On phones where Android blocks the GPU's sysfs files for apps, the HUD
reads the GPU without root (#1969). On Adreno 8xx that showed no GPU
clock and only the partial load from Vulkan timing (OnePlus 13, Adreno
830, reported by iamTooShaar).
- GPU clock: the clock job used the Adreno 6xx/7xx registers. The
register layout is now picked from the KGSL chip id: on 8xx
CP_ALWAYS_ON_COUNTER is 0x8e7 and the RBBM cycle countable is 1.
Unknown later generations read nothing.
- GPU load: hardware-scheduled kgsl kernels (every 8xx, and some 7xx
such as the AYANEO Pocket FIT's) never write the memstore ringbuffer
timestamps. When those stay 0, the reader samples the contexts' own
memstore entries (every context that has submitted work, up to 64,
contexts with new or running work first).
- HUD diagnostics report: a "GPU readers without root" line with the
reason each reader works or fails, "none" instead of "null", and the
per-core CPU % source named honestly.
Tested: iamTooShaar's OnePlus 13 shows GPU clock and load with the PR
test build (06be2c5). Pocket FIT standalone build of the final readers
against the kernel's sysfs numbers: idle 1-2 % vs 3-4 %, GPU job loop
25-29 % vs 28-30 %, clock 1050 MHz vs 1050 MHz. The review fixes after
the test build were not tested on the OnePlus 13.
DOOM (2016) asks for exactly two swapchain images and stops with "FATAL
ERROR: getSwapchainImagesKHR failed with error (VK_SUCCESS)" when it gets
more. Every game launch gets GAMESCOPE_WSI_MIN_IMAGE_COUNT=8 and
GAMESCOPE_WSI_ENSURE_MIN_IMAGE_COUNT=1, so gamescope's WSI layer gave it 8
images (3 with only the count removed) and the game crashed on start.
When DOOMx64vk.exe sits beside the launched program (DOOMx64.exe for a GOG
install, or DOOMx64vk.exe), both variables are removed and
vk_x11_strict_image_count=true is set, the option Mesa's driconf gives the
game. Other games keep the defaults. A per-game value in the environment
editor still wins.
Tested on an AYANEO Pocket FIT with DroidDeck 0.6.9 and the same values set
per game: the swapchain gets 2 images and the game reaches its main menu at
60 fps; before, it crashed on every launch.
* feat: manage several Steam libraries beside internal and the card
* fix: serialize Steam library registry edits, refuse folders inside a Games root, refresh games after changes
* fix: rename refreshes games, added x86-64 libraries join redist seeding and direct mapping
* chore(i18n): translate every missing string in all locales
Adds the 87 strings the bottom-screen Dashboard and in-game stats pages
introduced to all 32 locales, plus the 51 older settings strings Japanese
and the 50 Chinese locales were still missing. Every locale now has all
3456 translatable strings with the same placeholders as English.
Also translates 36 entries that were still English where the language
does not use the English word, and makes es-419 touch_app_id use the same
%1$s placeholder as English.
* chore(patches): drop comments from the lines downstream patches add
Removes comment lines and trailing comments from lines our patches add,
with hunk headers recounted. Lines that a later patch uses as context
and the upstream backport 0123 are left as they are.
The gamescope, FEX, proot and xodus series still apply to their pinned
sources, and the patched trees differ only in those comments.
test_hud_frames now finds the HUD writer by its first declaration.
* docs: credit Greenlight, Better xCloud, viture-v2, GameNative and DiRTTelemetryErrorFix
These were credited only in source headers and NOTICE files.
* chore: remove source comments and unused functions
Removes explanatory comments from first-party Kotlin, Java, C, C++,
Rust, TypeScript, Python, shell, Gradle, CMake and workflow files.
License and attribution headers stay, and vendored third-party trees,
generated files and license files are untouched. A shellcheck directive
is kept, and udevmon's fallthrough comments became fallthrough attributes.
Syntax trees and token streams match the previous sources apart from the
comments and Python docstrings. Module docstrings that argparse prints
are kept, and device-smoke.sh prints the same --help text from a heredoc.
Also removes 37 functions and 4 fields that nothing referenced.
