by command and if it does not work it could be removed from the loader the same way it was added, in that case it should be reversible

 

edit: i also made a diff patch file that i can provide so anyone could build that driver (its not best practice to just provide a binary), i will also provide patches for smaller things i did to 3.10 and 4.4 source to add some missing hardware (mvsas and killer E2400/E2500 in atheros)

Edited February 25, 2022 by IG-88

[cferra]

i suggested in a PM to remove the mlx driver(s) from the rp-loader and let take over the DSM's build-in drivers (when installing with just a mlx card present as nic that method is not going to work because in syno's rd.gz on the loader is no mlx driver)
!!! please provide the dmesg log from when @pocopico's driver was active maybe we can see whats going on - i think just mlx_core.ko as in @pocopico's driver repository is not enough, it also needs mlx4_en.ko (en - like ethernet, ib would be inifin band, so we dont need ib in most cases)
 
in synologys (external) driver also mlx4_core.ko depends on mlx_compat.ko, because of external mlx driver source
i can't build original kernels own mlx drivers with synology's kernel and kernel includes, looks they modded in a way that only builds with external source (or i have messed up my build system)
 
 

Ok - I had to go away for the weekend - let me see if I can do it remotely, if not I should be back Monday night.


Sent from my iPhone using Tapatalk

[smilenkovski]

1 hour ago, IG-88 said:

i also made a diff patch file

Where to download file? I have spare system to try it.

[Peter Suh]

1 hour ago, smilenkovski said:

Where to download file? I have spare system to try it.

 

Wouldn't this driver be available only when ig-88 and pocopico collaborate first?

Edited February 25, 2022 by Peter Suh

[smilenkovski]

1 hour ago, Peter Suh said:

 

Wouldn't this driver be available only when ig-88 and picopico collaborate first?

Probably I get it wrong. Sorry.

[yanjun]

 

7.1 test success.

[IG-88]

2 hours ago, cferra said:

Ok - I had to go away for the weekend - let me see if I can do it remotely, if not I should be back Monday night.

dont make to much out of it waiting a few days does not matter, just wanting to make sure we get feedback for the mellanox drivers as a lot of these cards are around

 

 

1 hour ago, Peter Suh said:

 

Wouldn't this driver be available only when ig-88 and picopico collaborate first?

a driver in tc might be the best but for just plain testing if it works and to collect some data the blunt method by just replacing the i915 in /lib/modules/ would work and does not need anything coded, only downside is the way back if it freezes/crashes the whole system, just not loading the driver is no problem as the system would still be coming up and you could remove the driver

 

i did have a kind of "nullify i915" loader for 6.2.x to remove i915 drivers after a failed attempt (we had a few cases of that with 6.2.2) that ended up in 6.2.3 loader removing i915 driver from /lib/modules/update/, but i guess that would not be working on a 7.0 installation

 

[IG-88]

6 minutes ago, yanjun said:

7.1 test success.

looks like synology still has to work in the protection before 7.1 comes out of beta ;-)))

[cferra]

 
7.1 test success.

Do anything special to get it working?


Sent from my iPhone using Tapatalk

[chickey]

On 2/25/2022 at 12:13 PM, IG-88 said:

i already compiled a i915 driver for 918+ having all gpu id's in it up to 12th gen (as long as it still is UHD xxx)

not tried it yet on my skylake test system and i can't on 10th or up as i dont have that (my main system is just coffee lake and that works ootb with 918+ default driver)

the main problem would be if the driver halt's the system on loading (as seen with some systems on 6.2.3 and the patched driver), then you will need to remove the new i915.ko in some way, its its done the dump and easy way as i suggested in the 6.2.3 thread you would have to test that on a non productive environment and just kill it if it goes wrong (can be a productive hardware if you just use a new usb and a empty disk) or you would need to access it as if you would do a recovery like booting up a recue linux assemble the raid1 (system) with mdadm, access the system partition and delete the i915.ko

 

it would need to be a plugin in @pocopico's tinycore that could be added manually by command and if it does not work it could be removed from the loader the same way it was added, in that case it should be reversible

 

edit: i also made a diff patch file that i can provide so anyone could build that driver (its not best practice to just provide a binary), i will also provide patches for smaller things i did to 3.10 and 4.4 source to add some missing hardware (mvsas and killer E2400/E2500 in atheros)

Would be keen to help test if needed at all.  I'm building a i5-9500t based system and have went with ds3622 as i've a 6 core CPU or would the 918+ actually be sufficient as my chip doesn't have hyperthreading?  Big plex user so hardware transcoding is a big draw for me.

