The previous fire-and-forget fix passed hwd's own INIT_NOTIFY fd to pcid,
but that fd had CLOEXEC set (by daemon::Daemon::new), so pcid inherited
a closed fd and panicked in PipeWriter::from_raw_fd.
Fix: create a new pipe in hwd before spawning pcid. Pass the write end
as INIT_NOTIFY with CLOEXEC cleared (via pre_exec). Drop the read end
immediately — pcid's daemon.ready() will get EPIPE, which is silently
ignored by the daemon library. This gives pcid a valid fd while still
being fully non-blocking from hwd's perspective.
Root cause: hwd used daemon::Daemon::spawn(pcid) which blocks waiting
for pcid's readiness signal. But pcid only signals readiness after
completing full PCI enumeration. On real Intel hardware with complex
ACPI tables, enumeration can hang (unresponsive device, AML deadlock),
causing pcid to never signal readiness, hwd to never signal its own
readiness, and init to stall the entire initfs phase.
Fix: replace blocking daemon::Daemon::spawn with std::process::Command::spawn
(fire-and-forget). hwd signals its own readiness immediately, allowing
init to continue the initfs phase regardless of pcid's enumeration progress.
pcid runs independently and registers the pci scheme when ready.
Also: promote pcid enumeration completion log from debug to info level.
Suppress EPIPE in SchemeDaemon::ready_with_fd() to eliminate broken pipe
errors from gpiod, i2cd, ucsid at boot. Handle ENODEV gracefully in audiod
when /scheme/audiohw is absent. Both fixes verified: QEMU boots to login
prompt with zero non-fatal errors, patch applies cleanly on clean rebuild.
- IOAPIC: enable full IOAPIC initialization on AMD/Intel bare metal,
dual GSI 0/2 timer mapping for platform compatibility, NMI handler
uses raw COM1 PIO writes to avoid mutex deadlock
- HPET: counter validation, graceful fallback to PIT when HPET missing
- PS/2: fix 0xFE RESEND handling in all MouseState variants, add
controller flush/self-test retry/aux port test from Linux 7.0
- ACPI: defer AML evaluation to avoid blocking initfs driver spawn
- VT chain: remove duplicate rootfs service files (inputd, vesad,
fbcond, getty) that were already handled by initfs phase 1 and the
legacy 30_console script from minimal.toml
- QEMU verified: boots to login prompt, 20 rootfs units (was 26),
single login prompt (was double), only 1 expected error (wifictl)
215 fixes across 33 Rust source files replacing unwrap/expect/panic
with graceful error handling in init, all boot-critical daemons,
and the six graphics driver packages. Fixes inverted scheduler
conditions_met() logic that prevented rootfs from mounting.
Both P2 patches were stale — generated against an older upstream HEAD whose
context lines shifted after redox.patch modified the same files. Regenerated
from scratch against the current upstream commit so they apply cleanly.
P2-boot-runtime-fixes: hwd I2C candidate logging, pcid-spawner initfs detach,
pcid sendfd PCI fd handoff (319 lines)
P2-acpi-i2c-resources: new acpi-resource shared decoder crate (688 lines),
acpid /scheme/acpi/resources/ endpoint, resources.rs re-export shim,
sleep.rs restore (1265 lines)
- Add /scheme/acpi/resources/<device> endpoint to acpid for _CRS evaluation
- Extract acpi-resource shared crate (917 lines) with ResourceDescriptor types
- Eliminate duplicate type definitions in 5 consumers (i2c-hidd, dw-acpi-i2cd,
intel-thc-hidd, i2c-gpio-expanderd, ucsid)
- Add P2-acpi-i2c-resources.patch (48KB) with all source changes
- Update ACPI-I2C-HID-IMPLEMENTATION-PLAN.md to reflect actual codebase state
Consolidate the active desktop path around redbear-full while landing the greeter/session stack and the runtime fixes needed to keep Wayland and KWin bring-up moving forward.
Keep the relibc compatibility work in tracked local patch carriers and align the recipe with the full durable patch stack so clean reapply and rebuild paths stay reproducible.
Keep the relibc overlay consistent so the generated stdlib header preserves C linkage for strtold and the existing toolchain can still satisfy stale C++ callers while it is refreshed.
Regenerate aggregate base patch to include ihdgd device ID additions
for modern Intel integrated GPUs (12th-14th Gen, Core Ultra, Arrow Lake).
This allows pcid-spawner to match and load ihdgd on current Intel laptops.
Create grub-install and grub-mkconfig scripts in local/scripts/ that
match GNU GRUB CLI conventions for users migrating from Linux. Support
standard switches: --target, --efi-directory, --bootloader-id,
--removable, -o/--output, --verbose, --help, --version. Unsupported
Linux options are accepted and ignored for script compatibility.
Also fix ESP FAT type: force FAT32 in both with_whole_disk and
with_whole_disk_ext4 (UEFI spec requires FAT32, fatfs auto-selects
FAT16 for partitions under 32 MiB). Fix --write-bootloader to export
GRUB EFI in GRUB mode. Fix CLI example in GRUB plan. Update AGENTS.md
and GRUB-INTEGRATION-PLAN.md with Linux-compatible CLI docs.
