| ecrn | ok, so I have to try something else | 01:14 |
|---|---|---|
| ecrn | it is android 12 base, and for some reason pulseaudio crashes with the module-droid-card and the like | 01:15 |
| ecrn | there is an alsa device detected with thousands of controls and no sample format supported | 01:16 |
| ecrn | so I guess direct support by alsa plugin is out of question | 01:16 |
| ecrn | another serious thing is kernel panic on bluetooth socket creation | 01:50 |
| ecrn | which I patched out just after starting the porting because it was impossible to proceed | 01:50 |
| ecrn | but kernel panic on socket() seems indicative of the quality of the driver | 01:51 |
| ecrn | or the kernel sources in general | 01:51 |
| ecrn | also the kernel is definitely not the one they used in the device despite it is labelled as such on the opensource.samsung.com | 01:52 |
| ecrn | also what kind of silliness is doing chmod 0777 on the kernel source dir | 04:00 |
| rinigus | @b100dian: no luck on my side either. I have compilation errors with Sony copyleft kernel, but that would need some more investigation. Used configs that were either started in the corresponding readme or pulled from kernel image itself | 05:19 |
| rinigus | Also | 05:19 |
| rinigus | Also tried kernelsu gki kernel, but that bootlooped. But patching boot image with kernelsu via their app worked. | 05:20 |
| rinigus | When extracting kernel and ramdisk from stock, I get way smaller total size than not image itself. Reassembling it back also leads to image that's way smaller than original. Haven't tried to boot it yet, will try tonight | 05:23 |
| rinigus | @b100dian: note that with aosp13 images, we have different os-level/patch level used to make them during aosp13 builds by mkbootimg. there could be some roll back protection kicking in. although then it should be on some level that doesn't prevent booting into aosp13. | 06:20 |
| rinigus | maybe small size of stock kernel+ramdisk is ok. when getting info regarding the image using avbtool, it reports similar size | 06:33 |
| rinigus | @b100dian: I managed to boot GKI kernel (kernelsu to be more specific). in my case, I repacked it as described at the end of https://github.com/sailfishos-sony-nagara/main/wiki/Extracting-data-from-stock-ROM to follow stock as close as possible | 17:30 |
| rinigus | interesting, even fastbootD works with it | 17:32 |
| rinigus | so, now we just have to get the kernel built that we could use | 17:33 |
| T42 | <b100dian> rinigus: nice! - is this a generic kernel? And is it supporting all the hardware after booting? | 17:46 |
| rinigus | don't know whether it supports all features. but camera, wifi, bt seem to work | 17:50 |
| rinigus | downloading gki sources right now and want to see how to change its config. but maybe sfos hybris changes are too large for vendor modules to be happy about it | 17:51 |
| T42 | <b100dian> very cool - audio included, right? | 18:04 |
| T42 | <b100dian> Not sure about hybris patches vs modules, what do you think would be the problem? Ah, I see, you re-used the modules from the system. But we can ship our own modules in rootfs | 18:04 |
| T42 | <b100dian> rinigus: looks like you're on a way to generic sfos image 😉 | 18:05 |
| rinigus | @b100dian: no, its mainly experimenting and getting ready for stock/lineage port :) | 18:06 |
| rinigus | re audio and others - will test a bit later. will try to spin up gki build and will be back after a break | 18:07 |
| rinigus | @b100dian: I am sure its not gonna work in terms of generic builds. otherwise it would have been done already | 18:08 |
| rinigus | @b100dian: audio, camera, gps, bt, wifi works. gsm didn't test | 20:06 |
| rinigus | that's with gki kernel | 20:06 |
| rinigus | modules are reused from vendor, but that seems to be the point of gki to offload hardware to vendor partitions | 20:07 |
| rinigus | fp works as well | 20:12 |
| T42 | <b100dian> ah, makes sense, thanks for reminding me. But this is great nonetheless! That you can boot with a such packaged kernel | 20:14 |
| T42 | <b100dian> So there's the fear that adding kernel configs might 1. change the vermagic (I now have 5.10.168-gki-g926524f66307-dirty) and 2. worse, change the ABI for modules | 20:15 |
| mal | rinigus: indeed interesting to know if existing modules in vendor_boot would work with a kernel with hybris patches, I think nobody has tried that before | 20:22 |
| mal | *hybris defconfig changes | 20:22 |
| T42 | <b100dian> According to this KMI (Kernel Module Interface) should be stable https://source.android.com/docs/core/architecture/kernel/generic-kernel-image#fragmentation-fix-gki | 20:23 |
| rinigus | @b100dian: vermagick? I am currently booting with vbmeta --disable... options | 20:24 |
| rinigus | re ABI - I think that's a main concern. as mal pointed out, if kernel config would change it so much that vendor_xxx cannot be used | 20:25 |
| T42 | <b100dian> No, I mean the version magic string in modules - if you run modino on one e.g,. modinfo /vendor/lib/modules/ipam.ko | 20:25 |
| T42 | <b100dian> A mismatch in that magic string would prevent them to load without 'force' | 20:26 |
| mal | some devices like C2 require very early loading of modules to get anything working so it really follows the way of having all hw features in vendor_boot | 20:26 |
| rinigus | permission denied :( . would have to flash gsi-lineage to test that. | 20:27 |
| mal | compared to sony which doesn't have any modules | 20:27 |
| mal | well at least not in vendor_boot | 20:27 |
| mal | or was it so that sonys don't have modules at all by default | 20:27 |
| rinigus | mal: I still have vendor_boot same as in stock. | 20:28 |
| mal | that is different question, how different is vendor kernel vs aosp kernel | 20:28 |
| rinigus | and seems like gki kernel builds cleanly for a change :) . when using simplified build instructions from here https://kernelsu.org/guide/how-to-build.html | 20:29 |
| rinigus | mal: don't know how different it is from aosp. but we couldn't boot stock with aosp boot images (5.10 and 5.15). | 20:31 |
| rinigus | maybe some avb data was different or something else. hard to say | 20:31 |
| T42 | <b100dian> Yeah, I have not tried very hard, just using an existing voot/vendor boot/vbmeta I had from previous weeks | 20:34 |
| rinigus | while I could build gki kernel, something was off in the image with it and it rebooted. so, before hybris config, would have to fix that. well, one step in a time and will continue tomorrow | 20:43 |
| rinigus | boot looped basically. | 20:43 |
| T42 | <b100dian> even with that manual packaging step? | 20:44 |
| rinigus | @b100dian: yes, with that manual packing. I think I used ramdisk from older gki kernel, they didn't match. Hopefully that's a reason. Afk now and could check it tomorrow. | 21:04 |
| T42 | <b100dian> got it, thanks. | 21:05 |
| rinigus | Kernel was also newer than the one used by kernelsu. As that worked, will have to look and check out the differences. Probably some silly mistake on my side | 21:10 |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!