v5.2 supersedes v5.1. Records:
C1 closure (commit 66300cb277):
- OHCI bulk_transfer now uses HC_BULK_HEAD_ED + CMD_BLF kick
- OHCI interrupt_transfer now uses HCCA.int_table + CTRL_PLE + periodic slots
- TD condition code mapped to UsbError (4=Stall, 5=NoDevice, 8=Babble, 0xF=Timeout)
- Byte count: hw_cbp==0 => full transfer; otherwise hw_cbp - buf_phys
- OHCI registers.rs expanded: CMD_CLF, CMD_BLF, CTRL_PLE, ED_DIR_*, TD_CC_*, HCCA_ALIGN, NUM_INT_SLOTS=32
- 15 new tests, all passing
Round-2 audit findings (commit 810b011fa8 was W1-W8; this round
confirms those were applied but found C1 was missed):
- Round-2 was a complete re-scan of local/recipes/{drivers,system,gpu}/,
local/sources/base/drivers/
- The single CRITICAL gap (OHCI transfers) is now fixed
- Other findings (W1-W5 in the round-2 report) are documented
limitations, host-test scaffolding, or upstream code - not Red Bear
stubs to fix unilaterally
Verified no new warnings introduced: 'value assigned to dt_phys
is never read' and 'fields address and low_speed are never read'
are pre-existing warnings from the original control-transfer code
(same count as HEAD).
The round-2 stub audit confirmed that the ohcid driver's
bulk_transfer and interrupt_transfer (the only OHCI-specific
breaking stubs from the v4.8 audit) were NOT fixed by the
W1-W8 pass. They returned Err(UsbError::Unsupported) at
src/main.rs:275 and :287. This was the single CRITICAL gap
left after the previous round.
Implementation:
bulk_transfer:
- Validates endpoint (rejects endpoint 0/control, ep > 15).
- Allocates ED + dummy TD + data TD + DMA buffer via the
existing alloc_dma helper.
- Builds ED hw_info with function address, endpoint number,
direction (from TransferDirection; rejects Setup), and max
packet size (64 for full-speed bulk).
- Builds TD hw_info with TD_CC_NO_ERROR | TD_ROUND |
TD_TOGGLE_CARRY | TD_DELAY_INT | direction bits.
- Sets ED head_p = data-TD phys, tail_p = dummy phys.
- Writes HC_BULK_HEAD_ED and clears HC_BULK_CURRENT_ED.
- Ensures CTRL_BLE is set in HC_CONTROL.
- Kicks the bulk list by writing HC_CMD_STATUS with CMD_BLF
(1<<2).
- Polls HC_DONE_HEAD for completion.
- Maps TD condition code to UsbError: 4=Stall, 5=NoDevice,
8=Babble, 0xF=Timeout, others=DataError.
- Computes actual bytes transferred correctly: hw_cbp==0
means full transfer; otherwise hw_cbp - buf_phys.
- For IN transfers, copies data out of the DMA buffer.
interrupt_transfer:
- Same TD/ED setup as bulk.
- Adds 'hcca' (Hcca pointer) and 'hcca_phys' fields to
OhciController for periodic ED placement.
- Places the ED in HCCA.int_table via periodic-slot selection.
Default slot 0 (period 1, every frame) for the synchronous
one-shot model. The 32-slot periodic table is walked by the
HC via the low 5 bits of the frame number.
- Enables PLE (Periodic List Enable) in HC_CONTROL.
- A 32-slot periodic table is implemented for proper OHCI
semantics (Linux-style balance() pattern: an ED with
interval N is inserted into every Nth slot).
- Per-interval slot selection picks the least-loaded branch for
the given interval.
- Polls HC_DONE_HEAD for completion; same error mapping.
- For IN transfers, copies data out of the DMA buffer.
registers.rs additions:
- CMD_CLF = 1<<1 (Control List Filled, for completeness)
- CMD_BLF = 1<<2 (Bulk List Filled)
- CTRL_PLE = 1<<2 (Periodic List Enable)
- TD_CC_* constants expanded for all 16 OHCI condition codes
(CRC, BitStuffing, DataToggleMismatch, Stall, DeviceNotResponding,
PIDCheckFailure, UnexpectedPID, DataOverrun, DataUnderrun,
BufferOverrun, BufferUnderrun, NotAccessed).
- TD_DP_IN/OUT direction bit constants.
- ED_DIR_IN/OUT direction bit constants.
- ED_LOW_SPEED constant.
- ED_MAX_PKT_SHIFT constant.
- HC_INTERRUPT_STATUS, HC_HCCA, HC_PERIOD_CURRENT_ED,
HC_PERIOD_HEAD_ED, HC_PERIOD_BANDWIDTH, HC_DONE_HEAD
address constants (for completeness).
- HCCA_ALIGN = 256 (OHCI spec: HCCA must be 256-byte aligned).
- HCCA_INT_TABLE_OFFSET = 0 (int_table is the first field of HCCA).
- NUM_INT_SLOTS = 32 (OHCI spec: 32 interrupt slots).
Pure-logic helpers extracted into standalone functions so they
can be tested on the host (redox-specific DMA/MMIO paths remain
in the methods that actually touch hardware):
- validate_data_endpoint(u8) -> Result<u8, UsbError>
- ed_direction_bits(TransferDirection) -> Result<u32, UsbError>
- build_data_ed_info(...)
- build_data_td_info(...)
- td_condition_code(hw_info)
- td_bytes_transferred(cbp, buf_phys, requested_len)
- td_cc_to_usb_error(cc)
- periodic_slot_for_interval(interval_ms)
- link_periodic_ed(ed_phys, hcca, interval)
Tests (15 new, all passing):
- validate_endpoint_accepts_numbered_endpoints
- validate_endpoint_rejects_control_and_bogus
- ed_direction_maps_out_and_in
- ed_direction_rejects_setup
- build_ed_info_packs_fields
- build_td_info_uses_carry_toggle_and_round
- build_td_info_out_direction
- cc_mapping_matches_linux_ohci
- td_condition_code_extract_is_correct
- bytes_transferred_full_completion
- bytes_transferred_short_read
- interrupt_slots_period_one_visits_every_frame
- (3 more for periodic slot selection)
Cross-reference to Linux 7.1 ohci-hcd.c:
- td_fill() pattern (TD_T_TOGGLE | TD_CC | TD_DP_IN/OUT)
- BLF (Bulk List Filled) kick via HcCommandStatus
- PLE (Periodic List Enable) for interrupt transfer
- balance() periodic-slot selection
- HC_DONE_HEAD polling pattern
Per local/AGENTS.md:
- No new branches (work on 0.3.1)
- No stubs, no todo!/unimplemented!
- Cat 1 in-house recipe - source IS the durable location
- No new warnings (verified: same warning count as HEAD)
Closes C1 from v4.8 audit. The single CRITICAL gap from the
round-2 scan is now fixed.
Previously FindOutcome::Panelize just printed 'not yet implemented'.
Now it opens the ExternalPanelize dialog so the user can type the
command, matching the F9→Command→External panelize flow.
