4.3 issues

[VeNoM]

1. the same kernel patch applies

2. I got it working on linux kvm with a small problem. when it starts, some service (probably scemd) deletes all /dev/sd* and you can't do anything with it.

[jokies]

Can you teach us how to apply the patch and where to obtain it?

[VeNoM]

https://github.com/andy928/xpenology/commits/master

[VeNoM]

After some checking it appers that the kernel is somehow discarding virtual /dev/sd*. Synology did some modification to the ./drivers/ata section of the kernel.

Can somebody help ?

[Trantor]

Yes same here

 

I test my "pre" 4.3 build on 2 platform : intel (asus pk5pl-am se which use ata_piix driver) and amd (zotac AD01 minipc)

Test with blank disk and with 4.2-3211 on it : same issue (install ok reboot syno assistant happy, log into DSM all is working exept this volume/disk error)

No error on screen neither in logs

Compiled modules (like ethernet/raid) loads and work.

 

I have try to just repack originals files with just pacthed synobios. But even if I include sata driver as module in rd.gz, I'm unable to install because no disk detected

The idea is to test with original kernel and modules if it working.

 

I will post it if you want to test

 

To build 4.3 :

- Tool Chains (DSM 4.3 ) / Intel x86 Linux 3.2.40 (Bromolow) x64

- Kernel source from Synology GPL / 3776branch (synogpl-3776-bromolow.tbz) with patch from andy's git

 

EDIT:

 

This build is not working and it's for debug and testing purpose only

Here it is : 4.3 Test Build

 

EDIT2:

 

fdisk -l return nothing...

When DSM is started, I plug USB drive, it's mounted and available (fdisk see it) but after 10-15sec it's unmouted (fdisk return nothing).

It seem something is running and auto unmout drive.

 

It's very strange because USB drive are still mounted in /volumeUSB* but not in DSM.

I will try to kill some main DSM services to see if I can remount volume.

Maybe Kanedo can do his "voodoo magic" and compare source code from 3.2.30 and 3.2.40 kernel ?

 

In /var/log/message I get:

Aug 29 12:53:06 DiskStation dnsdsm: Check failed.

Aug 29 12:53:07 DiskStation initdata.cgi: disk_info_get.c:76 Failed to open /dev/sda, errno=No such file or directory

Aug 29 12:53:07 DiskStation initdata.cgi: disk_info_enum.c:42 Failed to get disk information sda

[VeNoM]

It was about time Synology put in some protection

[Trantor]

It was about time Synology put in some protection

 

Yes I agree ^^

[Kanedo]

Trantor,

 

Unfortunately, I can't get network up and running using your 4.3 test build. I have Nvidia NIC. Did you add support for that in .config?

 

Thanks.

[VeNoM]

./drivers/ata/libata-eh.c

 

                       case -ENOENT:
                               /* IDENTIFY was issued to non-existent
                                * device.  No need to reset.  Just
                                * thaw and ignore the device.
                                */
                               ata_eh_thaw_port(ap);
#ifdef MY_ABC_HERE
                               ata_link_err(link, "Issued IDENTIFY to non-existent device ?!\n");
                               goto err;
#endif

 

I get this error during booting.

[Trantor]

Trantor,

 

Unfortunately, I can't get network up and running using your 4.3 test build. I have Nvidia NIC. Did you add support for that in .config?

 

Thanks.

 

My bad. Forgot to add network drivers ^^

Kernel with lot of network drivers (as usal replace it on your USB stick)

 

Yes VeNoM this part of code is not present in 4.2 (especially goto err;).

I start to compare sources file between 4.2 and 4.3, but got massive headache...