- Add grub package to redbear-full-grub.toml so make all works from
a clean tree (the installer needs grub.efi before it runs)
- Fix stat -f%z (macOS-only) to stat -c%s (Linux) in GRUB recipe
- Normalize bootloader config value to lowercase in install_inner so
bootloader = "GRUB" from config files is accepted
- Add bad-cluster marker (0x0FFFFFF7) check in fat_tool.py
_next_cluster to prevent potential infinite loops on degraded media
- Fix file handle leak in fat_tool.py if _read_bpb raises
- Clean up temp directory in fetch_bootloaders on error
- Update AGENTS.md with GRUB recipe and installer documentation
- Update GRUB plan with clean-build prerequisite note
When bootloader = "grub" is set, the installer now:
- Rejects unknown bootloader values (only "redox" and "grub" accepted)
- Enforces minimum efi_partition_size of 8 MiB for GRUB mode
- Fails with a clear error if grub.efi or grub.cfg are missing from
the GRUB package output instead of silently falling back
Adds --bootloader and --filesystem to installer help text. Fixes
fstools host build by symlinking local/ directory alongside recipes/
so the ext4-blockdev path dependency resolves correctly.
Extends the installer to write a GRUB chainload ESP layout when
bootloader = "grub" is set in config. Changes GeneralConfig, DiskOption,
fetch_bootloaders, with_whole_disk, with_whole_disk_ext4, and both CLI
binaries. When GRUB mode is active, the ESP contains GRUB as primary
(BOOTX64.EFI), grub.cfg, and the Redox bootloader as chainload target
(EFI/REDBEAR/redbear.efi).
5-second countdown with any-key cancel into manual mode selection. Defaults to 1280x720, falls back to EDID best resolution. UEFI only via get_key_timeout polling with BootServices.Stall.
AMD display driver: expanded DCN pipeline setup with plane/controller/stream mapping.
Intel driver: cleaned up module structure.
New interrupt module for MSI-X vector management across GPU drivers.
PCID config endpoint patch and Intel GPU TOML for automatic driver spawning.
Expanded redox_stubs with additional kernel API shims.
Rebuild base/acpid patch as comprehensive unified diff combining: DMAR iterator fix, FADT shutdown via PM1a/PM1b CNT_BLK with S5 sleep types from _S5 AML, ACPI reset register reboot with keyboard controller fallback (port 0x64, 0xFE), and power methods (_PS0/_PS3/_PPC). GenericAddress now supports memory-mapped and I/O port writes. Reboot wired into main.rs event loop with reboot_requested flag. All ivrs/mcfg stub references removed. Validated with git apply --check against upstream base source.
Add MadtLocalApicNmi (type 0x4), MadtLapicAddressOverride (type 0x5), and MadtLocalX2ApicNmi (type 0xA) structs with compile-time size assertions. Add enum variants and iterator cases for all three. Implement set_lvt_nmi() on LocalApic for both xAPIC (LINT0/LINT1 at offsets 0x350/0x360) and x2APIC (MSRs 0x835/0x836) with NMI delivery mode, polarity, and trigger mode from MADT flags. Process NMI entries in x86.rs MADT loop to configure per-CPU LVT NMI registers. Parse and log LAPIC address override (64-bit) for future use.
Add Sdt::validate_checksum() method that sums all bytes in the table and verifies the result is zero per ACPI spec. Call it during ACPI table iteration in init() — warn on invalid checksum but do not skip the table, to avoid breaking boot on firmware with slightly incorrect checksums.
The base patch referenced ivrs and mcfg modules that don't exist in the source tree, causing build failures. Removed the module declarations, imports, and init calls. MCFG is already handled by pcid (the PCI daemon). IVRS (AMD IOMMU) needs real implementation, not a stub. DMAR iterator fix and Dmar::init() call preserved.
The critical fix: local_apic.rs id() now returns (read(0x20) >> 24) for xAPIC mode instead of the raw register value. This was causing wrong APIC IDs on Intel, leading to misrouted IPIs, missed TLB shootdowns, and the page fault during context switch at switch.rs:317.
Also adds:
- Named ICR constants (ICR_INIT_ASSERT, ICR_STARTUP) with bit-layout comments
- Comment documenting x2APIC timeout limitation (cpu_id allocated before timeout check)
- All existing changes preserved (x2APIC MADT type 9, cpuid split, memory alignment, RSDP checksum, ICR pending wait, MADT entry length guard)
Derivative of Redox OS (https://www.redox-os.org) adding:
- AMD GPU driver (amdgpu) via LinuxKPI compat layer
- ext4 filesystem support (ext4d scheme daemon)
- ACPI fixes for AMD bare metal (x2APIC, DMAR, IVRS, MCFG)
- Custom branding (hostname, os-release, boot identity)
Build system is full upstream Redox with RBOS overlay in local/.
Patches for kernel, base, and relibc are symlinked from local/patches/
and protected from make clean/distclean. Custom recipes live in
local/recipes/ with symlinks into the recipes/ search path.
Build: make all CONFIG_NAME=redbear-full
Sync: ./local/scripts/sync-upstream.sh
vasilito