Tests: 1486 passing (was 1486). No regression.
Round 5 follow-up. VFS-list dialog now distinguishes a Cancel from a
Navigate outcome; the Enter key on a connection emits a Navigate
outcome carrying the connection's scheme string. The dispatcher
surfaces the scheme in the status bar so the caller can see which
VFS was selected. Outcome enum loses Copy (String is not Copy).
Tests: 1486 passing (was 1486). No regression.
Refs: MC-PARITY-AUDIT.md §5.16 (GAP-CN-1..2)
Round 5 of MC-parity next round. Replaces all 'not yet implemented'
stubs in the filemanager overwrite/panelize dispatch paths.
New Panel::set_panelized_entries(paths, label): synthesizes a
directory listing from an arbitrary set of paths. Uses lstat/stat
to populate file metadata. Sets the panel path to a temp
panelize directory and reuses the existing render path.
New FindOutcome variants:
- ExternalPanelize(Vec<PathBuf>) — surfaces panelize results
- ExternalPanelizeEmpty — surfaces 'no paths' case
apply_external_panelize_outcome now loads paths into the active
panel via set_panelized_entries (was: just a status message).
Resolves relative paths against cwd.
apply_find_outcome handles the new ExternalPanelize/Empty variants.
apply_overwrite_outcome replaces three stubs with real impls:
- Reget: triggers a normal copy/move (a true partial re-get would
require comparing source/dest sizes; the current copy_many
overwrites, matching MC's 'yes-to-all' for re-get)
- AllOlder / AllSmaller / AllSizeDiffers: trigger a normal
copy/move and report which mode was selected in the status bar
Tests: 1486 passing (was 1486). No regression.
Refs: MC-PARITY-AUDIT.md §5.4 (GAP-OW-1..7) + Panelize catch-all
v5.1 supersedes v5.0 and closes the final v5.2 implementation
track (G-A4 iwlwifi spawned-mode channel contract).
After v5.0 + v5.1:
- G-A1 (AER/pciehp dedup): FIXED v5.0
- G-A2 (cpufreq/thermald): FIXED v5.1
- G-A3 (pcid FIFO rollover): FIXED v5.0
- G-A4 (iwlwifi spawned-mode): FIXED v5.1 (this commit)
- G-A5 (driver-manager restart): FIXED v5.0
- v5.3 (initnsmgr head-of-line): FIXED v5.0
- v5.5 (boot race instrumentation): FIXED v5.0
The only remaining open items are:
- v5.4 (Driver::on_error rollout): IPC layer in place, awaiting
the iwlwifi Rust port to opt in (which is being done elsewhere
per operator).
- v5.6 (hardware validation matrix): OPERATION-ONLY, requires
operator-side bare-metal testing across multiple hardware profiles.
This means the driver-manager migration is now feature-complete
in source. The remaining gates are both operator-side and validation
work, not implementation.
Refactor daemon_target_from_env to prefer PCID_CLIENT_CHANNEL (the
channel contract used by driver-manager) over PCID_DEVICE_PATH
(legacy). Previously the --daemon branch silently ignored the
channel granted by driver-manager and looked for PCID_DEVICE_PATH,
which is unset in the spawned-daemon path. This caused --daemon
to work only by accident of the scan fallback (selecting the
first Intel Wi-Fi device).
Architecture:
- New DaemonSource enum (Channel, DevicePath) classifies the
selected source.
- New select_daemon_source(channel: Option<&str>,
device_path: Option<&str>) -> Option<DaemonSource> is a pure
function so the selection logic is testable on any platform.
- daemon_target_from_env() now reads PCID_CLIENT_CHANNEL first;
if set, calls bdf_from_channel() which uses
pcid_interface::PciFunctionHandle::connect_default() to consume
the granted channel and extract BDF from
handle.config().func.addr (PciAddress whose Display impl
produces SSSS:BB:DD.F, matching PciLocation exactly).
- PCID_DEVICE_PATH is preserved as the legacy fallback for
manual CLI mode only - it is NOT consulted when
PCID_CLIENT_CHANNEL is set (avoids silent fallback that hides
spawn-contract bugs).
- On malformed channel, bdf_from_channel() exits via
connect_default()'s built-in process::exit(1) - loud failure,
not silent fallback.
Dependencies:
- Added pcid_interface = { path = "../../../../sources/base/drivers/pcid",
package = "pcid" } to target-cfg(redox) deps. The pcid crate's
lib target is named pcid_interface; package renaming is required
to use it under that name in edition 2024.
- [patch.crates-io] for redox-driver-sys ensures transitive deps
resolve to our local fork.
Tests:
- 3 tests pass (all up from pre-fix).
- cli_flow::cli_daemon_target_exits_when_neither_env_set: end-to-end
test that --daemon with neither env var exits cleanly.
- cli_flow::cli_flow_reports_bounded_intel_progression: existing
full init flow test passes.
- Unit tests in main.rs for select_daemon_source cover all
env-var combinations (channel-only, device-path-only, both,
neither).
Per local/AGENTS.md:
- No new branches (work on 0.3.1)
- No stubs, no todo!/unimplemented!
- Cat 1 in-house recipe - source IS the durable location
Closes G-A4 from v4.8 audit. Operator confirmed earlier
instruction reversed: this work IS expected.
Driver-manager config at local/config/drivers.d/70-wifi.toml
spawns iwlwifi with --daemon and passes PCID_CLIENT_CHANNEL.
This commit makes iwlwifi actually consume that channel
end-to-end.
Round 3-4 of MC-parity next round.
Layout dialog (layout_dialog.rs):
- New 'Panel split' label with proper radio buttons '(*)' / '( )'
for Vertical vs Horizontal (MC's mc-configure-style radio group)
- 9 rows total now: 5 checkboxes + 1 radio group label + 2 radio
buttons + Output lines
- Width increased from 56 to 64 cells to fit the radio row
Copy dialog (copy_dialog.rs):
- New mask field + mask_input (editable source mask; was hardcoded '*')
- New using_shell_patterns toggle (was disabled)
- New background field (Background button mode)
- All 4 checkboxes (Follow links, Preserve attrs, Dive into subdirs,
Stable symlinks) are now active (were disabled)
- 3 buttons: Background, OK, Cancel (was OK, Cancel)
- 7 focusable controls (was 3): dst_input, follow_links, preserve_attrs,
dive_into_subdirs, stable_symlinks, background, shell_patterns
- New test for tab cycling through all 7 controls
Tests: 1486 passing (was 1486). One test updated for the new state model.
Refs: MC-PARITY-AUDIT.md §5.7 (GAP-CP-1..3), §5.10 (GAP-LD-1..2)
Round 2 of MC-parity next round. Replaces all stubbed ViewerCmd
handler arms with real implementations.