[IG-88]

3 minutes ago, chickey said:

Would be keen to help test if needed at all.  I'm building a i5-9500t based system and have went with ds3622 as i've a 6 core CPU or would the 918+ actually be sufficient as my chip doesn't have hyperthreading?  Big plex user so hardware transcoding is a big draw for me.

i uploaded the stuff here

https://xpenology.com/forum/topic/48743-drivers-extension-for-redpill-request-for-comments-drivers-packages/?do=findComment&comment=268029

read the red text, its not meant to be tested with a working live system as it can prevent the system from booting

[chickey]

1 minute ago, IG-88 said:

i uploaded the stuff here

https://xpenology.com/forum/topic/48743-drivers-extension-for-redpill-request-for-comments-drivers-packages/?do=findComment&comment=268029

read the red text, its not meant to be tested with a working live system as it can prevent the system from booting

I literally just saw your post after i posted mine!  My setup is not live and with only one disk so perfect for testing.  Many thanks for your hard work!

[Wraydroth]

Hi, 
I am trying to test/install 7.0.1 as DS3622xs+ at my setup:

 

MB - SuperMicro X10SDV-4C-7TP4F
CPU - XeonD 1518 (4core)
OnBoard controller - Broadcom 2116 (16x ports)
Onboard 4x SATA ports - disabled 
Additional controller - LSI 9211-8i (8 ports) 
Chassis for 24 drives hot-swap

 

I can build the loader for 3622+ and I can boot it. it will appear on Synology Assistant. 
But I am failing at installing the system because it says that it cant find any drives. 

 

I tried multiple variants of SataPortMap - but cant find any working config so the system can find drives. 

Based on thread I assume that drivers should be fine and mpt3sas should cover both controllers. 

 

Any idea what to try as SataPortMap to get it working with my setup? Or am I missing something else? 

 

Thanks for any help. 

[Peter Suh]

6 hours ago, Wraydroth said:

Hi, 
I am trying to test/install 7.0.1 as DS3622xs+ at my setup:

 

MB - SuperMicro X10SDV-4C-7TP4F
CPU - XeonD 1518 (4core)
OnBoard controller - Broadcom 2116 (16x ports)
Onboard 4x SATA ports - disabled 
Additional controller - LSI 9211-8i (8 ports) 
Chassis for 24 drives hot-swap

 

I can build the loader for 3622+ and I can boot it. it will appear on Synology Assistant. 
But I am failing at installing the system because it says that it cant find any drives. 

 

I tried multiple variants of SataPortMap - but cant find any working config so the system can find drives. 

Based on thread I assume that drivers should be fine and mpt3sas should cover both controllers. 

 

Any idea what to try as SataPortMap to get it working with my setup? Or am I missing something else? 

 

Thanks for any help. 

 

@Wraydroth

 

Hi, what kind of Redfill did you use?
Did you use the latest version image of TinyCore 0.4.5.2 by any chance?
I am also operating HBA (LSI 9211-8i) well in ds3622xs+ with mpt3sas driver.
When only HBA cards are used, the Sataportmap and DiskIdxMap settings are meaningless.
It doesn't matter if you delete it from the user_config.json or grub.cfg file at all.
LSI 9211-8i is IR mode?
If you have updated it in IT mode, has it been confirmed whether it is in normal update status?

 

P.S : Which controller's port number is the DSM installation HDD mounted on?

Edited March 2, 2022 by Peter Suh

[RacerX]

I can test Bare Metal or Promax VM , TrueNas Scale VM for mellanox

 

In the past I had no luck at all booting ds3615 with several different Mellanox 10gb cards  it had to move to Intel 720 to get it to work.  I regret that move.

 

I have the following ConnectX3

CX353A-FCBT  Single Port

CX353A-QCBT  Single Port

544+QSFP  Dual Port

 

I'm not clear how to install 7.01? do we have to use Tinyloader?

