Pro1x with SailfishOS

[ducksoup 110]

3 minutes ago, mosen said:

I flashed sfos as first action after unboxing. Maybe you changed something to the stock settings the Pro¹ X got delivered with?
If you are not keen to debug the issue, i'd just reflash to the 2.1.2 build factory image using EDL/QFIL to have the device in unboxing state again.

I certainly didn't do anything that would have changed the system partition, and if I'm not mistaken, system, boot_a, and userdata are the only real parts in play here, and the latter two are overwritten in the Sailfish flashing process.  It's curious indeed.

Right now I don't really want to undo the work I've put into Sailfish, but it will be interesting to see if anyone else runs into a similar problem.  Maybe it's just me, and the warning added to the first post is unnecessary.

[claude0001 1,341]

13 hours ago, matf said:

Meanwhile, my old container works [...]

Congrats. And very interesting desktop setup! I start to understand what you meant when you wrote it is optimised for operation via keybindings. 🙂

As you have the same setup on both devices (Pro1 and Pro1X) now, can you say anything about performance?

In the past we worried a lot about the Pro1X's slower CPU (after its re-design). As, in our Linux containers, we rely on all-software rendering (llvmpipe) there could be a significant difference e.g. in OpenGL benchmarks or video playing ... 

[EskeRahn 5,460]

2 hours ago, mosen said:

I successfully flashed, omitting the set active command.

I will highly recommend NOT omitting this tiny step. Remember that an OTA update swaps what is active between A/B so could easily become a mess
Alternatively flash the boot image to booth boot_a and boot_b, so it is in whatever is active.

[mosen 202]

Agreed, disclaimer, i was just curious and can not fully oversee the implications.
Checking the device with fastboot getvar all revealed slot_a to be active "anyway". So i was curious to try what happens without the set active command.
Note, the original ubuntu instructions even had a set active a before the boot image flash. Which is already omitted in the first posts instructions.

[matf-kabouik 414]

On 8/11/2022 at 10:57 AM, ducksoup said:

So far this is working pretty well, but I may have run into my first example of "not quite ready for primetime".  I'm trying to set up my personal VPN (OpenVPN) and while I can get it to connect just fine if I execute the openvpn client myself from a root shell, it will not connect via the (not especially helpful) UI.

journalctl reveals two errors possibly related:

Aug 11 11:37:09 Pro1-X connman-vpnd[548]: error setting gid -1 Invalid argument
Aug 11 11:37:09 Pro1-X connman-vpnd[548]: error setting gid list Invalid argument

This is perhaps beyond the scope of this thread, but just in case anyone else here is playing with OpenVPN on Sailfish, it would be good to hear of your success or failure.

Thanks for checking that. I don´t have access to my openvpn files where I am, but I'll be able to test that later.

 

On 8/11/2022 at 3:10 PM, claude0001 said:

Congrats. And very interesting desktop setup! I start to understand what you meant when you wrote it is optimised for operation via keybindings. 🙂

As you have the same setup on both devices (Pro1 and Pro1X) now, can you say anything about performance?

In the past we worried a lot about the Pro1X's slower CPU (after its re-design). As, in our Linux containers, we rely on all-software rendering (llvmpipe) there could be a significant difference e.g. in OpenGL benchmarks or video playing ... 

It's a bit early to tell, especially as I haven´t use my Pro1 in a long time because it had the ghost touches issue and I was too lazy to swap the screen at that time. Overall, it doesn´t feel snappier (no surprise), and might even be a tad slower but I need more time to confirm, and a proper comparison with the Pro1 really. In any case, the extra RAM can be useful for that usage, as I was often reaching 100% RAM use on the Pro1.

 

On 8/11/2022 at 3:52 PM, mosen said:

Agreed, disclaimer, i was just curious and can not fully oversee the implications.
Checking the device with fastboot getvar all revealed slot_a to be active "anyway". So i was curious to try what happens without the set active command.
Note, the original ubuntu instructions even had a set active a before the boot image flash. Which is already omitted in the first posts instructions.

I think the "boot_a" subcommand takes care of selecting which slot to flash boot.img on, does it not? Anyway, thanks for the tests, and glad to see some other wild SFOS Pro1x now!

Edited August 12, 2022 by matf

[mosen 202]

@matf Please add this warning to your first post. Users should backup their persist partition as early as possible.