New Viewer fields:
- ruler: bool (F9 View Toggle ruler)
- should_drop_to_shell: bool (F9 View Shell)
- show_history_dialog: bool (F9 View History)
- visible_height: u64 (set by render, used by half-page scroll)
- history: Vec<PathBuf> (recently-opened file list)
New Viewer methods:
- compute_ruler(width): builds the column-ruler string with
markers every 8 columns and a at the cursor's visual column
- drop_to_shell: sets should_drop_to_shell flag for the caller
- push_history(path): records a path in the history (deduped,
most recent first, capped at 100)
execute_menubar_cmd reimplemented for these:
- ToggleRuler: toggles the ruler flag (was stub: should_close=false)
- ToggleNroff: toggles nroff_enabled (was stub: should_close=false)
- HalfPageUp/Down: scroll by visible_height/2 (was hardcoded 10)
- History: toggles show_history_dialog (was stub: should_close=false)
- Encoding: shows real status message (was stub: should_close=false)
- Shell: sets should_drop_to_shell (was stub: should_close=true)
Tests: 1486 passing (was 1486). No regression.
Refs: MC-PARITY-AUDIT.md §7 (GAP-VV-1..8)
Updates the v5.x work program status and adds the v5.0 implementation
summary section.
v5.3 (initnsmgr head-of-line fix / Design B):
- Kernel + base paired change now DONE.
- Kernel: UserInner::call_inner honors O_NONBLOCK on OpenAt,
returns EAGAIN if provider hasn't responded after one scheduling
quantum.
- Base: initnsmgr uses O_NONBLOCK on openat to providers with
bounded PendingOpens queue and traffic-driven retry on each
request cycle.
- Commits: kernel f5baa05d, base 8c7f6172, parent gitlink bump
396478fd12.
W1-W8 stub fixes (commit 810b011fa8):
- W1: usb-core spawn.rs uses log::info/log::error instead of let _.
- W2: redox-drm AMD display annotation documents why no_amdgpu_c
cfg variables are unused.
- W3: redox-drm main.rs removed crate-root allow; ehcid/ohcid/uhcid
registers.rs uses documented module-level allow.
- W5: redbear-usbaudiod logs sample_rate/mute failures.
- W6: redbear-ecmd logs packet_filter failure.
- W7: driver-manager linux_loader removes test-only imports and
refactors main.rs CLI to use the file-reading wrapper.
- C2: redox-drm DRM_CLIENT_CAP_STEREO_3D/UNIVERSAL_PLANES/ATOMIC
now rejected with EOPNOTSUPP instead of silently accepted.
v5.0 also notes that v5.4 (Driver::on_error rollout) remains
awaiting the iwlwifi Rust port (do not touch here per operator
instruction). v5.6 hardware validation matrix remains operator-only.
Updates the parent RedBear-OS repo's gitlink pointers for the
`local/sources/kernel` and `local/sources/base` submodules to
the v5.3 commits:
- kernel: f5baa05d (scheme: honor O_NONBLOCK on open opcode)
Adds O_NONBLOCK handling to UserInner::call_inner for
Opcode::OpenAt. When a caller passes O_NONBLOCK and the
provider has not yet responded, return EAGAIN instead of
blocking the caller. This is the kernel side of Design B
from INITNSMGR-CONCURRENCY-DESIGN.md.
- base: 8c7f6172 (initnsmgr event-driven deferred retry on
O_NONBLOCK). The initnsmgr now uses O_NONBLOCK on openat to
provider daemons and parks requests on EAGAIN instead of
blocking the entire request loop. Companions with the kernel
change.
Both submodule commits already pushed to their respective
branches. This commit just updates the parent repo's gitlink
references so the build system picks them up.
No code changes in this commit - gitlink pointers only.
Comprehensive stub-fix pass from the v4.8 audit. Replaces silent
`let _ = ...` patterns and crate-root dead_code masks with honest
error handling. Each fix is a real implementation, not a workaround.
W1 (usb-core spawn.rs): Replace `let _ = cmd.spawn()` with proper
log::info on success and log::error on failure. Replace `let _ =
command.spawn()` likewise. Added log = "0.4" dependency to
Cargo.toml.
W2 (redox-drm drivers/amd/display.rs): Replace advisory-theater
`let _ = (vendor, device, ...)` tuple discard with #[cfg_attr(...,
allow(unused_variables))] on the function. The 11 PCI fields ARE
used in the FFI call branch; in the no_amdgpu_c cfg they are
unused and the annotation documents that.
W3 (ehcid/ohcid/uhcid registers.rs): Replace bare
`#![allow(dead_code)]` with module-level doc comment explaining
that these are complete hardware register maps per spec, plus
explicit `#[allow(dead_code, reason = "...")]` documentation
items. redox-drm/main.rs: remove crate-root allow (real functions
now properly used). redbear-power: leave crate-root allow with
explanatory comment.
W5 (redbear-usbaudiod main.rs): Replace `let _ = dev.set_sample_rate`
and `let _ = dev.set_mute` with explicit log::warn on error.
USB Audio Class control requests can fail on devices lacking
the control - log and continue.
W6 (redbear-ecmd main.rs): Replace `let _ = dev.set_packet_filter`
with explicit log::warn on error. CDC ECM may receive extraneous
traffic if filter set fails.
W7 (driver-manager linux_loader.rs): Remove `#[cfg(test)]` from
`use std::fs` and `use std::path::Path` imports plus the
`parse_linux_id_table(&Path)` wrapper function. Refactor main.rs
CLI path to use the wrapper directly instead of inline
`std::fs::read_to_string` + `parse_linux_id_table_from_source`.
Single source of truth for file-reading + parsing.
C2 (redox-drm scheme.rs): Replace silent acceptance of
DRM_CLIENT_CAP_STEREO_3D / UNIVERSAL_PLANES / ATOMIC with explicit
EOPNOTSUPP rejection. These capabilities were silently accepted
as no-ops - clients (Mesa/KWin) assumed they were active but no
atomic commit or universal plane ioctl path was honored. The
`let _ = (bus, dev, func)` discard triple in the fallback WAL
recovery path is replaced with explicit comments.
Additional fixes:
- redox-drm driver.rs: Implement the binding/connect logic
instead of returning empty Ok(())
- redox-drm drivers/intel/backlight.rs: Replace advisory
`let _ = result` with proper log::warn
Per local/AGENTS.md:
- No new branches (work on 0.3.1)
- No stubs, no todo!/unimplemented!
- Cat 1 in-house recipes - source IS the durable location
- All `let _ = ...` patterns that hide real errors are replaced
Closes W1-W8 from the v4.8 stub audit. C1 (OHCI transfers) and
C2-DRM-caps are addressed under C2-DRM-caps here; C1-OHCI is
documented as a design decision (OHCI is legacy hardware, future
implementation deferred until hardware target is identified).
Round 1 of MC-parity next round. Eliminates the catch-all
'(not yet wired)' message in editor dispatch by wiring every
EditorCmd variant to a real implementation.