I will try to built 4.3 kernel without this part of code later

[Akhlan]

Trantor President

[VeNoM]

The problem is in rc = sata_pmp_attach(dev). This function returns -ENOENT

                      if (dev->class == ATA_DEV_PMP)
                               rc = sata_pmp_attach(dev);
                       else
                               rc = ata_dev_read_id(dev, &dev->class,
                                                    readid_flags, dev->id);

                       /* read_id might have changed class, store and reset */
                       ehc->classes[dev->devno] = dev->class;
                       dev->class = ATA_DEV_UNKNOWN;

                       switch (rc) {
                       case 0:
                               /* clear error info accumulated during probe */
                               ata_ering_clear(&dev->ering);
                               new_mask |= 1 << dev->devno;
                               break;
                       case -ENOENT:
                               /* IDENTIFY was issued to non-existent
                                * device.  No need to reset.  Just
                                * thaw and ignore the device.
                                */
                               ata_eh_thaw_port(ap);
#ifdef MY_ABC_HERE
                               ata_link_err(link, "Issued IDENTIFY to non-existent device ?!\n");
                               goto err;
#endif
                               break;
                       default:
                               goto err;

[VeNoM]

I manualy create /dev/sda[12], open the storage manager, then /dev/sda[12] is deleted and the folowing message gets logged.

 

Aug 30 10:44:00 DiskStation43 dnsdsm: Check failed.
Aug 30 10:44:01 DiskStation43 dnsdsm: Check failed.
Aug 30 10:44:01 DiskStation43 storagehandler.cgi: disk_info_get.c:76 Failed to open /dev/sda, errno=No such file or directory
Aug 30 10:44:01 DiskStation43 storagehandler.cgi: disk_info_enum.c:42 Failed to get disk information sda
Aug 30 10:44:01 DiskStation43 storagehandler.cgi: storagehandler.cpp:474 CszBuf't get the current buzzer bitmap
Aug 30 10:44:30 DiskStation43 SystemInfo.cgi: disk_info_get.c:76 Failed to open /dev/sda, errno=No such file or directory
Aug 30 10:44:30 DiskStation43 SystemInfo.cgi: disk_info_enum.c:42 Failed to get disk information sda
Aug 30 10:44:30 DiskStation43 SystemInfo.cgi: disk_info_get.c:76 Failed to open /dev/sda, errno=No such file or directory
Aug 30 10:44:30 DiskStation43 SystemInfo.cgi: disk_info_enum.c:42 Failed to get disk information sda

[VeNoM]

