- Add intel_vtd.rs module with DMAR parsing, DRHD discovery,
register definitions, and basic unit initialization
- Update iommu daemon discovery to detect both AMD-Vi (IVRS)
and Intel VT-d (DMAR) units
- Update IommuScheme to track both amd_units and intel_units
- Intel VT-d init: version check, capability read, disable
translation, report supported features (QI, IR, EIM)
Full DMA remapping enablement (root table, context entries,
page tables, command buffer) remains as TODO for follow-up.
- drivers/acpid/src/cstate.rs: Evaluate _CST per processor, parse
Package-of-Packages into CStateInfo structs
- AcpiContext: add cstate_state field with refresh, add processor_names()
to scan _PR namespace
- acpid scheme: expose /scheme/acpi/cstates/<proc> read handles
- thermald: read /scheme/sys/cstate, set /scheme/sys/cstate_policy
to restrict to C1 when temp exceeds WARNING_TEMP
Works with kernel P25 cpuidle deep C-states.
Define 4 hardware target classes (AMD/Intel desktop/laptop),
per-target checklist, negative-result capture format, and
quick/full test procedures. Ready for bare-metal evidence.
Detect CPU vendor by probing MSRs (Intel IA32_THERM_STATUS vs AMD
TCTL MSR C0010293). Support both Intel Tjmax-based and AMD direct
temperature reading. Log detected vendor per CPU at startup.
Extend logd output thread to write logs to per-service files in
/var/log/<service>.log, with automatic size-based rotation (10 MB
threshold, 5 backup files). All logs also go to /var/log/system.log.
Backwards compatible with existing sink file descriptors.
Add RateLimitedLog to common::logger for per-message rate limiting with
"last message repeated N times" warnings. Add structured_log! macro for
key=value formatted logs. Update thermald to rate-limit the max-temp
summary line (30s interval) to reduce log volume.
Track the target CPU ID in InterruptVector, and log the interrupt type
(MSI-X/MSI/Legacy) and CPU affinity at allocation time in
pci_allocate_interrupt_vector. Add log_affinity() helper for drivers
to call after setup.
Add fan.rs module to acpid that discovers FAN* devices under \_TZ,
evaluates _FST for current speed level and RPM, and exposes them via
/scheme/acpi/fan/<name>/status. Update thermald to read and log fan
status alongside temperature sensors.
Replace the old hardcoded /scheme/acpi/thermal_zone/{n} paths with
proper discovery of /scheme/acpi/thermal/ zones and /scheme/coretemp/
CPU temperatures. Logs per-zone and per-CPU temps with max tracking.
Switch network drivers from legacy INTx to pci_allocate_interrupt_vector
which auto-prefers MSI-X > MSI > Legacy. rtl8139d and rtl8168d already
used this helper; e1000d and ixgbed were the remaining legacy-only NIC
drivers.
Implement CPU power-saving idle loop using x86 MONITOR/MWAIT:
- Add monitor(), mwait(), enable_and_mwait() to interrupt module
- Detect MWAIT availability via CPUID at boot
- Use MONITOR+MWAIT instead of STI+HLT when supported
- Expose /scheme/sys/cstate_policy for userspace control
- Add RdWr Kind variant to sys scheme for read+write files
Graceful init patches for fbcond, graphics scheme, smolnetd, vesad,
PCI interrupt allocation, BAR probing, common init, inputd fallback,
and dhcpd hard dependency ordering.
Implement full thermal zone backend in acpid:
- thermal.rs: Discover \_TZ_.TZ* zones, evaluate \_TMP, \_CRT, \_PSV,
\_AC0, \_TC1, \_TC2, \_TSP, \_TZP methods
- scheme.rs: Expose /scheme/acpi/thermal/ with per-zone temperature files
- acpi.rs: Add thermal_state and thermal_zone_names() to AcpiContext
Wired as P44 patch in base recipe.toml.