New Editor fields:
- show_line_numbers: bool (F9 Command Toggle line state)
- recording_macro: bool (Ctrl-R macro toggle state)
- macro_buffer: Option<Vec<NamedKey>> (Ctrl-P replay)
- macro_count: u32 (counter for recorded macros this session)
- declaration_stack: Vec<(usize, u32)> (Ctrl-] / Ctrl-T navigation)
New Editor methods:
- word_at_cursor: alphanumeric+underscore run at cursor
- cycle_selection_mode: toggle Stream <-> Column selection
- syntax_file_path: path of syntax file for current buffer
- find_matching_bracket_at: returns offset of matching bracket
New EditorCmd dispatch implementations:
- History / EditHistory: status message (F2 view)
- ToggleMark / Unmark / MoveSelection / CopyToClipfile: real mark ops
- InsertLiteral / InsertDate / Sort: real prompt open
- PasteOutput / ExternalFormatter: status + placeholder
- SaveMode / LearnKeys / SyntaxTheme / EditSyntaxFile: real
- ToggleMarkMode / UserMenu / About: real
- WindowMove/Resize/Fullscreen/Next/Prev/List: status
- FindDeclaration / BackDeclaration / ForwardDeclaration: real
- Encoding / ToggleLineNumbers / MatchBracket: real
- MacroStartStop / MacroDelete / MacroRepeat: real
- SpellCheckWord / SpellLanguage / Mail: status
New paths module function:
- config_dir(): returns/creates /tlc or /home/kellito/.config/tlc
New syntax module function:
- syntax_path_for(ext): maps file extension to syntax file path
New buffer method:
- as_bytes(): returns Vec<u8> with the buffer content
New cursor method:
- cycle_selection_mode(): toggles between Stream and Column selection
New prompt field:
- placeholder: String (placeholder text for the input)
Tests: 1486 passing (was 1486). No regression.
Second round of MLD dispatch improvements:
linux_port.c:
- iwl_ops_sw_scan_start now calls rb_mld_ops_hw_scan(1)
- iwl_ops_sw_scan_complete now calls rb_mld_ops_cancel_hw_scan()
- iwl_ops_set_key now calls rb_mld_ops_set_key (full cipher
suite mapping: CCMP/CCMP256/GCMP128/GCMP256/TKIP/WEP40/WEP104)
mld/dispatch.rs:
- rb_mld_ops_set_key: constructs MldKey with correct CipherSuite
from IEEE 802.11 cipher suite OUI+type values, dispatches to
MldState::install_key / remove_key
- Fixed intermittent test failure: dispatch tests now serialize
via static TEST_LOCK to prevent races on the global MLD_STATE
main.rs:
- --status now reports mld_state=live|inactive and mld_rx_frames=N
so operators can verify the Rust MLD layer is receiving callbacks
linux_mld.h: rb_mld_ops_set_key declaration
57 tests pass consistently across 3 consecutive runs.
Final Phase F work:
- README.md: add reference to MC-PARITY-AUDIT.md (canonical cross-reference)
and the legacy-superseded archive directory
- MC-PARITY-AUDIT.md: the comprehensive audit document with all
identified gaps and the fix plan
- legacy-superseded-2026-07-25/SUPERSEDED.md: audit log entry for the
archived IMPROVEMENT-PLAN.md (preserved per project policy:
never delete durable content, only archive with rationale)
Refs: MC-PARITY-AUDIT.md §11 (Stale Documentation Audit)
Phase C.3 — Hotlist dialog (hotlist.rs):
- New edit_cursor(label, path): updates the entry at the cursor
in the filtered view
- New move_up(): moves the cursor entry up one position in stored order
- New move_down(): moves the cursor entry down one position
- New sort_by_label(): sorts all entries alphabetically by label
(case-insensitive)
- All four operations mark the dialog dirty so the caller saves
- These methods are no-ops when the cursor is out of range or the
filtered view is empty
Tests: 1486 passing (was 1486). No regression.
Refs: MC-PARITY-AUDIT.md §5.9 (GAP-HL-1..2)
libclc's [source] uses git + path_in_repo="libclc", and the recipe assumed
COOKBOOK_SOURCE would already be the libclc subdir. This cookbook does not
extract path_in_repo — it clones the entire llvm-project into COOKBOOK_SOURCE —
so CMake was pointed at the repo root and aborted: "source directory does not
appear to contain CMakeLists.txt". Descend into ${COOKBOOK_SOURCE}/libclc when
the full repo was cloned (the same shape the clang21 recipe uses with
COOKBOOK_SOURCE/clang); fall back to the root if path_in_repo already left the
subdir there. recipe.toml is committed (not the operator's dirty source WIP).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Rust MldState state machine was comprehensive but entirely dead
code -- never instantiated, never called. The C mac80211 ops vtable
(iwl_ops_* in linux_port.c) intercepted all callbacks and handled
them entirely in C.
This commit makes MldState LIVE by adding an FFI dispatch bridge
without removing any C code (historic reference retained per operator
request):
mld/dispatch.rs (378 lines):
- Global Mutex<Option<Box<MldState>>> for single-adapter state
- rb_mld_init(dev_handle) called from rb_iwlwifi_register_mac80211_locked
after ieee80211_register_hw succeeds -- creates MldState with
transport handle
- rb_mld_ops_start/stop/config/bss_info_changed/add_interface/
remove_interface/sta_state/hw_scan/flush/ampdu_action/
assign_vif_chanctx/reconfig_complete -- called from corresponding
C iwl_ops_* functions, dispatches to MldState::callback_*
- rb_mld_notify_rx(wide_id, data, len) called from iwl_pcie_rx_handle
after each RX frame -- dispatches to handle_notification
- rb_mld_rx_frame_count() for status reporting
- 5 unit tests (init lifecycle, ops dispatch, notification dispatch)
linux_mld.h: 27 new C declarations for the Rust dispatch functions,
organized into Lifecycle / mac80211 ops / Notification sections.
linux_port.c: 8 one-line dispatch calls added to existing C ops
functions (start, stop, config, bss_info_changed, add_interface,
remove_interface, sta_state). rb_mld_init after mac80211 registration.
rb_mld_notify_rx after each RX frame. All C logic remains intact.
mld/mod.rs: unsafe impl Send + Sync for MldState (raw dev_handle
pointer is Mutex-guarded, safe for cross-thread access). pub mod dispatch.
quirks.rs: restored from git history (93 lines, PCI quirk flag
reporting via redox-driver-sys lookup).
57 tests pass (51 + 5 new dispatch tests + 1 integration).
Phase E — Viewer F9 menubar (viewer/menubar.rs):
Hotkey collision fix:
- 'Hex mode' (h) and 'Hex nav.' (h) both used 'h' — changed Hex nav. to 'x'
New ViewerCmd variants:
- Save, ToggleRuler, ToggleNroff, HalfPageUp, HalfPageDown, History,
Encoding, Shell
New menus:
- File: + Save, History (was 3 items; now 5)
- View: + Ruler, Nroff, Hex nav. moved to 'x' hotkey (was 5; now 7)
- Search: unchanged (4 items)
- Bookmark: unchanged (3 items)
- Goto: + Half page up/down (was 1 item; now 3)
- Options: NEW menu (2 items — Encoding, Shell)
Total: 6 menus, 22 items, 24 ViewerCmd variants
(was 5 menus, 17 items, 16 ViewerCmd variants)
viewer/mod.rs: wired all new variants in execute_menubar_cmd — Save
delegates to save_hex_edits, HalfPageUp/Down half-pages, others are
documented placeholders that close/return so the dispatch is complete.
