Integrate x86_64 desktop boot and extend VirtualBox validation

Register ext2 at boot, integrate mode-aware VFS opens and volume cleanup, and queue asynchronous transfers through a persistent worker.

Validation: Pascal compiler check and git diff --check passed. Full build blocked by Docker engine HTTP 500; runtime and Linux interoperability remain unverified.

.drone.yml

@@ -17,7 +17,6 @@ steps:
 
   - name: compile
     image: t3hn3rd/asuro-build:latest
-    pull: always
     depends_on:
       - build-image
     commands:

.gitignore

@@ -1,6 +1,8 @@
 /lib/*.ppu
 /lib/*.a
 /lib/*.o
+/lib/x86_64/
+/bin/x86_64/
 /release/*.svg
 /release/*.iso
 /bin/*.bin
@@ -15,12 +17,9 @@ localenv.json
 dockerout.txt
 AGENTS.md
 lessons_learnt.md
+/plan.md
 *.log
 wasuro
 src/core/core.version.pas
 site/*
 !site/.gitkeep
-/src/boot/dangling_units.inc
-/lib/dangling/
-.moderne
-CLAUDE.md

Dockerfile

@@ -1,5 +1,4 @@
-ARG RECIPES_IMAGE=t3hn3rd/moderne-asuro-recipes:0.1.1@sha256:eaf54188053511dad9c620a8fb98b0f50c41770cd49a0b3d71a0538386522011
+FROM ubuntu:latest
-FROM ${RECIPES_IMAGE}
 
 VOLUME ["/code"]
 
@@ -7,8 +6,8 @@ ENV DEBIAN_FRONTEND=noninteractive
 RUN dpkg --add-architecture i386
 RUN apt-get update && apt-get install -y \
 	curl dos2unix wget git make nasm binutils xorriso grub-pc-bin gcc gcc-multilib \
-	python3 python3-pip python3-pil && \
+	python3 python3-pip python3-pil fp-compiler && \
-	apt-get clean && rm -rf /var/lib/apt/lists/*
+	apt-get clean my room
 
 RUN pip3 install --no-cache-dir --break-system-packages "mkdocs>=1.6,<2" mkdocs-material
 
@@ -21,5 +20,6 @@ RUN curl -sL https://sourceforge.net/projects/freepascal/files/Linux/$FPC_VERSIO
 ADD https://raw.githubusercontent.com/fsaintjacques/semver-tool/master/src/semver /usr/bin/semver
 RUN chmod +x /usr/bin/semver
 WORKDIR /code
+# Sources are mounted at runtime; the CMD below normalizes build scripts.
 ENTRYPOINT ["/bin/bash", "-c"]
 CMD ["find toolchain -name '*.sh' -exec dos2unix {} + 2>/dev/null; find toolchain -name '*.sh' -exec chmod +x {} +; bash toolchain/compile.sh"]

Dockerfile.dockerignore

@@ -1,8 +0,0 @@
-**/.git
-**/.gradle
-**/build
-**/.moderne
-wasuro
-site
-lib
-release

compat/lmemorymanager.pas

@@ -22,12 +22,12 @@ type
     THeapPageHeader = memory.heap.THeapPageHeader;
 
 procedure init;
-function kalloc(size : uint32) : void;
+function kalloc(size : SizeUInt) : void;
-function klalloc(size : uint32) : void;
+function klalloc(size : SizeUInt) : void;
-procedure klfree(address : uint32);
+procedure klfree(address : PtrUInt);
-function kpalloc(address : uint32) : void;
+function kpalloc(address : PtrUInt) : void;
 procedure kfree(area : void);
-function lmm_total_free : uint32;
+function lmm_total_free : SizeUInt;
 function lmm_page_count : uint32;
 
 implementation
@@ -37,22 +37,22 @@ begin
     memory.heap.init;
 end;
 
-function kalloc(size : uint32) : void; inline;
+function kalloc(size : SizeUInt) : void; inline;
 begin
     kalloc := memory.heap.kalloc(size);
 end;
 
-function klalloc(size : uint32) : void; inline;
+function klalloc(size : SizeUInt) : void; inline;
 begin
     klalloc := memory.heap.klalloc(size);
 end;
 
