I have a similar issue as seen in #2049 with my Xiaomi Mi A2 (non-lite) which had stock Pie with Magisk 19.x (can't remember) running.
I wanted to apply the latest OTA update to Android 10 (security update from March 1st).
After uninstalling Magisk and applying the OTA, the phone boots the stock bootloader as expected.
Now I fastbooted into TWRP and extracted and copied both boot images.
I'm on slot b - so back in Android I patched the corresponding boot image via Magisk Manager and tried to fastboot it. The colored line starts to move which means the boot process itself works but it runs like that forever. I also tried to fastboot the extracted unpatched image which also works so there must be something wrong with the patched image.
stock boot image
patched boot image
Do you have any ideas on that?
_Originally posted by @SoftwareSchlosser in https://github.com/topjohnwu/Magisk/issues/2049#issuecomment-614779955_
It seems like TWRP can't handle the encryption anymore on Android 10. Maybe that is related?
You should figure out exactly what was the last version that worked.
It was 20.3 but that was on Pie and even the same version doesn't work since the update to Android 10.
I can't get magisk installed on los 17.1 , Razer phone. Ends in a boot loop.
Can you try on the latest Canary? The whole init logic was rewritten, so it might fix this.
So Magisk Manager (765d5d97 (289) (9)) offers me Magisk 20.4.05 (8ca5a048) after switching to Canary channel. If that's the right version then it unfortunately also doesn't work.
None of that makes sense. Manager is 283, and Magisk is 20405 right now.
Ok manager is 283 - that was a typo. For magisk I added two dots between the digits so if you remove them you get 20405. Isn't 20.4.05 the version number representation?
edit:
Sorry for closing and re-opening. Must have hit the button accidentally when putting away my phone.
Nope. It isn't represented by anything, it's just the version code. Anyway, guess it didn't work?
So "20.4.05" is not the version string for version code 20405? What a coincidence then.
Anyways it unfortunately doesn't work.
Correct. There is no version number for Canary builds anymore, only hashes and version codes.
I have similar issues. I have a Sony Xperia XA2 with LineageOS. After upgrading to LineageOS 17.1 (actually fresh install), I run into boot loops after installing Magisk. What helps is wiping Dalivk Cache, but only for one single time. So I have to do this now before every reboot. (Luckily TWRP works for me even with encryption...)
I have checked this with the latest Canary: 2468f5a6 (20406) for Magisk and 765d5d97 (283) for the Manager.
I cannot provide a logcat, since this seems to happen before adb gets initialised. I only see it "- waiting for device -"...
@SoftwareSchlosser can you try if wiping dalvik cache helps?
@SoftwareSchlosser can you try if wiping dalvik cache helps?
Just tried that (also with 2468f5a) but it didn't help :( A log would be nice but I can't get one via adb either.
i hate that this is an official forum for issues relating to magisk but all I get is closing my threads without offering solutions. i still have bootloops on infinix x604 (android 10) after flashing patched boot with both the latest stable and canary versions. i hope someone could at least point me in the right direction
Same here. Welcome to the club!
Den tors 30 apr. 2020 22:12victornworgu notifications@github.com skrev:
i hate that this is an official forum for issues relating to magisk but
all I get is closing my threads without offering solutions. i still have
bootloops on infinix x604 (android 10) after flashing patched boot with
both the latest stable and canary versions. i hope someone could at least
point me in the right direction—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
https://github.com/topjohnwu/Magisk/issues/2695#issuecomment-622083586,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/APEZVD5BKDICLFS2Q7WBX53RPHLRTANCNFSM4MKUTSWQ
.
This is so goddamn annoying. Bootloop on android 10, LineageOS 17.1 having tried both latest Canary and the normal 20.4 release.
+1 bootloops with havoc on mix2, need to go recovery and reboot 2 times for it to boot
for me infinite reload on los 17.1 on cheryl too. Entering the recovery and hitting reboot system now from THERE lets me boot to my sys
The solution doesn't seem to work for me, using Android Q SM-A600FN
Tried 99ef206 (20412) but it also didn't work. Why has this been closed? Was it supposed to be fixed?
Bootloop on latest Q (QQ2A.200501.001.B2) on Pixel 3 XL. logcat http://ix.io/2mMs
Tried both stable 20.4 and canary
Now I realise that I actually don't have to wipe the dalvik cache, just entering twrp is sufficient.
@topjohnwu This issue is not fixed, please reopen or point us to another issue where the bug is tracked. Cheers!