New local recipe coretempd reads IA32_THERM_STATUS MSR via the
new sys:msr scheme and exposes per-CPU temperatures via scheme:coretemp.
- Reads IA32_THERM_STATUS (0x19C) and IA32_TEMPERATURE_TARGET (0x1A2)
- Calculates Celsius from digital readout relative to TjMax
- Exposes /scheme/coretemp/ directory with per-CPU temperature files
- Added to redbear-mini.toml (inherited by redbear-full)
- Added 15_coretempd.service init file
Add deterministic S5 (soft-off) state derivation and structured error
handling to acpid. Derive S5 parameters once at startup (or retry at
shutdown if AML was not ready) instead of re-parsing the _S5 package
on every shutdown attempt. Replace unit-return set_global_s_state()
with ShutdownResult enum for proper error propagation and fallback
handling.
Changes:
- S5State struct caches SLP_TYPa/b, PM1a/b ports, derivation timestamp
- ShutdownError enum: MissingFadt, Pm1aZero, AmlNotReady, S5NotFound,
S5NotPackage, SlpTypNotInteger, S5WriteFailed
- ShutdownResult enum: Ok, FallbackReset, Err(ShutdownError)
- derive_s5_state() method with early init attempt and lazy fallback
- set_global_s_state() returns ShutdownResult instead of ()
- Early S5 derivation in AcpiContext::init() logs AML readiness status
- main.rs logs shutdown result for debugging
This is W2.1/W2.2 from ACPI-IMPROVEMENT-PLAN.md.
P21: Replace 67 panic-grade calls across 9 boot daemon files with
graceful error handling. Affected: ps2d, inputd, fbcond, fbbootlogd.
P22: Add x2APIC MADT fallback for processors with LocalApic entries
instead of LocalX2Apic entries. QEMU KVM boots now correctly detect
all vCPUs via zero-extended APIC ID fallback.
P23: Change 50_rootfs.service from requires_weak to requires on
40_drivers.target, ensuring redoxfs waits for disk drivers before
attempting filesystem mount. This fixes the boot race where rootfs
mount failed before drivers were ready, causing init to have no
userland services after switchroot.
ramfs@.service required randd as requires_weak, which doesn't enforce
readiness ordering. When ramfs called std::random before randd registered
/scheme/rand, it panicked with 'failed to generate random data'.
Changed requires_weak to requires so init waits for randd to register
its scheme before starting ramfs.
Also patched Rust stdlib sys/random/redox.rs to fall back to xorshift64
seeded from ASLR rather than panicking when /scheme/rand is unavailable.
This is a belt-and-suspenders fix: even with proper ordering, the stdlib
should not panic on missing entropy during early boot.
P18-5 had a duplicate acpi.rs RSDP probing hunk that reversed P5's
BIOS probing code, causing P19-acpid hunk #8 to fail. Removed the
duplicate, keeping only aml_physmem.rs hunks.
P19-init used log::warn!() but init in initfs has no log crate
dependency. Replaced with init_warn(&format!(...)) from the existing
color module.
All 58/58 patches validate. redbear-mini builds successfully.
All 58 base patches now pass repo validate-patches base.
- P18-9-msi-allocation-resilience: regenerated against P0-P18-8 baseline
with correct upstream content (deamon typo preserved for virtio-netd)
- P19-init-startup-hardening: regenerated against P0-P18-9 baseline
- P19-acpid-startup-hardening: regenerated against P0-P18-9 + P19-init
baseline with all 39 hunks in -U0 -w format (zero context lines)
x2APIC ICR mode fix cannot be a forward patch because P16-1-sipi-timing
references the ICR line as context. Modifying P1 (which introduces the line)
would require updating P16-1's offsets. Will address by consolidating into
an existing early patch in a future revision.