* chore: remove comments from the Syncthing integration and the latest main sources
* docs: credit Syncthing and the syncthing-android fork
* chore(i18n): translate the Syncthing strings in all locales
The 107 strings the Syncthing integration added are now in all 32 locales.
sync_web_path_locked names the Syncthing tab by its label.
* docs: credit @xXJSONDeruloXx for the HUD metric readers
The CPU, GPU load, thermal and battery readers come from their
PerformanceHudView pull requests #777, #877, #991, #1406 and #1573.
The FusionHUD credit line is unchanged.
* docs: credit the HUD metric readers to @xXJSONDeruloXx's work
* feat: KDE desktop as a bottom-screen mode beside the top-screen session
* test: bottom-desktop launch test sets DISPLAY instead of inheriting it
* fix: drop the bottom desktop's launch when the session stops while preparing it
* fix: bottom desktop restarts on a fresh surface, trackpad/touch toggle, loading throbber, DroidDeck's display driver
* fix: bottom desktop pauses with the session and opens the keyboard for a tapped text field
* fix: no bottom desktop guest starts or keeps running unfrozen while the session is paused
* fix: bottom desktop treats a freeze in progress as paused
* fix: only the session that began a freeze clears its pausing flag
* fix: bottom desktop launch is stored and cleared only for its own session generation
* fix: a stopped session tears down the bottom desktop it started even after a newer launch is prepared
* fix: a new session's bottom desktop replaces one an earlier session left running
* fix: release the bottom desktop's keyboard watcher when its compositor dies
* fix: bottom desktop starts its guest only for the launch its compositor was attached for
* fix: bottom desktop ignores compositor replies to an earlier session's attach
On some Adreno devices (the Retroid Pocket Duo) SELinux hides kgsl's sysfs
nodes from apps, so the HUD showed no GPU % and no GPU clock. Devices whose
nodes are readable keep using them; nothing changes there or on GPUs
without kgsl.
Where the nodes are blocked:
- GPU %: a native reader maps kgsl's memstore read-only (any kgsl client
may) and samples the ringbuffer start/end timestamps once a millisecond
while a HUD is shown. That covers all GPU work, Android's own
composition included.
- Backup for GPU %: an implicit Vulkan layer in the guest
(tools/gpuload) and the compositor time their own GPU work with
timestamp queries. It runs until the memstore stamps are seen moving,
and for the whole session if they never move.
- GPU clock: a small GPU job of our own reads the always-on and RBBM
cycle counters (libgpuclock).
The HUD diagnostics export lists which source each GPU value came from.
tools/gpuload/build.sh takes the Vulkan headers from build_local.sh's NDK
first and keeps NDK paths with spaces intact.
Codemasters EGO games (DiRT 3, DiRT Showdown, GRID, F1 ...) pick their thread plan by core count from system/hardware_settings_restrictions.xml, which only has plans for 2, 3, 4, 6 and 8+ cores. When DroidDeck leaves a phone's slow cores out Windows can see 5 (Retroid Pocket Duo), and DiRT 3 then puts every worker on one core and stops on a loading screen; Showdown closes at start.
When a game starts with WINE_CPU_TOPOLOGY set and its restrictions have no cpu rule for that count, droiddeck-game-env adds one copied from the nearest smaller plan (5 -> the game's 4-core plan), only if that plan's worker map file exists. The file as it was is kept as .droiddeck-original; later launches change nothing. Only ASCII digits count as a core count, and the edited file is staged under a unique name.
Tested on a Retroid Pocket Duo: with DiRT 3's shipped file, the plan is added at launch and the game loads to its menus and races.
A Dashboard mode for the second screen of dual-screen handhelds (developed on a Retroid Pocket Duo), the default on dual-screen devices. It opens on Chat at each Steam start and switches to In-game stats when a game starts.
- Downloads: Steam downloads and updates (SD-card libraries too), shader caches and DroidDeck store downloads.
- Notifications: other apps' notifications once access is allowed, kept in memory only.
- Screenshots: a game screenshot without the overlay, and a gallery of DroidDeck's and Steam's screenshots.
- Chat & voice: mic and game volume buttons, and Steam's web chat signed in with the Steam client's own session.
- In-game stats: FPS, CPU, GPU, RAM and battery plus the running game or Steam client type, and a racing dash from Codemasters, Forza and OutGauge games' data output, with War Thunder and Elite Dangerous layouts; speed in km/h or mph.
