mangoapp's VRAM figure on a Qualcomm driver comes only from the gpu_mem_total
trace event. The tracefs stand-in's event list now declares that event as
Android kernels do, and reading the stand-in's events answers with one
record carrying the total KGSL reports in /sys/class/kgsl/kgsl/page_alloc,
which every Adreno under Android has and the app can read.
Three things kept Valve's mangoapp from drawing the overlay in a session:
- GLFW 3.4 takes XDG_SESSION_TYPE=wayland with WAYLAND_DISPLAY set as
Wayland-only, and mangoapp needs gamescope's X server; the wrapper sets
XDG_SESSION_TYPE=x11.
- On a Qualcomm display driver it opens a tracefs instance for the GPU memory
total and aborts when that fails, which it always does for an Android app.
The session preload now places the instance in an empty directory of its own
when the real one cannot be made.
- gamescope sends it each frame's timings through a System V message queue,
and Android kernels have no System V IPC. sysv.c now emulates the queue
operations gamescope and mangoapp use, beside its semaphores and shared
memory.
The session's proot runs with -i <app uid>, and the rootfs names that uid
root, so the client calls itself root while geteuid() returns the app's
uid. The read-only retry required geteuid() == 0 and never ran: the
Proton Experimental (ARM64) update still failed with Missing file
permissions on wineopenxr.dll (2026-09-26 16:07, 2,471 read-only files).
Also recognise the client by the path it was started by: the kernel's
/proc/self/exe names proot's loader for every process in the session.
The client runs as proot's root, which the kernel does not see, and
leaves many content files read-only; an update that rewrites one was
refused (EACCES) and cancelled with "Missing file permissions"
(Proton Experimental (ARM64), wineopenxr.dll). For the client's own
processes, a refused write-open of a file the app owns now gets the
owner's write bit - on the file, or on the directory a new file goes
into - and one more try. Games and Wine keep ordinary permissions.
From Bannerlator's feat/linux-gamescope-runtime since this repo was cut:
- Session teardown (4509d788): proot was SIGKILLed, which never lets its
--kill-on-exit cleanup run; its tracees lived on with no tracer, every
seccomp-trapped syscall failing with ENOSYS, spinning at a full core for over
an hour. Now SIGTERM, a grace for proot's own cleanup, SIGKILL only if it
will not go, then a sweep of the tree it had — snapshotted first, each pid
checked against its start time.
- The Steam client's controller interposer is its own source
(fakeinput_steam.cpp, 28922e7c), separate from the one preloaded into Wine.
- The seeder runs on a beat so a prefix the client creates mid-session is
covered before its install scripts run (6ee86bff).
- GE-Proton and proton-cachyos as compatibility tools: the registrar
bannerlator-proton-extra and the compat script's adopt_extras (03867656).
- space.c in the session shim: a library's own free space, not the phone's.
The env-vs-args ordering bug (98d38931) does not apply here: nothing is added
after the script name.
Lifts Bannerlator's Linux/gamescope runtime out into an app of its own: the
Wayland compositor, the proot session, the runtime installer, the fake-evdev
controller path and the PulseAudio server, with none of the Wine, container or
store machinery around them.
One activity installs the runtime from the same winlator-contents catalog and a
second one runs the session: compositor on a SurfaceView, gamescope and Valve's
native arm64 client under proot, a physical pad republished into the fake-input
rings, and the session's milestones on screen until the first frame arrives.
xXJsonDeruloXx