- Add RawFb struct: direct framebuffer rendering via physmap
- Add RawTextScreen: simple text renderer using orbclient font
- Fallback in FbbootlogScheme::new() when V2GraphicsHandle fails
- Reads FRAMEBUFFER_ADDR/WIDTH/HEIGHT/STRIDE from bootloader env
- Scroll via ptr::copy on pixel rows, clear bottom line
- No DRM, no shadow buffer, no GPU required — like MS-DOS text mode
- Add common dependency to fbbootlogd Cargo.toml
- Derived line offsets from real pre-P20 vs post-P20 diff
- x86.rs: 3 hunks at @@ -446/-456/-468 converting hardcoded <<32 to
local_apic.x2-gated format with xAPIC <<56 fallback
- local_apic.rs: 1 hunk promoting debug! to info! for bootlog visibility
- Fix LocalX2Apic handler: use local_apic.x2 to select correct ICR
format (<<32 for x2APIC, <<56 for xAPIC) instead of hardcoded <<32
- Promote x2APIC/xAPIC detection from debug! to info! for bootlog
- Document build system durability in AGENTS.md: cardinal rule,
two-layer architecture, correct workflow, anti-patterns
P5: Enable PRIME export for GBM dumb buffers so virgl can import scanout\nbuffers via fd-to-handle.
P6: Add redox_probe_device_hw() that opens /scheme/drm/card0, resolves the\nDRI driver via PCI ID lookup (loader_get_driver_for_fd), and initializes the\nhardware path through dri2_load_driver_dri3 with __DRIimageLoaderExtension.\nIncludes fprintf(stderr) debug logging tagged [REDOX-EGL] for tracing the\nHW init sequence. Falls back to software rendering if HW probe fails.
Wire P1/P2 patches into recipe.toml. Source reflects the full ioctl bridge\nrewrite: VirtGPU NR mappings, EXECBUFFER/GET_CAPS special handlers,\nPCI info bridge for drmGetDevice2, and removal of the old per-ioctl C\nhandler functions in favor of unified redox_drm_simple_ioctl() dispatch.
P1: Replaces the old per-ioctl C handlers with a unified dispatch through\nredox_drm_simple_ioctl(), adds VirtGPU NR mappings (MAP, EXECBUFFER,\nGETPARAM, RESOURCE_CREATE, GET_CAPS, etc.), and special handlers for\nEXECBUFFER (inline command buffer) and GET_CAPS (variable response).
P2: Adds __redox__ blocks to drmParsePciBusInfo, drmParsePciDeviceInfo,\nand drmGetDevice2 using the DRM scheme GET_PCI_INFO ioctl to populate\nPCI device identification without /sys access.
Align 2D command opcodes, response opcodes, and 3D command opcodes with\nthe Linux UAPI definitions in include/uapi/linux/virtio_gpu.h. Enable\nVIRTIO_GPU_F_VIRGL feature negotiation during device initialization to\nactivate the virgl 3D rendering path.
VirtIO capability structures can start at non-page-aligned offsets within\nBARs. The kernel physmap requires page-aligned addresses. Align the physical\naddress down, adjust the size, and apply the page offset to the returned\npointer. Track the original map pointer/size separately for correct unmapping.
The previous BAR size calculation used !(inverted & mask) & mask which\nproduces incorrect results when the BAR has a non-zero base address.\nUse lowest-set-bit extraction (masked & (!masked + 1)) to correctly\ncompute the BAR size from the writable bit pattern read back from hardware.
Red Bear OS is a full fork. All sources must be available from git clone
with zero network access. Removed gitignore rules that excluded fetched
source trees under recipes/*/source/, local/recipes/kde/*/source/,
local/recipes/qt/*/source/, and vendor source trees.
Build artifacts (target/, build/, source.tar, *.o, *.so) remain excluded.
127291 files added — kernel, relibc, base, bootloader, pkgar, all KDE/Qt
frameworks, mesa, wayland, DRM drivers, and every other recipe source.
vasilito