proot: find tracees by pid, resolve deep paths at once, share clone3 … - #772
Merged
maxjivi05 merged 4 commits intoSep 25, 2026
Merged
Conversation
…threads' state Every stop looked its tracee up, and swept for terminated ones, by walking the list of all tracees - one per thread, so hundreds to thousands in a Steam session. Tracees are now hashed by pid, and the sweep runs only after a termination. Canonicalizing a guest path lstat()ed every component, with an ever longer host path. For a clean path under the rootfs binding alone, the directory holding the last component is now opened once and taken as canonical when the kernel's path for it is the one PRoot would build; anything else walks component by component as before. The final component of a call that does not follow it is no longer lstat()ed, its result was never used. clone3(2) was unknown, so the threads glibc 2.34+ creates with it got their own copy of the cwd and emulated heap: a chdir() in one thread was invisible to the others. Its flags are now read from struct clone_args.
…nd GameScope pad hotplug - Touchscreen mode tracks pointers by id and delays the press 50 ms, so a second finger starts a two-finger scroll or right-click tap instead of a click; cancel and detach release every held button. - Uncaptured mice move through one path with scaled relative deltas; fractional wheel values accumulate into whole notches; captured touchpads use position differences. - Relative mouse mode no longer delivers buttons and wheel twice in Wayland sessions. - D-pad keycodes set the D-pad (they set unused button bits before), and controller key repeats no longer move Android focus. - Guide: a tap reaches the guest (Steam opens its menu), a 2 s hold opens the WinNative menu. - GameScope: pad slots outlive devices, since the guest never learns of a pad added mid-game; a disconnect leaves the slot neutral for the next pad. - fakeinput waits at most 2 ms between ring checks (was 16 ms).
…ing for vsync Android batches joystick, captured-mouse and touch MOVE events to the next frame. The display activity now asks for unbuffered dispatch of joystick, trackball and position sources (re-asserted after focus changes, which reset it), and the on-screen controls ask for it per gesture when a control takes the touch. Measured on a RedMagic in GameScope (uinput write to fakeinput ring): pad axes 8.7 -> 1.3 ms median, on-screen stick 8.8 -> 0.9 ms median.
Covers the installer's first container and the Containers screen. The container's audio setting can still be changed afterwards.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
…threads' state
Every stop looked its tracee up, and swept for terminated ones, by walking the list of all tracees - one per thread, so hundreds to thousands in a Steam session. Tracees are now hashed by pid, and the sweep runs only after a termination.
Canonicalizing a guest path lstat()ed every component, with an ever longer host path. For a clean path under the rootfs binding alone, the directory holding the last component is now opened once and taken as canonical when the kernel's path for it is the one PRoot would build; anything else walks component by component as before. The final component of a call that does not follow it is no longer lstat()ed, its result was never used.
clone3(2) was unknown, so the threads glibc 2.34+ creates with it got their own copy of the cwd and emulated heap: a chdir() in one thread was invisible to the others. Its flags are now read from struct clone_args.