e1000d was the last NIC driver using legacy IRQ (irq.irq_handle()).
Migrated to pci_allocate_interrupt_vector which tries MSI-X first,
then MSI, then falls back to legacy INTx — matching rtl8168d, rtl8139d,
ihdad, ihdgd, and nvmed.
63-line patch at local/patches/base/P6-e1000d-msi-migration.patch,
symlinked and wired into recipes/core/base/recipe.toml.
interrupt.rs:
- InterruptRemapTable now owns optional DmaBuffer for self-allocated tables
- new_allocated(entry_count) constructor allocates physically contiguous
DMA memory via DmaBuffer::allocate, returns Result
- new(base_addr, size) still works for externally-provided tables
- private addr() helper replaces direct 'base' field access
- len_encoding() returns AMD-Vi log2-encoded IRT length for DTE entries
- physical_address() returns table base physical address
- Remove unused 'warn' and 'error' imports from log crate
amd_vi.rs:
- Use InterruptRemapTable::new_allocated instead of ::new for IRT init
- Cast len_encoding() from u64 to u8 for DeviceTableEntry::set_int_table_len
Verified: iommu crate compiles clean (0 errors, 0 warnings).
dma.rs: IommuDmaAllocator (145 lines)
- New struct wires existing IOMMU daemon (1003 lines) to existing DmaBuffer (261)
- allocate(): phys-contiguous alloc via scheme:memory, then MAP through IOMMU domain
- unmap(): sends UNMAP to IOMMU domain, releases IOVA
- Inlined IOMMU protocol constants — no new crate dependency
- encode_iommu_request/decode_iommu_response for scheme write/read cycle
Documentation updates:
- IMPLEMENTATION-MASTER-PLAN.md: K2 DMA/IOMMU section expanded from 3-line gap
list to full audit with component inventory, gap analysis, implementation plan
(D2.1-D2.5), Linux reference table. Added K2b thread/fork audit.
- CPU-DMA-IRQ-MSI-SCHEDULER-FIX-PLAN.md: Phase 1 (MSI) marked complete with
per-task status. Phase 2 (DMA) re-scoped from 'create' to 'wire' based on
audit. Phase 3 (scheduler) marked mostly done.
- IRQ-AND-LOWLEVEL-CONTROLLERS-ENHANCEMENT-PLAN.md: kernel MSI support noted
as materially strong with P8-msi.patch reference.
Audit findings:
- IOMMU daemon is solid: 1003-line lib.rs with full scheme protocol,
427-line amd_vi.rs, host-runnable tests. Needs wiring, not rewriting.
- DmaBuffer exists but is IOMMU-unaware — IommuDmaAllocator bridges this.
- relibc rlct_clone is correct for threads (shares addr space implicitly).
'3 IPC hops' claim is microkernel-architectural, not a real perf issue.
- No stale docs to archive at this time.
- Add GuiPrivate to Qt6 find_package in top-level CMakeLists.txt
- Add missing QElapsedTimer include in toolbarlayout.cpp
- Add network stub infrastructure in recipe (incomplete, Qt6Network
cross-compilation still needed for full build)
- udev-shim: replace .expect() with graceful errors (no more panic on Broken pipe)
- P4-initfs: remove duplicate sessiond (conflicted with config)
- accessibility/ime/keymapd: break instead of exit(1) on EBADF
- P6 driver patches rebased
- Docs: archive old reports, add implementation master plan
Base: fix P6-driver-new-modules.patch (ed format -> unified diff) for new
driver modules (ncq, itr, phy). P6-driver-main-fixes.patch now applies with
offset on current upstream source.
Relibc: remove stale P5-named-semaphores (upstream has stubs), add
P10-stack-size-8mb and P11-getrlimit-getrusage (per-process rlimit table,
sysconf integration, getdtablesize fix, null-pointer safety).
Kernel: consolidate 29 individual patches into single redbear-consolidated.patch.
Userutils: P5-redbear-branding replaces P4-login-rate-limit.
Recipe.toml changes now committed so they survive source resets.
MouseTx::handle() treated 0xFE (PS/2 RESEND) as an unknown response,
causing mouse init to fail on hardware where the mouse requests a
resend during the reset/command exchange. Now resends the current
command byte when the mouse returns 0xFE, matching the PS/2 protocol.