1. Attestation data in persist partition is unique to the device unit, therefore backup of persist is strongly recommended to people attempting to flash a custom OS.
2. The consequences of losing this data currently don't seem too critical, as the standard attestation checks like SafetyNet are passed on such devices. But they won't pass rarely used hardware attestation checks.
3. Warning, the rawprogram0.xml.bak from the EDL package will flash your persist partition with the generic image that will erase your attestation data.

The backed up persist partition will be needed in case you want to flash back to stock and use the device specific attestation keys contained in the persist image.
The catch is that the sensors "break" when flashing "back" from either UT or SFOS. And to restore those for stock android. A stock persist will be needed to flash.
In best case that is the one you backed up before starting your journey into alternate OS.

If you got EDL set up, its

edl r persist persist_backup.img

Using fastboot

fastboot fetch persist persist_backup.img

[EskeRahn 5,460]

25 minutes ago, mosen said:

Using fastboot

fastboot fetch persist persist_backup.img

With newest platform tools gives

fastboot: error: Unable to get max-fetch-size. Device does not support fetch command.

On both Pro1 and Pro1X
 

[mosen 202]

Yeah, i am getting it hard today...
I did use the edl command and just copied the fastboot command from XDA without testing...
There must be a way using adb. But i am to tired to get that right now.
Sorry for the confusion.

[ducksoup 110]

9 hours ago, mosen said:

If you got EDL set up, its

edl r persist persist_backup.img

Which EDL tool are you using with this phone?

[mosen 202]

https://github.com/bkerler/edl

Its dubbed edl.py colloquial.
You can also achieve the backup using the official Qualcomm QFIL software on wondows. But i have no experience with that.
Looking into the adb option today.

[ducksoup 110]

1 minute ago, mosen said:

https://github.com/bkerler/edl

Its dubbed edl.py colloquial.
You can also achieve the backup using the official Qualcomm QFIL software on wondows.

Excellent.  Thanks.

Windows?  What's that?  😂

[ducksoup 110]

Has anyone got the Jolla Mail app working?  When I try to set it up, either my own mail server or the one at my ISP, it always times out checking credentials.  tcpdump shows no traffic generated on any interface.

[ducksoup 110]

13 hours ago, mosen said:

If you got EDL set up, its

edl r persist persist_backup.img

Okay, I've successfully backed up my persist partition using the linked tool.  The edl tool produced a number of error-looking things on its output, but they seem to be non-fatal, and I eventually got...

Reading from physical partition 0, sector 8, sectors 8192
Progress: |██████████████████████████████████████████████████| 100.0% Read (Sector 0x2000 of 0x2000, ) 33.83 MB/s
Dumped sector 8 with sector count 8192 as persist_backup.img.

...which produced a 32 MB file containing an ext4 filesystem, the contents of which do appear to be the correct things.

A tip for anyone who has already removed Android from their device and thus can't use adb to boot into edl mode (gotta make this relevant to this thread!): Disconnect USB and power the phone completely off.  Hold down both volume up and volume down, and then while still holding volume up and down, plug in USB.  Keep holding volume up/down until you see a quick flash of the boot splash, followed by a black screen.  Finally release the volume rocker.  Your device should be in edl mode, although the screen will remain black.

 

Edited August 12, 2022 by ducksoup

[ducksoup 110]

On 8/7/2022 at 5:23 AM, matf said:

Still no luck with Waydroid despite my attempts, but theoretically this should be possible because there are unofficial images for Halium 11

Am I missing something here?  I followed the link to the image from your github issue, and it brings up a telegram page showing the zip file, and clicking on that zip seems to do nothing but open up a new tab to the same page.  How does one actually download the experimental image?

Unfortunately there doesn't seem to be any other source for it.

[brunoais 334]

23 hours ago, mosen said:

@matf Please add this warning to your first post. Users should backup their persist partition as early as possible.

1. Attestation data in persist partition is unique to the device unit, therefore backup of persist is strongly recommended to people attempting to flash a custom OS.
2. The consequences of losing this data currently don't seem too critical, as the standard attestation checks like SafetyNet are passed on such devices. But they won't pass rarely used hardware attestation checks.
3. Warning, the rawprogram0.xml.bak from the EDL package will flash your persist partition with the generic image that will erase your attestation data.

The backed up persist partition will be needed in case you want to flash back to stock and use the device specific attestation keys contained in the persist image.
The catch is that the sensors "break" when flashing "back" from either UT or SFOS. And to restore those for stock android. A stock persist will be needed to flash.
In best case that is the one you backed up before starting your journey into alternate OS.

If you got EDL set up, its

edl r persist persist_backup.img

Using fastboot

fastboot fetch persist persist_backup.img

I ran that edl command:

edl r persist persist_backup.img

but I got 8GB of data in multiple files, inside a directory called "memory".