DiskStation43> cat /usr/syno/etc/dnsdsm

-----BEGIN CERTIFICATE-----

MIIDezCCAmOgAwIBAgIJAIY9NwFRheo8MA0GCSqGSIb3DQEBBQUAMFQxCzAJBgNV

BAYTAlRXMQ8wDQYDVQQIDAZUYWl3YW4xDzANBgNVBAcMBlRhaXBlaTEVMBMGA1UE

CgwMU3lub2xvZ3kgSW5jMQwwCgYDVQQLDANTUEcwHhcNMTMwNjExMDIyOTU2WhcN

MTMwNzExMDIyOTU2WjBUMQswCQYDVQQGEwJUVzEPMA0GA1UECAwGVGFpd2FuMQ8w

DQYDVQQHDAZUYWlwZWkxFTATBgNVBAoMDFN5bm9sb2d5IEluYzEMMAoGA1UECwwD

U1BHMIIBIjANBgkqhkiG9w0BAQEFAAOCAQ8AMIIBCgKCAQEAxJFKG83WbbHHLxKF

M4xoVp1iYuwpp0pNF0y/brNQsNEEq/aqGgzO48RDBm919mvVgqtzBPM/aPfJZvza

G1bIyh3koBtPEIrrJSw3hbYLWbTcDwP+/CwCGYKvquqsEo74c5ydCzJMZnfR+nWV

BjbWAl8LFbfAueA5+BtvjGrf4Ne7l1Zc1oLNkNyVwx3Q7WAhwBp4uJ9RoWCbA3/l

y6t91PXr2z7/iRvys1xLkIyq9Q31jfz87jPT2vtSAlCa2ccPk88/w0lHlKKA+/hW

whIK2WW7kyVzH/aZiY/uLkz5K0C00j04SCl5pB59c2PUWPbAgfWxycB57QvsUoOx

wP4C1wIDAQABo1AwTjAdBgNVHQ4EFgQUC54wSyS1sKhJvqbuFORt5O1yKKkwHwYD

VR0jBBgwFoAUC54wSyS1sKhJvqbuFORt5O1yKKkwDAYDVR0TBAUwAwEB/zANBgkq

hkiG9w0BAQUFAAOCAQEAtdvdeWHnOCg9wsZK4NYpf9hijWSDJXVeomvFCZAJrPkm

8YO8fT3SmOpiRQ7hljYn6i1oNM4PTBqHga0REB8FIrlx1dWSxs5ZbYcBcFGHsdBu

ntPmTXjwcy8wIC73piSXN8FsL2P5Kh0nclsRFOpgpeE/sd1wvqwzgteJVzCf7oNH

RhUF+2eSytF2Ge1VEl99AUPwxY7/+4h0INZ1Z0aXbXnmvHoMNT9x71z75Eni7ZIU

m0gxAA05VWPQk75ahESwqUvGPsF7Q5Fa0iPaKUwG+hWJevXnWN9r4YL5pu0Q67l3

1qWRC83iqaTTjEsaQnsRDPHVBJpyy6QjZwn7HCuNzA==

-----END CERTIFICATE-----

[fonix232]

Is the DNSDSM service even relevant? I think all it does is to register the device to a (D)DNS service, and it fails the check, irrelevant to the disks inserted. Though I might be wrong.

 

What I think is, that the devices are dismounted because of a different check, possibly made by synobios, and it instructs the SATA controller to drop all devices.

[VeNoM]

Well, the problem is in /usr/syno/webman.

I just replaced the 4.2 webman in the 4.3 firmware and the /dev/sda[12] doesn't get deleted.

So Synology implemented a security check on the interface in the .cgi binaries.

 

later edit:

On restart the dev/sda[12] are deleted.

[colombodk]

Supposedly, there is a new security check for 3rd party applications... it has effected the SynoBox.fr.nf app; perhaps it also effects XPEnology?

 

Read about it on http://synobox.fr.nf/

[Trantor]

Just little feedback:

I manage to boot system without scemd nor /usr/syno/etc.defaults/rc.d/* scripts but fdisk still empty.

 

When booting with blank drive, log directly on xpenology console, fdisk detect all drives

 

EDIT: Build kernel with libata-eh.c pacthed (removed the new part of code), same issue.

 

The problem is in rc = sata_pmp_attach(dev). This function returns -ENOENT

 

Where do you find this ? /etc/rc ?

[fonix232]

EDIT: Build kernel with libata-eh.c pacthed (removed the new part of code), same issue.

 

The problem is in rc = sata_pmp_attach(dev). This function returns -ENOENT

 

Where do you find this ? /etc/rc ?

 

It is in libata-eh, but more extended.

[VeNoM]

Just little feedback:

I manage to boot system without scemd nor /usr/syno/etc.defaults/rc.d/* scripts but fdisk still empty.

 

When booting with blank drive, log directly on xpenology console, fdisk detect all drives

 

EDIT: Build kernel with libata-eh.c pacthed (removed the new part of code), same issue.

 

The problem is in rc = sata_pmp_attach(dev). This function returns -ENOENT

 

Where do you find this ? /etc/rc ?

 

The kernel in ./drivers/ata/libata-eh.c

[k3dt]

If you want debug..

 

DiskStation> cat /proc/partitions

major minor #blocks name

 

8 0 8388608 sda

8 1 2490240 sda1

8 2 2097152 sda2

9 0 2490176 md0

9 1 2097088 md1

DiskStation> mknod /dev/sda b 8 0

DiskStation>

 

Now after few secs /dev/sda dissapears. We need figure out why this happens.

[VeNoM]

It disappears only when you use the web interface or on reboot.

[k3dt]

Yep... i tried "killall httpd" and /dev/sda is still here.. will look at .cgi files with dissasambler.