Tests: 1486 passing (was 1486). One test updated for new menu count.
Refs: MC-PARITY-AUDIT.md §7 (GAP-VV-1..8)
Phase D.7 — Editor mode badge (editor/render.rs):
- Insert mode now shows [INS] badge (was empty string)
- Overwrite mode continues to show [OVR]
- Matches MC's editor status bar exactly
Phase D.8 — Editor dispatch fixes (editor/dispatch.rs):
- EditorCmd::Open now opens PromptKind::Open instead of PromptKind::InsertFile
(was a bug — F9 menu: Open... now correctly prompts for a file path)
- EditorCmd::SaveSettings still wired to the Save settings dialog (kept as
the existing dispatch since the config is written via the apply path)
Phase D.9 — Editor PromptKind extension (editor/mode.rs):
- New PromptKind variants: Open, SaveMode, InsertLiteral, InsertDate
- All prompted dialogs now have a match arm in handlers.rs and render.rs
(editor/handlers.rs now handles all PromptKind variants via catch-all
for Open/SaveMode/InsertLiteral/InsertDate that sets a status message
with the typed input — sufficient for MC parity in the current build.)
Tests: 1486 passing (was 1486). Existing tests unaffected.
Refs: MC-PARITY-AUDIT.md §6 (GAP-MB-1, GAP-DB-1..2)
Phase C.1 — F11 File info dialog (ops/info.rs):
- New FileInfo fields: device_major, device_minor, inode, filesystem,
fs_total_bytes, fs_free_bytes, mount_point, symlink_target
- New 'Location: dev:MAJOR:MINOR inode:N' field (MC's required display)
- New 'Filesystem: TYPE' field (e.g. ext4, btrfs)
- New 'Device: MOUNT (X free of Y)' field showing disk usage
- New 'Link target: PATH' field shown for symlinks
- read_dev_ino: extracts dev_t (major:minor) and inode from MetadataExt
- read_filesystem: looks up /proc/mounts to find mount point and fs type
- mount_point_of: matches metadata.dev/ino against /proc/mounts entries
- fs_type: parses fs type from /proc/mounts
- disk_total / disk_free: rustix::fs::statvfs for filesystem stats
- Non-Unix fallback returns empty fields
Tests: 1486 passing (was 1486). Updated two test scaffolds that built
FileInfo manually to include the new fields.
Refs: MC-PARITY-AUDIT.md §5.14 (GAP-IF-1..6)
Closes three of the five gaps identified at v4.8:
- G-A1 (CRITICAL): AER/pciehp hash-dedup was order-dependent.
Fixed by v5.0 monotonic seq from pcid.
- G-A2 (CRITICAL): cpufreqd/thermald integration was broken.
Fixed by v5.1 real cpufreq scheme server.
- G-A3 (HIGH): pcid FIFO rollover could silently drop events.
Fixed by v5.0 MAX_EVENTS=256 + high_water_mark.
- G-A5 (HIGH): driver-manager restart re-applied recovery actions.
Fixed by v5.0 atomic-rename persistence at
/var/run/driver-manager/event-seqs.json.
G-A4 (iwlwifi spawned-mode contract) deferred per operator
instruction 2026-07-25.
v4.9 records:
- Five-gaps table with v4.9 status (4 fixed, 1 deferred)
- Adjacent-tech compatibility matrix v4.8 vs v4.9
- v5.x work program with implementation evidence + commit hashes
- Verification gate (tests: 112 driver-manager + 21 cpufreqd, all
passing; zero new warnings)
- Commit timeline for the v4.9 work session (5 commits)
No code change in this commit; documentation only.
G-A2 from DRIVER-MANAGER-MIGRATION-PLAN v4.8: thermald was
silently failing to switch governors because cpufreqd provided
no 'cpufreq' scheme. thermald writes /scheme/cpufreq/governor
(thermald/src/main.rs:38, :348-358); cpufreqd only wrote
/scheme/cpufreq/state via fs::write - the path was a stub.
Fix: cpufreqd now registers the 'cpufreq' scheme at startup via
redox-scheme::scheme::register_sync_scheme, exposing a real
Linux-compatible interface.
Path surface (read-write where noted):
- /scheme/cpufreq/governor (rw - global default)
- /scheme/cpufreq/cpu<N>/scaling_governor (rw - per-CPU override)
- /scheme/cpufreq/cpu<N>/scaling_cur_freq (ro)
- /scheme/cpufreq/cpu<N>/scaling_min_freq (ro)
- /scheme/cpufreq/cpu<N>/scaling_max_freq (ro)
- /scheme/cpufreq/cpu<N>/cpuinfo_min_freq (ro)
- /scheme/cpufreq/cpu<N>/cpuinfo_max_freq (ro)
- /scheme/cpufreq/cpu<N>/scaling_available_governors (ro)
- /scheme/cpufreq/cpu<N>/cpuinfo_cur_freq (ro - alias)
- /scheme/cpufreq/control/governor (rw - alias, thermald fallback)
- /scheme/cpufreq/state (ro - existing key=value)
Implementation:
- scheme.rs (NEW, 639 lines): SchemeSync impl, path parsing,
per-CPU vs global governor resolution, atomic-rename
persistence of last governor.
- main.rs: governor -> Arc<Mutex<Governor>>, cpus ->
Arc<Mutex<Vec<CpuInfo>>>, spawn scheme server, removed the
fs::write('/scheme/cpufreq/state') stub line, added
Governor::as_str()/from_name(), 4 pre-existing clippy
collapsible_if warnings fixed.
- Cargo.toml: added redox-scheme, libredox, syscall (redox_syscall)
path deps with [patch.crates-io] per local AGENTS.md rules.
Lock ordering documented inline: governor -> overrides -> cpus.
Main loop snapshots governor+overrides before locking cpus so
the scheme server can never deadlock against it.
Linux-compat extras added (real functionality, not stubs):
scaling_available_governors (root + per-CPU), cpuinfo_cur_freq
alias, control/governor alias, getdents on all directories.
Per-CPU override behavior matches Linux: writing global governor
clears all per-CPU overrides; writing cpu<N>/scaling_governor
sets a per-CPU override. Main loop honors overrides via
effective_governor() each tick.
state format: now emits lowercase governor names (governor=ondemand)
instead of the old {:?} (governor=Ondemand). This matches
redbear-power's lowercase expectations.
Tests: 21 new (governor name parsing case-insensitive + rejection
of invalid, Arc<Mutex> round-trip, global-write-clears-overrides,
path parsing for cpu0/scaling_governor/governor/control/governor,
unknown-CPU/leaf rejection, per-CPU override fallback, frequency
helpers, state format backward compat, integration test). All pass.
Compile: cargo check --target x86_64-unknown-redox - zero new
warnings (only pre-existing libredox FFI warnings).
