Develop and refine SA3600(broadwellnk) and SA6400(epyc7002) thread

[Orphée]

1 hour ago, RedwinX said:

So you think it will not works with baremetal install too ? 

I dont know.

[RedwinX]

2 hours ago, Orphée said:

I dont know.

Works on chineese motherboard N5105 : 

Will try HW transcoding  

 

Edited March 24, 2023 by RedwinX

[RedwinX]

@Orphée Is there a link to download i915 module with dependencies? Don't have time to recompile it... Thx 

Edited March 24, 2023 by RedwinX

[IG-88]

1 hour ago, RedwinX said:

Is there a link to download i915 module with dependencies? Don't have time to recompile it... Thx 

the sa6400 contains the new i915 modules, look in the 3rd partition of the loader and you will see

i'm testing baremetal too, you should see something in /dev/dri just by booting into dsm

your cpu is jasper lake and should work wit the new i915 driver

[RedwinX]

6 minutes ago, IG-88 said:

the sa6400 contains the new i915 modules, look in the 3rd partition of the loader and you will see

i'm testing baremetal too, you should see something in /dev/dri just by booting into dsm

your cpu is jasper lake and should work wit the new i915 driver

Thx ! don't see it... Will try in a minute  

[IG-88]

12 hours ago, jimmmmm said:

I have found there is a modules named "pgdrv.ko" for realtek, you can remove it.

 

maybe not, as i could not find it in the loaders 3rd partition or /lib/modules/ on the installed system

at least i could unload that one and r8168 and then just load r8168 and it look ok

that driver seems related to firmware updates but where does it com from? not fix in syno's kernel so maybe rd.gz?

yes its in syno's original rd.gz so removing it will be not possible i think, i guess rd.gz is protected in the same way as the kernel so removing it might not work

 

found "initrd-dsm" on the 3rd partition of the loader and i think there we can see how its loaded

the question then why its not unloaded with rmmod as it should be

there is a oob.ROM in /.syno/patch/ (beside some other files (list see below)

 

...
#
# Update oob fw
#
OOB_FW_FILE="${Mnt}/oob.ROM"
UpdateOobFw()
{
        if [ 0 -ne $FsckFailed ]; then
                return 0
        fi
        if [ "$SupportOobCtl" != "yes" ]; then
                return 0
        fi
        if [ ! -f ${OOB_FW_FILE} ]; then
                return 0
        fi
        /sbin/insmod /lib/modules/pgdrv.ko
        /usr/syno/sbin/syno_oob_fw_upgrade /parta ${OOB_FW_FILE} /verify
        if [ 0 -eq $? ]; then
                echo "Update OOB FW A success!"
        else
                echo "Update OOB FW A failed! Try FW B"
                sleep 5
                /usr/syno/sbin/syno_oob_fw_upgrade /partb ${OOB_FW_FILE} /verify
                if [ 0 -eq $? ]; then
                        echo "Update OOB FW B success!"
                        sleep 5
                        # erase partA magic number, avoid boot into partA
                        /usr/syno/sbin/syno_oob_fw_upgrade /er 64
                else
                        echo "Update OOB FW B failed! Skip update"
                fi
        fi
        /bin/rm -f ${OOB_FW_FILE}
        /sbin/rmmod pgdrv.ko
}
 ...

 

/.syno/patch/

DiskCompatibilityDB.tar
GRUB_VER              
H2OFFT-Lx64           
VERSION               
bios.ROM              
checksum.syno         
diskaction.tgz        
expired_models        
grub_cksum.syno       
model.dtb             
nvme                  
oob.ROM               
platform.ini          
rd.gz                 
smartctl              
synofwupgrade         
updater               
zImage                

 

 

Edited March 24, 2023 by IG-88

[yanjun]

Actually, I have been testing SA6400 for a long time, and the initial version didn't encounter many problems until I tried to migrate my main machine to SA6400. The current issue I am facing is that when I use a combination of HBA card and SAS expander, the loading process of mpt3sas causes kernel magic. I don't have many clues to solve this problem.

