Keep-alive inactive by default: it froze the Mali display pipeline

The 16 ms keep-alive submits from a 5th thread, racing Slint's
render/present path and the four decode workers on the same wgpu
queue, hung the phone UI for seconds (51 decodes stalled ~6 s).
System survived (no reboot) but it read as a crash. Debug knob
.gallery_keep re-enables it.

Also: jsamp one-shot permission probes (the cgroup/PSI polling
loop flooded the SELinux audit subsystem at ~100 denials/s and
ANR'd the app); cgroup/PSI probing removed (the device has the
cgroup cpu controller disabled, so quota was ruled out anyway).
Sampling at 5 ms. HANDOFF 8.8 records the full findings.

HANDOFF.md

@@ -632,3 +632,40 @@ Result (Pixel 9 Pro, 120 Hz, 400 px): **zero mid-scroll renders > 4.4 ms**
 
 Perf file: `/sdcard/Download/.gallery_perf.log` (SCR/TICK/APLY/FR).
 Clear it by restarting the app (writer holds the fd).
+
+### 8.8 Jank-hunt round 2 + the keep-alive incident (2026-10-28)
+
+Follow-up on "still seeing jank" (post-GPU-texture). Instrumentation:
+in-app thread sampler (`src/jsamp.rs`, `/proc/self/task`, knob
+`.gallery_jsamp`), per-core cpufreq, cgroup/PSI probes. Findings:
+
+- **Cgroup CPU quota ruled out**: this device's unified hierarchy has the
+  `cpu` controller DISABLED (`cgroup.controllers` = `memory` only; no
+  `cpu.max` file) — no per-app bandwidth cap exists.
+- **`/proc` facts**: utime/stime are USER_HZ ticks (1/100 s), so
+  per-thread CPU deltas are 10 ms-quantized. The app's SELinux context
+  DENIES `/proc/stat`, `/proc/pressure/cpu`, and the unified cgroup files
+  (even the app's own cgroup), but allows `/proc/self/task/*` and
+  cpufreq `scaling_cur_freq`. **A denied read in a tight loop is an ANR:
+  each denial is an audit event; ~100/s froze the whole phone.**
+  jsamp now probes readability once and disables denied features forever.
+- **Real profile** (single verified 120 Hz swipe, 370 frames): steady
+  scroll = 1.5–2.5 ms frames, **zero spikes**. All 20–40 ms frames were
+  the first frame after an input pause (15–100 ms) — resume/warm-up
+  cost; ~820 MHz little-core clock during spikes vs 1.95 GHz between
+  (EAS downclock in the gap, not a cgroup/driver block). Swipe starts
+  after only 2 s pauses: zero spikes; only >10 s app idle pays the big
+  (43 ms) cold frame.
+- **KEEP-ALIVE INCIDENT (phone freeze)**: the `gputex::Keepalive`
+  thread (16 ms buffer-copy submits on Slint's wgpu queue from a 5th
+  thread) made the display pipeline FREEZE on Mali G715 during a 14-min
+  session — phone UI hung for seconds, 51 in-flight decodes stalled ~6 s
+  (workers also `write_texture` on the same queue). System did not
+  reboot (system_server/surfaceflinger 5-day uptime), but the user
+  experience was "crashing my phone". **Root lesson: do not run
+  concurrent submitters on Slint's Mali queue.** Keep-alive is now
+  INACTIVE BY DEFAULT (debug knob `.gallery_keep`), the jsamp sampler is
+  off (knob `.gallery_jsamp`), and both are REMOVE-after-hunt.
+- Status: mid-gesture 20–28 ms warm-up frames can still occur without
+  keep-alive (~1 per 10 s of scrolling); cold-start-after-idle frame is
+  40 ms. User finger-verification of the keep-alive-free build pending.

src/gputex.rs

@@ -22,7 +22,7 @@
 //! `upload_rgba` returns `None` and the caller delivers the RGBA8 bytes
 //! the old way (`Image::from_rgba8`).
 
