Manager crashes with SIGABRT on Samsung Knox: seapp_contexts lookup fails for spoofed process name (org.matrix.vector.manager)

[opassoca]

Steps to reproduce

1. Fresh update to Vector v2.2 on a Samsung device with Knox/OneUI.
2. Tap the manager notification/shortcut, or trigger the official launch intent directly:
   am start -a android.intent.action.MAIN -c org.lsposed.manager.LAUNCH_MANAGER -n com.android.shell/.BugreportWarningActivity
3. Process aborts immediately (SIGABRT), before any UI renders.

Expected behaviour

Manager UI opens normally (parasitic launch hosted inside com.android.shell).

Actual behaviour

Zygote aborts while specializing the process. logcat:

E SELinux: seapp_context_lookup_internal: No match for app with uid 2000, seinfo platform:privapp:targetSdkVersion=35:partition=system:complete, name org.matrix.vector.manager username shell appid 2000
E SELinux: selinux_android_setcontext: Error setting context for app with uid 2000 ...
F libc: Fatal signal 6 (SIGABRT) ... in tid ... (ch_zygote)
F DEBUG: Abort message: 'JNI FatalError called: (org.matrix.vector.manager) frameworks/base/core/jni/com_android_internal_os_Zygote.cpp:2179: selinux_android_setcontext(2000, 0, "platform:privapp:targetSdkVersion=35:partition=system:complete", "org.matrix.vector.manager") failed'

Confirmed identical whether triggered via the (stale) notification or the raw official launch intent, so this is not a shortcut-caching issue, it is the specialize call itself failing.

Root cause (as far as I traced it)

The parasitic manager specializes its hosted process under the synthetic identity org.matrix.vector.manager (uid 2000 / shell), apparently for stealth. The seapp_contexts of this ROM only has a name-matched rule for the real com.android.shell:

user=shell seinfo=platform domain=shell name=com.android.shell type=shell_data_file

There is no generic seinfo-only fallback that matches either: the seinfo string passed is the full compound platform:privapp:targetSdkVersion=35:partition=system:complete, which is not equal to any existing seinfo= value in the file. So the lookup for the fake name fails outright and Zygote hard-aborts instead of degrading gracefully.

Possibly the same underlying dependency as #483 ("reinstall com.android.shell cause this problem on Samsung Galaxy S10+(OneUI4.1)"): both seem to trace back to the parasitic manager depending on the identity/context of com.android.shell being resolvable, just triggered differently.

What I tried (did not work, documenting so it is not repeated)

Added a name-matched seapp_contexts entry via a KernelSU systemless module, mirroring the existing com.android.shell line with name=org.matrix.vector.manager. On this ROM, plat_seapp_contexts is read by init before SELinux relabeling of the magic-mounted overlay completes, so this caused a boot loop. Reverted (module removed), device stable again, crash unchanged. Not proposing this as a fix, just flagging that patching seapp_contexts via magic-mount is fragile at this boot stage on Knox.

Environment

• Device: Samsung Galaxy S10+ (SM-G973F), Exynos 9820, custom ROM (ExtremeROM Nexus)
• Root: KernelSU 1.0.9 + ZygiskNext 1.3.2
• Vector: v2.2 (versionCode 3080), updated from LSPosed v1.11.0
• Android 15 (API 35)