Bootloader hardcoded RedoxFS partition offset at 2 MiB, which fails when
efi_partition_size > 1 (redbear-full, redbear-grub use 16 MiB, placing
RedoxFS at LBA 34816 = 17 MiB). Added GPT partition table parser that
scans for Linux filesystem GUID (0FC63DAF) to find the actual offset.
Also removed invalid 'respawn = true' from 31_debug_console.service in
redbear-full.toml — init's service format does not support this field.
Verified: all three ISOs boot in QEMU UEFI and reach login prompt.
Add redbear-usb-storage-check in-guest binary that validates USB mass
storage read and write I/O: discovers /scheme/disk/ devices, writes a
test pattern to sector 2048, reads it back, verifies match, restores
original content. Updates test-usb-storage-qemu.sh with write-proof
verification step.
Includes all accumulated Red Bear OS work: kernel patches, relibc
patches, driver infrastructure, DRM/GPU, KDE recipes, firmware,
validation tooling, build system hardening, and documentation.
5-phase hardening to prevent silent file-layer collisions (the D-Bus
regression class):
Phase 1: lint-config-paths.sh + make lint-config in depends.mk
Phase 2: CollisionTracker in installer (content-hash comparison)
Phase 3: installs manifests in recipe.toml + validate-file-ownership.sh
Phase 4: validate-init-services.sh + make validate in disk.mk
Phase 5: documentation (AGENTS.md, BUILD-SYSTEM-HARDENING-PLAN.md)
Both redbear-mini and redbear-full build and validate clean.
66 declared install paths in base, zero conflicts.
The init system parser (serde deny_unknown_fields) rejects [service]
sections in .target files. The P4-initfs-dbus-services patch was
creating 05_boot_essential.target and 12_boot_late.target with a
no-op [service] section (cmd=/usr/bin/true). Removed those sections
so targets only contain [unit].
This eliminates "unknown field service, expected unit" init warnings
that could prevent proper service dependency resolution.
Oracle review found 3 gaps. All fixed:
1. Recipe #TODO updated from 'Always-permit stub' to 'Real UID-based policy'
2. init.d/20_polkit.service created
3. redbear-full.toml already has 14_redbear-polkit via [[files]] — verified
P5: redbear-polkit now enforces real authorization:
- is_authorized(uid, action_id) checks UID-based policy
- uid=0 (root) always authorized
- Other users checked against /etc/polkit-1/policy.toml
- Default: deny for unknown actions (fail-closed)
- Backend name changed from 'redbear-permit-all' to 'redbear-uid-policy'
- Default policy grants power/network/storage to root+user(1000)
P3-3: fbcond scrollback is now fully functional:
- text.rs: 1000-line ring buffer captures all console output
- scheme.rs: new Scrollback handle type, path/N/scrollback open
- main.rs: Scrollback match arm in event loop
- Users can read scrollback via:
cat /scheme/fbcon/2/scrollback
19/19 patches. base + base-initfs build.
Fixed common dependency path (../../common → ../common).
Added workspace member entry for drivers/thermald.
thermald now builds as part of base recipe.
19/19 patches. base + base-initfs build.
P3-3: fbcond scrollback — captures last 1000 lines of text output
in ring buffer, exposes via read_scrollback(). Patch created but
needs line number adjustment for clean application.
P3-5: thermal daemon source created at drivers/thermald/. Reads
ACPI thermal zone temperature, logs warnings >65°C, errors >80°C.
Needs Cargo.toml workspace integration and recipe.toml BINS entry.
Part of COMPREHENSIVE-FIX-PLAN-FINAL P3 implementation.
Changed <policy user="root"> to <policy user="0"> in D-Bus
system.conf.in to avoid getpwnam lookup failures on Redox.
Patch written but does not apply yet — tar source extraction
puts files under dbus-1.16.2/ which requires strip=2 or different
path prefix. Left as documented work-in-progress.
D-Bus daemon runs despite cosmetic user lookup warnings.
D-Bus daemon, sessiond, and seatd all running on redbear-full.
User lookup warnings are pre-existing Redox D-Bus config issue
(does not affect functionality).
Bootloader: alloc_zeroed_page_aligned no longer panics on OVMF
AllocatePages failure. Returns null on error.
Cookbook: pkgar falls back to repo/ dir when target pkgar missing.