-procedure klfree(address : uint32); inline;
+procedure klfree(address : PtrUInt); inline;
 begin
     memory.heap.klfree(address);
 end;
 
-function kpalloc(address : uint32) : void; inline;
+function kpalloc(address : PtrUInt) : void; inline;
 begin
     kpalloc := memory.heap.kpalloc(address);
 end;
@@ -62,7 +62,7 @@ begin
     memory.heap.kfree(area);
 end;
 
-function lmm_total_free : uint32; inline;
+function lmm_total_free : SizeUInt; inline;
 begin
     lmm_total_free := memory.heap.lmm_total_free;
 end;

doc/planning/memoryallocator.md

@@ -205,7 +205,7 @@ Key constants:
 - `DATA_OFFSET = $10000` — data area starts at 64 KB into each page
 - `LARGE_ALLOC_MAGIC = $FFFFFFFE` — sentinel for klalloc'd multi-page allocations
 
-Interrupt safety: `pushfd; cli; pop saved_flags` (stack-balanced) + `restore_if(saved_flags)`
+Interrupt safety: `pushfd; cli; pop saved_flags` (stack-balanced) + `arch.platform.cpu.unlock(saved_flags)`
 inline procedure that only restores IF via `sti`, avoiding dangerous EFLAGS restoration
 (IOPL, NT, VM, AC bits that caused Bad-TSS on IRET).
 
@@ -256,7 +256,7 @@ list, fixes Search_Page, calls `free_page_at_address`). Skips Root_Page.
 
 | Bug | Root Cause | Fix |
 |-----|-----------|-----|
-| Bad TSS exception during network processing | `pushfd`/`popfd` across separate asm blocks: FPC doesn't track ESP changes in asm, causing EFLAGS corruption (NT=1, IOPL wrong) → Bad TSS on IRET | Stack-balanced `pushfd; cli; pop local_var` + `restore_if()` that only does `sti` if IF was set |
+| Bad TSS exception during network processing | `pushfd`/`popfd` across separate asm blocks: FPC doesn't track ESP changes in asm, causing EFLAGS corruption (NT=1, IOPL wrong) → Bad TSS on IRET | Stack-balanced `pushfd; cli; pop local_var` + `arch.platform.cpu.unlock()` that only does `sti` if IF was set |
 | GPF from SSE MOVAPS in doublebuffer flush | `SIZE_PREFIX=4`→8 gave 8-byte alignment; `MOVAPS` needs 16-byte. `klalloc` returned `page+8` (not 16-aligned) | `SIZE_PREFIX=16` (4B size + 12B pad), klalloc header expanded to 16 bytes |
 | Pages never freed (historic) | `free_page` extracted wrong block index from PDE Address field | Fixed extraction to `Address SHR 10` — page freeing works for the first time |
 

doc/src/driver/storage/ctl/driver.storage.ctl.usb.md

@@ -1,49 +1,85 @@
-# driver.storage.ctl.usb
+<!--
+Copyright 2026 Aaron Hance
+Author: Aaron Hance <[email protected]>
 
-USB Mass Storage Class driver stub.
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
 
-## Overview
+http://www.apache.org/licenses/LICENSE-2.0
 
-This unit registers a USB class driver with the driver manager (`driver.mgr`) to claim USB Mass Storage Class interfaces (class `$08`, subclass `$06` SCSI, protocol `$50` Bulk-Only). When the USB stack enumerates a matching interface, the `load` callback is invoked. Currently the callback only logs a detection message; no I/O infrastructure is established.
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+-->
 
-Full Bulk-Only Transport (BOT) support — including Max LUN query, endpoint setup, CBW/CSW command exchange, and `driver.storage.mgr` device registration — is planned but not yet implemented.
+# USB mass storage
 
-## Boot Registration
+`driver.storage.ctl.usb` implements SCSI transparent commands over USB Mass
+Storage Bulk-Only Transport (interface class/subclass/protocol `08/06/50`).
 