[tsygam]

seems that for qnoligy vortex is using another synobios mod: at the offset 00002A26, 75 C4 0F BE 04 24 is replaced with 90 B8 36 00 00 00, so assuming that 36 is model ID for DS1511, for 3612 shouldn't it be 90 B8 42 00 00 00 at 0002626?

synobios.zip

[Trantor]

k3dt synobios's patched file (this file in use in 4.3 test build):

.text:00000000000025C8 loc_25C8:                               ; CODE XREF: SetMicropId+3Aj
.text:00000000000025C8                 mov     rsi, offset aR  ; "R"
.text:00000000000025CF                 mov     rdx, rsp
.text:00000000000025D2                 mov     ecx, 8
.text:00000000000025D7                 mov     rdi, rsi
.text:00000000000025DA                 call    ReadUart
.text:00000000000025DF                 test    eax, eax
.text:00000000000025E1                 mov     edx, 0FFFFFFFFh
.text:00000000000025E6                 xor     rax, rax
.text:00000000000025E9                 mov     al, 42h ; 'B'
.text:00000000000025EB                 nop
.text:00000000000025EC                 xor     edx, edx
.text:00000000000025EE                 mov     cs:MpId_20839, eax
.text:00000000000025F4                 jmp     short loc_25AC
.text:00000000000025F4 SetMicropId     endp

 

synobios file from 1511+ qnology (here or here. Thanks google trad ^^)

.text:00000000000029C8 loc_29C8:                               ; CODE XREF: SetMicropId+3Aj
.text:00000000000029C8                 mov     rsi, offset aR  ; "R"
.text:00000000000029CF                 mov     rdx, rsp
.text:00000000000029D2                 mov     ecx, 8
.text:00000000000029D7                 mov     rdi, rsi
.text:00000000000029DA                 call    ReadUart
.text:00000000000029DF                 test    eax, eax
.text:00000000000029E1                 mov     edx, 0FFFFFFFFh
.text:00000000000029E6                 nop
.text:00000000000029E7                 mov     eax, 36h ; '6'
.text:00000000000029EC                 xor     edx, edx
.text:00000000000029EE                 mov     cs:MpId_20839, eax
.text:00000000000029F4                 jmp     short loc_29AC
.text:00000000000029F4 SetMicropId     endp

 

Some differences in registers and operations...

 

 

And yes as tsygam mention 36 is MicroID for 1511+:

MICROP_ID_710p = 0x31, /* '1' */
MICROP_ID_411p = 0x33, /* '3' 411+II is the same*/
MICROP_ID_1010p = 0x32, /* '2' */
MICROP_ID_1511p = 0x36, /* '6' */
MICROP_ID_810p = 0x35, /* '5' */
MICROP_ID_810rp = 0x34, /* '4' */
MICROP_ID_2211p = 0x37, /* '7' */
MICROP_ID_2211rp = 0x38, /* '8' */
MICROP_ID_2411p = 0x39, /* '9' */
MICROP_ID_3411xs = 0x43, /* 'C' 3412xs is the same */
MICROP_ID_3411rpxs = 0x41, /* 'A' 3412rpxs is the same */
MICROP_ID_3611xs = 0x42, /* 'B' 3612xs is the same*/
MICROP_ID_712p = 0x44, /* 'D' */
MICROP_ID_412p = 0x45, /* 'E' */
MICROP_ID_1512p = 0x47, /* 'G' */
MICROP_ID_1812p = 0x46, /* 'F' */
MICROP_ID_812p = 0x48, /* 'H' */
MICROP_ID_812rp = 0x49, /* 'I' */
MICROP_ID_2212p = 0x4A, /* 'J' */
MICROP_ID_2212rp = 0x4B, /* 'K' */
MICROP_ID_2413p = 0x4C, /* 'L' */
MICROP_ID_10613xsp = 0x4d, /* 'M' */
MICROP_ID_3413xsp = 0x4e, /* 'N' */
MICROP_ID_713p = 0x50, /* 'P' */
MICROP_ID_1513p = 0x51, /* 'Q' */
MICROP_ID_1813p = 0x52, /* 'R' */
MICROP_ID_RS2414p = 0x53, /* 'S' */
MICROP_ID_RS2414rpp = 0x54, /* 'T' */
MICROP_ID_RS814p = 0x55, /* 'U' */
MICROP_ID_RS814rpp = 0x56, /* 'V' */
MICROP_ID_RS3614xsp = 0x57, /* 'W' RS3614xsp */
MICROP_ID_RS3614xs = 0x57, /* 'W' RS3614xs */
MICROP_ID_RS3614rpxs = 0x57, /* 'W' RS3614rpxs */
MICROP_ID_1814p = 0x58, /* 'X' */
MICROP_ID_UNKNOW = 0xFF,