-use std::sync::OnceLock;
+use std::sync::{Arc, OnceLock, atomic::AtomicBool};
 
 /// The wgpu 30.x crate re-exported by Slint (same version it renders with).
 pub use slint::wgpu_30::wgpu;
@@ -95,3 +95,70 @@ pub fn upload_rgba(rgba: &[u8], width: u32, height: u32) -> Option<wgpu::Texture
     );
     Some(texture)
 }
+
+/// GPU pipeline keep-alive.
+///
+/// INTENDED: keep the wgpu/Vulkan submission path warm so frames after a
+/// 30–80 ms input micro-pause don't pay 23–46 ms of wake-up.
+///
+/// **DISABLED BY DEFAULT (2026-10-28).** On the Pixel 9 Pro (Mali G715)
+/// the 16 ms submission stream from a 5th thread, racing Slint's own
+/// render/present path AND the four decode workers on the SAME queue, made
+/// the display pipeline freeze — the whole phone UI hung for seconds and
+/// 51 in-flight decodes stalled ~6 s. Concurrent submitters on one Mali
+/// queue are not safe in practice. Re-enable only via the debug knob
+/// `/sdcard/Download/.gallery_keep` (REMOVE after the hunt).
+pub struct Keepalive {
+    stop: Arc<AtomicBool>,
+}
+
+static KEEP_BUFS: OnceLock<(wgpu::Buffer, wgpu::Buffer)> = OnceLock::new();
+
+impl Keepalive {
+    /// Spawn the keep-alive thread. INACTIVE unless the debug knob
+    /// `/sdcard/Download/.gallery_keep` exists.
+    pub fn start() -> Self {
+        #[cfg(target_os = "android")]
+        if !std::path::Path::new("/sdcard/Download/.gallery_keep").exists() {
+            log::info!("gputex: keep-alive inactive (default off; .gallery_keep to enable)");
+            return Self { stop: Arc::new(AtomicBool::new(true)) };
+        }
+        let stop = Arc::new(AtomicBool::new(false));
+        let stop_t = stop.clone();
+        std::thread::Builder::new()
+            .name("gpu-keep".into())
+            .spawn(move || {
+                crate::prio::set_self_nice(10); // must never jank the UI thread
+                while !stop_t.load(std::sync::atomic::Ordering::Relaxed) {
+                    if let Some((device, queue)) = GPU.get() {
+                        let (src, dst) = KEEP_BUFS.get_or_init(|| {
+                            let desc = wgpu::BufferDescriptor {
+                                label: Some("gallery-keep"),
+                                size: 16,
+                                usage: wgpu::BufferUsages::COPY_SRC | wgpu::BufferUsages::COPY_DST,
+                                mapped_at_creation: false,
+                            };
+                            let a = device.create_buffer(&desc);
+                            let b = device.create_buffer(&desc);
+                            queue.write_buffer(&a, 0, &[0u8; 16]);
+                            (a, b)
+                        });
+                        let mut enc =
+                            device.create_command_encoder(&wgpu::CommandEncoderDescriptor::default());
+                        enc.copy_buffer_to_buffer(src, 0, dst, 0, 8);
+                        queue.submit(Some(enc.finish()));
+                    }
+                    std::thread::sleep(std::time::Duration::from_millis(16));
+                }
+            })
+            .ok();
+        log::info!("gputex: keep-alive started");
+        Self { stop }
+    }
+
+    /// Signal the thread to stop (does not join; drop is immediate).
+    pub fn stop(&mut self) {
+        self.stop.store(true, std::sync::atomic::Ordering::Relaxed);
+        log::info!("gputex: keep-alive stopped");
+    }
+}