Anything I might have done wrong?

[ducksoup 110]

 

Okay!  Got Debian Sid up and running in lxc, by fumbling my way through @matf's guide.  It works!  😄 

Edited August 12, 2022 by ducksoup

[matf-kabouik 414]

Regarding LXC, as discussed above, configuring keybindings is a key to make them productive on Pro1/Pro1x. However, at the moment, the Yellow modifiers do not send a keycode (nothing in "evdev_trace -t" on SFOS, and nothing in "xev" in Debian). Therefore, although they work and produce the yellow-printed characters when combined, they annot be used to configure custom system-wide keybindings. Yet, they are the most adapted for that use given that they are replicated on both sides of the keyboard and can therefore be combiened with pretty much every key, including other modifiers.

I warned Adam, who has already patched the kernel so that they can produce "FN" as keycode. This should be available in a next update but is in his boot.img, so flashing it over our Ubuntu Touch boot.img will be necessary. I'll report back when I do it in a few days after my trip. However I am not sure FN can be used as a modifier for WMs, I would rather make it Meta or Super, but we will see.

About backing up vanilla partitions, I'll try to sort the information later too and add the relevant paragraph in the guide. I have not backed up myself yet. 

[ducksoup 110]

94% battery when I went to bed, 52% when I woke up.  And this was with the Debian container frozen.  Probably need to work on that.  😆

Edited August 13, 2022 by ducksoup

[EskeRahn 5,460]

7 hours ago, matf said:

About backing up vanilla partitions, I'll try to sort the information later too and add the relevant paragraph in the guide. I have not backed up myself yet. 

Added a fairly easy way to backup partitions here

[VaZso 1,998]

9 hours ago, matf said:

However, at the moment, the Yellow modifiers do not send a keycode (nothing in "evdev_trace -t" on SFOS, and nothing in "xev" in Debian). Therefore, although they work and produce the yellow-printed characters when combined, they annot be used to configure custom system-wide keybindings.

Basically that was one of the things I was fighting for regarding kernel's keyboard driver of Pro1 (it should be the same for Pro1X)... others were not really understand why it can be important...

[therainingmonkey 2]

On 8/9/2022 at 1:51 PM, matf said:

Can you double check that the userdata image you flashed had the .simg extension (SFOS) and not just .img (UT)? Sorry if you already did, but that is what a reset UT after flashing would suggest.

Yep, I'm definitely flashing the sailfish .simg.
After reflashing the stock rom and trying again, the phone no longer boots into UT but hangs at the 'Powered by Android' screen. When It's connected to a computer it appears as a media device with the name 'Failed to boot'.

I haven't had any problems flashing UT or stock.

I might have done some strange things with fastboot in my frustration over the last few days, could anything there be the cause? I assumed flashing stock would restore everything to factory condition but I might be wrong there.

[matf-kabouik 414]

17 hours ago, therainingmonkey said:

Yep, I'm definitely flashing the sailfish .simg.
After reflashing the stock rom and trying again, the phone no longer boots into UT but hangs at the 'Powered by Android' screen. When It's connected to a computer it appears as a media device with the name 'Failed to boot'.

I haven't had any problems flashing UT or stock.

I might have done some strange things with fastboot in my frustration over the last few days, could anything there be the cause? I assumed flashing stock would restore everything to factory condition but I might be wrong there.

I can't promise I can help you but it would best if you could join the Discord server (see another thread here on the forum) so that the community can troubleshoot with you and try to diagnose your issue first, it'd be easier to interact quickly than on the forum.

[therainingmonkey 2]

7 hours ago, matf said:

it would best if you could join the Discord server

Is there an IRC/Matrix bridge, or some other way to connect?

There are too many issues with Discord for me to feel comfortable using it.

[matf-kabouik 414]

There is a Matrix bridge indeed but I don't remember the address, maybe some others here can comment (@mosenuses it for instance).

[Edit]: https://matrix.to/#/#fxtec:revreso.de

Edited August 15, 2022 by matf

[matf-kabouik 414]

Dynamic orientation works in LXC even with a tiling WM, thanks to qxcompositor (we were using qxdisplay with harbour-containers since 2020 because at the time qxcompositor didn't support portrait, but qxdisplay doesn't support dynamic changes so the update qxcompositor is now superior):

Useful mostly for other devices, but even on Pro1 it can be cool when just consuming content without the keyboard.

Edited August 15, 2022 by matf