-Registered with `boot.mgr` as `driver.storage.ctl.usb` at the `device` barrier. The barrier order places `device` after `bus`, ensuring `driver.bus.usb.core.init` has run before the storage class driver registers with the driver manager.
+It probes INQUIRY, TEST UNIT READY and READ CAPACITY(10), registers LUN 0 with
+the storage manager, and exposes READ(10), WRITE(10), REQUEST SENSE and
+SYNCHRONIZE CACHE(10). Existing partition discovery and filesystem mounting
+apply to the resulting USB disk. Writes complete only after cache synchronization.
 
-## Dependencies
+## Execution and ownership
 
-- `boot.mgr`
+An attachment task probes and registers each disk after scheduling starts; boot
+initialization never waits on a worker that cannot run yet.
 
-- `driver.mgr`
+Storage dispatch only publishes a request. One persistent `usb.storage` worker
-- `memory.heap`
+performs I/O in task context, with one command active per device and a reusable,
-- `driver.storage.types`
+page-aligned 4 KiB bounce buffer. Caller buffers are never submitted for DMA.
-- `io.syslog`
+Commands are split at the bounce-buffer boundary; all LBA/count and buffer-end
-- `debug.tracer`
+arithmetic is checked before submission. CBW/CSW wire sizes, tags, status,
-- `driver.bus.usb.core`
+residue, and actual transfer lengths are validated.
-- `driver.bus.usb.types`
-- `core.util`, `arch.x86.util`
 
-## Constants
+OHCI batches bulk data into spans of at most 4 KiB per transfer descriptor,
+using the endpoint's hardware toggle carry across packets. A 4 KiB full-speed
+read uses one data descriptor instead of 64. Short packets still halt the chain
+and report the actual byte count; interrupt endpoint packet sizing is unchanged.
 
-| Constant | Value | Description |
+The worker performs reset recovery (class reset, clear bulk-IN halt, clear
-|---|---|---|
+bulk-OUT halt) after malformed status or phase errors. Failed SCSI commands
-| `USB_CLASS_MASS_STORAGE` | `$08` | USB interface class for mass storage |
+request sense and return storage errors, including write protection.
-| `USB_SC_SCSI` | `$06` | SCSI transparent command set subclass |
-| `USB_PROTO_BBB` | `$50` | Bulk-Only (BOT) transport protocol |
 
-## Functions and Procedures
+On disconnect the disk is marked removed immediately. Accepted requests are
+completed with errors, and idle volumes are invalidated/unmounted by the worker;
+busy volumes are retried. Storage-manager records remain stable tombstones.
+Class-driver buffers and context are freed after pending hardware work drains.
 
-### init
+Host controllers currently have no cancellation API. A timed-out transfer is
+therefore quarantined with its private device snapshot and DMA memory. No further
+commands are issued to that device. If hardware never completes, those allocations
+remain retained until reboot; freeing them would permit DMA into reused memory.
+The timeout uses the PIT's actual frequency and a bounded polling fallback.
 
-```pascal
+## Supported scope
-procedure init;
-```
 
-Constructs a `TDeviceIdentifier` matching any USB device with `bInterfaceClass = $08`, `bInterfaceSubClass = $06`, `bInterfaceProtocol = $50`, and registers the `load` callback with `driver.mgr`.
+- LUN 0 on a single-interface device with one bulk-IN and one bulk-OUT endpoint.
+- Direct-access disks with power-of-two sectors from 512 through 4096 bytes.
+- Capacity representable by the existing 32-bit storage-sector API; the
+  READ CAPACITY(16) sentinel is rejected.
+- OHCI, EHCI and xHCI bulk transfers. OHCI/EHCI report actual data lengths and
+  terminate bulk reads on short packets.
 
-## Notes
+UAS, additional LUNs, composite/alternate-setting interfaces, UHCI storage,
+optical devices, and media replacement without USB reconnect are not supported.
+The current xHCI driver lacks endpoint-ring reset support: a halted bulk endpoint
+fails the device closed rather than attempting an unsafe ring reuse.
 