Possibly #2806 is a currently open similar issue. I won't reopen this one without @topjohnwu's input however.
i found the fix apparently magisk fails to mount fstab during boot so we can bypass that by adding mount in init.qcom.rc
add this in vendor/etc/hw/init.qcom.rc
` on fs
wait /dev/block/platform/soc/${ro.boot.bootdevice}
symlink /dev/block/platform/soc/${ro.boot.bootdevice} /dev/block/bootdevice
mount_all /vendor/etc/fstab.qcom`
Yeah won't work for those of us that can't run TWRP.
@osm0sis I mean this issue is older than the one you linked to. Why would that take precedence?
EDIT: Also that one was closed out for a completely unrelated reason.
I told you, John closed it so I won't reopen it without his further acknowledgement. Not sure how else to spell that out.
@topjohnwu this is still a problem. What can we do to help solving it?
i found the fix apparently magisk fails to mount fstab during boot so we can bypass that by adding mount in init.qcom.rc
add this in vendor/etc/hw/init.qcom.rc
` on fs
wait /dev/block/platform/soc/${ro.boot.bootdevice}
symlink /dev/block/platform/soc/${ro.boot.bootdevice} /dev/block/bootdevicemount_all /vendor/etc/fstab.qcom`
solved it for me. I'm on a Sony XA2 with lineageOS 17.1. The entry mentioned there was already part of vendor/etc/hw/init.qcom.rc, but with
mount_all /vendor/etc/fstab.qcom --early
Just deleted --early und rebooted like a charm. Will this be replaced after an OTA update? I'll check that on the build tomorrow...
I'm not on LOS but on Xiaomi stock firmware. How could I benefit from this kind of fix?
I've personally asked @topjohnwu to review this again, especially since the nature of the regression might now be revealed by the workaround you guys have found. I'll open this again in the meantime.
The linux under the hood can (should) be the same. Just search for the file and edit it, if it's available...
If someone could please try old Magisk canaries and tell us with which build the issue first happened that would also be helpful, since I don't think we ever figured that out here.
https://github.com/topjohnwu/magisk_files/commits/canary
Edit: Wait, nevermind, I guess we sort of got to the bottom of that, since once these devices updated to Android 10 they just never seemed to work on any Magisk version. Must be some edge case Magisk's init logic still doesn't cover.
The linux under the hood can (should) be the same. Just search for the file and edit it, if it's available...
If that means I would have to edit a file on the system partition of my phone then that's absolutely no option because unfortunately Xiaomis firmware updater refuses to do the update if the system partition has been altered in any way. I once did that and had to wipe and re-flash everything (incl. data) which was a real pain.
i found the fix apparently magisk fails to mount fstab during boot so we can bypass that by adding mount in init.qcom.rc
add this in vendor/etc/hw/init.qcom.rc
` on fs
wait /dev/block/platform/soc/${ro.boot.bootdevice}
symlink /dev/block/platform/soc/${ro.boot.bootdevice} /dev/block/bootdevicemount_all /vendor/etc/fstab.qcom`solved it for me. I'm on a Sony XA2 with lineageOS 17.1. The entry mentioned there was already part of vendor/etc/hw/init.qcom.rc, but with
mount_all /vendor/etc/fstab.qcom --early
Just deleted --early und rebooted like a charm. Will this be replaced after an OTA update? I'll check that on the build tomorrow...
This is the proper fix if you need it
https://review.lineageos.org/c/LineageOS/android_device_xiaomi_msm8998-common/+/276735
For some people this isn't an option. For example the Pixels can't use TWRP and thus cannot edit the filesystem outside of the OS.
So now we'll need to accommodate that Lineage patch too, most likely.🤦♂️
Pretty sure Pixels aren't broken, by the way.
I am using a Pixel... Granted I havent tried in a little bit. The first comment from me here is it bootlooping with a log on the Pixel 3 XL
I'll try the latest in a second.
Ok whatever it was on my device, it is working now. Thanks for the help.
i found the fix apparently magisk fails to mount fstab during boot so we can bypass that by adding mount in init.qcom.rc
add this in vendor/etc/hw/init.qcom.rc
` on fs
wait /dev/block/platform/soc/${ro.boot.bootdevice}
symlink /dev/block/platform/soc/${ro.boot.bootdevice} /dev/block/bootdevicemount_all /vendor/etc/fstab.qcom`solved it for me. I'm on a Sony XA2 with lineageOS 17.1. The entry mentioned there was already part of vendor/etc/hw/init.qcom.rc, but with
mount_all /vendor/etc/fstab.qcom --early
Just deleted --early und rebooted like a charm. Will this be replaced after an OTA update? I'll check that on the build tomorrow...This is the proper fix if you need it
https://review.lineageos.org/c/LineageOS/android_device_xiaomi_msm8998-common/+/276735
It doesn't seem to be merged, but why?
hello, i tried to root a Xiaomi Redmi 6 using magisk by paching the boot.img file, but the device didn't start after patching the file, what can I do to fix it?