src/jsamp.rs

@@ -0,0 +1,154 @@
+//! In-app thread sampler (jank hunt; REMOVE after).
+//!
+//! The phone is non-rooted, so an external sampler (shell uid) cannot read
+//! the app's /proc. This samples `/proc/self/task` from inside the app at
+//! 1 ms: per target thread, the scheduler state, wchan (what it waits in),
+//! and CPU-time delta — enough to decide, for any FR spike, whether the
+//! UI thread was ON-CPU doing work or WAITING (futex/GPU fence), and what
+//! the decode workers and the Mali driver threads (`mali-compiler` =
+//! shader compilation, `mali-event-hand`) were doing meanwhile.
+//!
+//! Enabled by `/sdcard/Download/.gallery_jsamp`; output goes to
+//! `/sdcard/Download/.gallery_jsamp.log` in the process's monotonic clock
+//! (same `now_ms` as SCR/TICK/FR, so windows line up exactly).
+#![cfg(target_os = "android")]
+
+use std::io::Write as _;
+use std::time::Duration;
+
+const KNOB: &str = "/sdcard/Download/.gallery_jsamp";
+const OUT: &str = "/sdcard/Download/.gallery_jsamp.log";
+
+pub fn start() {
+    if !std::path::Path::new(KNOB).exists() {
+        return;
+    }
+    std::thread::Builder::new()
+        .name("jsamp".into())
+        .spawn(move || {
+            // nice 0: at +10 the sampler starved exactly during the CPU
+            // contention we try to observe (no samples in the spike
+            // window).
+            crate::prio::set_self_nice(0);
+            // ONE-SHOT permission probes. CRITICAL: SELinux-denied reads
+            // must never be retried in a tight loop — each denial is an
+            // audit event, and ~100/s of them flooded the audit subsystem
+            // and ANR'd the whole app.
+            let can_task = std::fs::read_dir("/proc/self/task").is_ok();
+            let can_cpufreq = std::fs::read_to_string(
+                "/sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq",
+            )
+            .is_ok();
+            let _ = (can_task, can_cpufreq);
+            // append (not truncate): SELinux denies create+truncate on the
+            // fresh path; analysis takes the last lines.
+            let mut out = match std::fs::OpenOptions::new()
+                .create(true)
+                .append(true)
+                .open(OUT)
+            {
+                Ok(f) => std::io::BufWriter::new(f),
+                Err(e) => {
+                    log::error!("jsamp: open {OUT}: {e}");
+                    return;
+                }
+            };
+            // tid -> (comm, last_cpu_us); comm cached (read once per tid)
+            let mut last: std::collections::HashMap<u32, (String, u64)> =
+                std::collections::HashMap::new();
+            let mut skipped: std::collections::HashSet<u32> = std::collections::HashSet::new();
+            let mut tick = 0u32;
+            loop {
+                tick += 1;
+                if tick % 250 == 0 {
+                    last.clear(); // tids may change (mali pairs get recreated)
+                    skipped.clear();
+                }
+                let ts = crate::now_ms();
+                let dir = std::fs::read_dir("/proc/self/task");
+                if can_task {
+                    if let Ok(dir) = dir {
+                    for ent in dir.flatten() {
+                        let tid: u32 = match ent.file_name().to_string_lossy().parse() {
+                            Ok(t) => t,
+                            Err(_) => continue,
+                        };
+                        let known_interesting = last.get(&tid).map(|(c, _)| {
+                            c == "android_main"
+                                || c.starts_with("thumb-")
+                                || c.starts_with("mali-")
+                                || c == "slint vsync"
+                                || c == "gpu-keep"
+                        }).unwrap_or(false);
+                        if !known_interesting && !last.contains_key(&tid) && !skipped.contains(&tid) {
+                            let comm = match std::fs::read_to_string(format!("/proc/self/task/{tid}/comm")) {
+                                Ok(c) => c.trim().to_string(),
+                                Err(_) => continue,
+                            };
+                            let interesting = comm == "android_main"
+                                || comm.starts_with("thumb-")
+                                || comm.starts_with("mali-")
+                                || comm == "slint vsync"
+                                || comm == "gpu-keep";
+                            if !interesting {
+                                skipped.insert(tid); // cache the negative too
+                                continue;
+                            }
+                            last.entry(tid).or_insert_with(|| (comm, 0));
+                        }
+                        let (comm, prev) = match last.get(&tid) {
+                            Some(v) => (v.0.clone(), v.1),
+                            None => continue,
+                        };
+                        // stat: pid (comm) state ppid ... utime(14) stime(15)
+                        let stat = match std::fs::read_to_string(format!("/proc/self/task/{tid}/stat")) {
+                            Ok(s) => s,
+                            Err(_) => continue,
+                        };
+                        let close = match stat.rfind(')') {
+                            Some(i) => i,
+                            None => continue,
+                        };
+                        let fields: Vec<&str> = stat[close + 2..].split_whitespace().collect();
+                        let state = fields.first().copied().unwrap_or("?");
+                        // utime/stime are USER_HZ ticks (1/100 s) -> us
+                        let cpu_us = fields
+                            .get(11) // utime (0-based after comm: state=0 ppid=1 ... utime=11)
+                            .and_then(|s| s.parse::<u64>().ok())
+                            .unwrap_or(0)
+                            * 10_000
+                            + fields
+                                .get(12)
+                                .and_then(|s| s.parse::<u64>().ok())
+                                .unwrap_or(0)
+                                * 10_000;
+                        let wchan = std::fs::read_to_string(format!("/proc/self/task/{tid}/wchan"))
+                            .map(|w| w.trim().to_string())
+                            .unwrap_or_default();
+                        let delta = cpu_us.saturating_sub(prev);
+                        let _ =
+                            writeln!(out, "JS {ts:11.2} {tid} {comm} {state} {wchan} dcpu={delta}");
+                        last.insert(tid, (comm, cpu_us));
+                    }
+                }
+                }
+                // Per-core frequency: is the CPU in a low-clock state during
+                // a spike? (Readability probed once at start — never retry a
+                // denied path; the cgroup/PSI probes were removed after they
+                // ANR'd the app via audit flooding.)
+                if can_cpufreq {
+                    let freq0 = std::fs::read_to_string(
+                        "/sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq",
+                    )
+                    .ok()
+                    .map(|s| s.trim().to_string())
+                    .unwrap_or_default();
+                    let _ = writeln!(out, "JSCP {ts:11.2} freq0={freq0}");
+                }
+                let _ = out.flush();
+                std::thread::sleep(Duration::from_millis(5));
+            }
+        })
+        .ok();
+    log::info!("jsamp: sampling started");
+}