Resolves KDE build chain failure.
redbear-full: builds 4.0GB image+ISO, boots without crash.
getty converts numeric args to /scheme/fbcon/{N} paths.
Non-numeric args (like ttyS0) are treated as literal paths
which don't exist in Redox. Changed to:
- 29_activate_console: inputd -A 2
- 30_console: getty 2 → /scheme/fbcon/2 (VT2 login)
- 31_debug_console: getty 3 → /scheme/fbcon/3 (VT3 debug)
Serial console (ttyS0) has no standard scheme path in Redox.
Login prompt is on the framebuffer console (requires graphical QEMU
or VNC to view).
Boot process now includes:
- 25_serial_getty.service: getty on serial console (visible in QEMU -nographic)
- 29_activate_console.service: inputd -A 2 (activate VT2)
- 30_console.service: getty on VT2 (framebuffer console)
Fixed hunk counts (7→8, 8→9) for correct patch application.
Services use 'oneshot_async' type for fire-and-forget startup.
ZSH is the default shell for all user accounts (base.toml, mini, full, greeter).
Consolidated from all audits, QEMU testing, and implementation sessions.
P0: KDE build chain broken — libwayland pkgar staging race in cookbook.
Fix requires investigation of src/cook/package.rs pkgar path lookup.
P1: Graphical boot testing — dbus-daemon, sessiond, compositor, greeter.
P2: 7 remaining gaps — ion shell, shadow support, polkit, scrollback.
Includes cookbook tool investigation areas and total effort estimate
(~12 days per developer).
libwayland stage.pkgar missing when kf6-kitemviews builds during
make live CONFIG_NAME=redbear-full. This is a cookbook pkgar
staging race condition, not a code error.
redbear-mini (text-only) boots fully in QEMU with colored init
output, 25+ services, and login prompt on framebuffer console.
At initfs time, /var/log/ doesn't exist (rootfs not yet mounted).
Changed persistent_log from File to Option<File> with .ok() instead
of .unwrap_or_else(|| panic!()). If the file can't be opened, logging
continues without persistence — no crash.
QEMU verification: system boots through initfs→rootfs→switch_root→userland.
Colored init output visible. 25+ services start successfully.
P2-2: Login rate limiting (userutils/login.rs):
- Tracks consecutive failures, resets on success
- 3+ failures: exponential delay up to 30 seconds
- Applies to both password and blank-password login paths
P2-3: Network stack in initfs (base-initfs + service files):
- Added e1000d, rtl8168d to base-initfs BINS
- 60_smolnetd.service: network stack in initfs
- 61_dhcpd.service: DHCP client in initfs
- Network available before switch_root
Part of COMPREHENSIVE-FIX-AND-IMPROVEMENT-PLAN Phases P2.
Phase B1+B2 from BOOT-PROCESS-AUDIT:
- 45_usbscsid.service: USB mass storage driver in initfs (requires xhcid)
- 30_redox-drm.service: DRM/KMS display driver in initfs (requires hwd+pcid-spawner)
Both condition-architecture-gated to x86/x86_64.
Phase A1 from BOOT-PROCESS-AUDIT. The ACPI shutdown path now:
- Validates PM1a port is non-zero before writing
- Waits 3 seconds for power-off, then retries with PM1b+SLEEP_EN
- Falls back to keyboard controller reset (0x64=0xFE) on failure
- Handles SLP_TYPb correctly
- Removes fragile Pio::new()+write() without validation
logd now writes all log output to /var/log/system.log (5MB auto-rotation)
in addition to existing scheme listeners. Falls back to /tmp/logd-fallback.log
if /var/log is unavailable. Logs survive reboots for post-mortem analysis.
Part of Phase A2 (Boot Reliability) from BOOT-PROCESS-AUDIT-2026-05-03.
The 556MB monolithic redox.patch was impossible to manage, unreviewable,
blocked GitHub pushes, and could only grow. This commit:
- Moves all 64 absorbed patches from absorbed/ to active use in base/
- Removes the absorbed/ directory (consolidation history is now PATCH-HISTORY.md)
- Removes the redox.patch symlink from recipes/core/base/
- Fixes all recipe symlinks to point to active patches (not absorbed/)
- Patches are now individually wired, reviewable, and independently rebasable
The redox.patch mega-file is no longer needed — individual patches
are applied directly from the recipe.toml patches list.
vasilito