- Send dashboard data: turns a Codemasters game's data output on for the dash (per game or Automatic) and keeps the original setting to put back.
- DiRT 3 / DiRT Showdown: with their data output on they stop with error 10038 at race end; a ws2_32.dll stand-in (tools/dirt-ws2) is swapped into the prefix while the game runs, skipped on DroidDeck-Proton 11-11.10 which has the fix built in. Same behaviour as DiRTTelemetryErrorFix by RoccoC, based on Carsten Gottschling's fix.
About "telemetry": here it means a racing game's own live driving data (speed, gear, rpm, pedals), not app analytics. For In-game stats no information leaves the device and nothing is shared: the game sends its data to 127.0.0.1 on the same device, DroidDeck only listens there while the page is open and keeps the numbers in memory just long enough to draw them. DroidDeck sends no data about the user or their use of the app; logs leave the device only when the user taps Share logs.
* Run x86_64 Flatpak apps through FEX
* proot: run an execveat with an absolute path like execve
* Find x86_64 Flatpak installs and resolve sandbox paths like the sandbox does
* fix: a read-only file on a USB drive no longer fails to close the first time it is read
On the RedMagic Astra 2, Android's NTFS USB drive returns EIO from close() for a file opened read-only when that first read updates its access time; the update itself succeeds and the descriptor is released. The x86 loader under FEX treats a failed close as fatal, so every Linux game newly installed or updated on such a drive failed to start once (libfreetype.so.6: cannot close file descriptor: Input/output error). The session library's close() now reports success for EIO on read-only descriptors; errors on descriptors opened for writing are still returned.
* Check the read-write close still reports EIO
* fix(audio): DirectAudio no longer crackles where Android refuses low-latency output
On devices where Android won't give the game a low-latency output (AYN Thor,
Android 13), DirectAudio kept Wine's 10 ms period and 12 ms buffer. An
emulated game audio thread can't keep that pace there, so audio crackled.
The session now checks for low-latency output and, only when it is refused,
asks DirectAudio for a 20 ms period and a 60 ms buffer.
DirectAudio's settings are now AUDIO_DIRECT_* everywhere in DroidDeck and are
handed to the driver under the names it reads when DirectAudio is set up.
AUDIO_DIRECT_MS and AUDIO_DIRECT_PERIOD_MS can be set per game.
* fix(audio): game profile DirectAudio settings and the microphone-off setting reach the driver
* fix(desktop): video wallpapers no longer race ahead after DroidDeck returns from the background
* test: give the vrclient loader harness the clock syscall hook
* fix: a session in picture-in-picture draws at 30 fps instead of the full refresh
Since auto picture-in-picture became the default (#1727), pressing Home leaves the session running in a small window that takes no input. A game kept drawing at its full rate there: on a REDMAGIC Astra 2 Among Us in picture-in-picture drew 3.31 W, the same as playing it and about 1.05 W more than a suspended session. While in picture-in-picture the compositor now paces the session at 30 fps (or a lower limit already set). Mirroring to glasses keeps the glasses rate, and the display mode request ignores the picture-in-picture limit so the app in front is not slowed.
* fix: keep the no-argument output limit the glasses pacing tests read
* fix: hold scene renders to the output limit and keep frame generation off in picture-in-picture
A frame limit only paced dmabuf buffer releases, so a shared-memory client was still drawn on every vsync with its frame callbacks sent at the panel's rate. The vsync tick now waits until the limit's interval, less half a refresh, has passed since the last scene. Frame generation is disarmed while picture-in-picture is shown and armed again when it ends; frame generation requests are applied in order.
* Give the commit render harness the frame limit hold
* Hold scene renders to a frame limit deadline, also on the fallback timer, and keep frame generation for glasses mirroring in picture in picture
* fix(xbox): keep the game in its own folder on the short drive instead of at the drive root
* fix(xbox): SD card Xbox games can read their own package files
Xodus mounts a store library on the selected SD card at /mnt/droiddeck-stores/<card> and links the launch view to it, but game sessions only bound that library as /mnt/droiddeck-sd. The links were dangling inside the game, so MicrosoftGame.Config, appxmanifest.xml and XCurl.dll could not be read and GDK games stopped with "Failed to initialize GDK". Game sessions now bind the same store roots Xodus uses.