It doesn't seem to be merged, but why?
Probably because it's a commit to work around a Magisk bug, which is backwards; Magisk just needs to be fixed. :slightly_smiling_face:
I had this issue after upgrading my Mi Note 2 (scorpio) to LineageOS 17.1 too.
i found the fix apparently magisk fails to mount fstab during boot so we can bypass that by adding mount in init.qcom.rc
add this in vendor/etc/hw/init.qcom.rc
` on fs
wait /dev/block/platform/soc/${ro.boot.bootdevice}
symlink /dev/block/platform/soc/${ro.boot.bootdevice} /dev/block/bootdevicemount_all /vendor/etc/fstab.qcom`
Is this specific to MSM8998 devices? I couldn't find that file on mine which has a MSM8996 Pro
I had this issue after upgrading my Mi Note 2 (scorpio) to LineageOS 17.1 too.
i found the fix apparently magisk fails to mount fstab during boot so we can bypass that by adding mount in init.qcom.rc
add this in vendor/etc/hw/init.qcom.rc
` on fs
wait /dev/block/platform/soc/${ro.boot.bootdevice}
symlink /dev/block/platform/soc/${ro.boot.bootdevice} /dev/block/bootdevicemount_all /vendor/etc/fstab.qcom`Is this specific to MSM8998 devices? I couldn't find that file on mine which has a MSM8996 Pro
Yup it works on msm8996 too tested on my Nubia z11
Check system/vendor/etc/init/hw
Is this specific to MSM8998 devices? I couldn't find that file on mine which has a MSM8996 Pro
Yup it works on msm8996 too tested on my Nubia z11
Check system/vendor/etc/init/hw
Turns out I just didn't have vendor mounted.
Apparently those lines are already there, and it booted now?
I have no idea what changed to make this happen.
Hi all,
I have the same problem, the boot goes for ever, with oneplus 2.
None of the above workarounds worked.
Thanks
CB
@cbaptista Can you make your phone boot loop and provide last_kmsg via TWRP
(last_kmsg is the kernel logs of the previous boot so you need to get them via your recovery)
@Fox2Code I can, but only after next thursday . I will post them here then.
CB
@Fox2Code there is no /proc/last_kmsg.
The device work well on android 9 and older, but after android 10 it never worked.
Using latest version from release Magisk v20.4 the magisk manager doesn't recognizes magisk su device boot OK.
Using latest version 20418 the device doesn't finises the boot process.
It also looses adb root, so logcat is limited, I took some logs from 20418
logcat.txt
Can't run dmesg because no SU, but via TWRP I can cat /sys/fs/pstore/console-ramoops AKA dmesg
console-ramoops.txt
There is some odd Unmounting spamming on dmesg of data and cache, that doesn't happen on normal boot.
Anyone has some experience with this issue?
I have a similar issue as seen in #2049 with my Xiaomi Mi A2 (non-lite) which had stock Pie with Magisk 19.x (can't remember) running.
I wanted to apply the latest OTA update to Android 10 (security update from March 1st).
After uninstalling Magisk and applying the OTA, the phone boots the stock bootloader as expected.
Now I fastbooted into TWRP and extracted and copied both boot images.
I'm on slot b - so back in Android I patched the corresponding boot image via Magisk Manager and tried to fastboot it. The colored line starts to move which means the boot process itself works but it runs like that forever. I also tried to fastboot the extracted unpatched image which also works so there must be something wrong with the patched image.stock boot image
patched boot imageDo you have any ideas on that?
_Originally posted by @SoftwareSchlosser in #2049 (comment)_
It seems like TWRP can't handle the encryption anymore on Android 10. Maybe that is related?
I had the same issue it is solved now, i tried all the version 13, 15, 17, 19 ,20.4 and canary but the issue was solved when I used the version 20.0 of magisk.
If your phone is not booting after magisk flash or it is in bootloop, the only solution that I have found is that magisk 20.0 version works fine. Hope you find this helpful.

I had the same issue it is solved now, i tried all the version 13, 15, 17, 19 ,20.4 and canary but the issue was solved when I used the version 20.0 of magisk.
If your phone is not booting after magisk flash or it is in bootloop, the only solution that I have found is that magisk 20.0 version works fine. Hope you find this helpful.
I tried that version and even downgraded Magisk Manager to the version shown on your screenshot but it doesn't even display any boot- / splash logo. Just a black screen after the fastboot boot ... command. Also cleared the cache partitions in TWRP but that didn't help too. I can boot the unpatched boot image so there must be something wrong with the patched file.