Edited March 24, 2023 by yanjun

[IG-88]

On 3/24/2023 at 5:36 PM, IG-88 said:

maybe not, as i could not find it in the loaders 3rd partition or /lib/modules/ on the installed system

at least i could unload that one and r8168 and then just load r8168 and it look ok

that driver seems related to firmware updates but where does it com from? not fix in syno's kernel so maybe rd.gz?

yes its in syno's original rd.gz so removing it will be not possible i think, i guess rd.gz is protected in the same way as the kernel so removing it might not work

 

i mitigated it for now by adding the following to the beginning of /etc/rc.network and /etc.defaults/rc.network

rmmod pgdrv.ko
rmmod r8168.ko
insmod /usr/lib/modules/r8168.ko

 

edit: doing it that way will not work after a update where as the rc.network might be replaced wit a new version and it will again not use the realtek nic

Edited May 3, 2023 by IG-88

[RedwinX]

Finally works. The plex embeded app does not work. HW transcoding failed. But with PMS with docker, works perfectly. Migrated 923+ to SA6400 works too. 

M2 SSD has been detected, but cannot create volume with it (requirements not met). Can only create it in command line, to be detected after reboot. 

That's a great advanced. The kernel 5.10 looks very cool, and mor faster on my motherboard. Plex and docker are very faster than installed in 923+.

Thx for your incredible works!  

[IG-88]

1 hour ago, RedwinX said:

M2 SSD has been detected, but cannot create volume with it (requirements not met).

that the dsm default, m.2 nvme is only usable as cache, data volume will be some kind of mangling and can easily break on update (like fallback to unconfigured cache drive)

 

2 hours ago, RedwinX said:

and mor faster on my motherboard. Plex and docker are very faster than installed in 923+.

anything measurable or just how the gui feels like?

[RedwinX]

3 hours ago, IG-88 said:

that the dsm default, m.2 nvme is only usable as cache, data volume will be some kind of mangling and can easily break on update (like fallback to unconfigured cache drive)

That's wired, but not detected as cache device. Detected as m2 device, and after some research, it's appear that some model can create volume now with m2 device : https://kb.synology.com/fr-fr/DSM/tutorial/Which_models_support_M_2_SSD_storage_pool

 

3 hours ago, IG-88 said:

anything measurable or just how the gui feels like?

GUI feels like yeah, but performance of docker too. I launch like 6-7 containers with different apps, and now it makes less than 2min to launch thant 5min before. Same with analyse of my plex library. 10mins to scan vs 14mins. That's the same machine, with same RAM and same CPU. Nothing has changed

[IG-88]

9 minutes ago, RedwinX said:

That's wired, but not detected as cache device. Detected as m2 device, and after some research, it's appear that some model can create volume now with m2 device : https://kb.synology.com/fr-fr/DSM/tutorial/Which_models_support_M_2_SSD_storage_pool

ok, thats very recent then and its only for year 23 units and and 21/22 units for dsm 7.2 (and thats in beta riht now)

i guess if its supported in some units it might be possible to make it on others available too, so just wait for it or look here

https://xpenology.com/forum/topic/67961-use-nvmem2-hard-drives-as-storage-pools-in-synology/

 

24 minutes ago, RedwinX said:

GUI feels like yeah, but performance of docker too. I launch like 6-7 containers with different apps, and now it makes less than 2min to launch thant 5min before. Same with analyse of my plex library. 10mins to scan vs 14mins. That's the same machine, with same RAM and same CPU. Nothing has changed

 

sound like one more reason to test and looking forward to have it on the main unit, for now i'm just testing it (but as dva1622 screwed  my installation (just dsm config and packages, data are ok) i might switch soon to sa6400 too

 

 

[yanjun]

13 hours ago, RedwinX said:

That's wired, but not detected as cache device. Detected as m2 device, and after some research, it's appear that some model can create volume now with m2 device : https://kb.synology.com/fr-fr/DSM/tutorial/Which_models_support_M_2_SSD_storage_pool

 

13 hours ago, IG-88 said:

ok, thats very recent then and its only for year 23 units and and 21/22 units for dsm 7.2 (and thats in beta riht now)

i guess if its supported in some units it might be possible to make it on others available too, so just wait for it or look here

https://xpenology.com/forum/topic/67961-use-nvmem2-hard-drives-as-storage-pools-in-synology/

open it just run this command.
echo 1 > /run/synostorage/disks/nvme0n1/m2_pool_support

echo 1 > /run/synostorage/disks/nvme0n1/m2_pool_support

 