* fix(xbox): give the SD Steam library its own store path when another store root shares the card
externalRootsForXodus added the selected Steam library's Games folder with the card's guest path even when another store root on the same card already used it, so both bound to one destination. It now takes a path-derived suffix in that case, the same way externalRoots separates its own roots.
* perf(epic): keep the hidden launcher from loading its store and Fab pages, and compress it while games run
The hidden Epic Games Launcher that stays up for sign-in while a game runs pre-warmed its store and Fab pages on a display nobody sees. Under FEX each page used about one core and grew past a gigabyte, crashing and reloading in a loop. The hidden instance now skips the store pre-warm, gets a blank page in place of Fab, and is paged out to zRAM with the Steam client once a game has focus. The visible launcher is unchanged.
* test(zram): keep the scan fixture PIDs above the kernel PID range
* Use the second screen on dual-screen handhelds
Detect built-in second screens such as the Retroid Pocket Duo and AYN Thor.
The sections move to a menu on the bottom screen, L2 and R2 switch sections,
and text fields type on the bottom screen so the top keeps the full view.
Sessions open the bottom-screen trackpad and keyboard when they start.
Setup gains a Dual screen choice (Automatic, On, Off) with toggles for the
bottom menu and the in-game controls. Closing the bottom menu no longer ends
a running session.
* Type into Steam from the bottom screen
While the bottom-screen trackpad and keyboard controls are open in a Steam
session, Steam's own on-screen keyboard no longer covers the top screen.
The agent keeps Steam's text target and asks the app instead, which opens
the Android keyboard on the bottom screen and hands focus back to the game
when it closes. Modal and VR keyboards, and sessions without the bottom
controls, keep Steam's keyboard. The Show keyboard button now follows the
keyboard when it is closed another way.
* Open the bottom controls after the first frame and tidy dual-screen translations
* Move the in-game menu to the bottom screen on dual-screen handhelds
Fix controller input and focus on the second screen: keys and sticks reach the game after touching the bottom controls, closing the bottom controls returns focus to the game, B closes bottom typing, and the typing panel keeps a focusable window until focus has left it.
* Let B close the open layer of the session menu first and focus the selected bottom-screen mode
* Let the Retroid Pocket Duo Home double press open its quick menu during sessions
* Open the keyboard on the bottom screen for Steam text fields and hide it on Enter or when the field loses focus
* Keep controller focus on the top screen when a section is picked on the bottom screen
* Keep LB and RB on the store tabs and let Back return from a store to its tab
* Show the Adreno 722 as supported on the Retroid Pocket Duo
* Label the Steam integration suggestion's close button Close and cover that closing it is permanent
* Close the docked menu with its presentation, keep a focus picked during the stop delay, and ignore panel picks on process-settings pages
* perf(fex): answer raw stat, readlink and statx syscalls through the proot fast path
FEX issues newfstatat, readlinkat and statx as raw syscall() calls, which skipped the libc wrappers the fast path hooks and fell through to ptrace. Route them through libblfastpath the same way raw openat already is, and keep the fast path's own /proc/self/exe lookup off the hooked syscall().
* test(fex): keep the raw lookup test passing on kernels without statx
The session HUD is FusionHUD (Bannerlator's HUD) with every reading taken from
the same source, and worked out the same way, as the MangoHud build that runs
Steam's performance overlay in a DroidDeck session.
- gamescope patch 0125 (gamescope-3.16.29-p19) shares the frame times that
mangoapp reads (one per signalled commit of connector 0) in a ring named by
DROIDDECK_HUD_FRAMES; the app reads it once per display frame and counts the
compositor's frames when gamescope sends none.
- MangoHud's rules: FPS over periods of at least 500 ms, frame time = 1000/FPS,
200-frame graph on a fixed 0-50 ms scale; CPU, GPU, temperatures, RAM, swap,
VRAM and battery from MangoHud's sources. Zones 22/26/28 are typed as GPU and
DDR sensors, so Steam's overlay CPU temperature no longer includes them.
- Five sizes in a panel shaped like the DroidDeck mark, with a centred header,
one left edge, neon outline and FPS; Minimal is the default size. Settings on
a new "Performance HUD" drawer page, with Bannerlator's options.