src/lib.rs

@@ -76,6 +76,8 @@
 slint::include_modules!();
 
 pub mod db;
+#[cfg(target_os = "android")]
+pub mod jsamp;
 pub mod gputex;
 pub mod images;
 pub mod scan;
@@ -626,7 +628,7 @@ fn install_frame_logger(window: &Main) {
 /// mistaken for jank).
 static NOW_T0: std::sync::OnceLock<std::time::Instant> = std::sync::OnceLock::new();
 #[inline]
-fn now_ms() -> f64 {
+pub(crate) fn now_ms() -> f64 {
     let t0 = NOW_T0.get_or_init(std::time::Instant::now);
     t0.elapsed().as_secs_f64() * 1000.0
 }
@@ -1212,11 +1214,32 @@ fn android_main(app: slint::android::AndroidApp) {
     // (Start/Resume/Pause/Stop/SaveState/ConfigChanged/...) while Slint owns
     // the event loop. This is Slint's one first-class Android lifecycle hook:
     // M5 re-checks storage permission on Resume (user returned from Settings).
+    // Jank hunt (REMOVE after): in-app thread sampler, gated by knob.
+    jsamp::start();
+
+    // GPU pipeline keep-alive (gputex): started on Start/Resume, stopped
+    // on Pause/Stop (battery). INACTIVE BY DEFAULT — on Mali it froze the
+    // display pipeline (see gputex::Keepalive docs). Debug knob:
+    // .gallery_keep.
     slint::android::init_with_event_listener(app, |event| {
         use slint::android::android_activity::{MainEvent, PollEvent};
-        if let PollEvent::Main(MainEvent::Resume { .. } | MainEvent::Start { .. }) = event {
+        // Fn (not FnMut): mutable state in a static.
-            log::info!("gallery: app resumed");
+        static KEEPALIVE: std::sync::Mutex<Option<gputex::Keepalive>> =
-            // TODO(M5): storage::recheck_permission()
+            std::sync::Mutex::new(None);
+        match event {
+            PollEvent::Main(MainEvent::Resume { .. } | MainEvent::Start { .. }) => {
+                log::info!("gallery: app resumed");
+                if KEEPALIVE.lock().unwrap().is_none() {
+                    *KEEPALIVE.lock().unwrap() = Some(gputex::Keepalive::start());
+                }
+                // TODO(M5): storage::recheck_permission()
+            }
+            PollEvent::Main(MainEvent::Pause { .. } | MainEvent::Stop { .. }) => {
+                if let Some(mut k) = KEEPALIVE.lock().unwrap().take() {
+                    k.stop();
+                }
+            }
+            _ => {}
         }
     })
     .unwrap();