[jimmmmm]

On 3/26/2023 at 6:09 AM, RedwinX said:

That's wired, but not detected as cache device. Detected as m2 device, and after some research, it's appear that some model can create volume now with m2 device : https://kb.synology.com/fr-fr/DSM/tutorial/Which_models_support_M_2_SSD_storage_pool

 

GUI feels like yeah, but performance of docker too. I launch like 6-7 containers with different apps, and now it makes less than 2min to launch thant 5min before. Same with analyse of my plex library. 10mins to scan vs 14mins. That's the same machine, with same RAM and same CPU. Nothing has changed

 

Put the file to /lib64/libhwcontrol.so.1 and restart SA6400, you can create m2 storage pool.

 

This file is only for SA6400 7.1.1

 

libhwcontrol.so.1

Edited March 27, 2023 by jimmmmm

[blackmix]

I would like to thanks everyone for their work on the loader, I'm testing it and I'm going to move on to my main server, according to my tests it's much faster than my server, I uploaded some files and realized that it's very fast and doesn't increase the processing IO something that happen on my 3622xl. Soon I will migrate to the xeon server

[RedwinX]

Hi,

 

I've experienced some reboot at any time, without reasons. Cannot see everything on logs, just said that I need to make a data scrubing. I think this is link to the SA6400 loader. Any idea where I can found the log? In which file? 

[IG-88]

3 hours ago, RedwinX said:

Any idea where I can found the log? In which file? 

 

you can try "journalctl" in the console (q for quit)

files are in /var/log/

 

my test system (9th gen cpu with tested hardware) did not show unusual behavior but there is not much installed atm, 7 disks and some packages (and no internet access for the system -> synowedjat-exec[18930]: connection error while downloading payload. [35], synowedjat-exec[18930]: Error downloading payload, synowedjat-exec[18930]: Error execute payload, - no thanks not interested in that stuff atm)

 

there are some log messages indicating some minor problems with the loader

 

...
2023-03-31T20:21:51+02:00 test3 builtin-libhwcontrol_oob-syno_oob_set_nowtime[19380]: oob/oob_reg_read.c:46 Read OOB reg 292 failed, ret=-1
2023-03-31T20:21:51+02:00 test3 builtin-libhwcontrol_oob-syno_oob_set_nowtime[19380]: oob/oob_reg_write.c:51 Write OOB reg 292 failed, ret=-1
2023-03-31T20:21:52+02:00 test3 builtin-libhwcontrol_oob-syno_oob_set_nowtime[19380]: oob/oob_tools.c:107 CMAC RESET error
2023-03-31T20:21:52+02:00 test3 builtin-libhwcontrol_oob-syno_oob_set_nowtime[19380]: oob/oob_reg_read.c:46 Read OOB reg 292 failed, ret=-1
2023-03-31T20:21:52+02:00 test3 builtin-libhwcontrol_oob-syno_oob_set_nowtime[19380]: oob/oob_reg_write.c:51 Write OOB reg 292 failed, ret=-1
2023-03-31T20:52:38+02:00 test3 kernel: [174006.217158] <redpill/smart_shim.c:686> Expected to copy HDIO_DRIVE_CMD header of 4 bytes from 0000000000000000 - it failed
2023-03-31T20:52:38+02:00 test3 kernel: [174006.228753] <redpill/smart_shim.c:686> Expected to copy HDIO_DRIVE_CMD header of 4 bytes from 0000000000000000 - it failed
...
2023-03-31T23:16:27+02:00 test3 kernel: [182634.562114] <redpill/smart_shim.c:686> Expected to copy HDIO_DRIVE_CMD header of 4 bytes from 0000000000000000 - it failed
2023-03-31T23:16:27+02:00 test3 kernel: [182634.573636] <redpill/smart_shim.c:686> Expected to copy HDIO_DRIVE_CMD header of 4 bytes from 0000000000000000 - it failed
2023-03-31T23:16:27+02:00 test3 kernel: [182634.585204] <redpill/smart_shim.c:686> Expected to copy HDIO_DRIVE_CMD header of 4 bytes from 0000000000000000 - it failed
2023-03-31T23:16:27+02:00 test3 kernel: [182634.596803] <redpill/smart_shim.c:686> Expected to copy HDIO_DRIVE_CMD header of 4 bytes from 0000000000000000 - it failed
2023-03-31T23:17:18+02:00 test3 kernel: [182685.586770] <redpill/pmu_shim.c:234> Unknown 1 byte PMU command with signature hex="6b" ascii="k"
2023-03-31T23:17:18+02:00 test3 kernel: [182685.968973] <redpill/pmu_shim.c:234> Unknown 1 byte PMU command with signature hex="6b" ascii="k"
2023-03-31T23:17:18+02:00 test3 kernel: [182686.270364] <redpill/pmu_shim.c:234> Unknown 1 byte PMU command with signature hex="6b" ascii="k"
2023-03-31T23:17:19+02:00 test3 kernel: [182686.571662] <redpill/pmu_shim.c:234> Unknown 1 byte PMU command with signature hex="6b" ascii="k"
...

