fix: ext2 RwLock conversion for ARM64 CI boot test hang #188
Conversation
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
…d access The ARM64 CI boot test hung at stage 125 (fs_write_test) because the ext2 filesystem used a global spin::Mutex, forcing all operations — including exec, file reads, and directory listings — to take exclusive access. Under slow QEMU TCG emulation, a writing process holding the mutex blocked all concurrent operations, exhausting the CI timeout. Convert ROOT_EXT2 from Mutex<Option<Ext2Fs>> to RwLock<Option<Ext2Fs>> with separate accessor functions: - root_fs_read(): shared lock for exec, file reads, getdents, stat - root_fs_write(): exclusive lock for create, truncate, write, rename Refactor sys_open to use the read lock for non-modifying opens (no O_CREAT/O_TRUNC) and the write lock only when filesystem mutation is needed. Update all callers across syscall handlers, ARM64 syscall entry, main_aarch64, and test framework. Add x86-arm64 test comparison doc tracking parity progress (184 ARM64 stages vs 252 x86_64 stages) with gap analysis. Verified: ARM64 184/184 boot stages pass, x86_64 252/252 boot stages pass, zero compiler warnings on both architectures. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Fix ARM64 CI hang caused by spin::RwLock writer starvation: root_fs_write() now uses upgradeable_read().upgrade() which sets the UPGRADED bit to block new readers, guaranteeing writer forward progress. Also fix O_APPEND TOCTOU race by acquiring write lock once for the entire read-size-then-write operation in both handlers.rs and io.rs. Unify x86_64/ARM64 boot-stage test infrastructure: extract 163 shared userspace stage definitions into boot_stages.rs (single source of truth), add QemuConfig::for_boot_stages(arch) constructor, and replace duplicate boot_stages()/arm64_boot_stages() runners with unified run_boot_stages(arch). Reduces xtask/src/main.rs from 5372 to 2176 lines. x86_64: 252/252 stages pass. ARM64: 184/184 stages pass. Co-Authored-By: Ryan Breen <ryanbreen@gmail.com> Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…opy drawing Introduce sys_fbmmap (syscall 412) which maps a compact framebuffer buffer into userspace, enabling direct pixel writes with no per-pixel syscall overhead. Only a single flush syscall per frame is needed to sync the user buffer to VRAM. Kernel changes: - Implement fb_mmap allocation and process tracking in graphics.rs - Wire FBMMAP through dispatcher, handler, and ARM64 syscall entry - Flush now copies mmap'd user buffer to shadow buffer before VRAM sync Userspace changes: - Add fb_mmap() wrapper to libbreenix graphics module - Rewrite bounce demo to use mmap'd framebuffer with direct pixel writes, monotonic clock timing, and text rendering for FPS/ball count display Co-Authored-By: Ryan Breen <ryan@ryanbreen.com> Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
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.
Summary
spin::Mutextospin::RwLockto allow concurrent read access (exec, file reads, getdents, stat) while reserving exclusive access for writes (create, truncate, rename, link)sys_opento use read lock for non-modifying opens and write lock only for O_CREAT/O_TRUNCProblem
ARM64 CI boot test hung at stage 125 (
fs_write_test) because the ext2 filesystem used a globalspin::Mutex, forcing all operations to take exclusive access. Under slow QEMU TCG emulation in CI, a writing process holding the mutex blocked all concurrent exec/read/getdents operations, causing 21 stages to fail (163/184 passing).Fix
ROOT_EXT2: Mutex<Option<Ext2Fs>>→RwLock<Option<Ext2Fs>>root_fs()→root_fs_read()(shared) /root_fs_write()(exclusive)Test plan
🤖 Generated with Claude Code