- The HUD reads sensors and polls frames only while it is drawn (not while
hidden, in picture-in-picture or behind another app), keeps its place after
picture-in-picture, and the diagnostics export has its own reader.
Device-tested on the AYANEO Pocket FIT: same readings as Steam's overlay side
by side in Dead Cells; the HUD updates in a game, hides in picture-in-picture
and comes back in place, no crashes.
* fix: stop stale Android readbacks snapping Steam's brightness slider back during a held stick
* fix: let deploy_fast sign with the release env on macOS and Linux
* fix: honor DROIDDECK_SIGNING_ENV first and cover the env signer in tests
* fix: honor DROIDDECK_SIGNING_ENV first and cover the env signer in tests
* fix: retry held brightness readbacks, use a monotonic hold clock and resolve the signing env path
* fix: resolve the build-tools path before signing changes directory
* fix: pick the newest build-tools numerically and keep the env file from replacing them
* fix: only pick build-tools that have aapt, zipalign and apksigner
Libc-internal file operations bypass the existing preload wrappers and
still stop in PRoot. Patch recognized ARM64 SVC sites in libc and the
runtime loader, reuse the current translator for 20 path syscall
numbers, and return declined calls to PRoot. Enabled by default after
the tracer-key check; `PROOT_FP_PATCH_SVC=0` remains the troubleshooting
opt-out.
The hardened implementation patches 47 sites on Thor's glibc 2.43.
Relays preserve general/SIMD registers, flags, FP state and errno, and
provide unwind metadata back to the original libc frame. Text is patched
through `/proc/self/mem` after relay pages become executable.
Initialization skips multiple live threads and SVE/SME. Alternate signal
stacks and insufficient remaining stack use PRoot; main-stack growth
respects the current limit. No proroot code or binaries are imported.
Robust-list set/get sites deliberately remain with PRoot/the kernel.
Device tests reproduced stale state when the earlier private-TLS
implementation mixed patched and raw callers. Removing that shortcut
fixes coherence and preserves the host's actual return policy; the
earlier thread-creation gain is retired. SDK forwarding still enables
faccessat2 ENOSYS fallback and clone3 ENOSYS only before Android 15.
Successful clone3 is never executed from the C relay. Existing fsync
handling is unchanged. No in-process RET_TRAP filter is added, so
SIGSYS-mask/handler machinery is not applicable.
Version-2 build-ID/API tables cache relative patch sites in `/tmp`.
Every hit revalidates executable bounds, alignment, instructions and
syscall numbers. Invalid/unsafe files rescan; publication uses a private
temporary file and atomic rename. The PRoot dependency fetch also gains
a mirror matching the existing pinned checksum, resolving upstream
availability failures without changing versions.
Five alternating pairs using the final bundled library on Thor/Android
13:
| operation | wrappers only | hardened relays |
|---|---:|---:|
| libc-internal stat | 49.6 us | 6.0 us |
| libc-internal open/close | 50.4 us | 7.8 us |
| libc faccessat | 49.7 us | 22.0 us |
| libc getcwd | 54.9 us | 5.0 us |
| thread create/join | 136 us | 144 us |
| complete tiny exec/wait | 4.44 ms | 5.60 ms |
An earlier five-pair hardened-library run measured stat 50.1 to 5.9 us
and open/close 51.9 to 7.1 us. The metadata gains survive the safety
checks. Neither run shows a thread-creation gain, and tiny execs retain
roughly 0.9-1.2 ms overhead.
Three full Steam startup pairs yielded Home-ready medians of 47.02 s off
/ 45.25 s on, with overlapping 43-50 s ranges. Steam/CEF-only PSS
medians were 1,223.1 / 1,233.9 MiB; the roughly 11 MiB difference is
smaller than observed variation, but does not establish zero memory
cost. Startup/memory used one unchanged signed bundled-library APK
without overrides; the final stack-gap guard only increases distance
reserved above other mappings. No reliable whole-client startup
improvement is claimed.
Earlier fresh Half-Life SD installs took 350/341 s off and 326/327 s on.
These are historical results from the preceding robust-list revision,
not a post-hardening download retest. Network/shader scheduling varied,
allocation did not reliably improve, and the repeat baseline committed
faster. This PR demonstrates metadata acceleration; it does not
demonstrate a reliable fix for Steam's SD allocation/commit bottleneck.