 

[smilenkovski]

On 3/24/2023 at 11:17 AM, RedwinX said:

So you think it will not works with baremetal install too ? 

I've installed on 10th gen i3, baremetal and it's working fine except realtek  onboard card, after replacing with intel pci lan card, works O.K. Althugh realtek 8125B is working full speed. Plex transcoding hw.

[IG-88]

1 hour ago, smilenkovski said:

I've installed on 10th gen i3, baremetal and it's working fine except realtek  onboard card

it's stuck in firmware update mode (driver for update not unloaded), a small hack how to get it working is here

https://xpenology.com/forum/topic/68067-develop-and-refine-sa3600broadwellnk-and-sa6400epyc7002-thread/?do=findComment&comment=440597

atm i'm running my test system with just the realtek onboard

Edited April 1, 2023 by IG-88

[rok1]

Do we know what the max cpu is set to for the SA6400? I currently have a 10th gen CPU running as a 920+ with HT disabled, but if this will allow me to turn HT back on I'd make a test of it.

 

Found it myself: it's 12C/24T

Edited April 8, 2023 by rok1

[IG-88]

6 hours ago, rok1 said:

Do we know what the max cpu is set to for the SA6400? I currently have a 10th gen CPU running as a 920+ with HT disabled, but if this will allow me to turn HT back on I'd make a test of it.

 

Found it myself: it's 12C/24T

 

the kernel config look like this

CONFIG_NR_CPUS_RANGE_BEGIN=2
CONFIG_NR_CPUS_RANGE_END=512
CONFIG_NR_CPUS_DEFAULT=64
CONFIG_NR_CPUS=24

 

the "official" table from synology can be deceiving some times as the value will be about the max. config in this platform (one kernel config per platform)

as there is only sa6400 in epyc7002 it is 24 (at least for now, if they release something more potent in that platform the value can go up)

 

but i guess there might be performance inconsistencies with syno's old kernels/configs when it comes to E and P core's of newer intel cpu's

 

[rok1]

4 hours ago, IG-88 said:

 

but i guess there might be performance inconsistencies with syno's old kernels/configs when it comes to E and P core's of newer intel cpu's

 

 

I saw something similar regarding using a Denverton build (DVA3221) vs  DS920+ (Gemini Lake) the old Denverton would perform slower on cpu intensive tasks (openvpn for one)

[IG-88]

2 hours ago, rok1 said:

I saw something similar regarding using a Denverton build (DVA3221) vs  DS920+ (Gemini Lake) the old Denverton would perform slower on cpu intensive tasks (openvpn for one)

both are kernel 4, denverton 16 threads, geminilake 8 threads kernel limit

similar to HT and the core/thread limits it might be worth a try to disable E cores in bios

not sure how kernel 5.10 deals by default with that but it might be a better then kernel 4

maybe a hypervisor can help to only assign p cores to a dsm vm

[rok1]

How are yall installing on baremetal? I get no disk found when using the image here https://blog.jim.plus/blog/post/jim/synology-sa6400-with-i915.

 

 

[IG-88]

7 hours ago, rok1 said:

How are yall installing on baremetal?

nothing special, just a 9th gen cpu, B365 chipset and onboard sata (ahci)

what hardware did you try?