 

 

 

 

 

Edited March 2, 2022 by RacerX

[Wraydroth]

Hi @Peter Suh,

 

I used latest image for Tinycore from Pocopico - the UEFI one. 

The RAID controllers all running at IT mode - actually all works with Jun loader since I am running DSM-6.1.7 and it can see all drives. 

 

Shall I delete Sataportmap and diskidxmap from loader boot or put there zeros? 

 

I tried one recommendation with: 

sataportmap = 1

diskidxmap= 00

 

 

with same result - no disks found. 

[Peter Suh]

55 minutes ago, Wraydroth said:

Hi @Peter Suh,

 

I used latest image for Tinycore from Pocopico - the UEFI one. 

The RAID controllers all running at IT mode - actually all works with Jun loader since I am running DSM-6.1.7 and it can see all drives. 

 

Shall I delete Sataportmap and diskidxmap from loader boot or put there zeros? 

 

I tried one recommendation with: 

sataportmap = 1

diskidxmap= 00

 

 

with same result - no disks found. 

 

In fact, under the same conditions, I'm going to play IT MODE's H310 card's
The disk has not been found.
It is a recently purchased ASROCK E3C232D4U WORKSTATION MAINBOARD.
In addition, there was no problem with disk recognition under the same conditions on general boards.
I'm planning to test JUN Loader DS918+ DSM 6.2.3 first with the above E3C232D4U.
Even though it is recognized here, the fact that disk recognition is not possible in Redfill is likely to be another issue for the mpt3sas driver.
I will upload the test results after work in the evening.

Edited March 2, 2022 by Peter Suh

[yanjun]

1 hour ago, Wraydroth said:

Hi @Peter Suh,

 

I used latest image for Tinycore from Pocopico - the UEFI one. 

The RAID controllers all running at IT mode - actually all works with Jun loader since I am running DSM-6.1.7 and it can see all drives. 

 

Shall I delete Sataportmap and diskidxmap from loader boot or put there zeros? 

 

I tried one recommendation with: 

sataportmap = 1

diskidxmap= 00

 

 

with same result - no disks found. 

can you post the result of "fdisk -l"

[Orphée]

@yanjun For info, some other already raised it, but it seems the MAC address parameter is ignored on DS3622xs loader.

 

Usually if you set a different MAC address between VM host and VM guest (mac1 setting), the mac1 setting from loader overwrite original host MAC value.

But it does not work on DS3622xs loader. So I bet on baremetal, no matter what, the real Mac address will be taken anc mac1 value ignored.

 

It only happens with DS3622xs loader, whereas DS3615xs no issue.

[Peter Suh]

4 hours ago, Peter Suh said:

 

In fact, under the same conditions, I'm going to play IT MODE's H310 card's
The disk has not been found.
It is a recently purchased ASROCK E3C232D4U WORKSTATION MAINBOARD.
In addition, there was no problem with disk recognition under the same conditions on general boards.
I'm planning to test JUN Loader DS918+ DSM 6.2.3 first with the above E3C232D4U.
Even though it is recognized here, the fact that disk recognition is not possible in Redfill is likely to be another issue for the mpt3sas driver.
I will upload the test results after work in the evening.

 

@Wraydroth

I installed DSM 7.0.1 of DS3622xs+ with 9211-i4 (IT Mode) and 9211-i8 (IT Mode) in ASROCK E3C232D4U, and both disks were recognized normally.

There was no need to test the jun loader, so I didn't try it.

The peculiarity of 9211-i8 (Dell PercH310) is that PCIe pins B5 and B6 are covered like guides below with tape.

https://techmattr.wordpress.com/2016/04/11/updated-sas-hba-crossflashing-or-flashing-to-it-mode-dell-perc-h200-and-h310/

Edited March 2, 2022 by Peter Suh

[painkiller895]

Trying to build a DS920 Loader I don't see Geminilake platform version.

I'm using  redpill Tinycore ver 4.5.2 

Which platform do I use apllolake ? nope that didn't work....

guess I'll wait till they add it...

 

[OlivierG]

FYI. Successfully installed on Proxmox 7.1-10 with network adapter E1000. 

[yanjun]

RS4021xs+ now available