-- The `load` callback currently returns true unconditionally after logging a detection message.
+## Validation
-- Once full BOT support is implemented, `load` should query the Max LUN, allocate endpoint structures, and call `driver.storage.mgr.register_device` for each logical unit.
+
+`UnitTest` is called by the boot test suite and is available as `USBMSCTEST`.
+Tests use a memory-only mock host controller: wire layout, byte order, capacity
+limits, reads, writes, sense errors, short transfers, status stalls, reset ordering,
+and request bounds. They perform no disk I/O.
+
+For hardware/VM checks, boot with a disposable USB disk, confirm `USBMSC` discovery
+and filesystem mounting, test file read/write/read-back, then unplug while idle
+and during I/O. Do not run destructive disk tests against a disk containing data.
+
+Protocol: [USB-IF Bulk-Only Transport 1.0](https://www.usb.org/sites/default/files/usbmassbulk_10.pdf).

doc/src/driver/storage/driver.storage.mgr.md

@@ -1,3 +1,20 @@
+<!--
+Copyright 2026 Aaron Hance
+Author: Aaron Hance <[email protected]>
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+-->
+
 # driver.storage.mgr
 
 Physical storage device registry and I/O dispatch manager.
@@ -14,7 +31,7 @@ The I/O model operates as follows:
 - `storage_read` / `storage_write` are convenience wrappers that build a stack-local `TIORequest` and call `submit_io_wait`.
 - `storage_read_async` / `storage_write_async` build and submit a request with a caller-supplied callback, returning immediately.
 
-Legacy `storage_read_legacy` / `storage_write_legacy` procedure wrappers are retained for callers not yet migrated to the function-returning API.
+Blocking callers use `storage_read` / `storage_write` and check the returned `TError`.
 
 ## Dependencies
 
@@ -205,17 +222,42 @@ function storage_write_async(device : PStorage_Device; addr : uint32; sectors :
 
 Non-blocking write. Analogous to `storage_read_async`.
 
-### storage_read_legacy / storage_write_legacy
-
-```pascal
-procedure storage_read_legacy(device : PStorage_Device; addr : uint32; sectors : uint32; buffer : puint32);
-procedure storage_write_legacy(device : PStorage_Device; addr : uint32; sectors : uint32; buffer : puint32);
-```
-
-Drop-in procedure replacements for callers using the old void-returning API. Forward to `storage_read` / `storage_write` and discard the error code. Marked for removal once all callers are migrated.
-
 ## Notes
 
 - `dispatch_next` is an internal procedure called with interrupts disabled. It dequeues the head of the CFIFO, allocates a heap copy via `ioreq_copy` (for ISR safety), and invokes the appropriate `dispatchRead` or `dispatchWrite` on the device. If neither dispatch pointer is set, the request is failed with `eDeviceNotFound`.
 - The blocking path relies on pointer-dereferenced spin (`puint32(@wait.Done)^`) to prevent FPC from caching the `Done` flag in a register across the `hlt` instruction.
 - Devices without `dispatchRead`/`dispatchWrite` set (i.e., drivers not yet migrated to Phase 2) will have `submit_io` return `eDeviceNotFound`. Legacy drivers must be migrated before they can use the new I/O path.
+
+
+## Bounded performance path
+
+The manager merges up to 32 consecutive queued requests when operation, device,
+LBA range, and virtual buffer range are contiguous. It never searches past the
+queue head or delays a lone request to create a merge. Drivers opt in through
+`mergeMaxSectors`; SATA AHCI advertises 65,536 sectors, while other drivers default
+to zero (disabled). A merged command delivers one completion to each original
+request, in FIFO order. A command error fails every member conservatively.
+
+`storage_get_stats` takes an interrupt-protected snapshot of the device counters.
+`DISK info <index>` displays accepted/completed/failed logical requests, hardware
+commands, merged requests, queue-full rejections, queue and active high-water
+marks, successful read/write sectors, request-size buckets, queue/service ticks,
+and FAT metadata cache hits/misses. Queue ticks are accumulated per logical
+request; service ticks are accumulated per hardware command. Tick values use
+`Counters.c32`, not milliseconds. Counters are unsigned 32-bit values, start at
+device registration, and wrap; use snapshot differences over bounded benchmark
+intervals. Size buckets are <=8, 9..256, and >256 sectors.
+
+FAT32 blocking and async reads share a pooled transfer engine with batches of up
+to four outstanding reads, capped by device queue depth. Destination slices are
+disjoint. On any error, the engine drains every accepted request before releasing
+its context and reports only the contiguous successful byte prefix. Unaligned
+edges and odd DMA addresses use the context's private scratch sector. Writes
+remain sequential. Run maps store cumulative file-cluster offsets; an open-file
+cursor handles sequential locality and binary search handles random seeks.
+
+`STORPERF` runs memory-backed regression checks (also included in boot unit tests)
+for reverse completion order, failures, unaligned reads, run lookup, queue
+rejection, merging limits, and callback accounting. It does not access real disks
+or measure hardware throughput. Compare counters alongside `STORBENCH` and file
+benchmarks on contiguous and fragmented files for actual performance validation.