Hello again, this is the last line from my failing boot log:
mdss_fb_do_ioctl: unsupported ioctl (5401)
Does this help?
Initially reported by another user here (and I left a few comments): https://gitlab.com/LineageOS/issues/android/-/issues/2300
Essentially as soon as Magisk is installed on OnePlus 7 Pro on either stock OxygenOS or LineageOS 17.1 it either causes bootloop (most of the time was fixed by "Fix Recovery Bootloop" in TWRP for some reason) and WiFi/voice calls not working (LTE data still works). I had 20.4 running just fine until last update though.
It happens in such a way that uninstallation or re-flashing OS doesn't help, only complete reflashing of a bunch of partitions to completely stock phone state.
There have been no comments here for a while, is this fixed in nightly builds?
I always run Magisk Canary and Lineage 17.1 with TWRP 3.4.0 installed on my OP7Pro without issue.
Probably a stupid question, but readme mentions "Canary Magisk Manager", where can I get it?
Tap the download count badge for the Canary Manager, it'll grab everything else for you
Update to Canary didn't fix WiFi. There was sonething in /data left by Magisk 20.4 that was destroying WiFi on all slots right after first boot with /data restored from backup.
With everything wiped once again, fresh LineageOS 2020-08-09 installed on top + opengapps + Canary Magisk I can finally both have Magisk and working WiFi again. Yay!
Please release new stable version then, 20.4 clearly breaks Android 10 devices and Canary (in my case at least) does not.
Magisk doesn't touch anything except boot.img and put its own stuff in /data/adb (and /system/addon.d on custom ROMs). Any lasting effect basically can't logically be Magisk's fault.
But anyway, obviously another stable will happen, now stop polluting this thread.
I have a similar issue with my Xiaomi Redmi 9. I have tried Magisk 20.4, 18.4, 19.4, 20.1 but unfortunately all failed.
I firstly using Magisk 20.4 to root my device, but after flashing patched boot img, it turned into bootloop, so i have to flash back to stock boot image.Then i try other version again and again
This is the image and log:
Log
Patched Boot Image
BTW ,I cannot install TWRP in my phone because it did't support.
Add from #3099
Might be worth trying latest Canary again, SAR detection and init logic was rewritten. 👍
The device work well on android 9 and older, but after android 10 it never worked.
Using latest version from release Magisk v20.4 the magisk manager doesn't recognizes magisk su device boot OK.
Using latest version 20418 the device doesn't finises the boot process.
It also looses adb root, so logcat is limited, I took some logs from 20418
logcat.txtCan't run
dmesgbecause no SU, but via TWRP I cancat /sys/fs/pstore/console-ramoopsAKAdmesg
console-ramoops.txtThere is some odd
Unmountingspamming on dmesg of data and cache, that doesn't happen on normal boot.Anyone has some experience with this issue?
I fix my problem...
Regarding the release version 20.4 (20400), it doesn't work because it can't patch the dtb this devices was using a separated zImage and dtb , to fix this issue I add the kernel defconfig flags
CONFIG_BUILD_ARM_APPENDED_DTB_IMAGE=y
CONFIG_BUILD_ARM_APPENDED_DTB_IMAGE_NAMES=""
Removed on dt BoardConfig.mk
BOARD_KERNEL_SEPARATED_DT := true
Technically the canary version 20420 and up doesn't have this issue regarding the dtb (I assume the dtb issue was fixed on a version prior to 20420 that was the one I tested after 20400), but it will not boot the device just stay on the bootanimation screen for ever...
So to fix that issue I changed the data partition to latemount in the fstab file
https://github.com/fgl27/device_motorola_quark/commit/bec4069d85543b261e558a4200c60d91c62521b1
I did other things on the above, but what is need there is to add latemount to the data partition in fstab and change the mount process of fstab
on fs
mount_all /vendor/etc/fstab.qcom --early
on late-fs
mount_all /vendor/etc/fstab.qcom --late
So for some reason that I don't know magisk need a "delay" at init, as it is inside the data partition late mouting it resolves the issue for this device, but I do believed that with this information some improves on magisk can be made to address the problem.
regarding 20423, it doesn't work, the last that work for me is 20422... using 20423 the device boot but magisk manager just freezes and no app can request SU. I know that is a wipe version for Android 11 but extra info always help...
But no canary version is needed as with this changes that I made release version 20400 just work, as long you don't use separated dtb and zImage and late mount the data partition.
In doubt about my process let me know.
Try 20424, there was an selinux issue
Might be worth trying latest Canary again, SAR detection and init logic was rewritten. 👍
Thanks, Try It later.
Canary Fixed nothing for me, sadly.
@topjohnwu
Most helpful comment
Yeah won't work for those of us that can't run TWRP.