Validation on `2c3c135b` (production code unchanged since `4998b046`):
- Local release build, runtime/native-library checks, release signing
and four focused Android tests pass. [Final-head
CI](https://github.com/Droid-Deck/DroidDeck/actions/runs/40704837477)
passed: 1,431 Python/native tests (49 skips), 2,648 Android tests (zero
failures/skips), and 162 Rust tests. CodeRabbit is green on the same
head. The installed production APK forwards API 33 and patches 47 sites
with an unset enable flag and no override; Steam Home renders.
- Embedded and shared-PIC matrices pass on glibc 2.43; shared-PIC also
passes on separately loaded glibc 2.39. Coverage includes raw
results/errno/registers/FP state, path/flag/error and protected-pointer
cases, exact readlink boundaries, mixed raw/relay robust state,
concurrent threads, fork, thread churn, alternate-stack signals,
deferred/FIFO cancellation, signal siglongjmp recovery, forced unwind
with caller cleanup, and limited remaining stack.
- Four device test-process filters deny the memory patch file, relay
mapping, executable permission, or patch writes. Every case leaves
original syscall execution working. Host injection also verifies partial
installation after an individual site fails.
- Concurrent atomic symlink replacement passes 3,000 internal
stat/readlink iterations at default/zero TTL. Eight focused host tests
pass, covering instruction/branch limits, stack/limit fallback,
untrusted cache files and NDEBUG checks. API selection gives 45/47/46
sites for unknown/33/simulated-35; unset and explicit 1 agree, and 0
leaves sites untouched. This does not replace Android 15 hardware
testing.
- All bench tools compile with -DNDEBUG. Relative/absolute fixture
workloads agree, and thread/ABI/failure checks pass on Thor with those
builds.
- Earlier internal fts/realpath/NSS/dlopen and normalized read-path
equivalence pass. The write suite retains the existing wrapper SD
Unix-socket mismatch against PRoot.
CodeRabbit disposition: fixed relative fixture paths, prevented
benchmark checks disappearing under NDEBUG, corrected the
limited-fallback description, and narrowed memory-file denial so the
thread check still succeeds. The default-off recommendation was
addressed by fixing/testing the underlying guard and unwind issue while
retaining the requested default-on behavior. The readlink count
suggestion is intentionally not adopted: device boundary comparisons
reproduce current PRoot behavior, so changing that inherited policy
belongs in a separate correctness change.
This is not consequence-free instruction patching: constructor time,
private text copies and stack checks have costs. Android 15+ ARM64 and
16 KiB-page hardware, arbitrary asynchronous cancellation/non-local
exits, user-managed stacks, and attributable tracer-counter measurements
remain gaps. Loader calls before the constructor and arbitrary guest ELF
syscall sites remain outside scope.
Thor was restored to its original APK and bundled library, with app data
preserved and test settings, overrides, temporary files and version-2
test caches cleared.
Detailed evidence and revision boundaries: [proroot
analysis](https://github.com/Droid-Deck/DroidDeck/blob/2c3c135b52c3dd949ef5a99a67d8c12eba24406c/docs/development/proroot-analysis.md).
With no frame limit, AIO's Direct3D 7 (DirectDraw) test ran at exactly
144 fps on a 144 Hz panel: every frame waited one refresh (6.94 ms).
Wine asks for a swap interval of 1 on every OpenGL window, and WineD3D's
front-buffer drawing used by DirectDraw and Direct3D 1-7 never lowers
it; it only sets the interval in its present path. Mesa's Zink then
presents in FIFO.
droiddeck-game-env sets vblank_mode=0 for games launched through Proton
before the game's environment edits, so a value set in the launch
options or the editor wins and an editor Remove leaves Mesa's default.
Steam's own runs in compatdata/0 are unchanged. Steam's Frame Limit
still holds a game: gamescope's WSI layer presents every swapchain in
FIFO while its limiter is on.
On the AYANEO Pocket FIT, AIO Direct3D 7 went from 144 to 2,345 and
2,265 fps (1 % low 99 -> 620) with no limit; with the Quick Access
limit at 144 and 72 it held at 143.76 and 71.92.