doc/src/driver/storage/driver.storage.test.md

@@ -58,3 +58,111 @@ Registers the `STORTEST` and `DISKTEST` shell commands with `io.stdio`. Must be
 - `run_tests` exercises `makeAbsolutePathFrom`, `resolvePathFrom`, `changeDirectoryFrom`, `GetDirectoryListingFrom`, `newVirtualDirectory`, and `PathValid` including `.` and `..` traversal.
 - `run_disk_tests` is destructive and will irreversibly wipe the partition table of disk 0. It is never called automatically.
 - The `Assert` helper used in both procedures increments `passed` or `failed` counters and logs the test name on failure; passing tests produce no output to avoid noise.
+
+
+## Raw read benchmark
+
+`STORBENCH` is implemented and registered by `driver.storage.bench`. It performs
+read-only raw-device I/O; it does not invoke the destructive `DISKTEST` path.
+
+```text
+STORBENCH [disk] [MiB]
+STORBENCH list
+STORBENCH run [disk] [MiB] [KiB/request] [depth] [seq|random] [start_lba]
+STORBENCH compare [disk] [MiB]
+STORBENCH help
+```
+
+Manual (`run`) defaults are disk 0, 128 MiB, 1024 KiB/request, sequential reads, and a depth of
+`min(4, hardware depth)` (at least one). Examples:
+
+```text
+STORBENCH 0 128 64 4
+STORBENCH 0 128 4 16 random
+STORBENCH 0 256 1024 1 seq 2048
+STORBENCH compare 0 128
+```
+
+The comparison runs sequential reads at 4, 64, and 1024 KiB/request, each at depths
+1, 4, 16, and 32. The MiB argument applies to each of the twelve cases. Queue depth
+may exceed hardware depth to expose software queuing and request merging.
+
+Limits are 1..16384 MiB, 1..1024 KiB/request, and depth 1..32. Request size must be
+an exact number of device sectors. Buffers are allocated once per case, with a
+separate slice for every outstanding request and a maximum total size of 32 MiB.
+The transfer range is clamped to the device. Without an explicit LBA, reads start
+at sector 2048 when the full requested range fits there; otherwise they start at
+zero. Random mode uses a fixed xorshift seed and samples that range with replacement,
+so total bytes read are bounded but unique coverage is not guaranteed.
+
+Reports include successful bytes, elapsed microseconds, MiB/s, IOPS, accepted and
+completed request counts, peak outstanding requests, average/p95/max completion
+latency, and per-run storage-counter deltas. Latencies are measured from submission
+to callback. The p95 is an upper bound from logarithmic microsecond buckets, not an exact
+microsecond percentile. Runs shorter than 10 ms are excluded from suite speed comparisons.
+No console output occurs inside the measured loop. Any failure stops new work;
+already accepted DMA is drained before buffers are freed, and failed requests
+contribute no bytes or successful IOPS.
+
+Timing prefers the Hyper-V-compatible VM reference counter (100 ns ticks), detected
+through its Hv#1 interface and reference-counter privilege. It needs no calibration.
+Other platforms use an invariant TSC calibrated against two complete RTC seconds,
+with up to three attempts to tolerate transient RTC updates or scheduling delays.