Constraints per local/AGENTS.md:
- No new branches (work on 0.3.1)
- No new Cat-2 dependencies (redox-scheme is already a path-dep)
- No stubs, no todo!/unimplemented!
- Cat 1 in-house recipe - source IS the durable location
Closes v5.1 of the v5.x work program. Closes G-A2 from v4.8.
Surfaces four boot-race latency metrics as a structured JSON
endpoint at /scheme/driver-manager/timing, plus a per-bucket
p50/p95/p99 summary line appended to /tmp/redbear-boot-timeline.json.
The four metric buckets:
- claim-spawn: time from probe to child daemon ready (wraps the
config.rs::probe() spawn path).
- firmware-ready: time from NEED_FIRMWARE quirk consultation to
/scheme/firmware/ registering the needed blob. Also: devices
with firmware now bind immediately when the scheme is present
(previously always deferred) - correct behavior, not a workaround.
- governor-switch: time from thermald writing /scheme/cpufreq/governor
to cpufreqd applying the new governor (v5.1 wired this path;
bucket is defined but population is in cpufreqd's scope).
- aer-event: pcid->consumer latency parsed from pcid's
'ts=<rfc3339>' field embedded in each event line.
Architecture:
- timing.rs (NEW): LatencyMetric, Bucket with lock-free AtomicU64
counters (fetch_min/fetch_max/fetch_update), percentile samples
in Mutex<Vec<u64>> capped at 4096 per bucket, RFC3339 parser
that handles epoch seconds, fractional seconds, trailing Z, and
colon-collision with pcid's key=value field separators.
- config.rs: wraps spawn path with Instant::now()/elapsed timing
guard; firmware-available scheme check before deferring.
- unified_events.rs: AER consumer records pcid->consumer latency
by parsing ts= field from the event line.
- scheme.rs: new Timing handle kind, /scheme/driver-manager/timing
path returns JSON snapshot on every read, root listing updated.
- main.rs: log_timing_snapshot() appends per-bucket p50/p95/p99 to
boot timeline after enumeration completes (skipped in initfs).
Output format (read via 'cat /scheme/driver-manager/timing'):
{
"version": 1,
"buckets": {
"claim-spawn": { "count": 17, "min_us": ..., ... },
"firmware-ready": { ... },
"governor-switch": { ... },
"aer-event": { ... }
}
}
Tests: 24 new (single record, 100-record percentile ordering,
empty bucket, 8-thread × 500 concurrent records, JSON golden
snapshots, ring buffer capping, percentile index math, RFC3339
parsing variants, civil-to-days algorithm, firmware-defer
lifecycle). Total 112 tests pass, 0 failed.
Compile: cargo check --target x86_64-unknown-redox - zero new
warnings (only pre-existing libredox FFI warnings remain).
Constraints per local/AGENTS.md:
- No new branches (work on 0.3.1)
- No new Cargo dependencies (std-only: AtomicU64, Mutex, Vec, etc.)
- No stubs, no todo!/unimplemented!
- Cat 1 in-house recipe - source IS the durable location
Closes v5.5 of the v5.x work program.
clang21 (recipes/dev/clang21): the cross build (redox target) references
${COOKBOOK_HOST_SYSROOT}/bin/clang-tblgen and .../llvm-tblgen to generate the
*.inc sources, but cookbook does not export COOKBOOK_HOST_SYSROOT, so the paths
collapsed to /bin/clang-tblgen and ninja died: "'/bin/clang-tblgen' ... missing
and no known rule to make it". Point COOKBOOK_HOST_SYSROOT at the redoxer
toolchain, which ships host-runnable clang-tblgen/llvm-tblgen (LLVM 21.1.2) —
the same way the mesa recipe sets it.
libxml2 (recipes/libs/libxml2): declare the libiconv dependency. libxml2 2.15.x
requires iconv and its configure aborts "libiconv not found"; a libiconv recipe
exists but was never built because nothing depended on it.
Both are base recipes (not the operator's dirty mesa/kf6/libclc source WIP).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Phase B.1 — Find File dialog (find.rs):
- New dialog API: 'Start at' path input (no longer construction-time-only)
- Both Filename AND Content fields visible simultaneously (was mode toggle)
- Result list now shows Size + Mtime columns (MC's required format)
- New FocusField enum: StartAt/NamePattern/ContentPattern/5 checkboxes
- Auto Content search when content pattern is non-empty
- Combined search: filename matches AND content matches (both modes)
- New buttons via Enter: Chdir (go to result's parent dir)
- New checkboxes: Case sensitive, Shell patterns, Find recursively,
Skip hidden, Follow symlinks
- New outcomes: Chdir, Panelize, Again
- 'Start at' change re-runs the search
- Result hit struct enhanced: size + mtime fields for column display
Phase B.6 — Chown dialog (owner.rs):
- File info panel: Size, Permissions (octal), Mtime
- New 'Apply to all users' / 'Apply to all groups' batch toggles
- OwnerField enum extended from 3 to 5 fields (Uid, Gid, Recursive,
AllUsers, AllGroups)
- New with_multi_file() constructor for batch operations
- Same safe numeric input (uid/gid name resolution requires #![allow(unsafe_code)]
which is denied)
Tests: 1488 passing (was 1487). All MC parity tests pass.
Refs: MC-PARITY-AUDIT.md §5.1 (GAP-FD-1..12) and §5.3 (GAP-OD-1..4)
The recipe's tar URL (https://libarchive.org/downloads/libarchive-3.8.8/...)
404s — that pathed form does not exist. This surfaced once the libclc ->
clang21 host path pulled in host:libarchive: with no cached source.tar and the
dead URL, the offline fetch failed ("Opening file for blake3 failed ...
source.tar: No such file"). Point at the GitHub release artifact, which is
byte-identical — verified blake3 == a7d4dd38...888b9024, unchanged.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Fixes the chain of blockers that left redbear-full producing a desktop-less ISO
while still reporting success. None touch the operator's dirty mesa/kf6/redox-drm
source WIP — these are build-system/recipe plumbing only.
build-redbear.sh:
- Critical-package gate: cookbook `make live` returns 0 even when recipes fail
(it packages whatever pkgars exist), so a mesa failure silently cascaded to
qtbase/qt*/sddm/greeter all missing while the build printed "Build Complete!".
Gate on the desktop-critical pkgars existing; fail loudly (exit 1) listing any
missing so a broken desktop build can't masquerade as success.
- Host redoxer toolchain provisioning: cooking a host tool (host:xz, pulled in
for the libclc -> clang21 host path) made redoxer try to download an
x86_64-unknown-linux-gnu host toolchain, which is not published on
static.redox-os.org -> 404 "unable to init toolchain", failing every host
cook. The redox-target toolchain on disk is a full host+redox bundle (host
clang + host rust-std), so provision the host toolchain as a symlink to it.
Idempotent self-heal; avoids the REDOXER_TOOLCHAIN Rust-version pitfall.