+Calibration requires agreement within 1%; failures include the specific reason.
+Unlike delivered timer interrupts, this clock advances while interrupts are masked.
+Queue submission restores interrupts between requests. Filesystem caches are
+bypassed, but drive and virtual-machine/host caches may still affect results.
+Virtual or sparse disks can report throughput above physical-media speed.
+Device counter deltas may include concurrent I/O from other processes;
+benchmark-owned byte, request, and latency counters do not.
+
+`driver.storage.bench.UnitTest` runs memory-only parser, arithmetic, range,
+queue-depth, completion-order, and error-accounting tests at boot. It never reads
+or writes a real disk. The clock tests also verify that an advertised VM reference
+counter advances with interrupts masked; they never calibrate the RTC at boot.
+
+
+### Automatic suite and table output
+
+`STORBENCH` without arguments, or `STORBENCH auto [disk] [MiB]`, runs six
+read-only profiles. Auto defaults to disk 0 and **32 MiB per profile** (up to
+192 MiB total before device-size clamping). One or two numeric arguments select the auto suite: `STORBENCH 2 512` tests
+disk 2 with 512 MiB per profile. `STORBENCH list` lists disk indices, capacities,
+controllers, sector sizes, and hardware queue depths. Three or more numeric
+arguments retain the legacy manual syntax; `run` selects manual mode explicitly.
+
+| Profile | Request size | Queue depth |
+| --- | --- | --- |
+| Seq large | 1024 KiB | 1 |
+| Seq queued | 1024 KiB | queued |
+| Seq medium | 64 KiB | queued |
+| Random small | 4 KiB | 1 |
+| Random queued | 4 KiB | queued |
+| Random medium | 64 KiB | queued |
+
+Queued depth is the device hardware depth clamped to 4..16, allowing software
+queue testing even on serial devices. A request smaller than a device sector is
+increased to one sector. Each profile samples the same byte-span size, with the
+existing fixed random seed and range clamping rules.
+
+Auto and compare modes print an aligned ASCII table with request KiB, queue
+depth, MiB/s, IOPS, average latency, p95 latency upper bound, and status. Rows are
+printed outside the measured loop. Successful runs shorter than 10 ms show `-` metrics and
+`SHORT`, and are excluded from best-result and queue-gain summaries. Errors show
+`FAILED`, stop the suite after accepted DMA has drained, and leave remaining
+profiles unrun. The summary includes successful profiles, bytes actually read,
+measured time, best sequential throughput, peak random IOPS, and (in auto mode)
+queued/QD1 throughput ratios. These ratios describe the measured cases, including
+any controller-cache effects; they are not a hardware rating.
+
+```text
+STORBENCH
+STORBENCH list
+STORBENCH 2 512
+STORBENCH auto 0 64
+STORBENCH compare 0 32
+```
+
+Profiles reuse the same raw device range. Host/drive caches are not flushed, so
+later profiles may read warm data. These results are observed raw read throughput,
+not a claim of uncached physical-media performance.

doc/src/memory/memory.heap.md

@@ -19,7 +19,7 @@ Every allocation is preceded by a 16-byte prefix: 4 bytes for the unit count, 12
 
 Allocations larger than a single heap page's capacity are handled by `klalloc`, which maps contiguous virtual pages and stores a magic sentinel (`$FFFFFFFE`) in the header so `kfree` can identify them and call `klfree` instead of the normal bitmap path.
 
-`kalloc` and `kfree` disable interrupts for the duration of their critical sections using a stack-balanced `pushfd`/`cli`/`pop` sequence and restore the interrupt flag via `restore_if` on all exit paths.
+`kalloc` and `kfree` disable interrupts for the duration of their critical sections using a stack-balanced `pushfd`/`cli`/`pop` sequence and restore the interrupt flag via `arch.platform.cpu.unlock` on all exit paths.
 
 ## Dependencies
 
@@ -132,7 +132,7 @@ Returns the number of heap pages currently in the linked list.
 
 ## Notes
 
-The `restore_if` helper checks bit 9 (IF) of the saved EFLAGS word and issues `STI` only if interrupts were enabled on entry. It uses `STI` rather than `POPFD` to avoid restoring potentially dangerous EFLAGS bits (IOPL, NT, VM, AC) that could trigger a General Protection Fault or Bad-TSS on the next `IRET`.
+The `arch.platform.cpu.unlock` helper checks bit 9 (IF) of the saved EFLAGS word and issues `STI` only if interrupts were enabled on entry. It uses `STI` rather than `POPFD` to avoid restoring potentially dangerous EFLAGS bits (IOPL, NT, VM, AC) that could trigger a General Protection Fault or Bad-TSS on the next `IRET`.
 
 When `DEBUG_LMM` is defined, `kalloc` allocates one additional unit and writes `GUARD_MAGIC` at the end of the allocation. `kfree` verifies this sentinel before freeing; a mismatch triggers a GPF with a syslog error, catching buffer overruns at free time.
 

doc/toolchain/dangling_units.md

@@ -1,45 +0,0 @@
-# Automatic boot-unit inclusion
-
-Run the usual `docker compose build builder` and `docker compose run builder`
-from Asuro. Only this repository is required. Compose and Drone use the same
-Dockerfile and repository build context, producing `t3hn3rd/asuro-build:latest`.
-The Dockerfile inherits the published `t3hn3rd/moderne-asuro-recipes:0.1.1` image,
-pinned by digest. It already includes Java, Moderne CLI, the Pascal LST generator,
-and the registered recipe. The Asuro layer installs FPC and the kernel build tools.
-
-To upgrade tooling, publish the upstream images first and update `RECIPES_IMAGE`
-in the Dockerfile to the new tag and digest. Local tooling image testing can use
-`docker compose build --build-arg RECIPES_IMAGE=<local-tag> builder`.
-
-`ASURO_ARCH` selects the source architecture, defaulting to `x86`. For example:
-
-```sh
-export ASURO_ARCH=x86
-docker compose run --rm builder
-```
-
-Compose forwards this variable to the builder; Drone inherits it from its compile
-environment. Only `src/arch/$ASURO_ARCH` is added to FPC's unit search paths.
-Other architecture folders are excluded from LST generation and from the recipe's
-dependency graph and results. Common units outside `src/arch` remain eligible.
-The selected architecture folder must exist. This source selection does not
-change the existing i386 compiler target, assembly stub, or linker settings;
-an x86_64 port still needs those platform-specific changes.
-
-`compile_dangling.sh` runs after Wasuro synchronization and before FPC. It copies
-only `src/` and `compat/` into an isolated temporary analysis repository, generates
-LSTs with rewrite-pascal, runs the standalone Moderne recipe, exports its
-datatable, and writes `src/boot/dangling_units.inc`. Wasuro is not parsed.
-Datatable conversion is provided by the standalone project's Java exporter.
-Parse errors, recipe failures, and malformed datatables stop the build. The
-generated include is ignored by Git and removed before each generation attempt.
-
-Use `{$DEFINE NOT_DANGLING}` near a unit declaration to exclude an experimental
-or legacy unit from automatic inclusion. It can still be referenced normally.
-This is a file-level marker; conditional definitions are not evaluated. The
-analysis follows both interface and implementation uses, without expanding
-includes or evaluating conditional compilation.
-
-Moderne runs through its CLI normally; any license or tenant configuration
-required by your CLI installation must be provided to the container separately.
-No host credentials are copied into the image.

docker-compose.yml

@@ -1,8 +1,8 @@
 services:
   builder:
-    image: t3hn3rd/asuro-build:latest
     build: .
     environment:
       ASURO_ARCH: ${ASURO_ARCH:-x86}
+      ASURO_BOOT_TEST: ${ASURO_BOOT_TEST:-none}
     volumes:
       - .:/code

mkdocs.yml

@@ -283,7 +283,6 @@ nav:
     - Version Generation: toolchain/compile_vergen.md
     - LVGL Compilation: toolchain/compile_lvgl.md
     - Wasuro Pull: toolchain/compile_wasuro.md
-    - Automatic Boot Units: toolchain/dangling_units.md
     - Source List: toolchain/compile_sourcelist.md
     - Source Compilation: toolchain/compile_sources.md
     - Checksum: toolchain/compile_checksum.md
@@ -296,6 +295,7 @@ nav:
 