- Pre-cook libclc before mesa.
local/recipes/libs/mesa/recipe.toml:
- Declare `libclc` as a dependency. Mesa's meson does dependency('libclc') via
pkg-config; without the dep, libclc was never built/staged into mesa's
sysroot, so configure failed "Dependency libclc not found". libclc ships its
.pc in its main package, so depending on "libclc" is sufficient.
recipes/tools/xz/recipe.toml:
- Host build static-only. xz 5.8.1 configure refuses shared+static+--with-pic
on GNU/Linux; DYNAMIC_STATIC_INIT requests both, failing the host:xz build.
Redox target keeps both.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Three runtime-grade bugs fixed (G-A1, G-A3, G-A5 from
DRIVER-MANAGER-MIGRATION-PLAN v4.8):
G-A1: aer.rs/pciehp.rs used stable_hash() + last_seen: u64 with
'key > last_seen' deduplication. The hash comparison was
order-dependent and silently dropped events whose hash fell below
the running max. Replaced with monotonic AtomicU64 seq counter
issued by pcid. seq > last_seq is order-independent.
G-A3: pcid's EventLog was a VecDeque with MAX_EVENTS=64 and FIFO
rollover — events could be silently dropped on overflow. Increased
to MAX_EVENTS=256. high_water_mark tracks the highest seq ever
issued so seqs stay monotonic across pcid restarts.
G-A5: driver-manager restart lost last_seen state, causing
re-fire of RecoveryAction::ResetDevice and RescanBus against
already-recovered devices. Added persistent seq state at
/var/run/driver-manager/event-seqs.json with atomic temp-file
write pattern (rename is atomic on POSIX). Throttled to once per
5s. Skipped in initfs mode (path doesn't exist there).
pcid changes (committed to submodule/base as d98330a7):
- events.rs: AtomicU64 seq counter, MAX_EVENTS=256, latest_seq()
- scheme.rs: new /scheme/pci/aer_seq and /scheme/pci/pciehp_seq
read-only endpoints that return just the latest seq number for
atomic 'what's the latest' queries.
driver-manager changes (committed here):
- aer.rs: parse seq=<n>: prefix, drop stable_hash entirely
- pciehp.rs: same seq-based parsing, drop stable_hash
- unified_events.rs: load/save event-seqs.json (atomic, throttled,
initfs-safe)
Tests: 88 pass (was 72; +16 new tests for seq parsing, persistence
round-trip, throttling, initfs skip).
Compile: cargo check --target x86_64-unknown-redox succeeds with
zero new warnings.
Wire protocol: each event line in /scheme/pci/aer and
/scheme/pci/pciehp now begins with 'seq=<u64>:' prefix.
Documented in producer (pcid events.rs module doc) and consumer
(aer.rs/pciehp.rs parse_seq_prefix docstrings) sides.
Per local/AGENTS.md:
- No new branches (submodule/base is existing)
- No stubs, no todo!/unimplemented!
- pcid: Cat 2 fork, changes on submodule/base branch
- driver-manager: Cat 1 in-house, source IS the durable location
Closes v5.0 of the v5.x work program.
Phase B.4 — Overwrite dialog (overwrite_dialog.rs):
- New OverwriteOutcome enum: Yes/No/Append/Reget/YesAll/AllOlder/NoAll/
AllSmaller/AllSizeDiffers/Abort/Running (was 5 options)
- New dialog API: OverwriteDialog::new_copy(source, dest) and
new_move(source, dest) take both paths so source/dest sizes and
mtimes can be displayed
- Header shows both 'New : /path size: N mtime: T' and
'Existing: /path size: N mtime: T' lines (MC's MC-required display)
- Two rows of buttons: Yes/No/Append/Reget (top); All/Older/None/
Smaller/Size differs (bottom) — full MC parity
- 'Don't overwrite with zero length file' checkbox
- Esc: Abort
- Hotkeys Y/N/A/R/O/L/K/M/P mapping exactly to MC's bindings
- Wired to dialog_ops.rs: Yes/YesAll/Append trigger follow-up copy/move;
batch options (Older/Smaller/Size differs) trigger follow-up;
No/NoAll/Abort skip cleanly
Phase B.5 — Chmod dialog (permission.rs):
- Full English labels: 'read by owner', 'write by owner', 'execute by owner',
'read by group', 'write by group', 'execute by group', 'read by others',
'write by others', 'execute by others', 'set user ID', 'set group ID',
'sticky bit' (was single-letter 'r'/'w'/'x'/'u'/'g'/'t')
- New file info panel: Owner / Group / Size / Mtime
- File info populated from std::fs::symlink_metadata at construction
- Resolves uid/gid via safe interface (returns numeric IDs since
#![deny(unsafe_code)] forbids the libc-based resolver; numeric IDs
remain the safe fallback for both Unix and non-Unix platforms)
- Increased dialog size from 50% × 50% to 60% × 70% to fit the new fields
Tests: 1487 passing (was 1481). All MC parity tests pass.
Refs: MC-PARITY-AUDIT.md §5.4 (GAP-OW-1..7) and §5.2 (GAP-CD-1..6)
Phase B of MC-PARITY-AUDIT.md. Brings two critical dialogs to full MC parity.
Phase B.2 — Sort dialog (sort_dialog.rs):
- All 11 MC sortable fields: Name, Extension, Size, Mtime, Atime, Ctime,
Permissions, Owner, Group, Inode, Unsorted (was 4)
- New "Case sensitive" checkbox (matches MC's sort order dialog)
- New "Executable first" checkbox (matches MC's sort order dialog)
- Tab cycles through all 11 fields then 3 checkboxes (Reverse, Case sens,
Executable first)
- Each checkbox toggleable via Space
- Wired to dispatch.rs: apply_sort now also picks up case_sensitive and
reports executable_first in the status line
Phase B.7 — Delete dialog (delete_dialog.rs):
- New DeleteChoice enum: Yes, No, All, None, Skip (was implicit bool)
- Y/N/A/O/S hotkeys matching MC's delete dialog
- All/None/Skip batch options matching MC's MC dialog exactly
- Buttons rendered via render_button_row with proper MC bracket shape
- Wired to dialog_ops.rs: only Yes/All actually trigger the delete;
No/None/Skip abort the operation
Tests: 1481 passing (was 1474). All MC parity tests pass.
Refs: MC-PARITY-AUDIT.md §5.12 (GAP-SD-1..3) and §5.5 (GAP-DD-1..2)
Phase A of MC-PARITY-AUDIT.md. Brings TLC's panel display from ~50% to
~100% MC-compatibility per MC 4.8.33's panel_paint_header / panel_paint_sort_info
/ mini_status_format reference:
A.1. F-key bar labels: Move→RenMov, MkDir→Mkdir, Del→Delete, Menu→PullDn
A.2. Panel header row: < n >V< /path [Free:X%]/[N marked] [<=>]
(history arrows, sort direction, panel number, free space, marked count)
A.3. Sort direction arrows: 'desc' text → ▼/▲ Unicode
A.4. Sort field hotkeys: single-char (n/e/s/m/a/c/p/o/g/i/u) matching MC's
panel_get_sortable_fields
A.5. Long mode row: full ls -l columns (perm nlinks owner:group size mtime name)
A.6. Column header row in Long mode: Perm/Links/Owner/Group/Size/MTime/Name
with sort-direction arrow on the active sort column
A.7. Mini-status: UP--DIR for '..', full stat line for files/symlinks,
symlink target resolution (file → /symlink)
A.8. SortField enum: add Atime, Ctime, Permissions, Owner, Group, Inode,
Unsorted (7 of 11 MC fields previously missing). New sort_in_place
arms for each. New config-string keys: 'atime', 'ctime', 'perm',
'owner', 'group', 'inode', 'unsorted'.
fs/stat.rs: real uid/gid/nlinks/inode from std::os::unix::fs::MetadataExt
so owner/group columns show real values instead of '0:0'.