   - Planning:
     - Architecture: planning/architecture.md
+    - x86-64 Migration: planning/x86_64.md
     - USB: planning/USB.md
     - XHCI: planning/XHCI.md
     - FlatFS: planning/flatfs.md

readme.md

@@ -28,8 +28,7 @@ We welcome everyone to give building/breaking/fixing/shooting Asuro a go, feel f
 4. Install Git for Windows.
 5. Install VSCode & the listed plugins.
 6. Install VirtualBox (v7+).
-7. Clone this repository. The builder inherits the published Moderne/Pascal
+7. Clone this repository.
-   tooling image; see [boot-unit generation](doc/toolchain/dangling_units.md).
 8. Run the following command in the root of the repo to build the docker image:
     ```powershell
     docker compose build builder

src/app/app.base64.pas

@@ -22,7 +22,7 @@ unit app.base64;
 interface
 
 uses
-    io.stdio, core.util, arch.x86.util, core.strings, debug.tracer, core.enc.base64, memory.heap;
+    io.stdio, core.util, arch.platform.cpu, core.strings, debug.tracer, core.enc.base64, memory.heap;
 
 procedure init();
 

src/app/app.dhclient.pas

@@ -22,7 +22,7 @@ unit app.dhclient;
 interface
 
 uses
-    io.stdio, core.util, arch.x86.util, core.strings, debug.tracer, driver.net.proto.dhcp;
+    io.stdio, core.util, arch.platform.cpu, core.strings, debug.tracer, driver.net.proto.dhcp;
 
 procedure init();
 

src/app/app.diskcmd.pas

@@ -25,7 +25,7 @@ uses
     core.strings,
     io.stdio,
     debug.tracer,
-    core.util, arch.x86.util,
+    core.util, arch.platform.cpu,
     driver.storage.vol.mgr;
 
 var
@@ -150,6 +150,7 @@ var
     device   : PStorage_Device;
     volCount : uint32;
     freeSec  : uint32;
+    stats : TStorageStats;
 begin
     if paramCount(params) < 2 then begin
         println('Usage: DISK info <disk>');
@@ -191,6 +192,16 @@ begin
     else
         volCount := 0;
     print('Partitions:      '); printint(volCount); println('');
+    storage_get_stats(device,stats);
+    print('I/O submitted/completed/failed: '); printint(stats.Submitted); print(' / '); printint(stats.Completed); print(' / '); printint(stats.Failed); println('');
+    print('Commands / merged requests: '); printint(stats.Commands); print(' / '); printint(stats.Merged); println('');
+    print('Queue peak / active peak / queue full: '); printint(stats.QueuePeak); print(' / '); printint(stats.ActivePeak); print(' / '); printint(stats.QueueFull); println('');
+    print('Read / written sectors: '); printint(stats.ReadSectors); print(' / '); printint(stats.WriteSectors); println('');
+    print('Requests <=8 / <=256 / >256 sectors: '); printint(stats.SmallRequests); print(' / '); printint(stats.MediumRequests); print(' / '); printint(stats.LargeRequests); println('');
+    print('Queue ticks total / max: '); printint(stats.QueueTicks); print(' / '); printint(stats.MaxQueueTicks); println('');
+    print('Service ticks total / max: '); printint(stats.ServiceTicks); print(' / '); printint(stats.MaxServiceTicks); println('');
+    print('FAT cache hits / misses: '); printint(stats.FatCacheHits); print(' / '); printint(stats.FatCacheMisses); println('');
+
 end;
 
 { ---------- DISK WIPE ---------- }
@@ -243,7 +254,7 @@ begin
     { Allocate a zero buffer - 64 sectors at a time }
     bufSects := 64;
     buf := puint32(kalloc(bufSects * device^.sectorSize));
-    memset(uint32(buf), 0, bufSects * device^.sectorSize);
+    memset(PtrUInt(buf), 0, bufSects * device^.sectorSize);
 
     print('Wiping device ');
     printint(idx);