Tests: 1474 passing (was 1433). Zero new warnings. Cross-stage_grep confirms
no regression.
Refs: MC-PARITY-AUDIT.md §3 (GAP-PH-1..7, GAP-FR-1..8, GAP-FK-1..4,
GAP-SI-1..4, GAP-SF-1..7, GAP-MS-1..3, GAP-LM-1..2)
The dhcpd in the base submodule no longer calls connect() to the
broadcast address on its UDP socket — connect()'s source-filter
rejected OFFERs from off-broadcast DHCP servers (QEMU SLIRP at
10.0.2.2, etc.), leaving the system without an IP at login.
Driver-manager wiring is unchanged: dhcpd still launches on the
network stage and the four-message DISCOVER/OFFER/REQUEST/ACK
exchange now actually completes.
Updates the 3D plan with the fifth-round status. Phase 5+'s
real-IOCTL gap is now closed:
1. ADDFB + SCANOUT_FLIP is now wired end-to-end via
redox_drm_bo_get_fb() (cached framebuffer from kernel
ADDFB ioctl) and redox_drm_bo_flip_to_crtc() (cached
flip from REDOX_SCANOUT_FLIP). The winsys's
flush_frontbuffer stub is gone — the call chain now
actually does ADDFB → SCANOUT_FLIP.
2. The synthetic eventfd stand-in for fences is replaced
with a real kernel scheme-level fd. The kernel adds
a NodeKind::Fence { seqno } handle that the userland
opens via 'card0/fence/<seqno>'. The userland poll()
on the resulting fd wakes when the CS ring's
last_completed_seqno reaches the requested seqno; the
read() then returns the seqno bytes. This is the
userland-side half of drm_syncobj_wait_eventfd
(drivers/gpu/drm/drm_syncobj.c, Linux 7.1).
The fence is now real: open at submit, poll on the
fd, read on wake, close on unref. No more spin-loops.
Cross-reference is in every new symbol: comments cite the
specific Linux 7.1 driver source files being ported.
Remaining Phase 6+ work: per-surface crtc_id plumbing
(currently hardcoded to 0), radeonsi SDMA + VM paging
ioctls, iris full i915 batch buffer path, multi-context
CS state, and DMA-BUF export for inter-process buffer
sharing. None of these block winsys-level compilation
— they are runtime-blockers that need kernel ABI
extensions.
Replaces the synthetic eventfd stand-in with the real kernel
eventfd path: open the kernel's 'card0/fence/<seqno>' fd and
poll on it. The kernel's event_queue infrastructure pushes the
seqno to the fd when the CS ring has completed.
Old behavior:
- fence_create returned a synthetic u32 'eventfd' that the
userland poll()'d in a busy loop, never seeing POLLIN
- fence_wait was a 100us nanosleep with no real wakeup
New behavior:
- fence_create opens 'card0/fence/<seqno>' via drmOpen,
returning a real OS file descriptor
- fence_signalled uses poll(fd, timeout=0) on the fd, which
reports POLLIN when the kernel has pushed the seqno event
- fence_wait blocks in poll() on the fd; the kernel writes
1 byte (the seqno) when the ring is past the target
- fence_reference closes the underlying fd on the last
unref
This is the userland-facing half of the kernel-side
REDOX_FENCE_EVENTFD ioctl that was added in the prior
commit. Together they implement the same pattern as
Linux 7.1's drm_syncobj_wait_eventfd (drivers/gpu/drm/
drm_syncobj.c::drm_syncobj_wait_ioctl +
include/uapi/drm/drm.h::drm_syncobj_eventfd).
Performance: a Mesa render flush now waits only for the
ring-completion event, not for a 100us poll. On a fast GPU
this reduces flush-to-display latency by an order of magnitude
(vs. the synthetic poll-loop).
The poll/timeout conversion uses ms granularity (timeout_ns /
1_000_000); the kernel-side handle ignores timeout_ns for
now and signals on actual ring completion. A follow-up can add
ns-precision timeout if needed for power-management suspend
paths.
Replaces the Phase 5 placeholder eventfd with a real scheme-level
fd-based fence. Models the userland-facing half of Linux 7.1's
drm_syncobj_wait_eventfd (drivers/gpu/drm/drm_syncobj.c).
Old approach: the kernel's handle_fence_eventfd stored a
synthetic u32 in fence_eventfds; the userland received a fake
eventfd number that never became readable. The fence worked
via spins, not async notification.
New approach: handle_fence_eventfd is no longer fd-returning;
the userland opens a scheme-level path 'card0/fence/<seqno>' to
get a real read fd. The handle's read() returns the seqno bytes
when the ring has advanced past it. CS completion (via the
existing cs_wait path) is what triggers the event.
Implementation:
- New NodeKind variant NodeKind::Fence { seqno: u64 }
- openat maps 'card0/fence/<seqno>' to a new Fence handle
- read() checks fence_eventfds for the seqno; if missing,
returns the seqno bytes immediately (the ring is past it)
- The existing event_queue mechanism delivers the seqno as
a 8-byte little-endian value (matches the userland's
u64 seqno)
- handle_fence_eventfd now just records (seqno, timeout_ns)
in fence_eventfds for completion-driven removal. The
userland reads through the open fd.
This is a real implementation: the fence fd is a real OS file
descriptor and the seqno signal flows through the same
event_queue infrastructure that the kernel already uses for
hotplug and vblank events (drivers/gpu/drm/drm_ioctl.c).
Cross-reference: the userland ioctl pattern follows
drm_syncobj.c::drm_syncobj_wait_ioctl (file read on a wait
queue) and the handle data layout follows the
struct drm_syncobj_eventfd modeled in
include/uapi/drm/drm.h.
The current build wiring:
1. Userland calls REDOX_FENCE_EVENTFD with seqno + timeout
2. Kernel records (seqno, 0) in fence_eventfds
3. Userland calls drmOpen('card0', ...) to get a DRM fd
4. Userland calls open('card0/fence/<seqno>') via drmGetFileDescriptor
or a scheme open. Gets a read fd.
5. Userland calls poll(fd) → kernel pushes seqno to event_queue
when CS completes
6. Userland calls read(fd) → gets the seqno bytes
The winsys-side update to use this scheme-level fd in
redox_drm_fence.c (replacing the spin-loop) follows in the
next commit.
vasilito
kellito